Professional Documents
Culture Documents
Social Research
Qualitat ive and Quanftative Approaches
SECOND EDITIO N
w. Lawrence Neuman
University of Wisconsin- Whitewater
All rights reserved. No part of the material protected by this copyright notice may be reproduced or utilized in CHAPTER!
any form or by any means, electronic or mechanical, including photocopying, recording, or by any information
CHAPTER I
storage and retrieval system, without written permission from the copyright owner.
To obtain permission(s) to use material from this work, please submit a written request to Allyn and Bacon,
Permissions Department, 75 Arlington Street, Boston, MA 02116 or fax your request to 617-848-7320. PART TWCl
Between the time Website information is gathered and then published, it is not unusual for some sites to have CHAPTER ~
closed. Also, the transcription ofURLs can result in typographical errors. The publisher would appreciate
notification where these errors occur so that they may be corrected in subsequent editions. CHAPTER I
CHAPTER'
Cataloging-in-Publication data unavailable at press time.
CHAPTER 11
0-205-48437-9
PART THREE
CHAPTER 11
CHAPTER 12
CHAPTER 1]1
PART FOUl
CHAPTER 14
10 9 8 7 6 5 4 3 2 1 RRD-VA 10 09 08 07 06
PAR TON E Foundations
CHAPTER 1 Doing Social Research
CHAPTER 2 Theory and Social Research
CHAPTER 3 Ethics in Social Research
CHAPTER 4 Reviewing the Scholarly literature and Planning a Study
CHAPTER 5 Qualitative and Quantitative Measurement
CHAPTER 6 Qualitative and Quantitative Sampling
v
Doing Social Research
Introduction
Alternatives to Social Research
Authority
Tradition
Common Sense
Media Myths
Personal Experience
How Science Works
Science
The Scientific Community
The Scientific Method and Attitude
Journal Articles in Science
Steps in the Research Process
Dimensions of Research
Use of Research
Purpose of a Study
Time Dimension in Research
Data Collection Techniques
Conclusion
2 PART ONE / FOUNDATIONS
an exciting process of discovery, but it requires position of authority says it is true or because it
persistence, personal integrity, tolerance for am is in an authoritative publication, you are relying
biguity, interaction with others, and pride in do on authority as a basis for knowledge. Relying
ing quality work. on the wisdom of authorities is a quick, simple,
Reading this book cannot transform you and cheap way to learn something. Authorities
into an expert researcher, but it can teach you to often spend time and effort to learn something,
be a better consumer of research results, help you and you can benefit from their experience and
to understand how the research enterprise works, work.
and prepare you to conduct small research pro J There are also limitations to relying on au
jects. After studying this book, you will be aware thority. First, it is easyto overestimate the exper
of what research can and cannot do, and why tise of other people. You may assume that they
properly conducted research is important. are right when they are not. History is full of past
experts whom we now see as being misinformed.
For example, some "experts" of the past mea
sured intelligence by counting bumps on the
ALTERNATIVES TO SOCIAL
skull; other "experts" used bloodletting to try to
RESEARCH
cure diseases. Their errors seem obvious now,
Unless you are unusual, most of what you know but can you be certain that today's experts will
about the social world is not based on doing so not become tomorrow's fools? Second, authori
cial research. You probably learned most of what ties may not agree, and all authorities may not be
you know using an alternative to social research. equally dependable. Whom should we believe if
It is based on what your parents and other people authorities disagree? Third, authorities may
(e.g., friends, teachers) have told you. You also speak on fields they know little about or be plain
have knowledge based on your personal experi wrong. An expert who is very informed about
ences, the books and magazines you have read, one area may use his or her authority in an un
and the movies and television you have watched. related area. Also, using the halo effect (dis
You may also use plain old "common sense." cussed later), expertise in one area may spill over
More than a collection of techniques, social illegitimately to be authority in a totally different
research is a process for producing knowledge. It area. Have you ever seen television commercials
is a more structured, organized, and systematic where a movie star uses his or her fame as au
process than the alternatives that most of us use thority to convince you to buy a car?We need to
in daily life. Knowledge from the alternatives is ask: Who is or is not an authority?
often correct, but knowledge based on research An additional issue is the misuse of author
is more likely to be true and have fewer errors. ity. Sometimes organizations or individuals
Although research does not always produce per give an appearance of authority so they can con
fect knowledge, compared to the alternatives it is vince others to agree to something that they
much less likely to be flawed. Let us review the might not otherwise agree to. A related situation
alternatives before examining social research. occurs when a person with little training and ex
pertise is named as a "senior fellow" or "adjunct
scholar" in a private "think tank" with an im
Authority
pressive name, such as the Center for the Study
You have acquired knowledge from parents, of X or the Institute on Y Research. Some think
teachers, and experts as well as from books, tele tanks are legitimate research centers, but many
vision, and other media. When you accept are mere fronts created by wealthy special-inter
something as being true because someone in a est groups to engage in advocacy politics. Think
4 PART ONE / FOUNDATIONS
such beliefs has been growing over time as the tive to features that confirm what we think, but
entertainment media give the phenomenon ignore features that contradict it. For example, I TABLE 1-1
more prominence. 3 believe tall people are excellent singers. This may
be because of stereotypes, what my mother told
me, or whatever. I observe tall people and, with
Personal Experience
out awareness, pay particular attention to their
If something happens to you, if you personally singing. I look at a chorus or top vocalist and no
see it or experience it, you accept it as true. Per tice those who are tall. Without realizing it, I no
sonal experience, or "seeing is believing," has a tice and remember people and situations that
strong impact and is a powerful source of reinforce my preconceived ideas. Psychologists Authority
knowledge. Unfortunately, personal experience found that people tend to "seek out" and distort
can lead you astray. Something similar to an op their memories to make them more consistent
tical illusion or mirage can occur. What appears with what they already think."
true may actually be due to a slight error or dis A third error is premature closure. It often
tortion in judgment. The power of immediacy operates with and reinforces the first two errors.
and direct personal contact is very strong. Even Premature closure occurs when you feel you
knowing that, people fall for illusions. Many have the answer and do not need to listen, seek
people believe what they see or personally expe information, or raise questions any longer. Un
Traditiofl
rience rather than what very carefully designed fortunately, most of us are a little lazy or get a lit
research has discovered. tie sloppy. We take a few pieces of evidence or
The four errors of personal experience rein look at events for a short while and then think
force each other and can occur in other areas, as we have it figured out. We look for evidence to
well. They are a basis for misleading people confirm or reject an idea and stop when a small CommonSemr
through propaganda, cons or fraud, magic, amount of evidence is present. In a word, we
stereotyping, and some advertising. The most jump to conclusions. For example, I want to
frequent problem is overgeneralization; it occurs learn whether people in my town support Mary
when some evidence supports your belief, but Smith or Jon Van Horn for mayor. I ask 20 peo
you falsely assume that it applies to many other ple; 16 say they favor Mary, 2 are undecided, and Media Myth
situations, too. Limited generalization may be only 2 favor Jon, so I stop there and believe Mary
appropriate; under certain conditions, a small will win.
amount of evidence can explain a larger situa Another common error is the haloeffect; it is
tion. The problem is that many people general when we overgeneralize from what we accept as Personal
ize far beyond limited evidence. For example, being highly positive or prestigious and let its Experience
over the years, I have known five blind people. strong reputation or prestige "rub off' onto
All of them were very friendly. Can I conclude other areas. Thus, I pick up a report by a person
that all blind people are friendly? Do the five from a prestigious university, say Harvard or
people with whom I happened to have personal Cambridge University. I assume that the author
experience with represent all blind people? is smart and talented and that the report will be
The second error, selective observation, oc excellent. I do not make this assumption about a
curs when you take special notice of some people report by someone from Unknown University. I
or events and tend to seek out evidence that con form an opinion and prejudge the report and HOW SCI
firms what you already believe and ignore con may not approach it by considering its own mer Although it
tradictory information. People often focus on or its alone. How the various alternatives to social natWe,,~oi
observe particular cases or situations, especially research might address the issue of laundry is ..rhat sepaGIIII5
when they fit preconceived ideas. Weare sensi shown in Table 1.1. umllves tlbi. . . .
~ut the sociII
_________m.
TABLE 1.1 Alternatives to Social
CHAPTER 1 / DOING SOCIAL RESEARCH
Research
Science
The term science suggests an image of test tubes,
computers, rocket ships, and people in white lab
coats. These outward trappings are a part of sci
ence, especially natural science (i.e., astronomy,
Authority Experts say that as children,
biology, chemistry, geology, and physics,), that
females are taught to make,
deals with the physical and material world (e.g.,
select, mend, and clean clothing
plants, chemicals, rocks, stars, and electricity).
as part of a female focus on
physical appearance and on
The social sciences, such as anthropology, psy
caring for children or others in a chology, political science, and sociology, involve
family. Women do the laundry the study of people-their beliefs, behavior, in
based on their childhood teraction, institutions, and so forth. Fewer peo
preparation. ple associate these disciplines with the word
Tradition Women have done the laundry
science. Science is a social institution and a way
for centuries, so it is a
to produce knowledge. Not everyone is well in
continuation of what has formed about science. For example, a 2001 sur
happened for a long time. vey found that about only one-third of U.S.
adults could correctly explain the basics of
Common Sense Men just are not as concerned
science.f
about clothing as much as
women, so it only makes sense
Scientists gather data using specialized tech
that women do the laundry
niques and use the data to support or reject the
more often. ories. Data are the empirical evidence or
information that one gathers carefully accord
Media Myth Television commercials show
ing to rules or procedures. The data can be
women often doing laundry and
quantitative (i.e., expressed as numbers) or
enjoying it, so they do laundry
because they think it's fun.
qualitative (i.e., expressed as words, visual im
ages, sounds, or objects). Empirical evidence
Personal My mother and the mothers of
refers to observations that people experience
Experience all my friends did the laundry.
through the senses-touch, sight, hearing, smell,
My female friends did it for their
and taste. This confuses people, because re
boyfriends, but never the other
searchers cannot use their senses to directly ob
way around. It just feels natural
for the woman to do it.
serve many aspects of the social world about
which they seek answers (e.g., intelligence, atti
tudes, opinions, feelings, emotions, power, au
thority, etc.). Researchers have many specialized
techniques to observe and indirectly measure
HOW SCIENCE WORKS
such aspects of the social world.
Although it builds on some aspects of the alter
native ways of developing knowledge, science is
The Scientific Community
what separates social research. Social research
involves thinking scientifically about questions Science comes to life through the operation of
about the social world and following scientific the scientific community, which sustains the as
8 PART ONE / FOUNDATIONS
sumptions, attitudes, and techniques of science. vate industry in organizations such as the Na tJ5CS,. ~ ~11111111
The scientific community is a collection of people tional Opinion Research Center and the Rand vithin~
who practice science and a set of norms, behav Corporation. Most, however, work at the ap
iors, and attitudes that bind them together. It is a proximately 200 research universities and insti
~ofb"'"
~ OIl pnIlil:!lliil
professional community-a group of interacting tutes located in a dozen advanced industrialized ~. aalliliil
people who share ethical principles, beliefs and countries. Thus, the scientific community is ~h*-»-
values, techniques and training, and career paths. scattered geographically, but its members tend such as boIlaI~
For the most part, the scientific community in to work together in small clusters. search, ~
cludes both the natural and social sciences.? How big is the scientific community? This is one roodw1nt
Many people outside the core scientific
community use scientific research techniques. A
not an easy question to answer. Using the broad
est definition (including all scientists and those
its of the
teristics of
1l5I:'"
range of practitioners and technicians apply re in science-related professions, such as engi
search techniques that scientists developed and neers), it includes about 15 percent of the labor
refined. Many use the research techniques (e.g., force in advanced industrialized countries. A
a survey) without possessing a deep knowledge better way to look at the scientific community is
of scientific research. Yet, anyone who uses the to examine the basic unit of the larger commu
techniques or results of science can do so better nity: the discipline (e.g., sociology, biology, psy
if they also understand the principles and chology, etc.). Scientists are most familiar with a
processes of the scientific community. particular discipline because knowledge is spe
The boundaries of the scientific community cialized. Compared to other fields with ad
and its membership are defined loosely. There is vanced training, the numbers are very small. For ence or a rrxfti..
no membership card or master roster. Many example, each year, about 500 people receive
people treat a Ph.D. degree in a scientific field as Ph.D.s in sociology, 16,000 receive medical de
an informal "entry ticket" to membership in the grees, and 38,000 receive law degrees.
scientific community. The Ph.D., which stands A discipline such as sociology may have
for doctorate of philosophy, is an advanced about 8,000 active researchers worldwide. Most
graduate degree beyond the master's that pre researchers complete only two or three studies
pares one to conduct independent research. in their careers, whereas a small number of
Some researchers do not have Ph.D.s and not all highly active researchers conduct many dozens
those who receive Ph.D.s enter occupations in ofstudies. In a specialty or topic area (e.g., study
which they conduct research. They enter many of the death penalty, social movements, di
occupations and may have other responsibilities vorce), only about 100 researchers are very ac
(e.g., teaching, administration, consulting, clin tive and conduct most research studies.
ical practice, advising, etc.). In fact, about one Although research results represent what hu
half of the people who receive scientific Ph.D.s manity knows and it has a major impact on the
do not follow careers as active researchers. lives of many millions of people, only a small
At the core of the scientific community are number of people are actually producing most
researchers who conduct studies on a full-time new scientific knowledge.
or part-tirne basis, usually with the help of assis
tants. Many research assistants are graduate stu
The Scientific Method and Attitude
dents, and some are undergraduates. Working
as a research assistant is the way that most scien You have probably heard of the scientific
tists gain a real grasp on the details of doing re method, and you may be wondering how it fits
search. Colleges and universities employ most into all this. The scientific method is not one sin
members of the scientific community's core. gle thing; it refers to the ideas, rules, techniques,
Some scientists work for the government or pri and approaches that the scientific community
CHAPTER 1 / DOING SOCIAL RESEARCH 9
uses. The method arises from a loose consensus regularly reject half of the submissions. Thus,
within the community of scientists. It includes a several experienced researchers screen a journal
way of looking at the world that places a high article based on its merits alone, and publication
value on professionalism, craftsmanship, ethical represents the study's tentative acceptance by the
integrity, creativity, rigorous standards, and dili scientific community as a valid contribution to
gence. It also includes strong professional norms knowledge. Unlike the authors of articles for the
such as honesty and uprightness in doing re popular magazines found at newsstands, scien
search, great candor and openness about how tists are not paid for publishing in scholarly jour
one conducted a study, and a focus on the mer nals. In fact, they may have to pay a small fee to
its of the research itself and not on any charac help defray costs just to have their papers consid
teristics of individuals who conducted the study. ered. Researchers are happy to make their re
search available to their peers (i.e., other
scientists and researchers) through scholarly
Journal Articles in Science
journals. The article communicates the results of
Consider what happens once a researcher fin a study that a researcher might have devoted
ishes a study. First, he or she writes a detailed de years of his or her life to, and it is the way re
scription of the study and the results as a searchers gain respect and visibility among their
research report or a paper using a special format. professional peers. Likewise, the reviewers are
Often, he or she also givesan oral presentation of not paid for reviewing papers, but consider it an
the paper before other researchers at a confer honor to be asked to conduct the reviews and to
ence or a meeting of a professional association carry out one of the responsibilities of being in
and seeks comments and suggestions. Next, the the scientific community. The scientific commu
researcher sends several copies to the editor of a nity imparts great respect to researchers who
scholarly journal. Each editor, a respected re publish many articles in the foremost scholarly
searcher chosen by other scientists to oversee the journals because these researchers are directly
journal, removes the title page, which is the only advancing the scientific community's primary
place the author's name appears, and sends the goal-the accumulation of carefully developed
article to several reviewers. The reviewers are re knowledge. A researcher gains prestige and
spected scientists who have conducted studies in honor and a reputation as an accomplished re
the same specialty area or topic. The reviewers searcher through such publications.
do not know who did the study, and the author You may never publish an article in a schol
of the paper does not know who the reviewers arly journal, but you will probably read many
are. This reinforces the scientific principle of such articles. It is important to see how they are
judging a study on its merits alone. Reviewers a vital component in the system of scientific re
evaluate the research based on its clarity, origi search. Researchers actively read what appears in
nality, standards of good research methods, and the journals to learn about new research findings
advancing knowledge. They return their evalua and the methods used to conduct a study. Even
tions to the editor, who decides to reject the pa tually, the new knowledge is disseminated in
per, ask the author to revise and resubmit it, or textbooks, new reports, or public talks.
accept it for publication. It is a very careful, cau
tious method to ensure quality control.
The scholarly journals that are highly re
STEPS IN THE RESEARCH
Process
..
1I'lochnI..__
she clew I.A II
SocilII . . .
sizes. Btbr~
powerful corporate elites. A topic is too broad 1. Select Topic make~.
for conducting a study. This makes the next step resean:h~ . .
crucial. You must then narrow down the topic,
or focus the topic into a specific research ques
7. Inform
Others 2. Focus
need to"'"
vantages oJfall
tion for a study (e.g., "Are people who marry
~-~
Question
younger more likely to engage in physical abuse the ~])(5."
of a spouse under conditions of high stress than eachoftOor.
those who marry older?"). As you learn about a 6. Interpret ....c: --- Tbe&Jr"
topic and narrow the focus, you should review Data '"Jl.... 3. Design research is ~
past research, or the literature, on a topic or Study
search. The-llll
mi~~3
question. You also want to develop a possible
answer, or hypothesis, and theory can be impor next two
tant at this stage.
After specifying a research question, you
have to develop a highly detailed plan on how
you will carry-6·lit the study. Th.iS third step re
5. Analyze
Data
specific3
is incorpoI3lal
The .
da
mensions an:
quires that you de~gl~Q~14e_many practical de of a study and a~
tailsgfsl,ojpgJheJesearch (e.g., whether to use a ing an end. The seven steps are for one research you learn the ~
surveyor qualitative observing in the field, how project; it is one cycleof going through the steps howthep .
many subjects to use, etc.). It is only after com in a single study on a specific topic. want to inv
pleting the design stage that you are ready to Science is an ongoing enterprise that builds with certain
gather the data or evidence (e.g., ask people the on prior research and builds a larger, collectively lecting data. In
questions, record answers, etc.). Once you have created body of knowledge. Anyone study is a mensions of
very carefully collected the data, your next step is small part of the much larger whole of science. A understand the
to manipulate or analyze the data. This will help single researcher may be working on multiple
you see any patterns in it and help you to give research projects at once, or several researchers
meaning to or interpret the data (e.g., "People may collaborate on one project. Likewise, one Use of R.eselill1l
who marry young and grew up in families with project may result in one scholarly article or sev For over a
abuse have higher rates of physical domestic eral, and sometimes several smaller projects are Some reseal'
abuse than those with different family histo reported in a single article. entific, and
ries"). Finally, you must inform others by writing more activist. ~
a report that describes the study's background, oriented. This ~
how you conducted it, and what you discovered. searchers in the ~
DIMENSIONS OF RESEARCH
The seven-step process shown in Figure 1.1 tain friendly rdl
is oversimplified. In practice, you will rarely Three years after they graduated from college. from one wing ~
complete one step totally then leave it behind to Tim and Sharon met for lunch. Tim asked their careers. In_
move to the next step. Rather, the process is in Sharon, "So, how is your new job as a researcher concentrate on!
teractive in which the steps blend into each for Social Data, Inc.? What are you doing?" over the long •
other. What you do in a later step may stimulate Sharon answered, "Right now I'm working on studies to SONel
you to reconsider and slightly adjust your think an applied research project on day care quality in Those who CODIII
ing in a previous one. The process is not strictly which we're doing a cross-sectional survey to get dam ental natura
linear and may flow back and forth before reach descriptive data for an evaluation study." Sharon basic research..
CHAPTER 1 / DOING SOCIAL RESEARCH 11
touched on four dimensions of social research as Basic Research. Basic social research advances
she described her research on day care. fundamental knowledge about the social world.
Social research comes in several shapes and Basic researchers focus on refuting or support
sizes. Before you begin a study, you will need to ing theories that explain how the social world
make several decisions about the specific type of operates, what makes things happen, why social
research you are going to conduct. Researchers relations are a certain way, and why society
need to understand the advantages and disad changes. Basic research is the source of most new
vantages of each type, although most end up scientific ideas and ways of thinking about the
specializing in doing one type. We can think of world. Many nonscientists criticize basic re
the types as fitting into one of the categories in search and ask, "What good is it?" and consider
each of four dimensions of research. it to be a waste oftime and money. Although ba
The first dimension is a distinction in how sic research often lacks a practical application in
research is used, or between applied and basic re the short term, it provides a foundation for
search. The next is the purpose of doing research, knowledge that advances understanding in
or its goal, to explore, describe, or explain. The many policy areas, problems, or areas of study.
next two dimensions are more specific: how time Basic research is the source of most of the tools,
is incorporated into the study design, and the methods, theories, and ideas about underlying
specific data collection technique used. causes of how people act or think used by ap
The dimensions overlap, in that certain di plied researchers. It provides the major break
mensions are often found together (e.g., the goal throughs that significant advances in knowledge;
of a study and a data collection technique). Once it is the painstaking study of broad questions
you learn the dimensions, you will begin to see that has the potential of shifting how we think
how the particular research questions you might about a wide range of issues. It may have an im
want to investigate tend to be more compatible pact for the next 50 years or century. Often, the
with certain ways of designing a study and col applications ofbasic research appear many years
lecting data. In addition, being aware of the di or decades later. Practical applications may be
mensions of research will make it easier to apparent only after many accumulated advances
understand the research reports by others. in basic knowledge build over a long time pe
riod. For example, in 1984, Alec Jeffreys, a ge
neticist at the University of Leicester in England,
Use of Research
was engaged in basic research studying the evo
For over a century, science has had two wings. lution ofgenes. As an indirect accidential side ef
Some researchers adopt a detached, purely sci fect of a new technique he developed, he
entific, and academic orientation; others are discovered a way to produce what is now call hu
more activist, pragmatic, and interventionist man DNA "fingerprints" or unique markings of
oriented. This is not a rigid separation. Re the DNA of individuals. This was not his intent.
searchers in the two wings cooperate and main He even said he would have never thought ofthe
tain friendly relations. Some individuals move technique if DNA fingerprints had been his goal.
from one wing to another at different stages in Within 10 years applied uses of the technique
their careers. In simple terms, some researchers were developed. Today, DNA analysis is a widely
concentrate on advancing general knowledge used technique in criminal investigations.
over the long term, whereas others conduct
studies to solve specific, immediate problems. Applied Research. Applied socialresearch is de
Those who concentrate on examining the fun signed to address a specific concern or to offer
damental nature of social reality are engaged in solutions to a problem identified by an em
basic research. ployer, club, agency, social movement, or orga
-~----~~~~~------~--------------
nization. Applied social researchers are rarely wisely.Sometimes despite serious problems with
concerned with building, testing, or connecting a study's methodology and cautions from the re
to a larger theory, developing a long-term gen searchers, politicians use results to justify cutting
eral understanding, or carrying out a large-scale programs they dislike or to advance programs
investigation that might span years. Instead, they they favor. Because applied research often has
usually conduct a quick, small-scale study that immediate implications or involves controver
provides practical results for use in the short sial issues, it often generates conflict. One
term (i.e., next month or next year). For exam famous researcher, William Whyte (1984), en
ple, the student government of University X countered conflict over findings in his applied
wants to know whether the number of Univer research on a factory in Oklahoma and on
sity X students who are arrested for driving while restaurants in Chicago. In the first case, the
intoxicated or involved in auto accidents will de management was more interested in defeating a
cline if it sponsors alcohol-free parties next year. union than in learning about employment rela
Applied research would be most applicable for tions; in the other, restaurant owners really
this situation. sought to make the industry look good and did
People employed in businesses, government not want findings on the nitty-gritty of its oper
offices, health care facilities, social service agen ations made public.
cies, political organizations, and educational in Applied and basic researchers adopt differ
stitutions often conduct applied research and ent orientations toward research methodology
use the results in decision making. Applied re (see Table 1.2). Basic researchers emphasize high
search affects decisions such as the following: methodological standards and try to conduct
Should an agency start a new program to reduce near-perfect research. Applied researchers must
the wait time before a client receives benefits? make more tradeoffs. They may compromise
Should a police force adopt a new type of re scientific rigor to get quick, usable results, but
sponse to reduce spousal abuse? Should a politi compromise is never an excuse for sloppy re
cal candidate emphasize his or her stand on the search. Applied researchers try to squeeze re
environment instead of the economy? Should a search into the constraints of an applied setting tals, goveTnllDelnr,J
company market a skin care product to mature and balance rigor against practical needs. Such demonstrate the
adults instead of teenagers? balancing requires an in-depth knowledge of re doing. An es
The scientific community is the primary search and an awareness of the consequences of techniques .
consumer of basic research. The consumers of compromising standards. researchers. The
applied research findings are practitioners such decision makers.
as teachers, counselors, and social workers, or Types of Applied Research. There are many themselves, d
decision makers such as managers, agency ad specific types of applied research. Here, you will research. Also~·
ministrators, and public officials. Often, some learn about three major types: evaluation, ac a practical si .
one other than the researcher who conducted tion, and social impact assessment. Evaluation
the study uses the results. elude: Does a ~
Applied research results are less likely to en Evaluation Research Study. Evaluation research improve learning ~
ter the public domain in publications and may study is applied research designed to find out forcement pr~
be available only to few decision makers or prac whether a program, a new way of doing some spouse abuse? ~
titioners. This means that applied research find thing, a marketing campaign, a policy, and so employee prod~
ings often are not widely disseminated and that forth, is effective-in other words, "Does it measure the effedll
well-qualified researchers rarely get to judge the work?" The most widely used type of applied re or way of doing ~
quality of applied studies. search is evaluation research." This type of re research technique
The decision makers who use the results of search is widely used in large bureaucratic can be used, the e:IIj
an applied study mayor may not use them organizations (e.g., businesses, schools, hospi ally preferred. Prad
CHAPTER 1 ! DOING SOCIAL RESEARCH 13
----------------------~--.
TAB L E 1.2 Basic and Applied Social Research Compared
5. The driving goal is to contribute to basic, 4. The primary concern is with the ability to
theoretical knowledge. generalize findings to areas of interest to
sponsors.
6. Success comes when results appear in a
scholarly journal and have an impact on others 5. The driving goal is to have practical payoffs or
in the scientific community. uses for results.
6. Success comes when results are used by
sponsors in decision making.
tals, government, large nonprofit agencies) to icy or program may conduct evaluation research
demonstrate the effectiveness of what they are for their own information or at the request of
doing. An evaluation researcher does not use outside decision makers. The decision makers
techniques different from those of other social may place limits on the research by fixing
researchers. The difference lies in the fact that boundaries on what can be studied and by de
decision makers, who may not be researchers termining the outcome of interest. This often
themselves, define the scope and purpose of the creates ethical dilemmas for a researcher.
research. Also, their objective is to use results in Ethical and political conflicts often arise in
a practical situation.f evaluation research because people can have op
Evaluation research questions might in posing interests in the findings. The findings of
clude: Does a Socratic teaching technique research can affectwho gets or keeps a job, it can
improve learning over lecturing? Does a law-en build political popularity, or it may help pro
forcement program of mandatory arrest reduce mote an alternative program. People who are
spouse abuse? Does a flextime program increase personally displeased with the findings may at
employee productivity? Evaluation researchers tack the researcher or his or her methods.
measure the effectiveness of a program, policy, Evaluation research has several limitations:
or way of doing something and often use several The reports of research rarely go through a peer
research techniques (e.g., survey and field). If it review process, raw data are rarely publicly avail
can be used, the experimental technique is usu able, and the focus is narrowed to select inputs
ally preferred. Practitioners involved with a pol and outputs more than the full process hywhich
14 PART ONE I FOUNDATIONS
,
a program affects people's lives. In addition, de
cision makers may selectivelyuse or ignore eval
mestic violence that will be discussed shortly as
an explanatory study example (Cherlin et al., ., ,
uation findings. 2004) testified in the United States Senate. The
study findings and the testimony helped to alter
Action Research Study. Action research is ap marriage promotion provisions in a 2005 wel
plied research that treats knowledge as a form of fare reform law.?
power and abolishes the division between creat
ing knowledge and using knowledge to engage in Social Impact Assessment Research Study. A re
political action. There are several types of action searcher who conducts social impact assessment
research, but most share five characteristics: (1) (SIA) estimates the likely consequences of a
the people being studied actively participate in planned intervention or intentional change to
the research process; (2) the research incorpo occur in the future. It may be part ofa larger en- .
rates ordinary or popular knowledge; (3) the re vironmental impact statement required by gov
search focuses on issues of power; (4) the ernment agencies and used for planning and
research seeks to raise consciousness or increase making choices among alternative policies. He
awareness of issues; and (5) the research is tied or she forecasts how aspects of the social envi
directly to a plan or program of political action. ronment may change and suggests ways to miti
Action research tends to be associated with a so gate changes likely to be adverse from the point
cial movement, political cause, or advocacy for ofview of an affected population. Impacts are the
an issue. It can be conducted to advance a range difference between a forecast of the future with
of political positions. Some action research has the project or policy and without the project or
an insurgent orientation with goals of empower policy. For example, the SIA might estimate the
ing the powerless, fighting oppression and injus ability of a local hospital to respond to an earth
tice, and reducing inequality. Wealthy and quake, determine how housing availability for
powerful groups or organizations also sponsor the elderly will change if a major new highway is
and conduct action research to defend their sta built, or assessthe impact on college admissions
tus, position, and privileges in society. if students receive interest-free loans. Re
Most action researchers are explicitly politi searchers who conduct SIAs often examine a
cal, not value neutral. Because the primary goal range of social outcomes and work in an inter
is to affect sociopolitical conditions, publishing disciplinary research team to estimate the social
results in formal reports, articles, or books is sec outcomes. The outcomes include measuring
ondary. Most action researchers also believe that "quality of life" issues, such as access to health •
knowledge develops from direct experience, par care, illegal drug and alcohol use, employment
concerns.
ticularly the experience of engaging in sociopo opportunities, schooling quality, teen pregnancy
litical action. rates, commuting time and traffic congestion, • Create a g
For example, most feminist research is ac availability of parks and recreation facilities, picture ofe
tion research. It has a dual mission: to create so shopping choices, viable cultural institutions, • Formulate and ~
cial change by transforming gender relations and crime rates, interracial tensions, or social isola questions few' ~
to contribute to the advancement of knowledge. tion. There is an international professional asso • Generate new ~
A feminist researcher who studies sexual harass ciation for SIA research that advances SIA conjectures.. Of" ~
ment might recommend policy changes to re techniques and promotes SIA by governments, • Determine the til
duce it as well as to inform potential victims so corporations, and other organizations. conducting ~
they can protect themselves and defend their Social impact assessments are rarely re • Develop technici
rights. At times, researchers will explain study quired, but a few governments mandate them. measuring and Id
results in a public hearing to try to modify new For example, in New South Wales, Australia, a data.
policies or laws. The authors of a study on do registered club or hotel cannot increase the
CHAPTER 1 / DOING SOCIAL RESEARCH 15
number of poker machines unless the Liquor lence), economic (e.g., unemployment, bank
Administration Board in the Department Gam ruptcy, tourism expansion), and cultural im
ing and Racing approves an SIA for the club or pacts (e.g.,time awayfrom other leisure activity)
hotel. The SIA enables the board to assess the listed by their effect on all gamblers, problem
likely local community impact from increasing gamblers, the local community, and the
the number of poker machines. The format in region. 10
cludes a matrix that allows the board to identify
the social and economic impacts, positive and
Purpose of a Study
negative, financial or nonfinancial, quantified or
qualitative. In New Zealand, the Gambling Act If you ask someone why he or she is conducting
of 2003 requires an SIA before expanding gam a study, you might get a range of responses: "My
bling. In one 2004 study in New Zealand for the boss told me to"; "It was a class assignment"; "I
Auckland City Council, it noted that 90 percent was curious"; "My roommate thought it would
of New Zealand's adults gamble, 10 percent gam be a good idea." There are almost as many rea
ble regularly (once a week or more often), and sons to do research as there are researchers. Yet,
about 1 percent are problem gamblers, although the purposes of social research may be organized
this varies by age, income, and ethnicity. The into three groups based on what the researcher is
SIA recommended limiting the locations of new trying to accomplish-explore a new topic, de
gambling venues, monitoring their usage, and scribe a social phenomenon, or explain why
tracing the amount of gambling revenues that something occurs. Studies may have multiple
are returned to the community in various ways purposes (e.g., both to explore and to describe),
(e.g., clubs, trusts, etc.). It contained a matrix but one of three major purposes is usually dom
with social (e.g, arrests, divorce, domestic vio inant (see Box 1.2).
Exploration. Perhaps you have explored a new Description. Perhaps you have a more highly .-Idrca ......
topic or issue in order to learn about it. If the is developed idea about a social phenomenon and J*x. .~...,
sue was new or no researchers had written about want to describe it. Descriptive research presents pIIICdS;~. . .
it, you began at the beginning. In exploratory re a picture of the specific details of a situation, so IOpercaaai..
search, a researcher examines a new area to for cial setting, or relationship; it focuses on "how?"
D!e"a-_~"'''
.:ooditiom __
mulate precise questions that he or she can and "who?" questions: "How did it happen?"
address in future research. Exploratory research "Who is involved?" A great deal of social re pb:oatory~
may be the first stage in a sequence of studies. A search is descriptive. Descriptive researchers use ears are ......
researcher may need to conduct an exploratory most data-gathering techniques-surveys, field Cberiin, . . . .
study in order to know enough to design and ex research, content analysis, and historical-com ~1~~~
.t-~JPR
• 'iii
._
ecute a second, more systematic and extensive parative research. Only experimental research is usinga_~
study. It addresses the "what?" question: "What less often used. Much of the social research phvsical abase.. 1
is this social activity really about?" found in scholarly journals or used for making women ,.ith ~
Many higher-education officials are con policy decisions is descriptive. likely IIJ.alIT
cerned about college students' low retention Descriptive and exploratory research often The authors
rates, especially students from minority-disad blur together in practice. In descriptive research, abused~
vantaged social backgrounds. For example, of a researcher begins with a well-defined subject to resist 01"
Latinos who enroll in college, 80 percent leave and conducts a study to describe it accurately more likely 10
without receiving a degree. Officials seek ways to and the outcome is a detailed picture of the sub guilt, and low"
reduce dropouts and increase the chances that ject. The results may indicate the percentage of rionofh~.
students who begin college will stay until they people who hold a particular view or engage in sive experience l
earn a degree. Garza and Landeck (2004) con specific behaviors-for example, that 8 percent distance and a I
ducted an exploratory study of over 500 Latino ofparents physically or sexually abuse their chil commitments, ..
students at a college along the Texas-Mexico dren. A descriptive study presents a picture of rive data gatheraj
border who had dropped out. They wanted to types of people or of social activities. inthreeci~
learn the influencing factors and rationales in Stack, Wasserman, and Kern (2004) con nio-thev~
student decision making. The authors discovered ducted a descriptive study on pornography use perienced P'& l
that the primary factors and rationales were un on the Internet by people in the United States. married,~~
related to teaching quality or university services. They found that the greatest users were those weremost.~
Instead, the students who dropped out had been with weak social bonds. More specifically, the
womenWI3
overwhelmed by personal problems or had seri types of people who were adult users of pornog found the
ous difficulties with family or job responsibilities. raphy tended to be males with unhappy mar an adult were
Such factors were a major reason given by over riages and weak ties to organized religion. women who half
80 percent of the students who dropped out. Pornography users were also more likely to have less likely to lea1lll
Exploratory researchers tend to use qualita engaged in nonconventional sexual behavior ries of unstable. •
tive data and not be wedded to a specific theory (i.e., had an extramarital affair or engaged in
or research question. Exploratory research rarely paid sex) but not other forms of deviance, such
yields definitive answers. If you conduct an ex as illegal drug use. TimeDim~
I
ploratory study, you may get frustrated and feel An awareness of!
it is difficult because there are few guidelines to Explanation. When you encounter an issue mension will hdf
follow. Everything is potentially important, the that is well recognized and have a description of This is because diiI
steps are not well defined, and the direction of it, you might begin to wonder why things are the sues incorporate!
inquiry changes frequently. You need to be cre way they are. Explanatory research identifies the give a snapshot d
ative, open-minded, and flexible; adopt an in sources of social behaviors, beliefs, conditions, allow you to anal)
vestigative stance; and explore all sources of and events; it documents causes, tests theories, Other studies pI1l
information. and provides reasons. It builds on exploratory you follow eYed
CHAPTER 1 / DOING SOCIAL RESEARCH 17
and descriptive research. For example, an ex over several time points (longitudinal). Quanti
ploratory study discovers a new type of abuse by tative studies generally Look at many cases, peo
parents; a descriptive researcher documents that ple, or units, and measure limited features about
10 percent of parents abuse their children in this them in the form of numbers. By contrast, a
new way and describes the kinds of parents and qualitative study usually involves qualitative
conditions for which it is most frequent; the ex data and examines many diverse features of a
planatory researcher focuses on why certain par small number of cases across either a short or
ents are abusing their children in this manner. long time period (see Figure 1.2).
Cherlin, Burton, Hurt, and Purvin (2004) ex
plained instability in marriage or cohabitation Cross-Sectional Research. Most social re
using a woman's past experience with sexual or search studies are cross-sectional; they examine a
physical abuse. They tested the hypothesis that single point in time or take a one-time snapshot
women with a history of abuse would be less approach. Cross-sectional research is usually the
likely marry than those without such histories. simplest and least costly alternative. Its disad
The authors reasoned that those who were vantage is that it cannot capture social processes
abused have fewer social supports and resources or change. Cross-sectional research can be ex
to resist or avoid abusive partners, and they are ploratory, descriptive, or explanatory, but it is
more likely to harbor feelings of self-blame, most consistent with a descriptive approach to
guilt, and low self-esteem that inhibit the forma research. The descriptive study by Stack,
tion of healthy romantic relationships. An abu Wasserman, and Kern (2004) on pornography
sive experience also creates greater emotional use was cross-sectional, based on a national U.S.
distance and a hesitancy to make long-term survey conducted in 2000.
commitments. Using quantitative and qualita
tive data gathered in Low-income neighborhoods Longitudinal Research. Researchers using
in three cities-Boston, Chicago, and San Anto longitudinal research examine features of people
nio-they found that adult women who had ex or other units at more than one time. It is usually
perienced past abuse were less likely to be more complex and costly than cross-sectional
married, and those with multiple forms of abuse research, but it is also more powerful and infor
were most likelyto remain single, It appears that mative. Descriptive and explanatory researchers
women without a past history of abuse who use longitudinal approaches. Let us now look at
found themselves in an abusive relationship as the three main types of longitudinal research:
an adult were likely to withdraw from it, but time series, panel, and cohort.
women who had been abused as children were
less likely to leave and tended to enter into a se Time-Series Study. A time-series study is longi
ries of unstable, transitory relations. tudinal research in which a researcher gathers
the same type of information across two or more
time periods. Researchers can observe stability
Time Dimension in Research
or change in the features of the units or can track
An awareness of how a study uses the time di conditions over time. The specific individuals
mension will help you read or conduct research. may change but the overall pattern is clear. For
This is because different research questions or is example, there has been a nationwide survey of a
sues incorporate time differently. Some studies large sample of incoming freshman students
give a snapshot of a single, fixed time point and since 1966. Since it began, over 11 million stu
allowyou to analyzeit in detail (cross-sectional). dents at more than 1,800 colleges participated.
Other studies provide a moving picture that lets The fall 2003 survey of 276,449 students found
you follow events, people, or social relations many facts and trends, such as only 34 percent of
18 PART ONE / FOUNDATIONS
.Alint
Fie U R E 1.2 The Time Dimension in Social Research
CROSS-SECTIONAL: Observe a collection of people at one time.
February 2007
2005~2007
CHAPTER 1 I DOING SOCIAL RESEARCH 19
entering freshmen studied six or more hours per term panel studies can clearly show the impact
week. This was the lowest level since the ques of a particular life event. For example, Oesterle,
tion was asked in 1987 (when it was 47 percent). Johnson, and Mortimer (2004) examined panel
Yet, alcohol consumption was down. In 2003, data from a longitudinal study that began in
44.8 percent reported drinking beer, which rep 1988 with 1,000 ninth-grade students enrolled
resented a steady decline from 73.7 percent in in the St. Paul, Minnesota, public school district
1982. In 2003, freshmen were more interested in and looked at volunteering activities during late
keeping up with politics. The 33.9 percent who adolescence and young adulthood, covering
said it was very important to stay politically in nine years from age 18-19 (1992) to age 26-27
formed was up from a low of 28.1 percent in (2000). They found that volunteering at an ear
2000, and 22.5 percent said they discussed poli lier stage strongly affected whether one volun
tics regularly, up from 19.4 percent in 2002 teered at a later stage. Also, people who devoted
(which had been the highest since a low point in full time to working or parenting at an earlier
1993). These figures are still far lower than the stage (18-19 years old) were less likely to volun
60.3 percent who expressed an interest in politics teer at a later stage (26-27 years old) than those
in 1966, or the one-third who discussed politics whose major activity was attending school.
regularly in 1968. The importance of family has
steadily increased over the years, with 74.8 per Cohort Study. A cohort study is similar to a
cent of students calling it essential or very im panel study, but rather than observing the exact
portant. This is up from the low point of 58.8 same people, the study focuses on a category of
percent in 1977 when the question was first people who share a similar life experience in a
asked. However, religious involvement declined. specified time period. Researchers examine the
The percentage of students who attended reli category as a whole for important features and
gious services regularly was at its lowest level in focus on the cohort, or category, not on specific
35 years. In addition, the percent claiming individuals. Commonly used cohorts include all
"none" as a religious preference reached a record people born in the same year (called birth co
high of 17.6 percent, compared to a record low of horts), all people hired at the same time, and all
6.6 percent in 1966. Another trend over the past people who graduate in a given year. Unlike
two decades has been a steady growth in opposi panel studies, researchers do not have to find the
tion to the death penalty. Nearly one in three in exact same people for cohort studies; rather,
coming students advocated ending capital they need only to identify those who experienced
punishment. This is the highest score since 1980 a common life event. In a study of Generation X
(when it was 33.2 percent), although the percent in the United States, Andolina and Mayer (2003)
withholding an opinion was far higher earlier in focused on the cohort of people born between
time; it exceeded 60 percent in the 1970. 11 1967 and 1974. They compared 10 birth cohorts
at different time periods over several decades,
Panel Study. The panel study is a powerful type tracing questions across 24 years. The authors
of longitudinal research in which the researcher found that White Xers are distinct in their
observes exactly the same people, group, or or support for school racial integration and for
ganization across multiple time points. It is government action to enforce such efforts, com
more difficult to conduct than time-series re pared to other birth cohorts, but not in their at
search. Panel research is formidable to conduct titudes toward employment opportunities or
and very costly. Tracking people over time is of affirmative action. Despite greater general sup
ten difficult because some people die or cannot port than other cohorts for equality through in
be located. Nevertheless, the results of a well-de tegration, it does not extend to issues beyond the
signed panel study are very valuable. Even short- schoolyard.
20 PART ONE / FOUNDATIONS
researcher first identifies a body of material to in detail. The researcher gets to know personally
analyze (e.g.,books, newspapers, films, etc.) and the people being studied, may conduct open
then creates a system for recording specific as ended and informal interviews, and takes de
pects of it. The system might include counting tailed notes on a daily basis. After leaving the
how often certain words or themes occur. Fi field site, the researcher carefully rereads the
nally, the researcher records what was found in notes and prepares written reports. Field re
the material. He or she often measures informa search is used most often for exploratory and de
tion in the content as numbers and presents it as scriptive studies; it is rarely used for explanatory
tables or graphs. This technique lets a researcher research.
discover features in the content oflarge amounts
of material that might otherwise go unnoticed. Historical-Comparative Research. Historical
Researchers can use content analysis for ex comparative researchers examine aspects of social
ploratory and explanatory research, but primar life in a past historical era or across different cul
ily it is used for descriptive research. tures. Researchers who use this technique may
focus on one historical period or several, com
ExistingStatistics. In existingstatistics research, pare one or more cultures, or mix historical pe
a researcher locates previously collected infor riods and cultures. Like field research, a
mation, often in the form of government reports researcher combines theory building/testing
or previously conducted surveys, then reorga with data collection and begins with a loosely
nizes or combines the information in new ways formulated question that is refined during the
to address a research question. Locating sources research process. Researchers often gather a
can be time consuming, so the researcher needs wide array of evidence, including existing statis
to consider carefully the meaning of what he or tics and dowments (e.g., novels, officialreports,
she finds. Frequently, a researcher does not books, newspapers, diaries, photographs, and
know whether the information of interest is maps) for study. In addition, they may make di
available when he or she begins a study. Some rect observations and conduct interviews. His
times, the existing quantitative information con torical-comparative research can be exploratory,
sists of stored surveys or other data that a descriptive, or explanatory and can blend types.
researcher reexamines using various statistical
procedures. Existing statistics research can be
used for exploratory, descriptive, or explanatory
CONCLUSION
purposes, but it is most frequently used for de
scriptive research. This chapter gave you an overview of social re
search. You saw how social research differs from
Qualitative Data Collection Techniques. the ordinary ways of learning-knowing about
Techniques for qualitative data collection in the social world, how doing research is based on
clude field research and historical-comparative science and the scientific community, and about
research. several types of social research based on its di
mensions (e.g., its purpose, the technique used
Field Research. Most field researchers conduct to gather data, etc.). The dimensions of research
case studies looking at a small group of people loosely overlap with each other. The dimensions
over a length of time (e.g., weeks, months, of social research are a kind of "road map" to
years). A field researcher begins with a loosely help you make your way through the terrain of
formulated idea or topic, selects a social group social research. In the next chapter, we turn to
or natural setting for study, gains access and social theory. You read about it a little in this
adopts a social role in the setting, and observes chapter. In the next chapter, you will learn how
r
22 PART ONE I FOUNDATIONS
theory and research methods work together and social impact assessment study
about several types of theory. social research
survey research
time-series study
Key Terms
Introduction
What Is Theory?
Blame Analysis
The Parts of Theory
Concepts
Assumptions
Relationships
The Aspects of Theory
Direction of Theorizing
Range of Theory
Levels of Theory
Forms of Explanation
The Three Major Approaches to Social Science
Positivist Approach
Interpretive Approach
Critical Approach
The Dynamic Duo
Conclusion
23
24 PART ONE / FOUNDATIONS
INTRODUCTION
Suppose you want to make sense of the hostility
between people of different races. Trying to un
derstand it, you ask a teacher, who responds:
in mind three things about how social scientific
theories work. First, social theories explain re
curring patterns, not unique or one-time events.
For example, they are not good for explaining
why terrorists decided to attack New York's
World Trade Center on September 11, 2001, but
-....-
~- 1JIr
Dwkl
~*- ...
..
Most raciallyprejudiced people learn nega they can explain patterns, such as the conditions
tive stereotypes about another racial group that generally lead to increased levels of fear and
from their families, friends, and others in feelings of patriotism in a people. Second, social
their immediate surroundings. If they lack theories are explanations for aggregates, not par
sufficient intimate social contact with mem ticular individuals. Aggregates are collections of
bers of the group or intense information that many individuals, cases, or other units (e.g.,
contradicts those stereotypes,they remain businesses, schools, families, clubs, cities, na
prejudiced. tions, etc.). A social theory rarely can explain
why Josephine decided to major in nursing
This makes sense to you because it is consis rather than engineering, but it can explain why
tent with what you know about how the social females more than males in general choose nurs
world works. This is an example of a small-scale ing over engineering as a major. Third, social
social theory, a type that researchers use when theories state a probability, chance, or tendency
conducting a study. for events to occur, rather than state that one
What do you think of when you hear the event must absolutely follow another. For exam
word theory? Theory is one of the least well un ple, instead of stating that when someone is
derstood terms for students learning social sci abused as a child, that person will always later
ence. My students' eyelids droop if! begin a class
by saying, "Today we are going to examine the
theory of ..." The mental picture many students
have of theory is something that floats high
among the clouds. My students have called it "a
abuse his or her own children, a theory might
state that when someone experiences abuse dur
ing his or her childhood, that person will tend to
or is more likely to become an abusive parent
when an adult. Likewise, it might state that peo
.........
i
SF '&$.
.........
tangled maze of jargon" and "abstractions that ple who did not experience childhood abuse tie~
...-- .
F
are irrelevant to the real world." might become abusive parents, but they are less e
Contrary to these views, theory has an im likely to than someone who has experienced
portant role in research and is an essential ally abuse as a child.
for the researcher. Researchers use theory differ ...« - .
ently in various types of research, but some type
of theory is present in most social research. It is
less evident in applied or descriptive than in ba
WHAT IS THEORY?
sic or explanatory research. In simple terms, re
searchers interweave a story about the operation In Chapter 1, social theory was defined as a sys
of the social world (the theory) with what they tem of interconnected abstractions or ideas that
observe when they examine it systematically (the condenses and organizes knowledge about the
data). social world. It is a compact way to think of the
People who seek absolute, fixed answers for social world. People are constantly developing
a specific individual or a particular one-time new theories about how the world works.
event may be frustrated with science and social Some people confuse the history of social
theories. To avoid frustration, it is wise to keep thought, or what great thinkers said, with social
CHAPTER 2 / THEORY AND SOCIAL RESEARCH 25
theory. The classical social theorists (e.g., Almost all research involves some theory, so
Durkheim, Weber, Marx, and Tennies) played the question is less whether you should use the
an important role in generating innovative ideas. ory than how you should use it. Being explicit
They developed original theories that laid the about the theory makes it easier to read someone
foundation for subsequent generations of social else's research or to conduct your own. An
thinkers. People study the classical theorists be awareness of how theory fits into the research
cause they provided many creative and interre process produces better designed, easier to un
lated ideas at once. They radically changed the derstand, and better conducted studies. Most re
way people understood and saw the social world. searchers disparage atheoretical or "crude
We study them because geniuses who generate empiricist" research.
many original, insightful ideas and fundamen
tally shift how people saw the social world are
Blame Analysis
rare.
At times people confuse theory with a hunch Blameanalysis is a type of counterfeit argument
or speculative guessing. They may say, "It's only presented as if it were a theoretical explanation.
a theory" or ask, "What's your theory about it?" It substitutes attributing blame for a causal ex
This lax use of the term theory causes confusion. planation that is backed by supporting empirical
Such guessing differs from a serious social the evidence. Blame belongs to the realm of making
ory that has been carefully built and debated moral, legal, or ideological claims. It implies an
over many years by dozens of researchers who intention, negligence, or responsibility for an
found support for the theory's key parts in re event or situation (usually an unfavorable one).
peated empirical tests. A related confusion is It shifts the focus from Why did it occur? to
when what people consider to be a "fact" (i.e., Who is responsible? Blame analysis assumes
light a match in a gasoline-filled room and it will there is a party or source to which a fixed
explode) is what scientists call a theory (i.e., a amount of responsibility can be attached. The
theory of how combining certain quantities of goal of inquiry is to identify a responsible party.
particular chemicals with oxygen and a level of Often, some sources are exempted or shielded.
heat is likely to produce the outcome of explo This may be the injured party, members of a
sive force). People use simple theories without sympathetic audience, or a sacred value or
making them explicit or labeling them as such. principle.
For example, newspaper articles or television re Blame analysis clouds discussion because it
ports on social issues usually have unstated so confuses blame with cause; it gives an account
cial theories embedded within them. A news (or story) instead of a logical explanation with
report on the difficultyof implementing a school intervening causal mechanisms; and it fails to
desegregation plan will contain an implicit the explore empirical evidence for and against sev
ory about race relations. Likewise, political lead eral alternative causes. Blame analysis first pre
ers frequently express social theories when they sents an unfavorable event or situation. It could
discuss public issues. Politicians who claim that be a bank is robbed, a group is systematically
inadequate education causes poverty or that a paid less in the labor force, or traffic congestion
decline in traditional moral values causes higher is terrible in an urban area. It next identifies one
crime rates are expressing theories. Compared or more responsible parties, then it provides se
to the theories of social scientists, such layper lective evidence that shields certain parties or
sons' theories are less systematic, less well for sources (e.g., employment conditions, the
mulated, and harder to test with empirical choices available to the underpaid group, trans
evidence. portation policy, and land cost).'
26 PART ONE I FOUNDATIONS
- ....
..n
of everyday words in specialized social science Concepts have two parts: a symbol (word or
concepts can create confusion. term) and a definition. We learn definitions in
I do not want to exaggerate the distinction many ways. I learned the word height and its de
between concepts expressed as words and con finition from my parents. I learned it as I learned
cepts expressed as symbols. Words, after all, are to speak and was socialized to the culture. My
symbols, too; they are symbols we learn with parents never gave me a dictionary definition. I
language. Height is a concept with which you are learned it through a diffuse, nonverbal, informal
already familiar. For example, I can say the word process. My parents showed me many examples;
height or write it down; the spoken sounds and I observed and listened to others use the word; I
written words are part of the English language. used the word incorrectly and was corrected;
The combination ofletters in the sound symbol and I used it correctly and was understood.
izes, or stands for, the idea of a height. Chinese or Eventually, I mastered the concept.
Arabic characters, the French word hauteur, the This example shows how people learn con
German word hohe, the Spanish word altura cepts in everyday language and how we share
all symbolize the same idea. In a sense, a lan concepts. Suppose my parents had isolated me
guage is merely an agreement to represent ideas from television and other people, then taught
by sounds or written characters that people me that the word for the idea height was zdged. I
learned at some point in their lives. Learning would have had difficulty communicating with
concepts and theory is like learning a language. 3 others. People must share the terms for concepts e..-.,.u....
Concepts are everywhere, and you use them and their definitions if they are to be ofvalue. ftobtioR.
all the time. Height is a simple concept from Everyday life is filled with concepts, but !:JOUPS. or
everyday experience. What does it mean? It is many have vague and unclear definitions. Like aplSin _
easy to use the concept of height, but describing wise, the values, misconceptions, and experi social~_
the concept itself is difficult. It represents an ab ences of people in a culture may limit everyday meiatm
stract idea about physical relations. How would concepts. Social scientists borrow concepts from aDy remlloRillll&Jl
you describe it to a very young child or a crea everyday culture, but they refine these concepts meaning. f«
ture from a distant planet who was totally unfa and add new ones. Many concepts such as cept such as
miliar with it? A new concept from a social sexism, life-style, peer group, urban sprawl, and sociated
theory may seem just as alien when you en social class began as precise, technical concepts economic gr;
counter it for the first time. Height is a charac in social theory but have diffused into the larger city, revitalJ"tZal-"
teristic of a physical object, the distance from top culture and become less precise. noritiesy. ~
to bottom. All people, buildings, trees, moun We create concepts from personal experi Someco~
tains, books, and so forth have a height. We can ence, creative thought, or observation. The clas quantities. or CIJDII
CHAPTER 2 I THEORY AND SOCIAL RESEARCH 27
sical theorists originated many concepts. Exam concept are amount ofincome, temperature, den
ple concepts include family system, gender role, sity ofpopulation,yearsof schooling, and degree of
socialization, self-worth, frustration, and displaced violence. These are called variables, and you will
aggresSlOn. read about them in a later chapter. Other con
Some concepts, especially simple, concrete cepts express types of nonvariable phenomena
concepts such as book or height, can be defined (e.g., bureaucracy, family, revolution, homeless,
through a simple nonverbal process. Most social and cold). Theories use both kinds of concepts.
science concepts are more complex and abstract.
They are defined by formal, dictionary-type de Classification Concepts. Some concepts are
finitions that build on other concepts. It may simple; they have one dimension and vary along
seem odd to use concepts to define other con a single continuum. Others are complex; they
cepts, but we do this all the time. For example, I have multiple dimensions or many subparts.
defined heightas a distance between top and bot You can break complex concepts into a set of
tom. Top, bottom, and distance are all concepts. simple, or single-dimension, concepts. For ex
We often combine simple, concrete concepts ample, Rueschemeyer and associates (1992:43
from ordinary experience to create more ab 44) stated that democracy has three dimensions:
stract concepts. Height is more abstract than top (1) regular, free elections with universal suffrage;
or bottom. Abstract concepts refer to aspects of (2) an elected legislative body that controls gov
the world we do not directly experience. They ernment; and (3) freedom of expression and as
organize thinking and extend understanding of sociation. The authors recognized that each
reality. dimension varies by degree. They combined the
Researchers define scientific concepts more dimensions to create a set of types of regimes.
precisely than those we use in daily discourse. Regimes very low on all three dimensions are to
Social theory requires well-defined concepts. talitarian, those high on all three are democra
The definition helps to link theory with research. cies, and ones with other mixes are either
A valuable goal of exploratory research, and of authoritarian or liberal oligarchies.
most good research, is to clarify and refine con Classifications are partway between a single,
cepts. Weak, contradictory, or unclear defini simple concept and a theory." They help to orga
tions of concepts restrict the advance of nize abstract, complex concepts. To create a new
knowledge. classification, a researcher logically specifies and
combines the characteristics of simpler con
Concept Clusters. Concepts are rarely used in cepts. You can best grasp this idea by looking at
isolation. Rather, they form interconnected some examples.
groups, or concept clusters. This is true for con The idealtype is a well-known classification.
cepts in everyday language as well as for those in Ideal types are pure, abstract models that define
social theory. Theories contain collections of as the essence of the phenomenon in question.
sociated concepts that are consistent and mutu They are mental pictures that define the central
ally reinforcing. Together, they form a web of aspects of a concept. Ideal types are not explana
meaning. For example, if! want to discuss a con tions because they do not tell why or how some
cept such as urban decay, I will need a set of as thing occurs. They are smaller than theories, and
sociated concepts (e.g., urban expansion, researchers use them to build a theory. They are
economicgrowth, urbanization, suburbs, center broader, more abstract concepts that bring to
city, revitalization, mass transit, and racial mi gether several narrower, more concrete con
norities). cepts. Qualitative researchers often use ideal
Some concepts take on a range of values, types to see how well observable phenomena
quantities, or amounts. Examples of this kind of match up to the ideal model. For example, Max
28 PART ONE I FOUNDATIONS
Weber developed an ideal type of the concept used for a much broader range of specific time
bureaucracy. Many people use Weber's ideal type points and situations. More concrete concepts
(see Box 2.1). It distinguishes a bureaucracy are easy to recognize but apply to fewer situa
from other organizational forms (e.g., social tions. The concepts skin pigmentation, casting a
movements, kingdoms, etc.). It also clarifies crit ballotin an election, and age based on the date on
ical features of a kind of organization that people a birth certificate are less abstract and more con
once found nebulous and hard to think about. crete than the concepts racial group, democracy,
No real-life organization perfectly matches the and maturity. Theories that use many abstract
ideal type, but the model helps us think about concepts can apply to a wider range of social
and study bureaucracy. phenomena than those with concrete concepts.
An example of a theoretical relationship is: In
Scope. Concepts vary by scope. Some are creased size creates centralization, which in turn
highly abstract, some are at a middle level of ab creates greater formalization. Size, centralization,
straction, and some are at a concrete level (i.e., and formalization are very abstract concepts.
they are easy to directly experience with the They can refer to features of a group, organiza
senses such as sight or touch). More abstract tion, or society. We can translate this to say that
concepts have wider scope; that is, they can be as an organization or group gets bigger, author
ity and power relations within it become cen
tralized and concentrated in a small elite. The
elite will tend to rely more on written policies,
rules, or laws to control and organize others in
the group or organization. When you think ex
plicitly about the scope of concepts, you make a
theory stronger and will be able to communicate
• It is a continuous organization governed by a sys it more clearly to others.
tem of rules.
• Conduct is governed by detached, impersonal Assumptions
rules.
• There is division of labor, in which different of
fices are assigned different spheres of compe
tence.
Concepts contain built-in assumptions, state
ments about the nature of things that are not ob
servable or testable. We accept them as a
..........
..UP-"
('• ..:. . . . . r
Middle-Range Middle-Range
Theory Theory
Range of
personal encounters with the police; reports of Duneier (1999) used an inductive approach Social theories
police encounters by friends, family, or neigh in his study oflife on the sidewalk. He noted that source of the
bors; and noticing and interpreting news reports in much of social science, both quantitative sec the range at ~
about police activity. In these three areas, they ondary analysis research and qualitative field re are highly
predicted that non-Whites would interpret neg search, a researcher develops a theoretical cepts of - .
ative events or reports as strong evidence of seri understanding only after data have been col whole systems
ous and systematic police misconduct. By lected. He stated, "I began to get ideas from the tremelvab
constrast, Whites would tend to ignore or dis things I was seeing and hearing on the street" (p. building, verifJi
miss such events or reports or see them as iso 341). Many researchers who adopt an inductive connects th~
I
lated incidents. Data from a national survey of approach use grounded theory. Grounded theory ranges tog~
U.S. metropolitan areas (over 100,000 popula is part of an inductive approach in which a re boxes that fit ~
tion) supported predictions of the theory. searcher builds ideas and theoretical generaliza ian dolls. ~
tions based on closely examining and creatively
Inductive. If you use an inductive approach, thinking about the data (see Box 2.2). A re Empirical Gena
eralization is m.j
you begin with detailed observations of the
world and move toward more abstract general
izations and ideas. When you begin, you may
searcher creates grounded theory out of a
process of trying to explain, interpret, and ren
der meaning from data. It arises from trying to
ment and has
statement aboa
a,
have only a topic and a few vague concepts. As account for, understand, or "make sense of' the among two or ~
you observe, you refine the concepts, develop evidence. Duneier (1999:342) has suggested that very close to ell
empirical generalizations, and identify prelimi the process is similar to seeing many symptoms "More men thai!
nary relationships. You build the theory from and later arriving at a diagnosis (i.e., a story that a college major.t
the ground up. explains the source of the symptoms). tween gender ...
CHAPTER 2 / THEORY AND SOCIAL RESEARCH 31
Grounded theory is a widely used approach in qual generalizations by making comparisons across so
itative research. It is not the only approach and it is cial situations.
not used by all qualitative researchers. Grounded the Qualitative researchers use alternatives to
oryis "a qualitative research method that uses a sys grounded theory. Some qualitative researchers offer
tematic set of procedures to develop an inductively an in-depth depiction that is true to an informant's
derived theory about a phenomenon" (Strauss and worldview. They excavate a single social situation to
Corbin, 1990:24). The purpose of grounded the elucidate the micro processes that sustain stable social
ory is to build a theory that is faithful to the evi interaction. The goal of other researchers is to provide
dence. It is a method for discovering new theory. In a very exacting depiction of events or a setting. They
it, the researcher compares unlike phenomena with a analyze specific events or settings in order to gain in
view toward learning similarities. He or she sees mi sight into the larger dynamics of a society. Still other
cro-level events as the foundation for a more macro researchers apply an existing theory to analyze specific
level explanation. Grounded theory shares several settings that they have placed in a macro-level histor
goals with more positivist-oriented theory. It seeks ical context. They show connections among micro
theory that is comparable with the evidence that is level events and between micro-level situations and
precise and rigorous, capable of replication. and larger social forces for the purpose of reconstructing
generalizable. A grounded theory approach pursues the theory and informing social action.
..
Kalmijn. A theory of intermarriage patterns notes tend to interpret media reports about police mis ft_FFIEir
that young adults in modern society spend less time conduct as evidence of serious and systematic prob -.u5C!91 .
-. ....
in small, local settings, where family. religion. and lems with the police. By contrast. Whites have .,· 4
community all have a strong influence. Instead. different police encounters or interpret their en ~
young adults spend increasing amounts of time in counters and media reports about police actions .--
school settings. In these settings. especially in col more favorably. -~
CHAPTER 2 / THEORY AND SOCIAL RESEARCH 33
tion with Kalmijn's study of changing marriage in sociology and briefly describes the key con
partner selection (see also page 40). cepts and assumptions of each.
Sociologyand other social sciences have sev
eral major theoretical frameworks.f The frame
Levels of Theory
works are orientations or sweeping ways of
looking at the social world. They provide collec Social theories can be divided into three broad
tions of assumptions, concepts, and forms of ex groupings by the level ofsocial reality with which
planation. Frameworks include theories for they deal. Most of us devote the majority of our
many substantive areas (e.g., theories of crime, time to thinking about the micro level of reality,
theories of the family, etc.). Thus, there can be a the individuals we see and interact with on a day
structural functional theory, an exchange the by-day basis. Micro-level theory deals with small
ory, and a conflict theory of the family. Theories slices of time, space, or numbers of people. The
within the same framework share assumptions concepts are usually not very abstract.
and major concepts. Some frameworks are ori Brase and Richmond (2004) used a micro
ented more to the micro level; others focus more level theory about doctor-patient interactions
on macro-level phenomena (see Levels of The and perceptions. The theory stated that physican
ory next). Box 2.4 shows four major frameworks attire affects doctor-patient interactions. It sug-
Key Assumptions. Society is a system of interde Key Assumptions. People transmit and receive sym
pendent parts that is in equilibrium or balance. Over bolic communication when they socially interact.
time, society has evolved from a simple to a complex People create perceptions of each other and social
type, which has highly specialized parts. The parts of settings. People largely act on their perceptions.
society fulfill different needs or functions of the social How people think about themselves and others is
system. A basic consensus on values or a value sys based on their interactions.
tem holds society together.
Conflict Theory
Exchange Theory (also Rational Choice) Major Concepts. Power, exploitation, struggle, in
Major Concepts. Opportunities, rewards, approval, equality, alienation
balance, credit
Key Assumptions. Society is made up of groups that
Key Assumptions. Human interactions are similar to have opposing interests. Coercion and attempts to
economic transactions. People give and receive re gain power are ever-present aspects of human rela
sources (symbolic, social approval, or material) and tions. Those in power attempt to hold on to their
try to maximize their rewards while avoiding pain, ex power by spreading myths or by using violence if
pense, and embarrassment. Exchange relations tend necessary.
to be balanced. Ifthey are unbalanced, persons with
credit can dominate others.
34 PART ONE / FOUNDATIONS
gested that a patient makes judgments about a would produce gains because other workers and ilBrp IM
physican's abilities based on attire and that a pa government authorities would support their ac - C"
tient's trust-openness toward a physican is also af
fected. It said that perceptions of physican
authority increased with traditional professional
formal attire over informal attire, but that trust
tions.
Macro-level theoryconcerns the operation of
larger aggregates such as social institutions, en
tire cultural systems, and whole societies. It uses
xadlas
iaII . ._
~
rc:sc:udlm's
...
....
~OI" . .~. .
same- and opposite-gender models who were theory. He wanted to explain the conditions that tJ3III5il"
identified as being medical doctors, but who were led Black people to engage in protest to gain full good tadIa"
wearing different attire. Findings showed that a citizenship rights and he examined patterns ofna Thetwol!'fll:Mllll
white coat and formal attire are clearly superior to tional racial politics in three counties across two This~
makes iD'. . . .l11li
casual attire in establishing physican authority,
but it did not reduce trust-openness as expected.
Meso-level theorylinks macro and micro lev
centuries. His theory said that protest resulted in
an interaction between (1) race-based political
mobilization and (2) national government poli
Pt_
logical~'"
els and operates at an intermediate level. Theo cies of racial domination (i.e., apartheid in South will OCOII'. It..
ries of organizations, social movements, and Africa, Jim Crow laws in southern United States, and an e:q:J'-'-'
communities are often at this level. and no legalized race-based domination in prediction
Roscigno and Danaher (2001) used meso Brazil). Policies of racial domination developed diet. An e:IIJI...
level theory in a study on the 1930s labor move from practices of slavery, exploitation, and dis one outCOlDlC,
ment among southern textile workers. The crimination that justified White superiority. The predicted ~
researchers used a theory of movement subcul policies reinforced specific racial ideologies that it is less J)OIIICI*I
ture and political opportunity to explain grow shaped national development during the twenti ple are entnllOlllll
ing labor movement strength and increased eth century. A critical causal factor was how prediction.
strike activity among workers in one industry in
a region of the United States across several years.
national political elites used the legalized domina
tion of Blacks to reduce divisions among Whites.
A gam"'.
ence betweea
They expected strike activity to grow as the result In nations that had large regional or class divi enter a casiDo
of a strong movement subculture that carried a sions among Whites, national elites tried to predict the oar
message of injustice and a "political opportu increase White backing for the national govern ber on a nHllkll~
nity" or the expectation among people that col ment by creating legalized forms of racial domi may win a lot
lective action at a particular time would produce nation. Over time, such legalized domination officials realizr
positive results. Their study showed that a tech froze racial divisions, which promoted a sense of me. Yet, my
nological innovation (i.e., the spread of new ra racial identity and consciousness among Blacks. more inten:sti.
dio stations with songs and discussions of The strong sense of racial identity became a key Telling you.
working conditions and unfair treatment) con resource when Blacks mobilized politically to de more fascirJatiIDlll
tributed to the growth of a subculture of move mand full citizenship rights. Legalizedracial dom Here is
ment solidarity among the textile workers and ination also intensified the Blacks' protest and sun "rises" each
fostered self-identity as a worker who had com directed it against the national government as the at some time, ell
mon interests with the other textile workers. The societal institution that reinforced their experi clouds obscure ~
technological innovation and events in the polit ence of racial inequality. this so? One ~
ical environment (i.e., union organizers and carries the sun 4
speeches by the President of the United States) other expl~
Forms of Explanation
also created a political opportunity for the work ablaze, which ~
ers. The workers believed that collection action Prediction and Explanation. A theory's pri shoots it across ..
(i.e., strike) was necessary to achieve justice and mary purpose is to explain. Many people con these ancient exJ
CHAPTER 2 / THEORY AND SOCIAL RESEARCH 3S
fuse prediction with explanation. There are two probably accept involves a theory about the ro
meanings or uses of the term explanation. Re tation of the earth and the position of the sun,
searchers focus on theoretical explanation, a log the star of our solar system. In this explanation,
ical argument that tells why something occurs. It the sun only appears to rise. The sun does not
refers to a general rule or principle. These are a move; its apparent movement depends on the
researcher's theoretical argument or connec earth's rotation. Weare on a planet that both
tions among concepts. The second type ofexpla spins on its axis and orbits around a star millions
nation, ordinary explanation, makes something of miles away in space. All three explanations
clear or describes something in a way that illus make the same prediction: The sun rises each
trates it and makes it intelligible. For example, a morning. As you can see, a weak explanation can
good teacher"explains" in the ordinary sense. produce an accurate prediction. A good expla
The two types ofexplanation can blend together. nation depends on a well-developed theory and
This occurs when a researcher explains (i.e., is confirmed in research by empirical observa
makes intelligible) his or her explanation (i.e., a tions.
logical argument involving theory).
Prediction is a statement that something Causal Explanation. Causal explanation, the
will occur. It is easier to predict than to explain, most common type ofexplanation, is used when
and an explanation has more logical power than the relationship is one of cause and effect. We
prediction because good explanations also pre use it all the time in everyday language, which
dict. An explanation rarely predicts more than tends to be sloppy and ambiguous. What do we
one outcome, but the same outcome may be mean when we say cause? For example, you may
predicted by opposing explanations. Although say that poverty causes crime or that looseness in
it is less powerful than explanation, many peo morals causes an increase in divorce. This does
ple are entranced by the dramatic visibility of a not tell how or why the causal process works.
prediction. Researchers try to be more precise and exact
A gambling example illustrates the differ when discussing causal relations.
ence between explanation and prediction. If I Philosophers have long debated the idea of
enter a casino and consistently and accurately cause. Some people argue that causality occurs
predict the next card to appear or the next num in the empirical world, but it cannot be proved.
ber on a roulette wheel, it will be sensational. I Causality is "out there" in objective reality, and
may win a lot of money, at least until the casino researchers can only try to find evidence for it.
officials realize I am always winning and expel Others argue that causality is only an idea that
me. Yet, my method of making the predictions is exists in the human mind, a mental construc
more interesting than the fact that I can do so. tion, not something "real" in the world. This
Telling you what I do to predict the next card is second position holds that causality is only a
more fascinating than being able to predict. convenient way of thinking about the world.
Here is another example. You know that the Without entering into the lengthy philosophical
sun "rises" each morning. You can predict that debate, many researchers pursue causal relation
at some time, every morning, whether or not ships.
clouds obscure it, the sun will rise. But why is You need three things to establish causality:
this so? One explanation is that the Great Turtle temporal order, association, and the elimination
carries the sun across the sky on its back. An of plausible alternatives. An implicit fourth
other explanation is that a god sets his arrow condition is an assumption that a causal rela
ablaze, which appears to us as the sun, and tionship makes sense or fits with broader as
shoots it across the sky. Few people today believe sumptions or a theoretical framework. Let us
these ancient explanations. The explanation you examine the three basic conditions.
r 36 PART ONE I FOUNDATIONS
The temporal order condition means that a the effect. Most studies examine unidirectional
cause must come before an effect. This com relations. More complex theories specify recip
monsense assumption establishes the direction rocal-effect causal relations-that is, a mutual
of causality: from the cause toward the effect. causal relationship or simultaneous causality.
You may ask, How can the cause come after For example, studying a lot causes a student to
what it is to affect? It cannot, but temporal order get good grades, but getting good grades also
is only one of the conditions needed for causal motivates the student to continue to study. The
ity. Temporal order is necessary but not suffi ories often have reciprocal or feedback relation
cient to infer causality. Sometimes people make ships, but these are difficult to test. Some
the mistake of talking about "cause" on the basis researchers call unidirectional relations nonre
of temporal order alone. For example, a profes cursive and reciprocal-effect relations recursive.
sional baseball player pitches no-hit games when A researcher also needs an association for
he kisses his wife just before a game. The kissing causality. Two phenomena are associated if they
occurred before the no-hit games. Does that occur together in a patterned way or appear to
mean the kissing is the cause of the pitching per act together. People sometimes confuse correla
formance? It is very unlikely. As another exam tion with association. Correlation has a specific
ple, race riots occurred in four separate cities in technical meaning, whereas association is a more
1968, one day after an intense wave of sunspots. general idea. A correlation coefficient is a statisti
The temporal ordering does not establish a cal measure that indicates the amount of associ !
causal link between sunspots and race riots. M ation, but there are many ways to measure If a reseanl
ter all, all prior human history occurred before association. Figure 2.2 shows 38 people from a causal rela~
some specific event. The temporal order condi lower-income neighborhood and 35 people from searchers at1emJl
tion simply eliminates from consideration po an upper-income neighborhood. Can you see an measures of as5lIIj
tential causes that occurred later in time. association between race and income level? ten find an assed
It is not always easy to establish temporal More people mistake association for causal sociation e~
order. With cross-sectional research, temporal ity than confuse it with temporal order. For ex associated, burl
order is tricky. For example, a researcher finds ample, when I was in college,I got high grades on cause. It is a neal
that people who have a lot of education are also the exams I took on Fridays but low grades on tion. In other ..
less prejudiced than others. Does more educa those I took on Mondays. There was an associa but it is not ~
tion cause a reduction in prejudice? Or do highly tion between the day of the week and the exam Anassocialll
prejudiced people avoid education or lack the grade, but it did not mean that the day of the (i.e., everytime~
motivation, self-discipline, and intelligence week caused the exam grade. Instead, the reason also is) to s~~
needed to succeed in school? Here is another ex was that I worked 20 hours each weekend and ing exam gra~
ample. The students who get high grades in my was very tired on Mondays. As another example, association if 0lIIIl
class say I am an excellent teacher. Does getting the number of children born in India increased 1 C, whereas ~
high grades make them happy, so they return the until the late 1960s,then slowed in the 1970s. The were 6 Ds, 2 Cs".l
favor by saying that I am an excellent teacher number of U.S.-made cars driven in the United but the days ofd
(i.e., high grades cause a positive evaluation)? Or States increased until the late 1960s,then slowed not perfectly aSlI
am I doing a great job, so students study hard in the 1970s. The number of Indian children level associatioa
and learn a lot, which the grades reflect (i.e., born and the number of U.S. cars driven are as imperfect assodI
their learning causes them to get high grades)? It sociated: They vary together or increase and de Elimi~
is a chicken-or-egg problem. To resolve it, a re crease at the same time. Yet there is no causal searcher interea
searcher needs to bring in other information or connection. By coincidence, the Indian govern that the effect ..
design research to test for the temporal order. ment instituted a birth control program that not to som~
Simple causal relations are unidirectional, slowed the number of births at the same time ousness becausei
operating in a single direction from the cause to that Americans were buying more imported cars. that is actuallv •
CHAPTER 2 I THEORY AND SOCIAL RESEARCH 37
-------------------1.
•• 0
~
0t.." o
Q~ o~~~O o~'
0 0 0 •
o'o?~ n .'i ti
to
Q1f.t. ia 'ti
oO~'i .Q.~~ t
Q~ ~o o~'fat.
i ti"" Q~
0 0
Q~Q~
II
As I was driving home from the universityone day, I bias: the educational experience of students. It
heard a radio news report about gender and racial turns out that girls and boys take different numbers
bias in standardized tests. A person who claimedthat and types of mathematics courses in high school.
bias was a major problem said that the tests should Girls tend to take fewer math courses. Among the
be changed. Since I workin the field of education and girls who complete the same mathematics curricu
disdain racial or gender bias, the report caught my lum as boys, the gender difference dissolves. Like
attention. Yet, as a social scientist, I critically evalu wise, a large percentage of African Americans
ated the news story. The evidence for a bias charge attend racially segregated, poor-quality schools in
was the consistent pattern of higher scores in math inner cities or in impoverished rural areas. For
ematics for male high school seniors versus female African Americans who attend high-quality subur
high school seniors, and for European-background ban schools and complete the same courses, racial
students versus African American'students. Was the differences in test scores disappear. This evidence
cause of the pattern of different test scores a bias suggests that inequality in education causes test
built into the tests? score differences. Although the tests may have
When questioned by someone who had de problems, identifying the real cause implies that
signed the tests, the person charging bias lacked a changing the tests without first improvingor equal
crucial piece of evidence to support a claim of test izing education could be a mistake.
used a causal explanation; it said physican attire X leads to Y, X produces Y, X influences Y, X is Structural
causes certain types of patient perceptions. The related to Y, the greater X the higher Y. tum is used _ida
studybyWeitzer and Tuch (2004,2005) on po Here is a simple causal theory: A rise in un sequential. aocl
a
lice misconduct cited earlier used a causal ex employment causes an increase in child abuse. causal effect
planation. The cause was a person's group The subject to be explained is an increase in the balls lined up dJIIll
position and competitive pressure with other occurrence of child abuse. What explains it is a to bounce in ~
groups. These are causally linked to police en rise in unemployment. We "explain" the in
with spokes
counters, either directly or indirectly, and inter crease in child abuse by identifying its cause. A in which each
pretions of news reports, which differ by group complete explanation also requires elaborating researcher
position. The police encounters and the inter the causal mechanism. My theory says that when a set of int
pretations of news reports cause very different people lose their jobs, they feel a loss of self and relationshiplll
perceptions of police misconduct. We can re worth. Once they lose self-worth, they become he or she uses ~
state the logic in a deductive causal form: If the easily frustrated, upset, and angry. Frustrated lationships "'maIIIl
proposition is true, then we observe certain people often express their anger by directing vi lations within 4
. things in the empirical evidence. Good causal olence toward those with whom they have close reinforcing ~~
explanations identify a causal relationship and personal contact (e.g., friends, spouse, children, researcher ~
specify a causal mechanism. A simple causal ex etc.). This is especially true if they do not under tifies essential JIll
planation is: X causes Yor Yoccurs because of stand the source of the anger or cannot direct it ",-hole.
X, where X and Yare concepts (e.g., early mar toward its true cause (e.g., an employer, govern Structural ai
riage and divorce). Some researchers state ment policy, or "economic forces"). theory. Sanders.'
causality in a predictive form: If X occurs, then The unemployment and child abuse exam plained Asian iml
Yfollows. Causality can be stated in many ways: ple illustrates a chain of causes and a causal work theory. 11
CHAPTER 2 / THEORY AND SOCIAL RESEARCH 39
mechanism. Researchers can test different parts immigrants from the Philippines, Korea, Tai
of the chain. They might test whether unem wan, and China in Los Angeles and found that
ployment rates and child abuse occur together, social networks matched and sorted immigrants
or whether frustrated people become violent to with jobs. New immigrants with limited lan
ward the people close to them. A typical research guage and job skills sought employment either
strategy is to divide a larger theory into parts and with a co-ethnic employer or through informal
test various relationships against the data. social ties (i.e., they consulted experienced
Relationships between variables can be pos friends, relatives, and acquaintances and asked
itive or negative. Researchers imply a positive re them to be intermediaries). Network users ex
lationship if they say nothing. A positive panded job opportunities beyond employers in
relationship means that a higher value on the their own ethnic group. Thus, ethnic network
causal variable goJS with a higher value on the ties were "bridge ties" (i.e., they helped immi
effect variable. For example, the more education grants get jobs beyond their ethnic community
a person has, the longer his or her life expectancy by using co-ethnics who already made the tran
is. A negative relationship means that a higher sition to mainstream employment). Over time,
value on the causal variable goes with a lower as language and job skills improved, these im
value on the effect variable. For example, the migrants moved on to mainstream jobs. Immi
more frequently a couple attends religious ser grants lacking social ties, in limited networks, or
vices, the lower the chances of their divorcing who worked for co-ethnics found it difficult to
each other. In diagrams, a plus sign (+) signifies get a mainstream job. Thus, a person's network
a positive relationship and a negative sign (-) location, access to a large and diverse network,
signifies a negative relationship. and use of network ties is what facilitated ob
taining a mainstream job.
Structural Explanation. A structural explana Structural explanations are also used in se
tion is used with three types of theories: network, quence theory. The panel study on volun
sequential, and functional theories. Unlike a teerism by Oesterle, Johnson, and Mortimer
causal effect chain, which is similar to a string of (2004) discussed in Chapter 1 employs se
balls lined up that hit one another causing each quence theory. The authors used a "life course"
to bounce in turn, it is more similar to a wheel perspective in which the impact of an event
with spokes from a central idea or a spider web happening at one phase of a person's life differs
in which each strand forms part of the whole. A what it would have been if the same happened at
researcher making a structural explanation uses other phases, and early events generally shape
a set of interconnected assumptions, concepts, events in later phases. The authors noted that
and relationships. Instead of causal statements, the transition to adulthood is a critical stage
he or she uses metaphors or analogies so that re when a person learns new social roles and adult
lationships "make sense." The concepts and re expectations. They found that the amounts and
lations within a theory form a mutually types of volunteer activity in the last stage they
reinforcing system. In structural explanations, a observed (age 26-27) was strongly influenced
researcher specifiesa sequence of phases or iden by such activities at prior stages of a person's life
tifies essential parts that form an interlocked (age 18-19). People who volunteered at an early
whole. stage tended to volunteer at later stages. Those
Structural explanations are used in network who did not volunteer at an early stage or who
theory. Sanders, Nee, and Sernau (2002) ex devoted full time to working or parenting at
plained Asian immigrant job seeking with net other prior stages (18-19 years old) were less
work theory. They used interview data on likely to volunteer at a later stage (26-27 years
40 PART ONE I FOUNDATIONS
old). Thus, later events flowed from an inter In modern society, people spend time away
connected process in which earlier stages set a from small local settings in school settings. In
course or direction that pointed to specific these school settings, especially in college, they
events in a later stage. meet other unmarried people. Education is a
Additionally, structural explanations are major socialization agent in modern society.
used in functional theory.6 Functional theorists Increasingly, it affects a person's future earn
explain an event by locating it within a larger, ings, moral beliefs and values, and ways of
ongoing, balanced social system. They often use spending leisure time. This explains why there
biological metaphors. These researchers explain has been a trend in the United States for people
something by identifying its function within a to marry less within the same religion and in
larger system or the need it fulfills for the sys creasingly to marry persons with a similar level
tem. Functional explanations are in this form: "L of education. In traditional societies, the family
occurs because it serves needs in the system M." and religious organization served the function
Theorists assume that a system will operate to of socializing people to moral values and link
stay in equilibrium and to continue over time. A ing them to potential marriage partners who
functional theory of social change says that, over held similar values. In modern society, educa
time, a social system, or society, moves through tional institutions largely fulfill this function
developmental stages, becoming increasingly for the social system. -.r ......
differentiated and more complex. It evolves a fl.. i
specialized division oflabor and develops greater Interpretive Explanation. The purpose of an -.iIs~
individualism. These developments create interpretive explanation is to foster understand . . .GfII:II3IIIiiI
greater efficiency for the system as a whole. Spe ing. The interpretive theorist attempts to dis .-ar
cialization and individualism create temporary cover the meaning of an event or practice by
disruptions. The traditional ways of doing things placing it within a specific social context. He or
weaken, but new social relations emerge. The she tries to comprehend or mentally grasp the
system generates new ways to fulfill functions or operation of the social world, as well as get a feel
satisfy its needs. for something or to see the world as another per
Kalmijn (1991) used a functional explana son does. Because each person's subjective
tion to explain a shift in how people in the worldview shapes how he or she acts, the re
United States select marriage partners. He relied searcher attempts to discern others' reasoning
on secularization theory, which holds that on and view of things. The process is similar to
going historical processes of industrialization decoding a text or work of literature. Meaning
and urbanization shape the development of so comes from the context of a cultural symbol
ciety. During these modernization processes, system.
people rely less on traditional ways of doing Duneier's (1999) study of sidewalk life in
things. Religious beliefs and local community New York City discussed earlier in this chapter
ties weaken, as does the family's control over used an interpretive explanation. An interpretive
young adults. People no longer live their entire
lives in small, homogeneous communities.
Young adults become more independent from
their parents and from the religious organiza
explanation is also illustrated by Edelman,
Fuller, and Mara-Drita's (2001) study of how
companies adopted policies related to diversity
issues in the early 1990s-that is, affirmative ac
•......DI-.
....CIIfI.....
n
tions that formerly played a critical role in se
lecting marriage partners.
Society has a basic need to organize the way
people select marriage partners and find part
tion and equal opportunity. The authors exam
ined what managers said, or their rhetoric, about
diversity concerns. Rhetoric included various
statements about diversity made by professional
...........
.
&
'
...........
Me
ners with whom they share fundamental values. managers, business school professors, and con
CHAPTER 2 / THEORY AND SOCIAL RESEARCH 41
the growth of knowledge. They see multiple par Positivism sees social science research as funda
adigms as a sign of the immaturity or underde mentally the same as natural science research; it
velopment ofthe "science" in the social sciences. assumes that social reality is made up of objec
Some believe all social science researchers tive facts that value-free researchers can precisely
should embrace a single paradigm and stop us measure and use statistics to test causal theories.
ing alternatives to it. Large-scale bureaucratic agencies, companies,
Other social scientists accept the coexistence and many people in the general public favor a
of multiple paradigms. They recognize that this positivist approach because it emphasizes get
can be confusing and often makes communicat ting objective measures of "hard facts" in the
ing difficult among those who use a different ap form of numbers.
proach. Despite this, they argue that each social Positivists put a great value on the principle
science paradigm provides important kinds of of replication, even if only a few studies are repli
knowledge and insights, so to drop one would cated. Replication occurs when researchers or
limit what we can learn about the social world. others repeat the basics of a study and get iden
These social scientists note that no one definitely tical or very similar findings. Positivists em
can saywhich approach is "best" or even whether phasize replication and the ultimate test of
it is necessary or highly desirable to have only one knowledge. This is because they believe that dif
paradigm. So instead of closing off an approach ferent observers looking at the same facts will get
that offers innovative ways to study social life and the same results if they carefully specify their
gain insight into human behavior, they argue for ideas, precisely measure the facts, and follow the
keeping a diversity of approaches. standards of objective research. When many
In this section, we will look at three funda studies by independent researchers yield similar
mental paradigms or approaches used in social findings, confidence grows that we accurately
science. Each approach has been around for over captured the workings of social reality and there
150 years and is used by many highly respected fore scientific knowledge increases.
professional researchers. These approaches are If a researcher repeats a study and does not
unequal in terms of the number of followers, get similar findings, one or more offive possibil
quantity of new studies, and types of issues ad ities may be occurring: (1) the initial study was
dressed. Often, people who strongly adhere to an unusual fluke or based on a misguided un
one approach disagree with researchers who use derstanding of the social world; (2) important
another, or see the other approaches as being less conditions were present in the initial study, but
valuable or less "scientific" than their approach. no one was aware of their significance so they
Although adherents to each approach may use were not specified; (3) the initial study, or the
various research techniques, theories, and theo repeat of it, was sloppy-it did not include very The interpretree
retical frameworks, researchers who adopt one careful, precise measures; (4) the initial study, or its sees the idra
approach tend to favor certain research tech the repeat of it, was improperly conducted-re positivism,
niques, theories, or theoretical frameworks over searchers failed to closely follow the highest pretive re5leB1;h11
others. The three approaches are positivism, in standards for procedures and techniques, or qualita~
terpretive, and critical; each has internal divi
sions, offshoots, and extensions, but these are
failed to be completely objective; or (5) the re
peated study was an unusual fluke.
by science. This.
not just boJTOW4
the core ideas of the three major approaches. The positivist approach is nomothetic; it the natural sc:ie:IIII
means explanations use law or law-like princi necessary to ~
ples. Positivists may use inductive and deductive based on the uniq
.Positivist Approach
inquiry, but the ideal is to develop a general can really captunl
Positivism is the most widely practiced social sci causal law or principle then use logical deduc Most TeSeUlIIl
ence approach, especially in North America. tion to specify how it operates in concrete situa- approach adopt..
CHAPTER 2 / THEORY AND SOCIAL RESEARCH 43
tions. Next, the researcher empirically tests out view of social reality. This view holds that hu
comes predicted by the principle in concrete set man social life is based less on objective, hard,
tings using very precise measures. In this way, a factual reality than on the ideas, beliefs,and per
general law or principle covers many specificsit ceptions that people hold about reality. In other
uations. For example, a general principle says words, people socially interact and respond
that when two social groups are unequal and based as much, if not more, on what they believe
compete for scarce resources, in-group feelings to be real than what is objectively real. This
and hostility toward the other groups intensify, means that social scientists will be able to under
and the competing groups are likely to engage in stand social life only if they study how people go
conflict. The principle applies to sports teams, about constructing social reality. As people grow
countries, ethnic groups, families, and other so up, interact, and live their daily lives, they con
cial groupings. A researcher might deduce that tinuously create ideas, relationships, symbols,
in cities with high levels of interracial inequality, and roles that they consider to be meaningful or
when jobs become more scarce and thereby in important. These include things such as intimate
crease economic competition, each group will emotional attachments, religious or moral
express greater hostility about the other racial ideals, beliefsin patriotic values, racial-ethnic or
groups, and intergroup conflict (e.g., riots, gender differences, and artistic expressions.
demonstrations, violent attacks) will increase. Rarely do people relate to the objective facts of
The vast majority of positivist studies are reality directly; instead, they do so through the
quantitative, and positivists generally see the ex- filter of these socially constructed beliefs and
periment as the ideal way to do research. Posi perceptions. What positivists and many people
tivist researchers also use other quantitative view to be objective facts (e.g., a person's
research techniques, such as surveys or existing height), interpretive researchers say are only at
statistics, but tend to see them as approxima the trivial surface level of social life. Or, the
tions of the experiment for situations where an "facts" are images/categories that humans cre
experiment is impossible. Positivist researchers ated (i.e., I am two meters tall) and we "forget"
advocate value-free science, seek precise quanti that people originated the images/categories but
tative measures, test causal theories with statis now treat them as being separate from people
tics, and believe in the importance of replicating and objectively real.
studies. Interpretive researchers are skeptical of the
positivist attempts to produce precise quantita
tive measures of objective facts. This is because
Interpretive Approach
they view social reality as very fluid. For most
The interpretive approach is also scientific, but humans, social reality is largely the shifting per
its sees the idea of "scientific" differently from ceptions that they are constantly constructing,
positivism. Unlike the positivist approach, inter testing, reinforcing, or changing and that have
pretive researchers say that human social life is become embedded in social traditions or institu
qualitatively different from other things studied tions. For this reason, interpretive researchers
by science. This means that social scientists can tend to trust and favor qualitative data. They be
not just borrow the principles of science from lieve that qualitative data can more accurately
the natural sciences. Instead, they believe it is capture the fluid processes of social reality. In
necessary to create a special type of science, one addition, they favor interpretive over causal
based on the uniqueness of humans and one that forms of theory (see discussion earlier in this
can really capture human social life. chapter).
Most researchers who use an interpretive Interpretive researchers are not likely to
approach adopt a version of the constructionist adopt a nomothetic approach, but instead favor
44 PART ONE / FOUNDATIONS
an idiographic form of explanation and use in because it profoundly shapes much of human
ductive reasoning. Idiographic literally means action. TABLE 2.1
specific description and refers to explaining an The critical approach has an activist orien
aspect of the social world by offering a highly de tation and favors action research. Praxis is the
tailed picture or description of a specific social ultimate test of how good an explanation is in
setting, process, or type of relationship. For ex the critical approach. It is a blending of theory
Direction loci
ample, qualitative researchers do not see replica and concrete action; theory informs one about
tion as the ultimate test of knowledge. Instead, the specific real-world actions one should take Level Mici
they emphasize verstehen or empathetic under to advance social change, and one uses the expe Explanation c:.a.
standing. Verstehen is the desire of a researcher riences of engaging in action for social change to
~
Abstraction
to get inside the worldview of those he or she is reformulate the theory. All the approaches see a
studying and accurately represent how the peo mutual relationship between abstract theory and
ple being studied see the world, feel about it, and concrete empirical evidence, but the critical ap
act. In other words, the best test of good social proach goes further and tries to dissolve the gap
knowledge is not replication but whether the re between abstract theory and the empirical expe
searcher can demonstrate that he or she really riences of using the theory to make changes in Theory has .al
captured the inner world and personal perspec
tive of the people studied.
the world.
tra1 in
but its promineoq
applied-des
sic-explanatorv I"CS
descriptive r~
Critical Approach THE DYNAMIC DUO
cepts are often IDIJl
The critical approach shares many features with You have seen that theory and research are in to create zeneral I
an interpretive approach, but it blends an objec terrelated. Only the naive, new researcher mis searchers use thoIl
tive/materialist with a constructionist view ofso
cial reality. The key feature of the critical
approach is a desire to put knowledge into ac
takenly believes that theory is irrelevant to
research or that a researcher just collects the
data. Researchers who attempt to proceed with
on", and inS;
refine concepts, etl
Theon- does
tion and a belief that research is not value free. out theory may waste time collecting useless provisional and
~.
t;~_:~-~~~
Research is the creation of knowledge, and peo data. They easily fall into the trap of hazy and
ple regularly use knowledge to advance political vague thinking, faulty logic, and imprecise con
moral ends. This gives doing social research a cepts. They find it difficult to converge onto a
strong connection to political-moral issues. The crisp research issue or to generate a lucid ac rists toil to think.
researcher can decide to ignore and help those count of their study's purpose. They also find dIon has limia
with power and authority in society, or advance themselves adrift as they attempt to design or ica.nt progress is
social justice and empower the powerless. conduct empirical research. findings. I
Critical approach emphasizes the multilay The reason is simple. Theory frames how we The scienri6c
ered nature of social reality. On the surface level, look at and think about a topic. It gives us con ters theories
there is often illusion, myth, and distorted think cepts, provides basic assumptions, directs us to se:arc:hen ,,-ho
ing. The critical approach notes that people are the important questions, and suggests ways for me thoory to
often misled, are subject to manipulated mes us to make sense of data. Theory enables us to ~oi
sages, or hold false ideas. Yet, beneath the sur connect a single study to the immense base of mod:itY the
face level at a deeper, often hidden level lies knowledge to which other researchers con researchers
"real" objective reality. Part of the task of social tribute. To use an analogy, theory helps a re tearehin~
research is to strip away the surface layer of illu searcher see the forest instead of just a single dimr.:e thai: ~
sion or falsehood. Although a researcher wants tree. Theory increases a researcher's awareness ~modift-~
to see beyond this layer, he or she does not en of interconnections and of the broader signifi 'a:t them if irQ~
tirely ignore it. Such an outer layer is important cance of data (see Table 2.1). apne~
CHAPTER 2 I THEORY AND SOCIAL RESEARCH 45
----------.
Introduction
Why Be Ethical?
Scientific Misconduct
Unethical but Legal
Power Relations
Ethical Issues Involving Research Participants
Origins of Research Participant Protection
Physical Harm, PsychologicalAbuse, and Legal Jeopardy
Other Harm to Participants
Deception
Informed Consent
Special Populations and New Inequalities
Privacy, Anonymity, and Confidentiality
Mandated Protections of Research Participants
Ethics and the Scientific Community
Ethics and the Sponsors of Research
Whistle-Blowing
Arriving at Particular Findings
Suppressing Findings
Concealing the True Sponsor
Politics of Research
Value-Free and Objective Research
Conclusion
47
r
48 PART ONE / FOUNDATIONS
researcher is ultimately responsible to do what is others in Nazi Germany, and similar "medical
ethical in specific situations. experiments" to test biological weapons by
Japan in the 1940s. In these experiments, terrible
tortures were committed. For example, people
were placed in freezing water to see how long it
ETHICAL ISSUES INVOLVING
took them to die, people were purposely starved
RESEARCH PARTICIPANTS
to death, people were intentionally infected with
Have you ever been a participant in a research horrible diseases, and limbs were severed from
study? If so, how were you treated? More atten children and transplanted onto others.f
tion is focused on the possible negative effects of Such human rights violations did not occur
research on those being studied than any other
ethical issue, beginning with concerns about
biomedical research. Acting ethically requires
that a researcher balance the value of advancing
knowledge against the value of noninterference
only long ago. In a famous case of unethical re
search, the Tuskegee Syphilis Study, also known
as Bad Blood, the President of the United States
admitted wrongdoing and formally apologized
in 1997 to the participant-victims. Until the
.. ......
-
in the lives of others. Either extreme causes 1970s, when a newspaper report caused a scan
problems. Giving research participants absolute dal to erupt, the U.S. Public Health Service
rights of noninterference could make empirical sponsored a study in which poor, uneducated
research impossible, but giving researchers ab African American men in Alabama suffered and
a.dIIiaI"'lIIII
solute rights of inquiry could nullify partici died of untreated syphilis, while researchers
pants' basic human rights. The moral question studied the severe physical disabilities that ap
becomes: When, if ever, are researchers justified pear in advanced stages of the disease. The un
in risking physical harm or injury to those being ethical study began in 1929, before penicillin was
studied, causing them great embarrassment or available to treat the disease, but it continued
inconvenience, violating their privacy, or fright long after treatment was available. Despite their
ening them? unethical treatment of the people, the re
The law and codes of ethics recognize some searchers were able to publish their results for 40
clear prohibitions: Never cause unnecessary or years. The study ended in 1972, but a formal
irreversible harm to subjects; secure prior vol apology took another 25 years.!
untary consent when possible; and never unnec Unfortunately, the Bad Blood scandal is not
essarily humiliate, degrade, or release harmful unique. During the Cold War era, the U.S. gov
information about specific individuals that was ernment periodically compromised ethical re
collected for research purposes. In other words, search principles for military and political goals.
you should always show respect for the research In 1995, reports revealed that the government
participant. These are minimal standards and authorized injecting unknowing people with ra
are subject to interpretation (e.g., What does dioactive material in the late 1940s. In the 1950s,
unnecessary mean in a specific situation?). the government warned Eastman Kodak and
other film manufacturers about nuclear fallout
from atomic tests to prevent fogged film, but it
Origins of Research Participant
did not warn nearby citizens of health hazards.
Protection
In the 1960s, the U.S. army gave unsuspecting people in
Concern over the treatment of research partici soldiers LSD (a hallucinogenic drug), causing se comfort or~
pants arose after the revelation of gross viola rious trauma. Today, researchers widely recog comfort! 1M 4
tions of basic human rights in the name of nize these to be violations of two fundamental obedience stu4j
science. The most notorious violations were ethical principles: Avoid physical harm and ob Some sav ~
"medical experiments" conducted on Jews and tain informed consent. 5 knowledge gaia
CHAPTER 3 / ETHICS IN SOCIAL RESEARCH Sl
Physical Harm, Psychological Abuse, tential psychological harm that research partici
and Legal Jeopardy pants experienced. Others believe that the ex
treme stress and the risk of permanent harm
Social research can harm a research participant
were too great. Such an experiment could not be
in several ways: physical, psychological, and legal
conducted today because ofheightened sensitiv
harm, as well as harm to a person's career, repu
ity to the ethical issues involved.
tation, or income. Different types of harm are
Social researchers have created high levels of
more likely in other types ofresearch (e.g., in ex
anxiety or discomfort. They have exposed par
periments versus field research). It is a re
ticipants to gruesome photos; falsely told male
searcher's responsibility to be aware of all types
students that they have strong feminine person
of potential harm and to take specific actions to
ality traits; falsely told students that they have
minimize the risk to participants at all times. failed; created a situation of high fear (e.g.,
Physical Harm. Physical harm is rare. Even in smoke entering a room in which the door is
biomedical research, where the intervention into locked); asked participants to harm others;
a person's life is much greater, 3 to 5 percent of placed people in situations where they face social
studies involved any person who suffered any pressure to deny their convictions; and had par
harm.? A straightforward ethical principle is that ticipants lie, cheat, or steal." Researchers who
researchers should never cause physical harm. study helping behavior often place participants
An ethical researcher anticipates risks before be in emergency situations to see whether they will
ginning a study, including basic safety concerns lend assistance. For example, Piliavin and asso
(e.g., safe buildings, furniture, and equipment). ciates (1969) studied helping behavior in sub
This means that he or she screens out high-risk ways by faking someone's collapse onto the
subjects (those with heart conditions, mental floor. In the field experiment, the riders in the
breakdown, seizures, etc.) if great stress is in subway car were unaware of the experiment and
volved and anticipates possible sources of injury did not volunteer to participate in it.
or physical attacks on research participants or The only researchers who might even con
assistants. The researcher accepts moral and le sider conducting a study that purposely induces
gal responsibility for injury due to participation great stress or anxiety in research participants
in research and terminates a project immediately are very experienced and take all necessary pre
if he or she can no longer fully guarantee the cautions before inducing anxiety or discomfort.
physical safety of the people involved (see the The researchers should consult with others who
Zimbardo study in Box 3.1). have conducted similar studies and mental
health professionals as they plan the study. They
Psychological Abuse, Stress, or Loss of Self should screen out high-risk populations (e.g.,
Esteem. The risk of physical harm is rare, but those with emotional problems or weak hearts),
social researchers can place people in highly and arrange for emergency interventions or ter
stressful, embarrassing, anxiety-producing, or mination of the research if dangerous situations
unpleasant situations. Researchers want to learn arise. They must always obtain written informed
about people's responses in real-life, high-anxi consent (to be discussed) before the research
ety-producing situations, so they might place and debrief the people immediately afterward
people in realistic situations of psychological dis (i.e., explain any deception and what actually
comfort or stress. Is it unethical to cause dis happened in the study). Researchers should
comfort? The ethics of the famous Milgram never create unnecessary stress (i.e., beyond the
obedience study are still debated (see Box 3.1). minimal amount needed to create the desired ef
Some say that the precautions taken and the fect) or stress that lacks a very clear, legitimate
knowledge gained outweighed the stress and po research purpose. Knowing what "minimal
52 PART ONE / FOUNDATIONS
Stanley Milgram's obedience study (Milgram, 1963, later, in disguise, Humphreys used a deceptive story
1965, 1974) attempted to discover how the hor about a health survey to interview the subjects in
rors of the Holocaust under the Nazis could have oc their homes. Humphreys was careful to keep names
curred by examining the strength of social pressure in safety deposit boxes, and identifiers with subject
to obey authority, After signing "informed consent names were burned. He significantly advanced knowl
forms," subjects were assigned, in rigged random se edge of homosexuals who frequent "tearooms" and
lection,to be a "teacher" while a confederate was the overturned previous false beliefsabout them. There
"pupil," The teacher was to test the pupil's memory has been controversy over the study: The subjects
of word lists and increase the electric shock level if neverconsented; deception was used; and the names
the pupil made mistakes, The pupil was located in a could have been used to blackmail subjects, to end
nearby room, so the teacher could hear but not see marriages, or to initiate criminal prosecution.
the pupil. The shock apparatus was clearly labeled In the limbardo prison experiment (Zimbardo,
with increasing voltage. As the pupil made mistakes 1972,1973; Zimbardo et al., 1973, 1974), male
and the teacher turned switches, she or he also made students were divided into two role-playing groups:
noises as ifin severe pain.The researcher was present guards and prisoners. Before the experiment, volun
and made comments such as "You must go on" to teer students were given personality tests, and only
the teacher. Milgram reported, "Subjects were ob those in the "normal" range were chosen. Volunteers
served to sweat, tremble, stutter, bite their lips, signed up for two weeks, and prisoners were told that
groan and dig their fingernails into their flesh. These they would be under surveillance and would have 'It"
were characteristic rather than exceptional re some civil rights suspended, but that no physical
sponses to the experiment" (Milgram, 1963:375).
The percentage of subjects who wouldshock to dan
gerous levels was dramatically higher than expected.
Ethical concerns arose over the use of deception and
the extreme emotional stress experienced by
abuse was allowed. Ina simulated prison in the base
ment of a Stanford University building, prisoners
were deindividualized (dressed in standard uniforms
and called only by their numbers) and guards were
militarized (with uniforms, nightsticks, and reflective
.......- ..
..
•
~
-.
. .l1li£. . . . . .
subjects. sunglasses). Guards were told to maintain a reason filii
In Laud Humphreys's (Humphreys, 1975) tea able degree of order and served 8-hour shifts, while
room trade study (a study of male homosexual en prisoners were locked up 24 hours per day. Unex n
_e
-
counters in public restroorns), about 100 men were pectedly, the volunteers became too caught up in
&
observed engaging in sexualacts as Humphreys pre their roles. Prisoners became passive and disorga
tended to be a "watchqueen" (a voyeur and look nized, while guards became aggressive, arbitrary, and
out). Subjects were followed to their cars, and their dehumanizing. By the sixth day, Zimbardo called off
license numbers were secretly recorded. Names and the experiment for ethical reasons. The risk of per
addresses were obtained from police registers when manent psychological harm,and even physical harm,
Humphreys posed as a market researcher. One year was too great.
amount" means comes with experience. It is best sitive researchers reduces the chances of making
to begin with too little stress, risking a finding of an ethical misjudgment.
no effect, than to create too much. It is always Research that induces great stress and anx
wise to work in collaboration with other re iety in participants also carries the danger that
searchers when the risk to participants is high, experimenters will develop a callous or manip
because the involvement ofseveral ethically sen ulative attitude toward others. Researchers
CHAPTER 3 / ETHICS IN SOCIAL RESEARCH 53
have reported feeling guilt and regret after con undermining future social research. At the same
ducting experiments that caused psychological time, a researcher who failsto report illegal behav
harm to people. Experiments that place sub ior is indirectly permitting criminal behavior. He
jects in anxiety-producing situations may pro or she could be charged as an accessoryto a crime.
duce significant personal discomfort for the Cooperation with law-enforcement officials raises
ethical researcher. the question, Is the researcher a professional sci
entist who protects research participants in the
Legal Harm. A researcher is responsible for process of seeking knowledge, or a free-lance un
protecting research participants from increased dercover informant who is reallyworking for the
risk of arrest. If participation in research in police trying to "catch" criminals?
creases the risk of arrest, few individuals will
trust researchers or be willing to participate in Other Harm to Participants
future research. Potential legal harm is one crit
Research participants may face other types of
icism of Humphreys's 1975 tearoom trade study
harm. For example, a survey interview may cre
(see Box 3.1).
ate anxiety and discomfort if it asks people to re
A related ethical issue arises when a re
call unpleasant or traumatic events. An ethical
searcher learns of illegal activity when collecting
researcher must be sensitive to any harm to par
data. A researcher must weigh the value of pro
ticipants, consider precautions, and weigh po
tecting the researcher-subject relationship and
tential harm against potential benefits.
the benefits to future researchers against poten
Another type of harm is a negative impact
tial serious harm to innocent people. The re
on the careers, reputations, or incomes of re
searcher bears the cost of his or her judgment.
search participants. For example, a researcher
For example, in his field research on police, Van
conducts a survey of employees and concludes
Maanen (1982:114-115) reported seeing police
that the supervisor's performance is poor. As a
beat people and witnessing illegal acts and irreg
consequence, the supervisor loses her job. Or, a
ular procedures, but said, "On and following
researcher studies homeless people living on the
these troublesome incidents I followed police
street. The fmdings show that many engage in
custom: I kept my mouth shut."
petty illegal acts to get food. As a consequence, a
Field researchers in particular can face
city government"cracks down" on the petty ille
difficult ethical decisions. For example, when
gal acts and the homeless people can no longer
studying a mental institution, Taylor (1987) dis
eat. What is the researcher's responsibility? The
covered the mistreatment and abuse of inmates
ethical researcher considers the consequences of
by the staff. He had two choices: Abandon the
research for those being studied. The general
study and call for an immediate investigation, or
goal is not to cause any harm simply because
keep quiet and continue with the study for sev
someone was a research participant. However,
eral months, publicize the findings afterwards,
there is no set answer to such questions. A re
and then become an advocate to end the abuse.
searcher must evaluate each case, weigh poten
After weighing the situation, he followed the lat
tial harm against potential benefits, and bear the
ter course and is now an activist for the rights of
responsibility for the decision.
mental institution inmates.
In some studies, observing illegal behavior
Deception
may be central to the research project. If a re
searcher covertly observes and records illegal Has anyone ever told you a half-truth or lie to
behavior, then supplies the information to law-en get you to do something? How did you feel
forcement authorities, he or she isviolating ethical about it? Social researchers follow the ethical
standards regarding research participants and is principle of voluntary consent: Never force any
r
54 PART ONE I FOUNDATIONS
_ .
explicitly agree to participate. A person's right unethical. Even those who accept covert research
not to participate can be a critical issue when as ethical in certain situations say that it should "iripmtt""
ever the researcher uses deception, disguises the be used only when overt observation is impossi
research, or uses covert research methods. ble. Whenever possible, the researcher should
Social researchers sometimes deceive or lie inform participants of the observation immedi
to participants in field and experimental re ately afterwards and give them an opportunity
search. A researcher might misrepresent his or to express concerns.
her actions or true intentions for legitimate Deception and covert research may increase
methodological reasons. For example, if partici mistrust and cynicism as well as diminish public
pants knew the true purpose, they would modify respect for social research. Misrepresentation in
their behavior, making it impossible to learn of field research is analogous to being an under
their real behavior. Another situation occurs cover agent or government informer in nonde
when access to a research site would be impossi mocratic societies. The use of deception has a
ble if the researcher told the truth. Deception is long-term negative effect. It increases distrust
never preferable if the researcher can accomplish among people who are frequently studied and
the same thing without using deception. makes doing social research more difficult in the
Experimental researchers often deceive sub long term.
jects to prevent them from learning the hypoth
esis being tested and to reduce "reactive effects"
Informed Consent
(see Chapter 8). Deception is acceptable only if a
researcher can show that it has a clear, specific A fundamental ethical principle of social re
methodological purpose, and even then, the re search is: Never coerce anyone into participat
searcher should use it only to the minimal de ing; participation must be voluntary at all times.
gree necessary. Researchers who use deception Permission alone is not enough; people need to
should always obtain informed consent, never know what they are being asked to participate in 2. A statement
misrepresent risks, and always explain the actual so that they can make an informed decision. Par ated with
conditions to participants afterwards. You might ticipants can become aware of their rights and 3. A guarantee
ask, How can a researcher obtain prior informed what they are getting involved in when they read
consent and still use deception? He or she can and sign a statement giving informed consent
itYOfrec~
4. The identi
describe the basic procedures involved and con an agreement by participants stating they are where to
ceal only specific information about hypotheses willing to be in a study and they know some rights or .
being tested. thing about what the research procedure will in 5. AstatemerrtuJ
Sometimes field researchers use covert ob volve. untary and GIRl
servation to gain entry to field research settings. Governments vary in the requirement for out penatty 1
6. A statement of~
In studies of cults, small extremist political sects, informed consent. The U.S. federal government
illegal or deviant behavior, or behavior in a large does not require informed consent in all re be used ;
public area, it may be impossible to conduct re search involving human subjects. Nevertheless,
search if a researcher announces and discloses researchers should get written informed consent
7. A statement aIj
provided to ~
her or his true purpose. If a covert stance is not unless there are good reasons for not obtaining it involved 1
essential, a researcher should not use it. If he or (e.g., covert field research, use of secondary data,
8. An offer to ~
she does not know whether covert access is nec etc.) as judged by an institutional review board
essary, then a strategy of gradual disclosure may (IRB) (see the later discussion ofIRBs).
CHAPTER 3 / ETHICS IN SOCIAL RESEARCH SS
minimal conditions: (1) a legal guardian grants long as it meets three conditions: it is attached to
written permission and (2) the researcher fol a clear educational objective, the students have a
lows all standard ethical principles to protect choice of research experience or an alternative
participants from harm. For example, a re activity, and all other ethical principles of re
searcher wants to conduct a survey of high search are followed.
school students to learn about their sexual be
havior and drug/alcohol use. If the survey is con Avoid Creating New Inequalities. Another
ducted on schoolproperty, school officials must type of harm occurs when one group of people is
give officialpermission. For any research partic denied a service or benefit as a result of partici
ipant who is a legal minor (usually under 18 pating in a research project. For example, a re
years old), written parental permission is searcher might have a new treatment for people
needed. It is best to ask permission from each with a terrible disease, such as acquired immune
student, as well. deficiency syndrome (AIDS). To determine the
The use of coercion to participate can be a effects of the new treatment, half the group is
tricky issue, and it depends on the specifics of a randomly chosen to receive the treatment, while
situation. For example, a convicted criminal others receive nothing. The design may clearly
faces the alternative of imprisonment or partici show whether the treatment is effective, but par
pation in an experimental rehabilitation pro ticipants in the group who receive no treatment
gram. The convicted criminal may not believe in may die. Of course, those receiving the treat
the benefits of the program, but the researcher ment may also die, until more is known about
may believe that it will help the criminal. This is whether it is effective. Is it ethical to deny people
a case of coercion. A researcher must honestly who have been randomly assigned to a study
judge whether the benefits to the criminal and to group the potentially life-saving treatment?
society greatly outweigh the ethical prohibition What if a clear, definitive test of whether a treat
on coercion. This is risky. History shows many ment is effective requires that one study group
cases in which a researcher believed he or she receive no treatment?
was doing something "for the good of' someone A researcher can reduce creating a new in
in a powerless position (e.g., prisoners, students, equality among research participants when the
homosexuals), but it turned out that the "good" outcome has a major impact on their survival or
actually was for the researcher or a powerful or quality oflife in three ways.First, the people who
ganization in society, and it did more harm than do not receive the "new, improved" treatment
good to the research participant. continue to receive the best previously accept
You may have been in a social science class able treatment. In other words, instead of deny
in which a teacher required you to participate as ing all assistance, they get the best treatment
a subject in a research project. This is a special available prior to the new one being tested. This
case of coercion and is usually ethical. Teachers ensures that people will not suffer in absolute
have made three arguments in favor of requiring terms, even if they temporarily fall behind in rel
student participation: (1) it would be difficult ative terms. Second, researchers can use a
and prohibitively expensive to get participants crossover design, which is when a study group
otherwise, (2) the knowledge created from re that gets no treatment in the first phase of the
search with students serving as subjects benefits experiment becomes the group with the treat
future students and society, and (3) students will ment in the second phase, and viceversa. Finally,
learn more about research by experiencing it di the researcher continuously monitors results. If
rectly in a realistic research setting. Of the three it appears early in the study that the new treat
arguments, only the third justifies limited coer ment is highly effective, the researcher should
cion. This limited coercion is acceptable only as offer it to those in the control group. Also, in
CHAPTER 3 / ETHICS IN SOCIAL RESEARCH 57
college students and ask many questions on a obligated them to destroy the records rather
questionnaire, including age, sex, religion, and than give them to government officials.
hometown. The sample contains one 22-year-old
Jewishmale born in Stratford, Ontario. With this
information, you could find out who the specific
individual is and how he answered very personal
Confidentiality can sometimes protect re
search participants from legal or physical harm.
In a study of illegal drug users in rural Ohio,
Draus and associates (2005) took great care to ..
.............
- • .-IIiai. .
......
~
..........-.
questions, even though his name was not directly protect the research participants. They con
recorded on the questionnaire. ducted interviews in large multiuse buildings, -
avoided references to illegal drugs in written
Confidentiality. Even if a researcher cannot documents, did not mention of names of drug . . . .asAr
6
protect anonymity, he or she alwaysshould pro dealers and locations, and did not affiliate with 6r "
tect participant confidentiality. Anonymity drug rehabilitation services, which had ties to iaaJIs; . .
means protecting the identity of specific individ law enforcement. They noted, "We intentionally
uals from being known. Confidentiality can avoided contact with local police, prosecutors,
include information with participant names at or parole officers" and "surveillance of the pro
tached, but the researcher holds it in confidence ject by local law enforcement was a source of
or keeps it secret from public disclosure. The re concern" (p. 169). In other situations, other
searcher releasesdata in a way that does not per principles may take precedence over protecting
mit linking specific individuals to responses and research participant confidentiality. For exam
presents it publicly only in an aggregate form ple, when studying patients in a mental hospital,
(e.g., as percentages, statistical means, etc.). a researcher discovers that a patient is preparing
A researcher can provide anonymity with to kill an attendant. The researcher must weigh
out confidentiality, or vice versa, although they the benefit of confidentiality against the poten
usually go together. Anonymity without confi tial harm to the attendant.
dentiality occurs if all the details about a specific Social researchers can pay high personal
individual are made public, but the individual's costs for being ethical. Although he was never
name is withheld. Confidentiality without accused or convicted of breaking any law and he
anonymity occurs if detailed information is not closely followed the ethical principles of the
made public, but a researcher privately links in American Sociological Association, Professor
dividual names to specific responses. Rik Scarce spent 16 weeks in a Spokane jail for
Attempts to protect the identity of subjects contempt of court because he refused to testify
from public disclosure has resulted in elaborate before a grand jury and break the confidentiality
procedures: eliciting anonymous responses, us of social research data. Scarce had been studying
ing a third-party custodian who holds the key to radical animal liberation groups and had already
coded lists, or using the random-response tech published one book on the subject. He had in
nique. Past abuses suggest that such measures terviewed a research participant who was sus
may be necessary. For example, Diener and pected ofleading a group that broke into animal
Crandall (1978:70) reported that during the facilities and caused $150,000 damage. Two
1950s,the U.S. State Department and the FBIre judges refused to acknowledge the confidential
quested research records on individuals who had ity of social research data."
been involved in the famous Kinsey sex study. A special concern with anonymity and con
The Kinsey Sex Institute refused to comply with fidentiality arises when a researcher studies
the government. The institute threatened to de "captive" populations (e.g., students, prisoners,
stroy all records rather than release any. Eventu employees, patients, and soldiers). Gatekeepers,
ally,the government agenciesbacked down. The or those in positions of authority, may restrict
moral duty and ethical code of the researchers access unless they receive information on sub
CHAPTER 3 / ETHICS IN SOCIAL RESEARCH 59
jects.?For example, a researcher studies drug use pact of research procedures on human partici
and sexual activity among high school students. pants and applies ethical guidelines by reviewing
School authorities agree to cooperate under two research procedures at a preliminary stage when
conditions: (1) students need parental permis first proposed. Some forms of research, educa
sion to participate and (2) school officials get the tional tests, normal educational practice, most
names of all drug users and sexually active stu nonsensitive surveys, most observation ofpublic
dents in order to assist the students with coun behavior, and studies of existing data in which
seling and to inform the students' parents. An individuals cannot be identified are exempt
ethical researcher will refuse to continue rather from institutional review boards.
than meet the second condition. Even though
the officials claim to have the participants' best
interests in mind, the privacy of participants will
ETHICS AND THE SCIENTIFIC
be violated and they could be in legal harm as a
COMMUNITY
result of disclosure. If the school officials really
want to assist the students and not use re Physicians, attorneys, family counselors, social
searchers as spies, they could develop an out workers, and other professionals have a code of
reach program of their own. ethics and peer review boards or licensing regu
lations. The codes formalize professional stan
dards and provide guidance when questions
Mandated Protections of Research
arise in practice. Social researchers do not pro
Participants
vide a service for a fee, they receive limited ethi
Many governments have regulations and laws to cal training, and rarely are they licensed. They
protect research participants and their rights. In incorporate ethical concerns into research be
the United States, legal restraint is found in rules cause it is morally and socially responsible, and
and regulations issued by the U.S. Department to protect social research from charges of insen
of Health and Human Services Office for the sitivity or abusing people. Professional social sci
Protection from Research Risks.Although this is ence associations have codes of ethics that
only one federal agency, most researchers and identify proper and improper behavior. They
other government agencies look to it for guid represent a consensus of professionals on ethics.
ance. The National Research Act (1974) estab All researchers may not agree on all ethical is
lished the National Commission for the sues, and ethical rules are subject to interpreta
Protection of Human Subjects in Biomedical tion, but researchers are expected to uphold
and Behavioral Research, which significantly ex ethical standards as part of their membership in
panded regulations and required informed con a professional community.
sent in most social research. The responsibility Codes of research ethics can be traced to the
for safeguarding ethical standards was assigned Nuremberg code adopted during the Nurem
to research institutes and universities. The De berg Military Tribunal on Nazi war crimes held
partment of Health and Human Services issued by the Allied Powers immediately after World
regulations in 1981, which are still in force. Fed War II. The code, developed as a response to the
eral regulations follow a biomedical model and cruelty ofconcentration camp experiments, out
protect subjects from physical harm. Other rules lines ethical principles and rights of human sub
require institutional review boards (IREs) at all jects. These include the following:
research institutes, colleges, and universities to
review all use of human subjects. An IRB is a • The principle of voluntary consent
committee ofresearchers and community mem • Avoidance of unnecessary physical and
bers that oversees, monitors, and reviewsthe im mental suffering
r 60 PART ONE I FOUNDATIONS
,...
--..........
• The principle that the results should be for
the good of society and unattainable by any
gain.
• Some form of informed consent is highly recom
_an b..,
The principles in the Nuremberg code dealt • Honor all guarantees of privacy, confidentiality,
and anonymity.
with the treatment of human subjects and fo
cused on medical experimentation, but they be • Do not coerce or humiliate subjects. • ...a.........
came the basis for the ethical codes in social • Use deception only if needed, and always accom .td~
............
....
research. Similar codes of human rights, such as pany it with debriefing. Jdilsrthr=.-_
the 1948 Universal Declaration of Human ...._~~
• Use the research method that is appropriate to a
Rights by the United Nations and the 1964 Dec topic.
laration of Helsinki, also have implications for Z
social researchers. Box 3.3 lists some of the basic • Detect and remove undesirable consequences to
research subjects.
principles of ethical social research.
Professional social science associations have • Anticipate repercussions of the research or publi
committees that review codes of ethics and hear cation of results.
about possible violations, but there is no formal • Identify the sponsor who funded the research.
policing of the codes. The penalty for a minor vi
• Cooperate with host nations when doing compar
olation rarely goes beyond a letter of complaint.
ative research.
If laws have not been violated, the most extreme
penalty is the negative publicity surrounding a • Release the details of the study design with the
well-documented and serious ethical violation. results.
The publicity may result in the loss of employ • Make interpretations of results consistent with
ment, a refusal to publish the researcher's find the data.
ings in scholarlyjournals, and a prohibition from • Use high methodological standards and strive for
receiving funding for research-in other words, accuracy.
banishment from the community of professional
• Do not conduct secret research.
researchers.
Codes of ethics do more than codify think
ing and provide individual researchers with
guidance; they also help universities and other
institutions defend ethical research against
abuses. For example, after interviewing 24 staff the university and demanded to know who on
members and conducting observations, a re their staff had talked to the researcher, with im
searcher in 1994 documented that the staff at the plications that there might be reprisals. The uni
Milwaukee Public Defenders Office were seri versity administration defended the researcher
ously overworked and could not effectivelypro and refused to release the information, citing
vide legal defense for poor people. Learning of widely accepted codes that protect human re
the findings, top officials at the office contacted search participants. 10
CHAPTER 3 / ETHICS IN SOCIAL RESEARCH 61
dents with those of students many years ago. gives the sponsors whatever they want, even if it
Many teachers, school principals, superinten is ethically wrong, or a professional who is oblig
dents, and school boards pressured for a type of ated to teach, guide, or even oppose sponsors in
result that would allow them to report to par the service of higher moral principles.
ents and voters that their school district was A researcher should ask: Why would spon
"above average."l2 sors want the social research conducted if they
are not interested in using the findings or in the
Limits on How to Conduct Studies. Is it ethi truth? The answer is that some sponsors are not
cally acceptable for a sponsor to limit research by interested in the truth and have no respect for
defining what a researcher can study or by limit the scientific process. They see social research
ing the techniques used? Sponsors can legiti only as "a cover" to legitimate a decision or prac
mately set some conditions on research tice that they plan to carry out, but use research
techniques used (e.g., survey versus experiment) to justify their action or deflect criticism. They
and limit costs for research. However, the re abuse the researcher's professional status and
searcher must follow generally accepted research
methods. Researchers must give a realistic ap
undermine integrity of science to advance their
own narrow goals. They are being deceitful by
_...a_...
.. .
praisal of what can be accomplished for a given
level of funding. The issue of limits is common
in contract research, when a firm or government
agency asks for work on a particular research
trying to "cash in" on social research's reputa
tion for honesty. When such a situation occurs,
an ethical researcher has a moral responsibility
to expose and stop the abuse.
.. .. '
'--.-
project. There is often a tradeoff between quality
and cost. Plus, once the research begins, a re
Suppressing Findings
searcher may need to redesign the project, or
costs may be higher. The contract procedure What happens if you conduct a study and the
......-
.....
makes midstream changes difficult. A researcher findings make the sponsor look bad, then the
may find that he or she is forced by the contract sponsor does not want to release the results? i 5 "rl
to use research procedures or methods that are This is a common situation for many applied re g "
...-..-...
less than ideal. The researcher then confronts a searchers. For example, a sociologist conducted
dilemma: complete the contract and do low a study for a state government lottery commis ~to
.-s
....' "
01' to
quality research, or fail to fulfill the contract and sion on the effects of state government-spon
lose money and future jobs. sored gambling. After she completed the report, "
A researcher should refuse to continue a but before releasing it to the public, the commis dcKmne.a ..
study if he or she cannot uphold generally ac sion asked her to remove sections that outlined potitial appai. .
cepted standards of research. If a sponsor de the many negative social effects of gambling and who hadpm.-"
mands a biased sample or leading survey to eliminate her recommendations to create so -asdlargOO
questions, the ethical researcher should refuse to cial services to help the anticipated increase of ports to ~
cooperate. If a legitimate study shows a spon compulsive gamblers. The researcher found her documented
sor's pet idea or project to be disaster, a re self in a difficult position and faced two conflict mdglobal-'"
searcher may anticipate the end of employment ing values: do what the sponsor requested and Inspol-1IllII
or pressure to violate professional research stan paid for, or reveal the truth to the public but ~ CODlibl"
dards. In the long run, the sponsor, the re then suffer the consequencesrP bt?ginning the
searcher, the scientific community, and society Government agencies may suppress scien effect. It mae
in general are harmed by the violation of sound tific information that contradicts official policy ...i thout such .a
research practice. The researcher has to decide or embarrasses high officials. Retaliation against researchers ...no
whether he or she is a "hired hand" who always social researchers employed by government do so. Altenlali.
CHAPTER 3 / ETHICS IN SOCiAl RESEARCH 63
agencies who make the information public also sponsor's criticism and hostility and release the
occurs. In 2004, leading scientists, Nobel laure findings over the sponsor's objections. Most re
ates, leading medical experts, former federal searchers prefer the first choice, since the second
agency directors, and university chairs and pres one may scare away future sponsors.
idents signed a statement voicing concern over Social researchers sometimes self-censor or
the misuse of science by the George W. Bush ad delay the release of findings. They do this to pro
ministration. Major accusations included su tect the identity of informants, to maintain ac
pressing research findings and stacking scientific cess to a research site, to hold on to their jobs, or
advisory committees with ideologically commit to protect the personal safety of themselves or
ted advocates rather than impartial scientists. family members. IS This is a less disturbing type
Other complaints included limiting the public of censorship because it is not imposed by an
release studies on auto-saftey data, negative data outside power. It is done by someone who is
about pharmaceuticals, and studies on pollu close to the research and who is knowledgeable
tion. These involved industries that were major about possible consequences. Researchers shoul
political campaign supporters of the administra der the ultimate responsibility for their research.
tion. Additional criticisms appeared over re Often, they can draw on many different re
moving a government fact sheet citing studies sources but they face many competing pressures,
that showed no relationship between abortions as well.
and breast cancer, removing study results about
positive effects of condom use in pregnancy pre
Concealing the True Sponsor
vention, holding back information on positive
aspects of stem cell research, and requiring re Is it ethical to keep the identity of a sponsor se
searchers to revise their study findings on dan cret? For example, an abortion clinic funds a
gers of arctic oil drilling and endangered species study on members of religious groups who op
so they would conform to the administration's pose abortion, but it tells the researcher not to
political agenda. An independent 2005 survey of reveal to participants who is funding the study.
460 biologists who worked for Fisheries Service The researcher must balance the ethical rule that
found that about one-third said they were di it is usually best to reveal a sponsor's identity to
rected to suppress findings for nonscientific rea participants against both the sponsor's desire for
sons or to inappropriately exclude or alter confidentiality and reduced cooperation by par
technical information from an official scientific ticipants in the study. In general, an ethical re
document. In June 2005, it was discovered that a searcher will tell subjects who is sponsoring a
political appointee without scientific training study unless there is a strong methodological
who had previously been an oil industry lobbyist reason for not doing so. When reporting or pub
was charged with editing official government re lishing results, the ethical mandate is very clear:
ports to play down the research findings that A researcher must alwaysreveal the sponsor who
documented linkages between such emissions provides funds for a study.
and global warming. 14
In sponsored research, a researcher can ne
gotiate conditions for releasing findings prior to
POLITICS OF RESEARCH
beginning the study and sign a contract to that
effect. It may be unwise to conduct the study Ethics largely address moral concerns and stan
without such a guarantee, although competing dards of professional conduct in research that
researchers who have fewer ethical scruples may are under the researcher's control. Political con
do so. Alternatively, a researcher can accept the cerns also affectsocial research, but many are be
64 PART ONE / FOUNDATIONS
1~"\.\.~\\\.~ '-~"\.\.\."'t~\~\ "'t~%~'O.."'t,-\\.~"'t~.. \.\\.~ ~~\\.\.\.~ ~\ 'l'l'O.."\.\.\. \.~ ~"'t~\.~,-\. ~"'t 'O..~'l'O.."\.\.'-~ \.~~\."'t ~~\\.\.\.'-'O..\
research usually involve actions by organized ad fmancial position, and lear social researchers
vocacy groups, powerful interests in society, might yield findings showing that their actions
governments, or politicians trying to restrict or are harmful to the public or some sectors of
control the direction of social research. Histori society. And third, some people in society do not
cally, the politieal influence over social research respect the ideals of science to pursue truth/
has included preventing researchers from con knowledge and instead view scientific research
ducting a study, cutting off or redirecting funds only as cover for advancing private interests (see
for research, harassing individual researchers, Box 3.4).
censoring the release of research findings, and
...• ...-...
using social research as a cover or guise for
covert government intelligence/military actions.
VALUE-FREE AND OBJECTIVE
\~J1
~
~ I I r I .
Attempts at control seem motivated by a fear mean following clear and publicly accepted re ......aIIIIII......
that researchers might discover something dam search procedures and not haphazard, personal JIIIlIPIr ~G, . . .
aging if they have freedom of inquiry. This ones.
shows that free scientific inquiry is connected to The three approaches to social science that
fundamental political ideals of open public de you read about in Chapter 2 hold different posi
bate, democracy, and freedom of expression. tions on the importance of value-free, objective
The attempts to block and steer social re research. Positivism puts a high value on such
search have three main reasons. First, some peo research. An interpretive approach seriously
ple defend or advance positions and knowledge questions whether it is possible, since human
that originate in deeply held ideological, politi values/beliefs pervade all aspects of human ac
cal, or religious beliefs, and fear that social re tivities, including research. Instead of eliminat
searchers might produce knowledge that ing values and subjective dimension, it suggestsa
contradicts them. Second, powerful interests relativist stance-no single value position is bet
CHAPTER 3 I ETHICS IN SOCIAL RESEARCH 65
Michael Burawoy (2004, 2005) distinguished clients. Both rely on professional social science for
among four ideal types of social research: policy, pro theories, bodies of knowledge, and techniques for
fessional, critical, and public. The aim of public soci gathering and analyzing data. Critical social science,
ology (or social science, more generally) is to enrich as was discussed in Chapter 2, emphasizes demysti
public debate over moral and political issues by in fying and raising questioning about basic conditions.
fusing such debate with social theory and research. The primary audience for professional and critical
Public sociology frequently overlaps with action-ori social science are members of the scientific commu
ented research. Burawoy argued that the place of so nity, whereas the main audience for public and policy
cial research in society centers on how one answers research are nonexperts and practitioners. Both crit
two questions: Knowledge for whom? and Knowl ical and public social science seek to infuse a moral,
edge for what? The first question focuses on the value dimension into social research and they try to
sources of research questions and how results are generate debates over moral-political values. Profes
used. The second question looks at the source of re sional and policy social science are less concerned
search goals. Are they handed down by some exter about debates over moral or value issues and may
nal sponsor or agency or are they concerned with avoid them. Instead, their focus is more on being ef
debates over larger societal political-moral issues? fective in providing advances to basic knowledge or
Public social science tries to generate a conversation specific solutions to practical problems. Both public
or debate between researchers and public. By con and policy social science are applied research and
strast, policy social science focuses on finding solu have a relevance beyond the community of scientific
tions to specific problems as defined by sponsors or researchers.
ter than any other. A critical approach also ques or her own value position explicit, reflect care
tions value-free research, but sees it often as a fully on reasons for doing a study and the proce
sham. dures used, and communicate in a candid, clear
Value free means free of everyone's values manner exactlyhow the study was conducted. In
except those of science, and objective means fol this way, other researchers see the role of a re
lowing established rules or procedures that some searcher's values and judge for themselves
people created, without considering who they whether the values unfairly influenced a study's
represent and how they created the rules. In findings.
other words, a critical approach sees all research Even highly positivist researchers who ad
as containing some values, so those who claim to vocate value-free and objective studies admit a
be value free are just hiding theirs. Those who limited place for some personal, moral values.
follow an interpretive and critical approach and Many hold that a researcher's personal, moral
reject value-free research do not embrace sloppy position can enter when it comes to deciding
and haphazard research, research procedures what topic to study and how to disseminate
that follow a particular researcher's whims, or a findings. Being value free and objective only
study that has a foregone conclusion and auto refers to actually conducting the study. This
matically supports a specific value position. means that you can study the issues you believe
They believe that a researcher should make his to be important and after completing a study
r
66 PART ONE / FOUNDATIONS
U.S. Fish and Wildlife Service" (February 9, ber 10, 2004). James Glanz, "Scientists Say Ad
2005)." Specific Examples ofthe Abuse of Science ministration Distorts Facts," New York Times
www.ucsusa.org/global_environment/rsiJpage.ef (February 19, 2004). Dylan O. Krider, "The Politi
m?pageID=1398, downloaded August 3, 2005. cization of Science in the Bush Administration,"
"Summary of National Oceanic & Atmospheric Skeptic Vol. 11, Number 2 (2004) at www. Skep
Administration Fisheries Service Scientist Survey" tic.com. C. Orstein, "Politics Trumps Science in
by Union of Concerned Scientists (June 2005). E. Condom Fact Sheet," New York Times (December
Shogren, "Researchers Accuse Bush ofManipulat 27, 2002). "Scientist Says Officials Ignored Advice
ing Science," LosAngeles Times (July 9, 2004). Jef on Water Levels," Washington Post (October 29,
frey McCracker, "Government Bans Release of 2002).
Auto-Safety Data," Detroit Free Press (August 19, 15. See Adler and Adler (1993).
2004). Garddiner Harris, "Lawmaker Says FDA 16. See Neuman (2003, Chapter 16) for a discussion
Held Back Drug Data," New York Times (Septem of political issues in social research.
CHAPTEI 4
68
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 69
' CHA"~
fessionallibrarians. Nonetheless, distinguishing and only then if a library pays for a special on Scholarly join
among types of publications is essential to build line subscription service. once a year or 35 i
on a body of research. One of the best ways to Once you locate a scholarly journal that re pear four to six
learn to distinguish among types of publications ports on social science research studies, you need Sociological Ouars
is to read many articles in scholarly journals. to make sure that a particular article presents the To assist in loca
The number of journals varies by field. Psy results of a study, since the journal may have scholars have des
chology has over 400 journals, whereas sociol other types of articles. It is easier to identify scholarly journals
ogy has about 250 scholarly journals, political quantitative studies because they usually have a issue is assigned a
science and communication have slightly fewer methods or data section and charts, statistical sue number. This
than sociology, anthropology-archaeology and formulas, and tables of numbers. Qualitative re locate an article. S
social work have about 100, urban studies and search articles are more difficult to identify, and details such as aut!
women studies have about 50, and there are many students confuse them with theoretical es is called an article
about a dozen journals in criminology. Each says, literature review articles, idea-discussion ographies. When
publishes from a few dozen to over 100 articles a essays, policy recommendations, book reviews, begins with volUIJI
year. and legal case analyses. To distinguish among increasing the D1
Many, but not all, scholarly journals may be these types requires a good grasp of the varieties most journals faa
viewed via the Internet. Usually, this is limited to of research as well as experience in reading many enough exception
selected years and to libraries that paid special articles. tention to citation
subscription fees. A few Internet services provide Your college library has a section for schol nals, each volume!
full, exact copies of scholarly journal articles arly journals and magazines, or, in some cases, issue with volumi
over the Internet. For example, JSTOR provides they may be mixed with books. Look at a map of means that the .iOll
exact copies, but only for a small number of library facilities or ask a librarian to find this sec 52 years. Most, bI
scholarly journals and only for past years. Other tion. The most recent issues, which look like thin publishing cycle it
Internet services, such as EBSCO HOST, offer a paperbacks or thick magazines, are often physi Most journals
full-text version of recent articles for a limited cally separate in a "current periodicals" section. by issue. The first
number of scholarly journals, but they are not in This is done to store them temporarily and make gins with page 1, OIl
the same format as a print version of an article. them available until the library receives all the is throughout the ell
This can make it impossible to find a specific sues of a volume. Most often, libraries bind all is first page of volun
page number or see an exact copy of a chart. It is sues of a volume together as a book before Most journals hal
best to visit the library and see what a full-print adding them to their permanent collections. and a table of cons
version of the scholarly article looks like. An Scholarly journals from many different title, the author's
added benefit is that it makes it easy for you to fields are placed together with popular maga page on which till
browse the Table of Contents of the journals. zines. All are periodicals, or serials in the jargon as few as 1 or 2 ..
Browsing can be very useful for generating new of librarians. Thus, you will find popular maga have 8 to 18 artidl
ideas for research topics, seeing an established zines (e.g., Time, Road and Track, Cosmopolitan, long. The article
topic in creative ways, or learning how to expand and Atlantic Monthly) next to journals for as summaries on til
an idea into new areas. Only a tiny handful of tronomy, chemistry, mathematics, literature, grouped together!
new Internet-only scholarly journals, called e and philosophy as well as sociology, psychology, Many librarie
journals, present peer-reviewed research studies social work, and education. Some fields have copies of older jOil
(e.g., Sociological Research Online, Current Re more scholarly journals than others. The "pure" they retain only.
search in Social Psychology, andlournal of World academic fields usually have more than the "ap hundreds of schol
Systems Research). Eventually, the Internet for plied" or practical fields such as marketing or so fields, with each ~
mat may replace print versions. But for now, 99 cial work. The journals are listed by title in a card Only the large res
percent of scholarly journals are available in catalog or a computerized catalog system. Li of them. You IruIII
print form and about one-third of these are also braries can provide you with a list of the period photocopy of ~
available in a full-text version over the Internet icals to which they subscribe. through an interl
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 73
Scholarly journals are published as rarely as which libraries lend books or materials to other
once a year or as frequently as weekly. Most ap libraries. Few libraries allow people to check out
pear four to six times a year. For example, recent issues of scholarly journals. You should
Sociological Quarterly appears four times a year. plan to use these in the library. Some, not all,
To assist in locating articles, librarians and scholarly journals are available via the Internet.
scholars have developed a system for tracking Once you find the periodicals section, wan
scholarly journals and the articles in them. Each der down the aisles and skim what is on the
issue is assigned a date, volume number, and is shelves. You will see volumes containing many
sue number. This information makes it easier to research reports. Each title of a scholarly journal
locate an article. Such information-along with has a call number like that of a regular library
details such as author, title, and page number book. Libraries often arrange them alphabeti
is called an article's citation and is used in bibli cally by title. Because journals change titles, it
ographies. When a journal is first published, it may create confusion if the journal is shelved
begins with volume 1, number 1, and continues under its original title.
increasing the numbers thereafter. Although
most journals follow a similar system, there are Citation Formats. An article's citation is the
enough exceptions that you have to pay close at key to locating it. Suppose you want to read the
tention to citation information. For most jour study by Weitzer and Tuch (2005) on percep
nals, each volume is one year. Ifyou see a journal tions of police misconduct discussed in Chapter
issue with volume 52, for example, it probably 2. Its citation is as follows:
means that the journal has been in existence for
52 years. Most, but not all, journals begin their Weitzer, Ronald, and Steven Tuch. 2005.
publishing cycle in January. "Racially Biased Policing: Determinants of
Most journals number pages by volume, not Citizen Perceptions." Social Forces
by issue. The first issue of a volume usually be 83:1009-1030.
gins with page 1, and page numbering continues
throughout the entire volume. For example, the This tells you that you can find the article in
first page ofvolume 52, issue 4, may be page 547. an issue of Social Forces published in 2005. The
Most journals have an index for each volume citation does not provide the issue or month, but
and a table ofcontents for each issue that lists the it gives the volume number, 83, and the page
title, the author's or authors' names, and the numbers, 1009 to 1030.
page on which the article begins. Issues contain There are many ways to cite the literature.
as few as 1 or 2 articles or as many as 50. Most Formats for citing literature in the text itself
have 8 to 18 articles, which may be 5 to 50 pages vary, with the internal citation format of using
long. The articles often have abstracts, short an author's last name and date of publication in
summaries on the first page of the article or parentheses being very popular. The full citation
grouped together at the beginning of the issue. appears in a separate bibliography or reference
Many libraries do not retain physical, paper section. There are many styles for full citations of
copies of older journals. To save space and costs, journal articles, with books and other types of
they retain only microfilm versions. There are works each having a separate style. When citing
hundreds ofscholarly journals in most academic articles, it is best to check with an instructor,
fields, with each costing $50 to $2,500 per year. journal, or other outlet for the desired format.
Only the large research libraries subscribe to all Almost all include the names of authors, article
of them. You may have to borrow a journal or title, journal name, and volume and page num
photocopy of an article from a distant library bers. Beyond these basic elements, there is great
through an interlibrary loan service, a system by variety. Some include the authors' first names,
74 PART ONE I FOUNDATIONS
others use initials only. Some include all authors, scriptions and complex theoretical or philo
others give only the first one. Some include in sophical discussions usually appear as books. Fi FIGURE 4.1
formation on the issue or month of publication, nally, an author who wants to communicate to
The oldest jouma/4
others do not (see Figure 4.1). scholarly peers and to the educated public may ginity pledges by_
Citation formats can get complex. Two ma write a book that bridges the scholarly, academic 2001 issue (nu~
jor reference tools on the topic in social science style and a popular nonfiction style. journal's 106th ~
are Chicago Manual of Style, which has nearly 80 Locating original research articles in books logical Review (ASRJj
pages on bibliographies and reference formats, can be difficult because there is no single source 1
and American Psychological Association Publica listing them. Three types of books contain col ASRStyle
tion Manual, which devotes about 60 pages to lections of articles or research reports. The first is Bearman, Peter ~
the topic. In sociology, the American Sociological designed for teaching purposes. Such books, Americanj~
Review style, with 2 pages of style instructions, is called readers, may include original research re
widely followed. ports. Usually, articles on a topic from scholarly APA Style
1
journals are gathered and edited to be easier for Bearman, P., and Bit
Books. Books communicate many types of in nonspecialists to read and understand. journalof~
formation, provoke thought, and entertain. The second type of collection is designed for
Other Styles 1
There are many types of books: picture books, scholars and may gather journal articles or may
textbooks, short story books, novels, popular fic contain original research or theoretical essayson Bearman, P., and H.~
tion or nonfiction, religious books, children's a specific topic. Some collections contain articles ofSociology 1
books, and others. Our concern here is with from journals that are difficult to locate. They Bearman, Peter and
those books containing reports of original re may include original research reports organized "Promising t:he~
search or collections of research articles. Li around a specialized topic. The table of contents Bearman, P. and BUll
braries shelve these books and assign call lists the titles and authors. Libraries shelve these can joumal of~
Bearman, Peter and
numbers to them, as they do with other types of collections with other books, and some library
"Promising 1:he.
books. You can find citation information on catalog systems include them.
(4):859-912..1
them (e.g., title, author, publisher) in the li Citations or references to books are shorter
Bearman, P. and H. III
brary's catalog system. than article citations. They include the author's
joumalof~
It is not easy to distinguish a book that re name, book title, year and place of publication, Peter Bearmanand ~
ports on research from other books. You are and publisher's name. joumalof~
more likely to find such books in a college or
I
university library. Some publishers, such as uni Dissertations. All graduate students who re
versity presses, specialize in publishing them. ceive the Ph.D. degree are required to complete
1
Nevertheless, there is no guaranteed method for a work of original research, which they write up example, Dis~
identifying one without reading it. as a dissertation thesis. The dissertation is bound dissertations wnh ~
Some types of social research are more likely and shelved in the library of the university that versities. This ~
to appear in book form than others. For exam granted the Ph.D. About half of all dissertations contains an al>str.iIl
ple, studies by anthropologists and historians are are eventually published as books or articles. Be can borrow IIlO5t ~
more likely to appear in book-length reports cause dissertations report on original research, loan from the ~
than are those of economists or psychologists. they can be valuable sources of information. university permits •
Yet, some anthropological and historical studies Some students who receive the master's degree
are articles, and some economic and psycholog conduct original research and write a master's Governmmt~
ical studies appear as books. In education, social thesis, but fewer master's theses involve serious ment of the c~
work, sociology, and political science, the results research, and they are much more difficult to lo other nations, state,
oflong, complex studies may appear both in two cate than unpublished dissertations. ments, the Lnited I
or three articles and in book form. Studies that Specialized indexes list dissertations com tional agencies SUI
involve detailed clinical or ethnographic de pleted by students at accredited universities. For sponsor studies _
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 75
---------------------.
ASRStyle
Bearman, Peter and Hannah BUckner. 2001. "Promising the Future: Virginity Pledges and First Intercourse."
American Journal ofSaciology 106:859-912.
APAStyle
Bearman, P., and BUckner, H. (2001). Promising the future: Virginity pledges and first intercourse. American
Journal of Sociology 106, 859-912.
Other Styles
Bearman, P., and H. Buckner. "Promising the Future: Virginity Pledges and First Intercourse," American Journal
of Sociology 106 (2001 ), 859-912.
Bearman, Peter and Hannah Buckner, 2001.
"Promising the future: Virginity pledges and first Intercourse." Am.J. af Sociol. 106:859- 91 2.
Bearman, P. and Buckner, H. (2001). "Promising the Future: Virginity Pledges and First Intercourse." Ameri
canJournal of Sociology 106 (January): 859-912.
Bearman, Peter and Hannah Buckner. 2001 .
"Promising the future: Virginity pledges and first Intercourse." American Journal of Sociology 106
(4):859-912.
Bearman,P.and H. BUckner. (2001 ). "Promising the future: Virginity pledges and first intercourse." American
Journal of Sociology 106,859-912.
Peter Bearman and Hannah BUckner, "Promising the Future: Virginity Pledges and First Intercourse," American
Journal of Sociology 106, no. 4 (2001): 859-912.
example, Dissertation Abstracts Internationallists search. Many college and university libraries
dissertations with their authors, titles, and uni have these documents in their holdings, usually
versities. This index is organized by topic and in a special "government documents" section.
contains an abstract of each dissertation. You These reports are rarely found in the catalog sys
can borrow most dissertations via interlibrary tem. You must use specialized lists of publica
loan from the degree-granting university if the tions and indexes, usually with the help of a
university permits this. librarian, to locate these reports. Most college
and university libraries hold only the most fre
Government Documents. The federal govern quently requested documents and reports.
ment of the United States, the governments of
other nations, state- or provincial-level govern Policy Reports and Presented Papers. A re
ments, the United Nations, and other interna searcher conducting a thorough review ofthe lit
tional agencies such as the World Bank, all erature will examine these two sources, which
sponsor studies and publish reports of the re are difficult for all but the trained specialist to
76 PART ONE / FOUNDATIONS
CHArTEI~
obtain. Research institutes and policy centers Design a Search. After choosing a focused re
(e.g., Brookings Institute, Institute for Research search question for the review, the next step is to able Yia con. .'"
search process,
on Poverty, Rand Corporation, etc.) publish pa plan a search strategy. The reviewer needs to de .-\hsb:a:n or
pers and reports. Some major research libraries cide on the type of review, its extensiveness, and . . . bamjl-al. . . .
purchase these and shelve them with books. The the types of materials to include. The key is to be .IIowa~.
only way to be sure ofwhat has been published is
to write directly to the institute or center and re
careful, systematic, and organized. Set parame
ters on your search: how much time you will de
- .- 1"
quest a list of reports. vote to it, how far back in time you will look, the -~~-.
4b.-m_*
Each year, the professional associations in
academic fields (e.g., sociology, political science,
minimum number of research reports you will
examine, how many libraries you will visit, and
* -iiwjew
Sa - f
; Cz::srI
psychology) hold annual meetings. Thousands so forth.
of researchers assemble to give, listen to, or dis Also, decide how to record the bibliographic
cuss oral reports of recent research. Most of citation for each reference you find and how to ,
these oral reports are available as written papers
to those attending the meeting. People who do
not attend the meetings but who are members of
take notes (e.g., in a notebook, on 3 X 5 cards, in
a computer file). Develop a schedule, because
several visits are usually necessary. You should --..-h ."5.
cIs-'''_-=AI
.......
MIMI.'
the association receive a program of the meeting, begin a file folder or computer file in which you II~--.I
listing each paper to be presented with its title, can place possible sources and ideas for new
..* ........
~--*
........
pers are later published as articles. The papers Locate Research Reports. Locating research
may be listed in indexes or abstract services (to reports depends on the type of report or "outlet"
be discussed). of research being searched. As a general rule, use
multiple search strategies in order to counteract
the limitations of a single search method.
How to Conduct a Systematic
Literature Review
Articles in Scholarly Journals. As discussed ear
Define and Refine a Topic. Just as a researcher lier, most social research is published in schol
must plan and clearly define a topic and research arly journals. There are dozens ofjournals, many
question when beginning a research project, you going back decades, each containing many arti
need to begin a literature review with a clearly cles. The task of searching for articles can be for
defined, well-focused research question and a midable. Luckily, specialized publications make
plan. A good review topic should be as focused the task easier.
as a research question. For example, "divorce" You may have used an index for general
or "crime" is much too broad. A more appro publications, such as Reader's Guide to Periodical
priate review topic might be "the stability of Literature. Many academic fields have "ab
families with stepchildren" or "economic in stracts" or "indexes" for the scholarly literature
equality and crime rates across nations." If you (e.g., Psychological Abstracts, Social Sciences In ,
conduct a context review for a research project, dex, Sociological Abstracts, and Gerontological
• »
. . c.- ..
..
-..
._*-..r..
it should be slightly broader than the specific re Abstracts). For education-related topics, the Ed
~.1
search question being tested. Often, a researcher ucational Resources Information Center (ERIC)
will not finalize a specific research question for a system is especially valuable. There are over 100 . . . . . . ._ ...._1
study until he or she has reviewed the literature. such publications. You can usually find them in
The review helps bring greater focus to the re the reference section of a library. Many ab ~_-.I
...a .....1III1
search question. stracts or index services as well as ERICare avail- ..· . .11
CHAPTER 4 I REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 77
able via computer access, which speeds the most computer-based searches and consider
search process. several synonyms. The computer's searching
Abstracts or indexes are published on a reg method can vary and most only look for a key
ular basis (monthly, six times a year, etc.) and word in a title or abstract. If you choose too few
allow a reader to look up articles by author name words or very narrow terms, you will miss a lot
or subject. The journals covered by the abstract of relevant articles. If you choose too many
or index are listed in it, often in the front. An in words or very broad terms, you will get a huge
dex, such as the Social Sciences Index, lists only number of irrelevant articles. The best way to
the citation, whereas an abstract, such as learn the appropriate breadth and number of
Sociological Abstracts, lists the citation and has a keywords is by trial and error.
copy of the article's abstract. Abstracts do not In a study I conducted on how college stu
give you all the findings and details of a research dents define sexualharassment(Neuman, 1992),
project. Researchers use abstracts to screen arti I used the following keywords: sexual harass
cles for relevance, then locate the more relevant ment, sexual assault, harassment, gender equity,
articles. Abstracts may also include papers pre gender fairness, and sexdiscrimination. I later dis
sented at professional meetings. covered a few important studies that lacked any
It may sound as if all you have to do is to go of these keywords in their titles. I also tried the
find the index in the reference section of the li keywords college student and rape, but got huge
brary or on the Internet and look up a topic. Un numbers of unrelated articles that I could not
fortunately, things are more complicated than even skim.
that. In order to cover the studies across many There are numerous computer-assisted
years, you may have to look through many issues search databases or systems. A person with a
of the abstracts or indexes. Also, the subjects or computer and an Internet hook-up can search
topics listed are broad. The specificresearch ques some article index collections, the catalogs ofli
tion that interests you may fit into several subject braries, and other information sources around
areas. You should check each one. For example, the globe if they are available on the Internet.
for the topic of illegal drugs in high schools, you All computerized searching methods share a
might look up these subjects: drug addiction, similar logic, but each has its own method of op
drug abuse, substance abuse, drug laws, illegal eration to learn. In my study, I looked for
drugs, high schools, and secondary schools. Many sources in the previous seven years and used five
of the articles under a subject area will not be rel computerized databases of scholarly literature:
evant for your literature review. Also, there is a 3 Social Science Index, CARL (Colorado Area Re
to 12-month time lag between the publication of search Library), Sociofile, Social Science Citation
an article and its appearance in the abstracts or Index, and PsychLit.
indexes. Unless you are at a major research li Often, the same articles will appear in mul
brary, the most useful article may not be available tiple scholarly literature databases, but each
in your library. You can obtain it only by using an database may identify a few new articles not
interlibrary loan service, or it may be in a foreign found in the others. For example, I discovered
language that you do not read. several excellent sources not listed in any of the
The computerized literature search works computerized databases that had been published
on the same principle as an abstract or an index. in earlier years by studying the bibliographies of
Researchers organize computerized searches in the relevant articles.
several ways-by author, by article title, by sub The process in my study was fairly typical.
ject, or by keyword. A keyword is an important Based on my keyword search, I quickly skimmed
term for a topic that is likely to be found in a ti or scanned the titles or abstracts of over 200
tle. You will want to use six to eight keywords in sources. From these, I selected about 80 articles,
, 78 PART ONE I FOUNDATIONS
reports, and books to read. I found about 49 of though you may not use some and later erase
the 80 sources valuable, and they appear in the them. Do not forget anything in a complete bib
bibliography of the published article. liographic citation, such as a page number or the
name of the second author; you will regret it
Scholarly Books. Finding scholarly books on a later. It is far easier to erase a source you do not
subject can be difficult. The subject topics of li use than to try to locate bibliographic informa
brary catalog systems are usually incomplete and tion later for a source you discover that you need
too broad to be useful. Moreover, they list only or from which you forgot one detail.
books that are in a particular library system, al I recommend creating two kinds of Source
though you may be able to search other libraries Files, or divide a master file into two parts: Have
for interlibrary loan books. Libraries organize File and Potential File. The Have File is for
books by call numbers based on subject matter. sources that you have found and for which you
Again, the subject matter classifications may not have already taken content notes. The Potential
reflect the subjects of interest to you or all the File is for leads and possible new sources that
subjects discussed in a book. Once you learn the you have yet to track down or read. You can add
system for your library, you will find that most to the Potential File anytime you come across a
books on a topic will share the main parts of the new source or in the bibliography of something
call number. In addition, librarians can help you you read. Toward the end ofwriting a report, the
locate books from other libraries. For example, Potential File will disappear while the Have File
the Library of Congress National Union Catalog will become your bibliography.
lists all books in the U.S. Library of Congress. Li Your note cards or computer documents go
brarians have access to sources that list books at into the Content File. This file contains substan
other libraries, or you can use the Internet. tive information of interest from a source, usu
There is no sure-fire way to locate relevant ally its major findings, details of methodology,
books. Use multiple search methods, including a definitions of concepts, or interesting quotes. If
look at journals that have book reviews and the you directly quote from a source or want to take
bibliographies of articles. some specific information from a source, you
need to record the specific page number(s) on
which the quote appears. Link the files by
Taking Notes
putting key source information, such as author
As you gather the relevant research literature, it is and date, on each content file.
easy to feel overwhelmed by the quantity of in
formation, so you need a system for taking notes. What to Record. You will find it much easier Photoc~
The old-fashioned approach is to write notes to take all notes on the same type and size of pa will save you ~
onto index cards. You then shift and sort the note per or card, rather than having some notes on sure that you 1'~
cards, place them in piles, and so forth as you sheets of papers, others on cards, and so on. Re you can make ~
look for connections among them or develop an searchers have to decide what to record about an several warnings •
outline for a report or paper. This method still article, book, or other source. It is better to err in tocopying can be c
works. Today, however, most people use word the direction of recording too much rather than search. Second, be
processing software and gather photocopies or too little. In general, record the hypotheses laws. U.S. cop~
printed versions of many articles. tested, how major concepts were measured, the for personal resea
As you discover sources, it is a good idea to main findings, the basic design of the research, record or photooq
create two kinds of files for your note cards or the group or sample used, and ideas for future all citation inform
computer documents: a Source File and a study (see Box 4.2). It is wise to examine the re tire articles can be ~
Content File. Record all the bibliographic infor port's bibliography and note sources that you eral different par1I
mation for each source in the Source File, even can add to your search. used. Finally, unk
CHAPTER 4 I REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 79
Critically reading research reports requires sources. The Internet continues to expand and
skills that take time and practice to develop. De change at an explosive rate. FIGURE 4.2
spite a peer-review procedure and high rejection The Internet has been a mixed blessing for
rates, errors and sloppy logic slip in. Read care social research, but it has not proved to be the
fully to see whether the introduction and title re panacea that some people first thought it might Bearman.~_
ally fit with the rest of the article. Sometimes, be. It provides new and important ways to find and Fnt
titles, abstracts, or the introduction are mislead information, but it remains one tool among oth 00.4).
ing. They may not fully explain the research pro ers. It can quickly make some specific pieces of
ject's method and results. An article should be information accessible. For example, from my
logically tight, and all the parts should fit to home computer, I was able to go to the U.S. Fed
gether. Strong logical links should exist between eral Bureau of Prisons and in less than three
parts of the argument. Weak articles make leaps minutes locate a table showing me that in 1980,
in logic or omit transitional steps. Likewise, arti 139 people per 100,000 were incarcerated in the
cles do not always make their theory or ap United States, whereas in 2004 (the most recent
proach to research explicit. Be prepared to read data available), it was 486 per 100,000. The In
the article more than once. (See Figure 4.2 on ternet is best thought of as a supplement rather Since 1993...
taking notes on an article.) than as a replacement for traditional library re
search. There are "up" and "down" sides to us
~
miIion tJ!IeI!5
fected the -
ing the Internet for social research:
What a Good Review Looks Like rlOIlJ*c Iti it
An author should communicate a review's pur The UpSide p6edge~"
~~
pose to the reader by its organization. The wrong
1. The Internet is easy, fast, and cheap. It is
way to write a review is to list a series of research ~sj.
widely accessible and can be used from many lo
reports with a summary of the findings of each.
cations. This near-free resource allows people to
This fails to communicate a sense of purpose. It
find source material from almost anywhere-lo
reads as a set of notes strung together. Perhaps
cal public libraries, homes, labs or classrooms,
the reviewer got sloppy and skipped over the im
or anywhere a computer is connected to the In
portant organizing step in writing the review.
ternet system. Also, the Internet does not close;
The right way to write a review is to organize
it operates 24 hours a day, seven days a week.
common findings or arguments together. A well
With minimal training, most people can quicldy
accepted approach is to address the most imp or- .
perform searches and get information on their
tant ideas first, to logically link statements or
computer screens that would have required
findings, and to note discrepancies or weaknesses
them to take a major trip to large research li
in the research (see Box 4.3 for an example).
braries a few years ago. Searching a vast quantity
of information electronically has always been
easier and faster than a manual search, and the
USING THE INTERNET FOR Internet greatly expands the amount and variety
SOCIAL RESEARCH of source material. More and more information
(e.g., Statistical Abstract of the United States) is
The Internet (see Box 4.4) has revolutionized
available on the Internet. In addition, once the
how social researchers work. A mere decade ago,
information is located, a researcher can often
it was rarely used; today, most social researchers
store it electronically or print it at a local site.
use the Internet regularly to help them review
the literature, to communicate with other re 2. The Internet has "links" that provide ad
searchers, and to search for other information ditional ways to find and connect to many other
CHAPTER 4 I REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 81
------------------1..
FIG U R E 4.2 Example of Notes on an Article
FULL CITATION ON BIBLIOGRAPHY (SOURCE FILE)
I Bearman, Peter, and Hannah Buckner. 2001. "Promising the Future: Virginity Pledges
and First Intercourse." American Journal of Sociology 106:859-91 2. (January, issue
no. 4).
Since 1993, the Southern Baptist Church sponsored a movement among teens
whereby the teens make a public pledge to remain virgins until marriage. Over 2.5
million teens have made the pledge. This study examines whether the pledge af
fected the timing of sexual intercourse and whether pledging teens differ from
non pledging teens. Critics of the movement are uncomfortable with it because
pledge supporters often reject sex education, hold an overly romanticized view of
marriage, and adhere to traditional gender roles.
Hypothesis
Adolescents will engage in behavior that adults enjoy but that is forbidden to
them based on the amount of social controls that constrain opportunities to en
gage in forbidden behavior. Teens in nontraditional families with greater freedom
and less supervision are more likely to engage in forbidden behavior (sex). Teens
in traditional families and who are closer to their parents will delay sexual activ
ity. Teens closely tied to "identity movements" outside the family will modify be
, havior based on norms the movements teach.
Method
Data are from a national health survey of U.S. teens in grades 7-12 who were in
public or private schools in 1994-1995. A total of 90,000 students in 141
schools completed questionnaires. A second questionnaire was completed by
20,000 of the 90,000 students. The questionnaire asked about a pledge, im
portance of religion, and sexual activity.
Findings
The study found a substantial delay in the timing of first intercourse among
pledgers. Yet, the effect of pledging varies by the age of the teen. In addition,
pledging only works in some social contexts (i.e., where it is at least partially a so
cial norm). Pledgers tend to be more religious, less developed physically, and from
more traditional social and family backgrounds.
82 PART ONE / FOUNDATIONS
.iII.
students. Benson and Thomson's study in Social
Problems (1982) lists many problems created by tke, and Padgitt, 1983; Benson and Thomson, ...ldMde.lBt
sexual harassment. In their excellent book, The Lech 1982; Dziech and Weiner, 1990). For example,
erous Professor, Dziech and Weiner (1990) give a Adams, Kottke, and Padgitt (1983) noted that 13
long list of difficulties that victims have suffered. percent of women students said they avoided taking
Researchers study the topic in different ways. a class or working with certain professors because of
Hunter and McClelland (1 991) conducted a study the risk of harassment.
of undergraduates at a small liberal arts college. They Research into campus sexual harassment has
had a sample of 300 students and students were taken several approaches. In addition to survey re
given multiple vignettes that varied by the reaction of search, many have experimented with vignettes or
the victim and the situation. Jaschik and Fretz presented hypothetical scenarios (Hunter and Mc
(1991) showed 90 women students at a mideastern Clelland, 1991 ;Jaschikand Fretz, 1991; Popovich et .
university a videotape with a classic example of sex al., 1987; Reilley, Carpenter, Dull, and Barlett, Uliwitdlt:d.. ......t. .
ual harassment by a teaching assistant. Before it was 1982; Rossi and Anderson, 1982; Valentine-French ~ c::uwit!dBll
labeled as sexual harassment, few women called it and Radtke, 1 989; Weber-Burdin and Rossi, 1982). c!a5e ~ c::u.-.[iiI"
that. When asked whether it was sexual harassment, Victim verbal responses and situational factors ap lha ..cMdts
cmc'~.... ..GII1rt1
98 percent agreed. Weber-Burdin and Rossi (1982) pear to affect whether observers label a behavior as D5l. _
replicated a previous study on sexual harassment, harassment. There is confusion over the application
of a sexual harassment label for inappropriate behav
only they used students at the University of Massa ~"'.-.c
chusetts. They had 59 students rate 40 hypotheti ior. For example, [aschik and Fretz (1 991) found that
cal situations. Reilley, Carpenter, Dull, and Bartlett only 3 percent of the women students shown a
(1982) conducted a study of 250 female and 150 videotape with a classic example of sexual harass
male undergraduates at the University of California ment by a teaching assistant initially labeled it as
at Santa Barbara. They also had a sample of 52 fac sexual harassment. Instead, they called it "sexist,"
ulty. Both samples completed a questionnaire in "rude," "unprofessional," or "demeaning." When
asked whether it was sexual harassment, 98 percent
3. . . . . . .
which respondents were presented vignettes of sex
agreed. Roscoe and colleagues (1987) reported ......-1
ual-harassing situations that they were to rate.
similar labeling difficulties. dkrt. h pnJri*:IlI
Popovich and Colleagues (1986) created a nine
item scale of sexual harassment. They studied 209
nw:ion ~
kKlg disuncn
Sk3doi~
scndolIba
sources of information. Many websites, home link indicator (usually a button or a highlighted month. the ill'.- ."I
pages, and other Internet resource pages have word or phrase). This connects people to more 00II:k ar DO COiL
"hot links" that can call up information from re information and provides "instant" access to tiom 00 .-0<.
lated sites or sources simply by clicking on the cross-referenced material. Links make embed wt:I3l apw:eus c.
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 83
Contrary to popular belief, the Internet has not inal materials and read them for ideas or to build
made all information free and accessible to
everyone. Often, what is free is limited, and
fuller information is available only to those who
on them. Also, it is easy to copy, modify, or dis
tort, then reproduce copies of a source. For ex
ample, a person could alter a text passage or a
........
w-
.......
'
pay. In fact, because some libraries redirected photo image then create a new webpage to dis
funds to buy computers for the Internet and cut seminate the false information. This raises issues .... , d
the purchases for books and paper copies of doc
uments, the Internet's overall impact may have
about copyright protection and the authenticity
of source material.
i
,.
actually reduced what is available for some users. There are few rules for locating the best sites
on the Internet-ones that have useful and
3. Finding sources on the Internet can be
truthful information. Sources that originate at
very difficult and time consuming. It is not easy
universities, research institutes, or government
to locate specific source materials. Also, different
agencies usually are more trustworthy for re
"search engines" can produce very different re
search purposes than ones that are individual
sults. It is wise to use multiple search engines
home pages of unspecified origin or location, or
(e.g., Yahoo, Excite, and Google), since they
that a commercial organization or a political/so
work differently. Most search engines simply
cial issue advocacy group sponsors. In addition
look for specific words in a short description of
to moving or disappearing, many webpages or
the webpage. This description may not reveal the
sources fail to provide complete information to
full content of the source, just as a title does not
make citation easy. Better sources provide fuller
fully tell what a book or article is about. In addi
or more complete information about the author,
tion, search engines often come up with tens of
date, location, and so on.
thousands of sources, far too many for anyone to
As you prepare a review of the scholarly lit
examine. The ones at the "top" may be there be
erature and more narrowly focus a topic, you
cause they were recently added to the Internet or
should be thinking about how to design a study.
because their short description had several ver
The specifics of design can vary somewhat de
sions of the search word. The "best" or most rel
pending on whether your study will primarily
evant source might be buried as the 150th item
found in a search. Also, one must often wade
through a lot of commercials and advertise
ments to locate "real" information.
employ a quantitative-deductive-positivist ap
proach or a qualitative-inductive-interpretive/
critical approach. The two approaches have a -
~ .........~
great deal in common and mutually comple iR IioIralID
4. Internet sources can be "unstable" and ment one another, but there several places where Reseu:dItu
difficult to document. After one conducts a "branches in the path" of designing a study di ~com ...."
search on the Internet and locates webpages with verge depending on the approach you adopt. \Jlber. but thr
information, it is important to note the specific iitrles are IJ:BJhUlilJl
"address" (usually it starts http://) where it re dion to uo:ll.ln1Ci1111
sides. This address refers to an electronic file sit ~.:.anbe
QUALITATIVE AND
ting in a computer somewhere. If the computer
file is moved, it may not be at the same address
QUANTITATIVE ORIENTATIONS
TOWARD RESEARCH lftarand..J
~cbers::l
the two styles comes from the nature ofthe data. other researchers have trod. Alternatively, it may
Softdata, in the form of impressions, words, sen be a new path into unknown territory where few
tences, photos, symbols, and so forth, dictate others have gone, and without signs marking the
different research strategies and data collection direction forward.
techniques than hard data, in the form of num In general, quantitative researchers follow a
bers. Another difference is that qualitative and more linear path than do qualitative researchers.
quantitative researchers often hold different as A linear research path follows a fixed sequence of
sumptions about social life and have different steps; it is like a staircase leading in one clear di
objectives. These differences can make tools rection. It is a way of thinking and a wayoflook
used by the other style inappropriate or irrele ing at issues-the direct, narrow, straight path
vant. People who judge qualitative research by that is most common in western European and
standards of quantitative research are often dis North American culture.
appointed, and vice versa. It is best to appreciate Qualitative research is more nonlinear and
the strengths each style offers. . cyclical. Rather than moving in a straight line, a
To appreciate the strengths of each style,it is nonlinearresearch path makes successive passes
conduct research on something that interests Typical research questions for qualitative
you. researchers include: How did a certain condition
All research begins with a topic but a topic is or social situation originate? How is the condi
only a starting point that researchers must nar tion/situation maintained over time? What are
row into a focused research question. Qualita the processes by which a condition/situation 1. Examine the liusl
tive and quantitative researchers tend to adopt changes, develops, or operates? A different type excellent 5OUra!
They are usuaty
different approaches to turn a topic to a focused of question tries to confirm existing beliefs or as
ficity and s~
research question for a specific study. Qualita sumptions. A last type of question tries to dis
onthefollo~
tive researchers often begin with vague or un cover new ideas.
clear research questions. The topic emerges Research projects are designed around re a. Replicate a pi
slowly during the study. The researchers often search problems or questions. Before designing a or with slignt
combine focusing on a specific question with the project, quantitative researchers focus on a spe b. Explore uneII
process of deciding the details of study design cific research problem within a broad topic. For previous resI
that occurs while they are gathering data. By example, your personal experience might sug c. Follow sugge
contrast, quantitative researchers narrow a topic gest labor unions as a topic. "Labor unions" is a research at 11
into a focused question as a discrete planning topic, not a research question or a problem. In d. Extend an eIIii
step before they finalize study design. They use it any large library, you will find hundreds of new topic Of" ~
as a step in the process of developing a testable books and thousands of articles written by soci e. Challenge ml
hypothesis (to be discussed later) and to guide ologists, historians, economists, management lationship. .
the study design before they collect any data. officials, political scientists, and others on f. Specify the ill
The qualitative research style is flexible and unions. The books and articles focus on different linking relatiG
encourages slowly focusing the topic throughout aspects of the topic and adopt many perspectives
2. Talkoverideas"
a study. In contrast to quantitative research, only on it. Before proceeding to design a research
a small amount of topic narrowing occurs in an project, you must narrow and focus the topic. a. Ask people •
the topic for
early research planning stage, and most of the An example research question is, "How much
have thoughl
narrowing occurs after a researcher has begun to did U.S. labor unions contribute to racial in
collect data. equality by creating barriers to skilled jobs for b. Seek out thea
The qualitative researcher begins data gath African Americans in the post-World War II from yours aI
ering with a general topic and notions of what period?" research quei
will be relevant. Focusing and refining contin When starting research on a topic, ask your 3. Apply to a s~
ues after he or she has gathered some of the data self: What is it about the topic that is of greatest a. Focus the tOIl
and started preliminary analysis. Qualitative re interest? For a topic about which you know little, riod or time II
searchers use early data collection to guide how first get background knowledge by reading b. Narrow the 111
they adjust and sharpen the research question(s) about it. Research questions refer to the rela ographic urjt
because they rarely know the most important is tionships among a small number of variables. c. Consider w+W
sues or questions until after they become fully Identify a limited number of variables and spec people/units.
immersed in the data. Developing a focused re ify the relationships among them. are differerxa
search question is a part of the data collection A research question has one or a small num 4. Define the aim Of'"i
process, during which the researcher actively re ber of causal relationships. Box 4.5 lists some
a. Will the re~
flects on and develops preliminary interpreta ways to focus a topic into a research question.
ploratory. ~
tions. The qualitative researcher is open to For example, the question, "What causes di
unanticipated data and constantly reevaluates b. Will the ~
vorce?" is not a good research question. A better
search?
the focus early in a study. He or she is prepared research question is, "Is age at marriage associ
to change the direction of research and follow ated with divorce?" The second question sug
new lines of evidence. gests two variables: age of marriage and divorce.
CHAPTER 4 I REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 87
family income in many different nations is al 4.1). In addition, researchers tend to adopt a dif ,
most impossible to answer because information ferent language and approach to study design,
on burglary and income is not collected or avail which we will consider next.
able for most countries. Some questions require
the approval of authorities (e.g., to see medical
records) or involve violating basic ethical princi QUALITATIVE DESIGN ISSUES
ples (e.g., causing serious physical harm to a per
..
The Language of Cases and Contexts
son to see the person's reaction). The expertise
or background of the researcher is also a limita
tion. Answering some research questions in
Qualitative researchers use a language of cases
and contexts, examine social processes and cases
-~
s-r
--.-.-.....
volves the use of data collection techniques,
statistical methods, knowledge of a foreign lan
in their social context, and look at interpreta
tions or the creation of meaning in specific set
.-a.,-.r
guage, or skillsthat the researcher may not have. tings. They try look at social life from multiple ~QII-"
.AIr.......
Unless the researcher can acquire the necessary points of view and explain how people construct a aJ
training or can pay for another person's services, identities. Only rarely do they use variables or
the research question may not be practical.
In summary, styles of qualitative and quan
titative researchers have much in common, but
the researchers often differ on design issues,
test hypotheses, or try to convert social life into
numbers.
Qualitative researchers see most areas and
activities of social life as being intrinsically qual
.-
.
such as taking a linear or nonlinear research path itative. To them, qualitative data are not impre
and developing a research question (see Table cise or deficient; they are highly meaningful.
---~
--------------------~
TABLE 4.1 Quantitative Reasearch versus Qualitative Research
.. c
A a 5
1IIr'
Test hypothesis that the researcher begins with. Capture and discover meaning once the researcher
Concepts are in the form of distinct variables. Concepts are in the form of themes, motifs,
Measures are systematically created before data Measures are created in an ad hoc manner and are
researcher.
Data are in the form of numbers from precise Data are in the form of words and images from
Theory is largely causal and is deductive. Theory can be causal or noncausal and is often
a . ._ .
inductive.
Procedures are standard, and replication is Research procedures are particular, and replication
T
charts and discussing how what they show relates generalizations from evidence and organizing data
Instead of trying to convert social life into vari car's license number before proceeding? After
ables or numbers, qualitative researchers borrow radioing the car's location, does the officer ask
ideas from the people they study and place them the motorist to get out of the car sometimes, but
within the context of a natural setting. They ex in others casually walk up to the car and talk to
amine motifs, themes, distinctions, and ideas in the seated driver? When data collection and the
stead of variables, and they adopt the inductive orizing are interspersed, theoretical questions
approach of grounded theory. arise that suggest future observations, so new
Some people believe that qualitative data are data are tailored to answer theoretical questions
"soft," intangible, and immaterial. Suchdata are that came from thinking about previous data.
so fuzzy and elusive that researchers cannot re
ally capture them. This is not necessarily the
The Context Is Critical
case. Qualitative data are empirical. They in
volve documenting real events, recording what Qualitative researchers emphasize the social
people say (with words, gestures, and tone), ob context for understanding the social world. They
serving specific behaviors, studying written doc hold that the meaning of a social action or state
uments, or examining visual images. These are ment depends, in an important way, on the con
all concrete aspects of the world. For example, text in which it appears. When a researcher
some qualitative researchers take and closely removes an event, social action, answer to a
scrutinize photos or videotapes of people or so question, or conversation from the social con
cial events. This evidence is just as "hard" and text in which it appears, or ignores the context,
physical as that used by quantitative researchers social meaning and significance are distorted.
to measure attitudes, social pressure, intelli Attention to social context means that a
gence, and the like. qualitative researcher notes what came before or
what surrounds the focus of study. It also im
plies that the same events or behaviors can have
Grounded Theory
different meanings in different cultures or his
A qualitative researcher develops theory during torical eras. For example, instead of ignoring the
the data collection process. This more inductive context and counting votes across time or cul
method means that theory is built from data or tures, a qualitative researcher asks: What does
grounded in the data. Moreover, conceptualiza voting mean in the context? He or she may treat
tion and operationalization occur simultane the same behavior (e.g.,voting for a presidential
ously with data collection and preliminary data candidate) differently depending on the social
analysis. It makes qualitative research flexible context in which it occurs. Qualitative re
and lets data and theory interact. Qualitative re searchers place parts of social life into a larger
searchers remain open to the unexpected, are whole. Otherwise, the meaning of the part may
willing to change the direction or focus of a re be lost. For example, it is hard to understand
search project, and may abandon their original what a baseball glove is without knowing some
research question in the middle of a project. thing about the game of baseball. The whole of
A qualitative researcher builds theory by the game-innings, bats, curve balls, hits-gives
making comparisons. For example, when a re meaning to each part, and each part without the
searcher observes an event (e.g., a police officer whole has little meaning.
confronting a speeding motorist), he or she im
mediately ponders questions and looks for sim
The Case and Process
ilarities and differences. When watching a police
officer stop a speeder, a qualitative researcher In quantitative research, cases are usually the
asks: Does the police officer always radio in the same as a unit of analysis, or the unit on which
90 PART ONE I FOUNDATIONS
variables are measured (discussed later). Quan tative researcher gives meaning by rearranging,
titative researchers typically measure variables of examining, and discussing textual or visual data
their hypotheses across many cases. For exam in a way that conveys an authentic voice, or that
ple, if a researcher conducts a survey of 450 indi remains true to the original understandings of
viduals, each individual is a case or unit on the people and situations that he or she studied.
which he or she measures variables. Qualitative Instead of relying on charts, statistics, and
researchers tend to use a "case-oriented ap
proach [that] places cases, not variables, center
stage" (Ragin, 1992:5). They examine a wide va
riety of aspects of one or a few cases.Their analy
displays of numbers, qualitative researchers put
a greater emphasis on interpreting the data.
Their data are often "richer" or more complex
and full of meaning. The qualitative researcher
..- .
ses emphasize contingencies in "messy" natural interprets to "translate" or make the originally
settings (i.e., the co-occurrence of many specific gathered data understandable to other people.
factors and events in one place and time). Expla The process of qualitative interpretation moves
nations or interpretations are complex and may through three stages or levels.
be in the form of an unfolding plot or a narrative A researcher begins with the point of view
story about particular people or specific events. of the people he or she is studying, and the re
Rich detail and astute insight into the cases re searcher wants to grasp fully how they see the
place the sophisticated statistical analysis of pre world, how they define situations, or what
cise measures across a huge number of units or
cases found in quantitative research.
The passage of time is integral to qualitative
research. Qualitative researchers look at the se
quence of events and pay attention to what hap
pens first, second, third, and so on. Because
things mean to them. A first-order interpretation
contains the inner motives, personal reasons,
and point of view of the people who are being
studied in the original context. As the researcher
discovers and documents this first-order inter
pretation, he or she remains one step removed
~
•
..
•
qualitative researchers examine the same case or from it. The researcher offers a second-order in
set of cases over time, they can see an issue terpretation, which is an acknowledgment that
evolve, a conflict emerge, or a social relationship however much a researcher tries to get very
develop. The researcher can detect process and close and "under the skin" of those he or she is
causal relations. studying, a researcher is still "on the outside
In historical research, the passage of time looking in." In the second-order interpretation,
may involve years or decades. In field research, the researcher tries to elicit an underlying co
the passage of time is shorter. Nevertheless, in herence or sense of overall meaning in the data.
both, a researcher notes what is occurring at dif To reach an understanding of what he or she
ferent points in time and recognizes that when sees or hears, a researcher often places the data
something occurs is often important. into a context of the larger flow of events and
behaviors. A qualitative researcher will often
move to the third step and link the understand
Interpretation
ing that he or she achieved to larger concepts,
Interpretation means to assign significance or a generalizations, or theories. The researcher can ~lhe.-;JI_rlll
coherent meaning to something. Quantitative share this broader interpretation with other
and qualitative researchers both interpret data, people who are unfamiliar with the original
but they do so in different ways. A quantitative data, the people and events studied, or the social
researcher gives meaning by rearranging, exam situations observed by the researcher. This level
ining, and discussing the numbers by using of meaning translates the researcher's own un
charts and statistics to explain how patterns in derstanding in a way that facilitates communica
the data relate to the research question. A quali tion with people who are more distant from the
CHAPTER 4 I REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 91
original source, and it represents a third-order able. It describes the intensity or strength of at
interpretation. tachment to attitudes, beliefs, and behaviors as
sociated with the concept of masculine within a
culture. "Married" is not a variable; it is an at
tribute of the variable "marital status." Related
QUANTITATIVE DESIGN ISSUES ideas such as "number of years married" or
The language of Variables and "depth of commitment to a marriage" are vari
ables. Likewise, "robbery" is not a variable; it is
Hypotheses
an attribute of the variable "type of crime."
Variation and Variables. The variable is a "Number of robberies," "robbery rate,"
central idea in quantitative research. Simply de "amount taken during a robbery," and "type of
fined, a variable is a concept that varies. Quanti robbery" are all variables because they vary or
tative research uses a language of variables and take on a range of values.
relationships among variables. Quantitative researchers redefine concepts
In Chapter 2, you learned about two types of interest into the language of variables. As the
of concepts: those that refer to a fixed phenom examples of variables and attributes illustrate,
enon (e.g., the ideal type of bureaucracy) and slight changes in definition change a nonvariable
those that vary in quantity, intensity, or amount into a variable concept. As you saw in Chapter 2,
(e.g., amount of education). The second type of concepts are the building blocks of theory; they
concept and measures of the concepts are vari organize thinking about the social world. Clear
ables. Variables take on two or more values. concepts with careful definitions are essential in
Once you begin to look for them, you will see theory.
variables everywhere. For example, gender is a
variable; it can take on two values: male or fe Types ofVariables. Researchers who focus on
male. Marital status is a variable; it can take on causal relations usually begin with an effect, then
the values of never married single, married, di search for its causes. Variables are classified into
vorced, or widowed. Type of crime committed is three basic types, depending on their location in
a variable; it can take on values of robbery, bur a causal relationship. The cause variable, or the
glary, theft, murder, and so forth. Family income one that identifies forces or conditions that act
is a variable; it can take on values from zero to on something else, is the independent variable.
billions of dollars. A person's attitude toward The variable that is the effect or is the result or
abortion is a variable; it can range from strongly outcome of another variable is the dependent
favoring legal abortion to strongly believing in variable. The independent variable is "indepen
antiabortion. dent of' prior causes that act on it, whereas the
The values or the categories of a variable are dependent variable "depends on" the cause.
its attributes. It is easy to confuse variables with It is not always easy to determine whether a
attributes. Variables and attributes are related, variable is independent or dependent. Two
but they have distinct purposes. The confusion questions help you identify the independent
arises because the attribute of one variable can variable. First, does it come before other vari
itself become a separate variable with a slight ables in time? Independent variables come be
change in definition. The distinction is between fore any other type. Second, if the variables
concepts themselves that vary and conditions occur at the same time, does the author suggest
within concepts that vary. For example, "male" that one variable has an impact on another vari
is not a variable; it describes a category of gender able? Independent variables affect or have an im
and is an attribute of the variable "gender." Yet, pact on other variables. Research topics are often
a related idea, "degree of masculinity," is a vari phrased in terms of the dependent variables be
r
cause dependent variables are the phenomenon punishment for criminal acts. A multicause ex
to be explained. For example, suppose a re planation usually specifies the independent vari
searcher examines the reasons for an increase in able that has the greatest causal effect.
the crime rate in Dallas, Texas; the dependent A complex theoretical explanation contains
variable is the crime rate. a string of multiple intervening variables that are
A basic causal relationship requires only an linked together. For example, family disruption
independent and a dependent variable. A third causes lower self-esteem among children, which
type of variable, the intervening variable, appears causes depression, which causes poor grades in
in more complex causal relations. It comes be school, which causes reduced prospects for a
tween the independent and dependent variables
and shows the link or mechanism between them.
Advances in knowledge depend not only on doc
umenting cause-and-effect relationships but
good job, which causes a lower adult income.
The chain of variables is: family disruption (in
dependent), childhood self-esteem (interven
ing), depression (intervening), grades in school
.....
4.
. . .MiI....
also on specifying the mechanisms that account (intervening), job prospects (intervening), adult
for the causal relation. In a sense, the intervening income (dependent).
variable acts as a dependent variable with respect Two theories on the same topic may have
to the independent variable and acts as an inde different independent variables or predict differ
pendent variable toward the dependent variable. ent independent variables to be important. In
For example, French sociologist Emile addition, theories may agree about the indepen
Durkheim developed a theory of suicide that dent and dependent variables but differ on the
specified a causal relationship between marital intervening variable or causal mechanism. For
status and suicide rates. Durkheim found evi example, two theories say that family disruption
dence that married people are less likely to com causes lower adult income, but for different rea
mit suicide than single people. He believed that sons. One theory holds that disruption encour
married people have greater social integration ages children to join deviant peer groups that are
(i.e., feelings of belonging to a group or family). not socialized to norms of work and thrift.
He thought that a major cause of one type of sui Another emphasizes the impact of the disrup
cide was that people lacked a sense of belonging tion on childhood depression and poor acade **.IQS ......
to a group. Thus, his theory can be restated as a mic performance, which directly affect job
three-variable relationship: marital status (inde performance.
pendent variable) causes the degree of social in A single research project usually tests only a
tegration (intervening variable), which affects small part of a causal chain. For example, a re
suicide (dependent variable). Specifying the search project examining six variables may take
chain of causality makes the linkages in a theory the six from a large, complex theory with two
clearer and helps a researcher test complex ex dozen variables. Explicit links to a larger theory
planations.' strengthen and clarify a research project. This
Simple theories have one dependent and applies especiallyfor explanatory, basic research,
one independent variable, whereas complex the which is the model for most quantitative re
ories can contain dozens of variables with multi search.
ple independent, intervening, and dependent
variables. For example, a theory of criminal be
Causal Theory and Hypotheses
havior (dependent variable) identifies four inde
pendent variables: an individual's economic The Hypothesis and Causality. A hypothesis is
hardship, opportunities to commit crime easily, a proposition to be tested or a tentative state
membership in a deviant subgroup of society ment of a relationship between two variables.
that does not disapprove of crime, and lack of Hypotheses are guesses about how the social
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 93
------------------~
FIG U R E 4. 3
FOUNDATIONS
1966
.. CHAPTE
hypotheses. Thes
A new one is developed.
iJ 9 0 8
I I
L~~J L~~J
I I
Null hypothesis t
tion that research
ship, so hypothesi
In 2006, 3 hypotheses are in contention, but from 1966 to 2006, 11 hypotheses were considered, and over make fmding a rei
time, 8 of them were rejected in one or more tests. researcher who 1
proach only dimIl
evidence supports
cept the null h~
tive evidence is more significant because the hy For example, a man stands on a street cor the tested relatioa
pothesis becomes "tarnished" or "soiled" if the ner with an umbrella and claims that his um plies that the altes
evidence fails to support it. This is because a hy brella protects him from falling elephants. His the other hand, iI
pothesis makes predictions. Negative and dis hypothesis that the umbrella provides protec dence to reject till
confirming evidence shows that the predictions tion has supporting evidence. He has not had a ternative hypo~
are wrong. Positive or confirming evidence for a single elephant fallon him in all the time he has researcher cannes
hypothesis is less critical because alternative hy had his umbrella open. Yet, such supportive ev by testing the nul
potheses may make the same prediction. A re idence is weak; it also is consistent with an alter the alternative h~
searcher who finds confirming evidence for a native hypothesis-that elephants do not fall null hypothesis til
prediction may not elevate one explanation over from the sky. Both predict that the man will be evidence, the argu
its alternatives. safe from falling elephants. Negative evidence esis can grow strOl
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 95
for the hypothesis-the one elephant that falls Many people find the null hypothesis to be
on him and his umbrella, crushing both-would confusing. Another way to think of it is that the
destroy the hypothesis for good. scientific community is extremely cautious. It
Researchers test hypotheses in two ways: a prefers to consider a causal relationship to be
straightforward way and a null hypothesis way. false until mountains of evidence show it to be
Many quantitative researchers, especially exper true. This is similar to the Anglo-American legal
imenters, frame hypotheses in terms ofa null hy idea of innocent until proved guilty. A re
pothesis based on the logic of the disconfirming searcher assumes, or acts as if, the null hypothe
hypotheses. They test hypotheses by looking for sis is correct until reasonable doubt suggests
evidence that will allow them to accept or reject otherwise. Researchers who use null hypotheses
the null hypothesis. Most people talk about a hy generally use it with specific statistical tests (e.g.,
pothesis as a way to predict a relationship. The r-test or F-test). Thus, a researcher may say there
null hypothesis does the opposite. It predicts no is reasonable doubt in a null hypothesis if a sta
relationship. For example, Sarah believes that tistical test suggests that the odds of it being false
students who live on campus in dormitories get are 99 in 100. This is what a researcher means
higher grades than students who live off campus when he or she says that statistical tests allow
and commute to college. Her null hypothesis is him or her to "reject the null hypothesis at the
that there is no relationship between residence .01 level of significance."
and grades. Researchers use the null hypothesis
with a corresponding alternative hypothesis or
Aspects of Explanation
experimental hypothesis. The alternative hy
pothesis says that a relationship exists. Sarah's Clarity about Units and Levels ofAnalysis. It
alternative hypothesis is that students' on-cam is easy to become confused at first about the
pus residence has a positive effect on grades. ideas of units and levels of analysis. Neverthe
For most people, the null hypothesis ap less, they are important for dearly thinking
proach is a backward way of hypothesis testing. through and planning a research project. All
Null hypothesis thinking rests on the assump studies have both units and levels ofanalysis, but
tion that researchers try to discover a relation few researchers explicitly identify them as such.
ship, so hypothesis testing should be designed to The levels and units of analysis are restricted by
make finding a relationship more demanding. A the topic and the research question.
researcher who uses the null hypothesis ap A level of analysis is the level of social reality
proach only directly tests the null hypothesis. If to which theoretical explanations refer. The level
evidence supports or leads the researcher to ac of social reality varies on a continuum from mi
cept the null hypothesis, he or she concludes that cro level (e.g., small groups or individual
the tested relationship does not exist. This im processes) to macro level (e.g., civilizations or
plies that the alternative hypothesis is false. On structural aspects ofsociety). The level includes a
the other hand, if the researcher can find evi mix of the number of people, the amount of
dence to reject the null hypothesis, then the al space, the scope of the activity, and the length of
ternative hypotheses remain a possibility. The time. For example, an extreme micro-level
researcher cannot prove the alternative; rather, analysis can involve a few seconds of interaction
by testing the null hypotheses, he or she keeps between two people in the same small room. An
the alternative hypotheses in contention. When extreme macro-level analysis can involve billions
null hypothesis testing is added to confirming of people on several continents across centuries.
evidence, the argument for an alterative hypoth Most social research uses a level of analysis that
esis can grow stronger over time. lies between these extremes.
96 PART ONE / FOUNDATIONS
The level of analysis delimits the kinds of hand, a study that compares the amounts differ
assumptions, concepts, and theories that a re ent colleges spend on their football programs
searcher uses. For example, I want to study the would use the organization (the college) as the
topic of dating among college students. I use a unit of analysis because the spending by colleges
micro-level analysis and develop an explana is being compared and each college's spending is
,
tion that uses concepts such as interpersonal recorded.
contact, mutual friendships, and common in Researchers use units of analysis other than
terests. I think that students are likely to date individuals, groups, organizations, social cate
someone with whom they have had personal gories, institutions, and societies. For example, a
contact in a class, share friends in common,
and share common interests. The topic and fo
cus fit with a micro-level explanation because
they are targeted at the level of face-to- face in
researcher wants to determine whether the
speeches of two candidates for president of the
United States contain specific themes. The re
searcher uses content analysis and measures the
themes in each speech of the candidates. Here,
~ ...•--.
............
teraction among individuals. Another example
topic is how inequality affects the forms ofvio the speech is the unit of analysis. Geographic
lent behavior in a society. Here, I have chosen a units of analysis are also used. A researcher in
more macro-level explanation because of the terested in determining whether cities that have f T
topic and the level of social reality at which it a high number of teenagers also have a high rate E
operates. I am interested in the degree of in of vandalism would use the city as the unit of
equality (e.g., the distribution of wealth, prop analysis. This is because the researcher measures
erty, income, and other resources) throughout the percentage of teenagers in each city and the
a society and in patterns of societal violence amount of vandalism for each city.
(e.g., aggression against other societies, sexual The units of analysis determine how a re
assault, feuds between families). The topic and searcher measures variables or themes. They also
research question suggest macro-level concepts correspond loosely to the level of analysis in an
and theories. explanation. Thus, social-psychologicalor micro
The unit of analysis refers to the type of unit levelsof analysisfit with the individual as a unit of
a researcher uses when measuring. Common analysis, whereas macro levels of analysis fit with
units in sociology are the individual, the group the social category or institution as a unit. Theo
(e.g., family, friendship group), the organization ries and explanations at the micro level generally
(e.g., corporation, university), the social cate refer to features of individuals or interactions
gory (e.g., social class, gender, race), the social among individuals. Those at the macro levelrefer
institution (e.g., religion, education, the family), to social forces operating across a society or rela
and the society (e.g., a nation, a tribe). Although tions among major parts of a society as a whole.
the individual is the most commonly used unit Researchers use levels and units of analysis
of analysis, it is by no means the only one. Dif to design research projects, and being aware of
ferent theories emphasize one or another unit of them helps researchers avoid logical errors. For
analysis, and different research techniques are example, a study that examines whether colleges
..._.......
associated with specific units of analysis. For ex in the North spend more on their football pro
ample, the individual is usually the unit of analy grams than do colleges in the South implies that
sis in survey and experimental research. a researcher gathers information on spending by
As an example, the individual is the unit of college and the location of each college. The unit
analysis in a survey in which 150 students are of analysis-the organization or, specifically, the
asked to rate their favorite football player. The college-flows from the research problem and
individual is the unit because each individual tells the researcher to collect data from each
student's response is recorded. On the other college.
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 97
Researchers choose among different units to the unit on which you collect data (see Box
or levels of analysis for similar topics or research 4.7).
questions. For example, a researcher could con
duct a project on the topic of patriarchy and vi Example. Tomsville and Ioansville each have
olence with society as the unit of analysis for the about 45,000 people living in them. Tomsville
research question, "Are patriarchal societies has a high percentage of upper-income people.
more violent?" He or she would collect data on Over half of the households in the town have
societies and classify each society by its degree of family incomes of over $200,000. The town also
patriarchy and its level of violence. On the other has more motorcycles registered in it than any
hand, if the research question was "Is the degree other town of its size. The town of Ioansville has
of patriarchy within a family associated with vi many poor people. Half its households live be-
olence against a spouse?" the unit of analysis
could be the group or the family, and a more mi
cro level of analysis would be appropriate. The
researcher could collect data on families by mea
suring the degree of patriarchy within different
families and the level of violence between Researchers have criticized the famous study Suicide
spouses in these families. The same topic can be ([1897] 1951) by Emile Durkheim for the ecologi
addressed with different levels and units of calfallacy of treating group data as though they were
analysis because patriarchy can be a variable that individual-level data. In the study, Durkheim com
describes an entire society, or it can describe so pared the suicide rates of Protestant and Catholic
cial relations within one family. Likewise, vio districts in nineteenth-century western Europe and
lence can be defined as general behavior across a explainedobserved differences as due to differences
society, or as the interpersonal actions of one between people's beliefsand practices in the two re
spouse toward the other. ligions. He said that Protestants had a higher suicide
rate than Catholics because they were more individ
Ecological Fallacy. The ecological fallacy arises ualistic and had lower social integration. Durkheim
from a mismatch of units of analysis. It refers to and early researchers only had data by district. Since
a poor fit between the units for which a re people tended to reside with others of the same re
searcher has empirical evidence and the units ligion, Durkheim used group-level data (i.e., region)
for which he or she wants to make statements. It for individuals.
is due to imprecise reasoning and generalizing Later researchers (van Poppel and Day, 1996)
beyond what the evidence warrants. Ecological reexamined nineteenth-century suicide rates only
fallacy occurs when a researcher gathers data at with individual-level data that they discovered for
some areas. They compared the death records and
a higher or an aggregated unit of analysis but
looked at the official reason of death and religion,
wants to make a statement about a lower or
but their results differed from Durkheim's. Appar
disaggregated unit. It is a fallacy because what
ently, local officials at that time recorded deaths dif
happens in one unit of analysis does not always
ferently for people of different religions. They
hold for a different unit of analysis. Thus, if a re recorded "unspecified" as a reason for death far more
searcher gathers data for large aggregates (e.g., often for Catholics because of a strong moral prohi
organizations, entire countries, etc.) and then bition against suicide among Catholics. Ourkheim's
draws conclusions about the behavior of indi larger theory may be correct, yet the evidence he
viduals from those data, he or she is commit had to test it was weakbecause he used data aggre
ting the ecological fallacy. You can avoid this gated at the group level while trying to explain the
error by ensuring that the unit of analysis you actions of individuals.
use in an explanation is the same as or very close
98 PART ONE / FOUNDATIONS
low the poverty line. It also has fewer motorcy the ecological fallacy or reductionism. They
cles registered in it than any other town its size. make a mistake about the data appropriate for a
But it is a fallacy to say, on the basis of this infor research question, or they may seriously over
mation alone, that rich people are more likely to generalize from the data. t
own motorcycles or that the evidence shows a You can make assumptions about units of Suppose )QI JIIid1
relationship between family income and motor analysis other than the ones you study empiri
cycle ownership. The reason is that we do not
know which families in Tomsville or ]oansville
cally. Thus, research on individuals rests on as
sumptions that individuals act within a set of Americ:GIf1'Kr~
CHi Rit;I«s &.
jOOty• .....
own motorcycles. We only know about the two social institutions. Research on social institu
la...lfti~
variables-average income and number of mo tions is based on assumptions about individual
torcycles-for the towns as a whole. The unit of behavior. We know that many micro-level units ~liIII:~
analysis for observing variables is the town as a
whole. Perhaps all of the low- and middle-in
form macro-level units. The danger is that it is
easy to slide into using the causes or behavior of ~":::~~.
.
to~r:!66
come families in Tomsville belong to a motorcy micro units, such as individuals, to explain the
cle club, and not a single upper-income family
belongs. Or perhaps one rich family and five
actions of macro units, such as social institu
tions. What happens among units at one level
ucation~
1964.
_.~
rD.
poor ones in ]oansville each own motorcycles. does not necessarily hold for different units of
resultofw~
In order to make a statement about the relation
ship between family ownership of motorcycles
analysis. Sociology is a discipline that rests on
the fundamental belief that a distinct level of so
foremost eM..,
and family income, we have to collect informa cial reality exists beyond the individual. Expla This says::
tion on families, not on towns as a whole. nations of this level require data and theory that
go beyond the individual alone. The causes,
U.S. race ret_..
independmt
Reductionism. Another problem involving forces, structures, or processes that exist among If you know
mismatched units of analysis and imprecise rea macro units cannot be reduced to individual seeaproblela.
soning about evidence is reductionism, also behavior. its successes ~
called the fallacy ofnonequivalence (see Box 4.8). one individual
This error occurs when a researcher explains Example. Why did World War I occur? You and guide a
macro-level events but has evidence only about may have heard that it was because a Serbian The idea of iI
specific individuals. It occurs when a researcher shot an archduke in the AustroHungarian Em force is reduced 1D
observes a lower or disaggregated unit of analysis pire in 1914. This is reductionism. Yes, the as cial phel1OlneflOrl'-i
but makes statements about the operations of sassination was a factor, but the macro-political are the actions of
higher or aggregated units. It is a mirror image of event between nations-war-cannot be re (marches. cOtn
the mismatch error in the ecological fallacy. A duced to a specific act of one individual. If it sit-ins, rioting.
advancing a shaRd
researcher who has data on how individuals be could, we could also say that the war occurred
have but makes statements about the dynamics because the assassin's alarm clock worked and
1
of macro-level units is committing the error of woke him up that morning. If it had not worked, l
reductionism. It occurs because it is often easier there would have been no assassination, so the but individual ~
to get data on concrete individuals. Also, the op alarm clock caused the war! The event, World Thus, it is ~ ~
eration ofmacro-level units is more abstract and War I, was much more complex and was due to out at about that ~
nebulous. As with the ecological fallacy, you can many social, political, and economic forces that had not occurred. j
avoid this error by ensuring that the unit of came together at a point in history. The actions I
analysis in your explanation is very close to the of specific individuals had a role, but only a mi Spuriousness. ~
one for which you have evidence. nor one compared to these macro forces. Indi variables spurious.
Researchers who fail to think precisely viduals affect events, which eventually, in Researchers get eIj
about the units of analysis and those who do not combination with larger-scale social forces and found a spurious ~
couple data with the theory are likely to commit organizations, affect others and move nations, show that what a(I
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 99
Suppose you pick up a book and read the following: The movement's ideology, popular mobilization, pol
itics, organization, and strategy are absent. Related
American race relations changed dramatically during the macro-level historical events and trends that may
Civil Rights Era of the 796 as. Attitudes among the ma have influenced the movement (e.g., Vietnam War
jority, white population shifted to greater tolerance as protest, mood shift with the killing of John F.
laws and court rulings changed across the nation. Op Kennedy, African American separatist politics, African
portunities that had beenlegally and officially closed to American migration to urban North) are also ignored.
all but the white population-in the areas of housing, This error is not unique to historical explanations.
jobs, schooling, voting rights, and so on-were opened Many people think only in terms of individual actions
to people of allraces. From the Brown vs. Board of Ed and have an individualist bias, sometimes called
ucation decision in 7955, to the Civil Rights Act of methodological individualism. This is especially true in
7964, to the Waron Poverty from 7966 to 7968, a the extremely individualistic U.S. culture. The error is
new, dramatic outlook swept the country. This was the that it disregards units of analysis or forces beyond
result of the vision, dedication, and actions of America's the individual. The errorof reductionism shifts expla
foremost civil rights leader, Dr. Martin Luther King Jr. nation to a much lower unit of analysis. One could
continue to reduce from an individual's behavior to
This says: dependent variable = major change in biological processes in a person, to micro-level neu
U.S. race relations over a 10- to 13-year period; rochemical activities, to the subatomic level.
independentvariable = King's vision and actions. Most people live in "social worlds" focused on lo
If you know much about the civil rights era, you cal, immediate settings and their interactions with a
see a problem. The entire civil rights movement and small set of others, so their everyday sense of reality
its successes are attributed to a single individual. Yes, encourages seeing social trends or events as individ
one individual does make a difference and helps build ual actions or psychological processes. Often, they
and guide a movement, but the movement is missing. become blind to more abstract, macro-level enti
The idea of a social-political movement as a causal ties-social forces, processes, organizations, institu
force is reduced to its major leader. The distinct so tions, movements, or structures. The idea that all
cial phenomenon-a movement-is obscured. Lost social actions cannot be reduced to individuals alone
are the actions of hundreds of thousands of people is the core of sociology. In his classic work Suicide,
(marches, court cases, speeches, prayer meetings, Emile Durkheim fought methodological individualism
sit-ins, rioting, petitions, beatings, etc.) involved in and demonstrated that larger, unrecognized social
advancing a shared goal and the responses to them. forces explain even highly individual, private actions.
but individual actions alone are not the cause. and a more complex relation exists. Any associ
Thus, it is likely that a war would have broken ation between two variables might be spurious,
out at about that time even if the assassination so researchers are cautious when they discover
had not occurred. that two variables are associated; upon further
investigation, it may not be the basis for a real
Spuriousness. To call a relationship between causal relationship. It may be an illusion, just
variables spurious means that it is false, a mirage. like the mirage that resembles a pool of water on
Researchers get excited if they think they have a road during a hot day.
found a spurious relationship because they can Spuriousness occurs when two variables ap
show that what appears on the surface is false pear to be associated but are not causally related
100 PART ONE I FOUNDATIONS
•
"
tl.
a. Initial relationship
"..~
. . . . . .14
.....
~~icIIII
POSITIVE ASSOCIATION are relevant for many topics (e.g., the causes of
crime or the reasons for war or child abuse). 6r...~,..
b. Addition of the missing true causal factor
-
Example 1. Some people say that taking illegal
Parents Nearsighted drugs causes suicide, school dropouts, and vio
lent acts. Advocates of "drugs are the problem"
Real Cause
position point to the positive correlations be
tween taking drugs and being suicidal, dropping
Night-Light Usage ~-------~I Nearsighted ---J out of school, and engaging in violence. They ar
gue that ending drug use will greatly reduce sui
SPURIOUS ASSOCIATION
cide, dropouts, and violence. Others argue that
many people turn to drugs because of their emo
tional problems or high levelsof disorder of their
because an unseen third factor is the real cause communities (e.g., high unemployment, unsta
(see Box 4.9). The unseen third or other variable ble families, high crime, few community services,
is the cause of both the independent and the de lack of civility). The people with emotional prob
pendent variable in the apparent but illusionary lems or who live in disordered communities are
relationship and accounts for the observed asso also more likely to commit suicide, drop out, and
ciation. In terms of conditions for causality, the engage in violence. This means that reducing 52.
unseen factor is a more powerful alternative emotional problems and community disorder
explanation. will cause illegal drug use, dropping out, suicide,
You now understand that you should be and violence all to decline greatly. Reducing drug
wary of correlations or associations, but how can taking alone will have only a limited effect be-
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 101
cause it ignores the root causes. The "drugs are treated differently because of it, such as having
the problem" argument is spurious because the different job opportunities and housing choices.
initial relationship between taking illegal drugs Discriminated-against people who are in some
and the problems is misleading. The emotional racial categories find limits in their housing
problems and community disorder are the true choices. This means they get separated or
and often unseen causal variables. grouped together in undesirable areas. Poor
housing gets combined with unequal schooling,
Example 2. In the United States and Canada, such that the lowest-quality schools are located
we observe an empirical association between in areas with the least desirable housing. Since
students classifiedas being in a non-White racial the relationship between school quality and test
category and scoring lower academic tests scores is very strong, students from families liv
(compared to students classifed as in a White ing in less desirable housing areas with low-qual
category). The relationship between racial classi ity schools get lower test scores.
fication and test scores is illusionary, because a We can now turn from the errors in causal
powerful and little-recognized variable is the explanation to avoid and more to other issues
true cause of both the racial classification and involving hypotheses. Table 4.2 provides a re
the test scores (see Figure 4.4). In this case, the view of the major errors.
true cause operates directly on the independent
variable (racial classification) but indirectly
From the Research Question to
through an intervening process on the depen
Hypotheses
dent variable (test scores). A belief system that is
based on classifyingpeople as belonging to racial It is difficult to move from a broad topic to hy
groups and assigning great significance to super potheses, but the leap from a well-formulated
ficial physical appearance, such as skin color, is research question to hypotheses is a short one.
the basis of what people call "race." Such a belief Hints about hypotheses are embedded within a
system also is the basis for prejudice and dis good research question. In addition, hypotheses
criminatory behavior. In such a situation, people are tentative answers to research questions (see
are seen as belonging to different races and Box4.10).
Spurious Association
Students treated as belonging to Lower scores on
----------------------------------------~
a racial social category ("White" academic tests
or "Non-White") based on for non-Whites
Discrimination against
superficial physical appearance than for Whites
non-Whites in jobs
and housing
~,
Real
Cause
lower-quality schools
102 PART ONE I
---------------~-----
TAB L E 4.2
FOUNDATIONS
Ecological Fallacy The empirical observations are at too New York has a high crime rate. Joan
high a level for the causal relationship lives in New York. Therefore, she
that is stated. probably stole my watch.
Reductionism The empirical observations are at too Because Steven lost his job and did not
Iowa level for the causal relationship buy a new car, the country entered a
that is stated. long economic recession.
Spuriousness An unseen third variable is the actual Hair length is associated with TV
cause of both the independent and programs. People with short hair prefer
dependentvariab~. watching football; people with long hair
prefer romance stories. (Unseen: Gender)
Set of Variables, Not Questions • Are the children raised in poverty households
Methodological
Technique
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 103
Consider an example research question: "Is the chances that the marriage will end in divorce,
age at marriage associated with divorce?" The unless it is a marriage between members of a
question contains two variables: "age at mar tight-knit traditional religious community in
riage" and "divorce." To develop a hypothesis, a which early marriage is the norm."
researcher asks, "Which is the independent vari Formulating a research question and a hy
able?" The independent variable is "age at mar pothesis do not have to proceed in fixed stages.A
riage" because marriage must logically precede researcher can formulate a tentative research
divorce. The researcher also asks, "What is the question, then develop possible hypotheses; the
direction of the relationship?" The hypothesis hypotheses then help the researcher state the re
could be: "The lower the age at time of marriage, search question more precisely. The process is
the greater the chances that the marriage will interactive and involves creativity.
end in divorce." This hypothesis answers the re You may be wondering: Where does theory
search question and makes a prediction. Notice fit into the process of moving from a topic to a
that the research question can be reformulated hypothesis I can test? Recall from Chapter 2 that
and better focused now: "Are couples who theory takes many forms. Researchers use gen
marry younger more likely to divorce?" eral theoretical issues as a source of topics. The
Severalhypotheses can be developed for one ories provide concepts that researchers turn into
research question. Another hypothesis from the variables as well as the reasoning or mechanism
same research question is: "The smaller the dif that helps researchers connect variables into a
ference between the ages of the marriage part research question. A hypothesis can both answer
ners at the time of marriage, the less likely that a research question and be an untested proposi
the marriage will end in divorce." In this case, tion from a theory. Researchers can express a hy
the variable "age at marriage" is specified pothesis at an abstract, conceptual level or
differently. restate it in a more concrete, measurable form.
Hypotheses can specify that a relationship Examples of specific studies may help to il
holds under some conditions but not others. For lustrate the parts of the research process. For ex
example, a hypothesis states: "The lower the age amples of three quantitative studies, see Table
of the partners at time of marriage, the greater 4.3; for two qualitative studies, see Table 4.4.
Study Citation (using Goar, Carla and Jane Musick, Mark, John Lauzen, Martha M. and
ASA format style) Sell. 2005. "Using Task Wilson, and William David M. Dozier. 2005.
Definition to Modify Bynum. 2000. "Race "Maintaining the Double
Racial Inequality Within and Formal Standard: Portrayals of
Task Groups" Volunteering: The Age and Gender in
Sociological Quarterly Differential Effects of Popular Films." Sex Roles
46:525-543. Class and Religion" 52: 437-446.
Social Forces 78:
1539-70.
Main Hypothesis Groups exposed to For Whites and Blacks, As with past popular
Tested instructions that social class and religion U.s. films and in other
suggest complex and affect whether a person popular mass media, a
TA8I.E 4_4
diverse skills are volunteers in different double standard still
required to complete a
task will show less racial
inequality in their
ways. exists.
.....0
...... ~
5
operations to complete
a task than groups
without such
instructions.
Main Independent Whether groups were Social class, religious The age and gender of
Variable(s) told they were to a attendance, race. major film characters.
complete a complex
task that requires
diverse skills or not. ~
Main Dependent The amount of Whether a person said Whether a character has
Variable(s) time/involvement by he or she volunteered a leadership role, high
people of different for any of five occupational status, and
races to resolve a group organizations (religious, goals. .
task. education, political or
labor, senior citizen, or
•
local).
Unit of Analysis Mixed race task group Individual adult The movie
CHAPTER 4 I REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY lOS
------------------------1..
TAB L E 4. 3 (Continued)
--------------------......,11'
Study Citation (using ASA Lu, Shun and Gary Fine. 1995. Molotch, Harvey, William
format style) "The Presentation of Ethnic Freudenburg, and Krista Paulsen.
Authenticity: Chinese Food as a 2000. "History Repeats Itself,
Social Accomplishment" but How? City Character, Urban
Sociological Quarterly Tradition, and the
36:535-53. Accomplishment of Place."
American Sociological Review
65:791-823.
Methodological Technique Field Research Historical-Comparative Research
Topic The ways ethnic cultures are The ways cities develop a distinct
displayed within the boundaries urban "character,"
of being acceptable in the United
States and how they deploy
cultural resources.
Research Question How do Chinese restaurants Why did the California cities of
present food to balance, givinga Santa Barbara and Ventura,
feeling of cultural authenticity which appear very similar on the
and yet satisfying non-Chinese surface, develop very different
U.S. customers? characters?
(continued)
106 PART ONE / FOUNDATIONS
---------------------.
-
~.-.
IiIIE
~
s.-
,.....
exotic and authentic eating
experience.
cultural, and social factors
combine to create distinct 1DIr:III:'
Q 5
"~ -~
cultural-economic places. Similar
forces can have opposite results
depending on context.
.....
Social Context or Field Site Chinese restaurants, especially The middle part of California's
four in Athens, Georgia. Pacific coast over the past 1 00
4,
[
years.
---
. . . . . . . . . .1
• .t*
CONCLUSION
In this chapter, you encountered the ground
focus the research. The style that a researcher
uses will depend on the topic he or she selects,
the researcher's purpose and intended use of
.......
.:.--*
. . . . .IIIIIIiI
.....-
1
work to begin a study. You saw how differences study results, the orientation toward social sci ' 1
in the qualitative and quantitative styles or ap ence that he or she adopts, and the individual re .-.i 1tI
proaches to social research direct a researcher to searcher's own assumptions and beliefs.
prepare for a study differently. All social re Quantitative researchers take a linear path
searchers narrow their topic into a more specific, and emphasize objectivity. They tend to use ex
focused research question. The styles of research plicit, standardized procedures and a causal ex
suggest a different form and sequence of deci planation. Their language of variables and
sions, and different answers to when and how to hypotheses is found across many other areas of
CHAPTER 4 / REVIEWING THE SCHOLARLY LITERATURE AND PLANNING A STUDY 107
science. The process is often deductive with a se strengths and limitations of each. The ultimate
quence ofdiscrete steps that precede data collec goal is to develop a better understanding and ex
tion: Narrow the topic to a more focused planation of events in the social world. This
question, transform nebulous theoretical con comes from an appreciation of the value that
cepts into more exact variables, and develop one each style has to offer.
or more hypotheses to test. In actual practice, re
searchers move back and forth, but the general
process flows in a single, linear direction. In ad
dition, quantitative researchers take special care Key Terms
to avoid logical errors in hypothesis develop
ment and causal explanation. abstract
Qualitative researchers follow a nonlinear alternative hypothesis
path and emphasize becoming intimate with the attributes
details of a natural setting or a particular cul citation
tural-historical context. They use fewer stan dependent variable
dardized procedures or explicit steps, and often ecologicalfallacy
devise on-the-spot techniques for one situation first-order interpretation
or study. Their language of cases and contexts hypothesis
directs them to conduct detailed investigations independent variable
of particular cases or processes in their search for intervening variable
authenticity. They rarely separate planning and level of analysis
design decisions into a distinct pre-data collec linear research path
tion stage, but continue to develop the study de literature review
sign throughout early data collection. The nonlinear research path
inductive qualitative style encourages a slow, null hypothesis
flexible evolution toward a specific focus based reductionism
on a researcher's ongoing learning from the second-order interpretation
data. Grounded theory emerges from the re spuriousness
searcher's continuous reflections on the data. third-order interpretation
Too often, the qualitative and quantitative unit of analysis
distinction is overdrawn and presented as a rigid universe
dichotomy. Adherents of one style of social re variable
search frequently judge the other style on the ba
sis of the assumptions and standards of their
own style. The quantitative researcher demands Endnotes
to know the variables used and the hypothesis
tested. The qualitative researcher balks at turn 1. For a discussion of the "logic of the disconfirm
ing humanity into cold numbers. The challenge ing hypothesis," see Singleton and associates
for the well-versed, prudent social researcher is (1988:456---460).
to understand and appreciate each style or ap 2. See Bailey (1987:43) for a discussion.
proach on its own terms, and to recognize the
I
INTRODU~
Youmay~'"
to measure u.IiiI
Qualitative and to measure r.IIII:iII~
measure "'~.
Quantitative Measurement Reportsto~
social reseu~
explanation. ~
Introduction
Why Measure?
Quantitative and Qualitative Measurement
Parts of the Measurement Process .::emed*-
Quantitative Conceptualization and Operationalization ~raI.d~
disIinct *p ill
Qualitative Conceptualization and Operationalization
pDor1D cilia
Reliability and Validity
Reliability and Validity in Quantitative Research
cioIl km_"''''
drdr- tiw iIpI~"
Reliability and Validity in Qualitative Research then aaIIr
Relationship between Reliability and Validity .-daua-"
cxptCWII ill
Other Uses of the Terms Reliable and Valid
Q+-
A Guide to Quantitative Measurement
_ _ 'WIrIY. . iI-."
Levels of Measurement
Specialized Measures: Scales and Indexes
~ ... alii"
Index Construction
The Purpose
Weighting
Missing Data
.... .. .......
...._- -
c
:.-.
.......
Scales
The Purpose
** i ' "-..III. . . .
•
Logic of Scaling
~.
Commonly Used Scales
- - . ... i11
Conclusion
"_IIIIIIIIiI.,.
~- t -
....
...
_d.~_~_-
'_Ii::::.
108
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 109
We also measure the nonphysical world in observed (e.g., attitudes, ideology, divorce rates,
everyday life, but usually in less exact terms. deviance, sex roles, etc.). Like the natural scien
We are measuring when we say that a restau tist who invents indirect measures of the "invis
rant is excellent, that Pablo is really smart, that ible" objects and forces of the physical world, the
Karen has a negative attitude toward life, that social researcher devises measures for difficult
Johnson is really prejudiced, or that the movie to-observe aspects of the social world. QvaliMiw
last night had a lot of violence in it. However, bctioRcbta
such everyday judgments as "really preju . . ID05l. of
diced" or "a lot of violence" are imprecise, lion.. Tbe cpaI....
QUANTITATIVE AND
vague measures.
Measurement also extends our senses. The
QUALITATIVE MEASUREMENT ~
-.d lCDlI:JiII'JlIIII
m"
astronomer or biologist uses the telescope or the Both qualitative and quantitative researchers use .tID and 0II:II"
microscope to extend natural vision. In contrast careful, systematic methods to gather high-qual ~ftftllO
••~~••
to our senses, scientific measurement is more ity data. Yet, differences in the styles of research CIIl the pI'1laS
sensitive, varies less with the specific observer, and the types of data mean they approach the
.and yields more exact information. You recog measurement process differently. The two ap
nize that a thermometer givesmore specific, pre proaches to measurement have three distinctions.
cise information about temperature than touch One difference between the two styles in
can. Likewise, a good bathroom scale gives you volves timing. Quantitative researchers think
more specific, constant, and precise information about variables and convert them into specific .~aR51_"
about the weight of a 5-year-old girl than you get actions during a planning stage that occurs be aaJOCqJl.-'
by lifting her and calling her "heavy" or "light." fore and separate from gathering or analyzing a IleChniqIE. •
Social measures provide precise information data. Measurement for qualitative researchers ~~OI''''~
about social reality.
In addition, measurement helps us observe
what is otherwise invisible. Measurement ex
occurs during the data collection process.
A second difference involves the data itself.
Quantitative researchers develop techniques that
....RIIIIk.
Waw .....
tends human senses. It lets us observe things that can produce quantitative data (i.e., data in the . . . . ",,- iif'
were once unseen and unknown but were pre form of numbers). Thus, the researcher moves Q 5. ..
dicted by theory.
Before you can measure, you need a clear
idea about what you are interested in. For exam
from abstract ideas to specific data collection
techniques to precise numerical information
produced by the techniques. The numerical in
...........
~-~
~~ .....
difference is how the two styles make such link definition. A conceptual definition is a definition
ages. Quantitative researchers contemplate and in abstract, theoretical terms. It refers to other
reflect on concepts before they gather any data. ideas or constructs. There is no magical way to
They construct measurement techniques that turn a construct into a precise conceptual defin
bridge concepts and data. ition. It involves thinking carefully, observing
Qualitative researchers also reflect on ideas directly, consulting with others, reading what
before data collection, but they develop many, if others have said, and trying possible definitions.
not most, of their concepts during data collec How might I develop a conceptual defini
tion. The qualitative researcher reexamines and tion of the construct prejudice? When beginning
evaluates the data and concepts simultaneously to develop a conceptual definition, researchers
and interactively. Researchers start gathering often rely on multiple sources-personal experi
data and creating ways to measure based what ence and deep thinking, discussions with other
they encounter. As they gather data, they reflect people, and the existing scholarly literature. I
on the process and develop new ideas. might reflect on what I know about prejudice,
ask others what they think about it, and go the li
brary and look up its many definitions. As I
gather definitions, the core idea should get
PARTS OF THE MEASUREMENT
clearer, but I have many definitions and need to
PROCESS
sort them out. Most definitions state that preju
When a researcher measures, he or she links dice is an attitude about another group and in
a concept, idea, or construct 1 to a measure (i.e., volves a prejudgment, or judging prior to getting
a technique, a process, a procedure, etc.) by specific information.
which he or she can observe the idea empirically. As I think about the construct, I notice that
Quantitative researchers primarily follow a de all the definitions refer to prejudice as an atti
ductive route. They begin with the abstract idea, tude, and usually it is an attitude about the
I
t
follow with a measurement procedure, and end
with empirical data that represent the ideas.
Qualitative researchers primarily follow an in
ductive route. They begin with empirical data,
members of another group. There are many
forms of prejudice, but most are negative views
about persons of a different racial-ethnic group.
Prejudice could be about other kinds of groups
follow with abstract ideas, relate ideas and data, (e.g.,people of a religion, ofa physical stature, or
and end with a mixture of ideas and data. Actu from a certain region), but it is always about a
ally,the process is more interactive in both styles collectivity to which one does not belong. Many
of research. As a quantitative researcher devel constructs have multiple dimensions or types, so
ops measures, the constructs become refined I should consider whether there can be different
and clearer, and as the researcher applies the types of prejudice-racial prejudice, religious
measures to gather data, he or she often adjusts prejudice, age prejudice, gender prejudice, na
the measurement technique. As a qualitative re tion prejudice, and so forth.
searcher gathers data, he or she uses some preex I also need to consider the units of analysis
isting ideas to assist in data collection, and will that best fit my definition of the construct. Prej
then mix old with new ideas that are developed udice is an attitude. Individuals hold and express
from the data. attitudes, but so might groups (e.g., families,
Both qualitative and quantitative researchers clubs, churches, companies, media outlets). I
use two processes: conceptualization and opera need to decide, Do I want my definition of prej
tionalization in measurement. Conceptualizati udice to include only the attitudes of individuals
on is the process of taking a construct and or should it include attitudes held by groups, or
refining it by giving it a conceptual or theoretical ganizations, and institutions as well? Can I say,
112 PART ONE I FOUNDATIONS
I
Quantitative researchers developed several ways ceptualizes a variable, giving it a clear concep
to rigorously link abstract ideas to measurement tual definition. Next, he or she operationalizes it
procedures that will produce precise quantita by developing an operational definition or set
tive information about empirical reality. of indicators for it. Last, he or she applies the
,- Figure 5.1 illustrates the measurement indicators in the empirical world. The links
process for two variables that are linked together from abstract constructs to empirical reality al
in a theory and a hypothesis. There are three lev low the researcher to test empirical hypotheses.
els to consider: conceptual, operational, and em Those tests are logically linked back to a con
pirical. At the most abstract level, the researcher ceptual hypothesis and causal relations in the
is interested in the causal relationship between world of theory.
two constructs, or a conceptual hypothesis. At the A hypothesis has at least two variables, and
level of operational definitions, the researcher is the processes of conceptualization and opera
interested in testing an empirical hypothesis to tionalization are necessary for each variable. In
determine the degree of association between in the preceding example, prejudice is not a
dicators. This is the level at which correlations, hypothesis. It is one variable. It could be a de
statistics, questionnaires, and the like are used. pendent variable caused by something else, or it
The third levelis the concrete empirical world. If co.uld be an independent variable causing
the operational indicators of variables (e.g., something else. It depends on my theoretical
questionnaires) are logically linked to a con explanation.
struct (e.g., racial discrimination), they will cap We can return to the quantitative study by
ture what happens in the empirical social world Weitzer and Tuch on perceptions of police bias
and relate it to the conceptual level. and misconduct discussed in Chapter 2 for an
Conceptualization
Conceptual Definition
Conceptualization
Conceptual Definition
} Level of
Theory
I
Operationalization
I
Operationalization
} Operational
Level
Tested Empirical
Hypothesis } Empirical
Level
114 PART ONE / FOUNDATIONS
example of how researchers conceptualize and sense" or organize the data and one's prelimi Just as qlWJli
operationalize variables. It is an explanatory nary ideas. ates from a rizid 4
study with two main variables in a causal hy As the researcher gathers and analyzes qual followed by ~
pothesis. The researchers began with the itative data, he or she develops new concepts, tual interaction. 1
conceptual hypothesis: Members of a nondomi formulates definitions for the concepts, and from beyond thd
nant racial group are more likely than a domi considers relationships among the concepts. ting. Qualitaliw:l
nant racial group to believe that policing is Eventually, he or she links concepts to one an howthe~
racially biased, and their experience with polic other to create theoretical relationships that may the researcher's.
ing and exposure to media reports on police or may not be causal. Qualitative researchers and CODcep6
racial bias increase the perception of racial bias. form the concepts as they examine their qualita emerged ciuriIIc
They conceptualized the independent variable, tive data (i.e., field notes, photos and maps.his qualitatiw
dominant racial group, as White and the non torical documents, etc.). Often, this involves a tually intel1l:kpc-t
dominant group as non-White subdivided into researcher asking theoretical questions about the '\Tecan~
Black and Hispanic. The researchers conceptual data (e.g., Is this a case of class conflict? What is tionalizatioo ill
ized the dependent variable, racially biased polic the sequence of events and could it be different?
of law bv Ed~"
ing, as unequal treatment by the police ofWhites Why did this happen here and not somewhere cussed in ......_----..0
and non-Whites and racial prejudice by police else?). de\-doped~
officers. The researchers operationalized the in A qualitative researcher conceptualizes by began ""ith ~
dependent variable by self-identification to a
survey question about race. They operationalized
developing clear, explicit definitions of con
structs. The definitions are somewhat abstract
r.nions CIIIa".
bte I~.
the dependent variable by using four sets ofsur and linked to other ideas, but usually they are tiutfums_
vey questions: (l) questions about whether po also closely tied to specific data. They can be ex *¥",~mr -
I.
lice treat Blacks better, the same, or worse than pressed in the words and concrete actions of the .ooIies-.l
Whites, and the same question with Hispanics people being studied. In qualitative research,
tDh-~
substituted for Blacks; (2) questions about conceptualization is largely determined by the en -
...........
hoods; (3) a question about whether there is process for qualitative research significantly
racial-ethnic prejudice among police officers in differs from that in quantitative research and -~
6r-.::;........
the city; and (4) a question about whether police often precedes conceptualization. A researcher
are more likely to stop some drivers because they forms conceptual definitions out of rudimen
are Black or Hispanic. tary "working ideas" that he or she used while
making observations or gathering data. Instead
Qualitative Conceptualization and of turning refined conceptual definitions into a
set of measurement operations, a qualitative re
Operationalization
searcher operationalizes by describing how
6 S
Conceptualization. The conceptualization specific observations and thoughts about the
process in qualitative research also differs from data contributed to working ideas that are the
that in quantitative research. Instead of refining basis of conceptual definitions and theoretical
abstract ideas into theoretical definitions early in concepts.
the research process, qualitative researchers re Operationalization in qualitative research is
fine rudimentary "working ideas" during the an after-the-fact description more than a before
data collection and analysis process. Conceptu the-fact preplanned technique. Almost in a re
alization is a process of forming coherent theo verse of the quantitative process, data gathering
retical definitions as one struggles to "make occurs with or prior to full operationalization.
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 115
research matches with the constructs researchers Otherwise, it is impossible to determine which
use to understand it. concept is being "indicated." For example, the FIGURE 5.Z
Qualitative and quantitative researchers indicator of a pure chemical compound is more
want reliable and valid measurement, but be reliable than one in which the chemical is mixed A
yond an agreement on the basic ideas at a gen with other material or dirt. In the latter case, it is
eral level, each style sees the specifics of
reliability and validity in the research process
difficult to separate the "noise" of other material
from the pure chemical.
r.....
I.
V. . . . . .
differently.
Increase the Level of Measurement. Levels of
Reliability and Validity in
measurement are discussed later. Indicators at
higher or more precise levels of measurement
Quantitative Research
are more likely to be reliable than less precise
Reliability. As just stated, reliability means de measures because the latter pick up less detailed
pendability. It means that the numerical results information. If more specific information is
produced by an indicator do not vary because of measured, then it is less likely that anything
characteristics of the measurement process or other than the construct will be captured. The
measurement instrument itself. For example, I
get on my bathroom scale and read my weight. I
get off and get on again and again. I have a reli
able scale if it gives me the same weight each
general principle is: Try to measure at the most
precise level possible. However, it is more diffi
cult to measure at higher levelsof measurement.
For example, if I have a choice of measuring
~
icnLt
...pm'"
willi
ota d:dia~
time-assuming, ofcourse, that I am not eating, prejudice as either high or low, or in 10 cate mapau..c..1
drinking, changing clothing, and so forth. An gories from extremely low to extremely high, it read~~
unreliable scale will register different weights would be better to measure it in 10 refined ~Iq.1Od5
each time, even though my "true" weight does categories. ~ wwl'_lI:a"
not change. Another example is my car
speedometer. IfI am driving at a constant slow Use Multiple Indicators of a Variable. A third .:omn~
speed on a level surface, but the speedometer way to increase reliability is to use multiple indi iti.:atiom.. ... I
needle jumps from one end to the other, my cators, because two (or more) indicators of the or-edmiL!IWIP
speedometer is not a reliable indicator of how same construct are better than one. Figure 5.2 il ~dcridrs_
fast I am traveling. lustrates the use of multiple indicators in hy ~OIl_<III'I"'.
pothesis testing. Three indicators of the one ~(uItipIr
How to Improve Reliability. It is rare to have independent variable construct are combined
perfect reliability. There are four ways to in into an overall measure,A, and two indicators of ataOWl, '
crease the reliability of measures: (1) clearlycon a dependent variable are combined into a single the COII5I:IWI
ceptualize constructs, (2) use a precise level of measure, B. own uin..:
......
-allIL
measurement, (3) use multiple indicators, and For example, I create three indicators of the ,;pJi"Stioo . . ~
(4) use pilot-tests. variable, racial-ethnic prejudice. My first indica but~_...
tor is an attitude question on a survey. I ask re ~ .S!'5I .
Clearly ConceptualizeAll Constructs. Reliability search participants their beliefs and feelings lDieaSWak..t.
increases when a single construct or subdimen about many different racial and ethnic groups.
"'oor~
sian of a construct is measured. This means For a second indicator, I observe research par
developing unambiguous, clear theoretical defi ticipants from various races and ethnic groups
nitions. Constructs should be specified to elimi interacting together over the course of three
nate "noise" (i.e., distracting or interfering days. I look for those who regularly either (1)
information) from other constructs. Each mea avoid eye contact, appear to be tense, and sound
sure should indicate one and only one concept. cool and distant; or (2) make eye contact, appear
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 117
----------------------..
A B
Independent
Variable Measure
------~~~~~-----_.
Association?
Dependent
Variable Measure
relaxed, and sound warm and friendly as they in a hypothesis-testing situation. This takes more
teract with people of their same or with people time and effort.
of a different racial-ethnic group. Last, I create The principle of using pilot-tests extends to
an experiment. I ask research participants to replicating the measures other researchers have
read the grade transcripts, resumes, and inter used. For example, I search the literature and find
view reports on 30 applicants for five jobs measures of prejudice from past research. I may
youth volunteer coordinator, office manager, want to build on and use a previous measure if it
janitor, clothing store clerk, and advertising ac is a good one, citing the source, of course. In ad
count executive. The applicants have many qual dition, I may want to add new indicators and
ifications, but I secretly manipulate their racial compare them to the previous measure.
or ethnic group to see whether a research partic
ipant decides on the best applicant for the jobs Validity. Validity is an overused term. Some
based on an applicant's race and ethnicity. times, it is used to mean "true" or "correct."
Multiple indicators let a researcher take There are several general types of validity. Here,
measurements from a wider range of the content we are concerned with measurement validity.
of a conceptual definition. Different aspects of There are also several types of measurement va
the construct can be measured, each with its lidity. Nonmeasurement types of validity are dis
own indicator. Also, one indicator (e.g., one cussed later.
question on a questionnaire) may be imperfect, When a researcher says that an indicator is
but several measures are less likely to have the valid, it is valid for a particular purpose and def
same (systematic) error. Multiple indicator inition. The same indicator can be valid for one
measures tend to be more stable than measures purpose (i.e., a research question with units of
with one item. analysis and universe) but less valid or invalid
for others. For example, the measure of preju
Use Pretests, Pilot Studies, and Replication. dice discussed here might be valid for measuring
Reliability can be improved by using a pretest or prejudice among teachers but invalid for mea
pilot version of a measure first. Develop one or suring the prejudice of police officers.
more draft or preliminary versions of a measure At its core, measurement validity refers to
and try them before applying the final version in how well the conceptual and operational defini
118 PART ONE I FOUNDATIONS
tions mesh with each other. The better the fit, t-Content Validity. Content validity is a special similar construe
the greater the measurement validity. Validity is type of face validity. It addresses the question, Is similar results,
more difficult to achieve than reliability. We the full content of a definition represented in a
cannot have absolute confidence about validity, measure? A conceptual definition holds ideas; it
but some measures are more valid than others. is a "space" containing ideas and concepts. Mea
The reason we can never achieve absolute valid sures should represent all ideas or areas in the
ity is that constructs are abstract ideas, whereas conceptual space. Content validity involves
indicators refer to concrete observation. This is three steps. First, specify fully the entire content
the gap between our mental pictures about the in a construct's definition. Next, sample from all
world and the specific things we do at particular areas of the definition. Finally, develop an indi
times and places. Validity is part of a dynamic cator that taps all of the parts of the definition.
process that grows by accumulating evidence An example of content validity is my defin
over time. Without it, all measurement becomes ition offeminism as a person's commitment to a
meaningless. set of beliefs creating full equality between men
and women in areas of the arts, intellectual pur
Three Types ofMeasurement Validity suits, family, work, politics, and authority rela
Face Validity. The easiest to achieve and the tions. I create a measure of feminism in which I
most basic kind of validity is face validity. It is a ask two survey questions: (1) Should men and
judgment by the scientific community that the women get equal pay for equal work and (2)
indicator really measures the construct. It ad Should men and women share household tasks?
dresses the question, On the face of it, do people My measure has low content validity because the
believe that the definition and method of mea two questions ask only about pay and household
surement fit? It is a consensus method. For ex tasks. They ignore the other areas (intellectual
ample, few people would accept a measure of pursuits, politics, authority relations, and other
college student math ability using a question aspects of work and family). For a content-valid
that asked students: 2 + 2 = ? This is not a valid measure, I must either expand the measure or
measure of college-level math ability on the face narrow the definition.
of it. Recall that in the scientific community, as Criterion Validity. Criterion validity uses some
pects of research are scrutinized by others. See standard or criterion to indicate a construct ac
Table 5.1 for a summary of types of measure curately. The validity of an indicator is verified
ment validity. by comparing it with another measure of the
same construct that is widely accepted. There are
----------.
similar construct, it is logical for them to yield qualitative researchers apply the principles dif
similar results. ferently in practice.
OdIerths
~
~
~ 1"dilI ,...
criticisms to determine whether the evidence A diagram might help you see the relation
they have is real or they believe it to be. Qualita ship between reliability and validity. Figure 5.3 1IIM:s 1IIr
tive researchers adhere to the core principle of illustrates the relationship between the concepts lks.-r......
validity, to be truthful (i.e., avoid false or dis by using the analogy of a target. The bull's-eye
torted accounts). They try to create a tight fit represents a fit between a measure and the defi
between their understanding, ideas, and state nition of the construct.
ments about the social world and what is actu Validity and reliability are usually comple
ally occurring in it. mentary concepts, but in some situations they
--------------------.
.
-..ilc......
..
_
, 7
*-r-.r
.....
.
...
_--,
_
..
.
conflict with each other. Sometimes, as validity External Validity. Externalvalidity is used pri
increases, reliabilityis more difficult to attain, and marily in experimental research. It is the ability
vice versa. This occurs when the construct has a to generalize findings from a specific setting and
highly abstract and not easily observable defini small group to a broad range of settings and peo
naire give reliable measures, but there is a danger Statistical Validity. Statistical validity means
oflosing the subjective essence of the concept. that the correct statistical procedure is chosen
and its assumptions are fully met. Different sta
Other Uses of the Terms Reliable tistical tests or procedures are appropriate for
and Valid different conditions, which are discussed in text
books that describe the statistical procedures.
Many words have multiple definitions, includ
All statistics are based on assumptions about
ing reliability and validity. This creates confusion
the mathematical properties of the numbers be
unless we distinguish among alternative uses of
ing used. A statistic will be invalid and its results
the same word.
nonsense if the major assumptions are violated.
Reliability. We use reliability in everyday lan For example, to compute an average (actually the
guage. A reliable person is one who is depend mean, which is discussed in a later chapter), one
able, stable, and responsible; a reliable car is cannot use information at the nominal level of
dependable and trustworthy. This means the measurement (to be discussed). For example,
person responds in similar, predictable ways in suppose I measure the race of a class of students.
different times and conditions; the same can be I give each race a number: White > 1, African
said for the car. In addition to measurement re American> 2, Asian > 3, others =: 4. It makes no
liability, researchers sometimes say a study or its sense to say that the "mean" race of a class of stu
results are reliable. By this, they mean that the dents is 1.9 (almost African American?). This is a
method of conducting a study or the results misuse of the statistical procedure, and the re
from it can be reproduced or replicated by other sults are invalid even if the computation is cor
researchers. rect. The degree to which statistical assumptions
can be violated or bent (the technical term is
Internal Validity. Internal validity means robustness) is a topic in which professional statis
there are no errors internal to the design of the ticians take great interest.
research project. It is used primarily in experi
mental research to talk about possible errors or
alternative explanations of results that arise de
A GUIDE TO QUANTITATIVE
spite attempts to institute controls. High inter
MEASUREMENT
nal validity means there are few such errors. Low
internal validity means that such errors are Thus far, you have learned about the principles
likely. of measurement, including the principles of reli
122 PART ONE / FOUNDATIONS
ability and validity. Quantitative researchers have Four Levels ofMeasurement with less preosa
developed ideas and specialized measures to help Precision and Levels. The idea oflevels of mea tion and then l1li
them in the process of creating operational defi surement expands on the difference between
nitions that will be reliable and valid measures continuous and discrete variables and organizes Distinguishing olI
and yield numerical data for their variable con types of variables for their use in statistics. The levels from 10\<111:
structs. This section ofthe chapter is a brief guide four levels of measurement categorize the degree sion are noJ:lliDl
to these ideas and a few of the measures. of precision of measurement. Each level gives ,
Deciding on the appropriate level of mea (see Table 5.~ l.ll
surement for a construct often creates confu that there is a dill
Levels of Measurement
sion. The appropriate level of measurement for a religion: Prot~
Levels of measurement is an abstract but impor variable depends on two things: (1) how a con racial heritage: .4
tant and widely used idea. Basically, it says that struct is conceptualized and (2) the type of indi panic, other). 01
some ways a researcher measures a construct are cator or measurement that a researcher uses. ference, plus tiwl
at a higher or more refined level, and others are The way a researcher conceptualizes a ranked (e.g., letre
crude or less precisely specified. The level of construct can limit how precisely it can be mea measures: Stro~
measurement depends on the way in which a sured. For example, some of the variables listed Strongly DisagmI:;
construct is conceptualized-that is, assump
tions about whether it has particular characteris
earlier as continuous can be reconceptualized as
discrete. Temperature can be a continuous vari
the first two
distance~
do."
tics. The level of measurement affects the kinds able (e.g., degrees, fractions of degrees) or it can Celsius teIl1J>eT3t'
of indicators chosen and is tied to basic assump be crudely measured with discrete categories no, 125). AlQnn.
tions in a construct's definition. The way in (e.g., hot or cold). Likewise, age can be continu val measures; thII
which a researcher conceptualizes a variable lim ous (how old a person is in years, months, days, score. Ratio JDeaSI
its the levels of measurement that he or she can hours, and minutes) or treated as discrete cate levels do, plus~
use and has implications for how measurement gories (infancy, childhood, adolescence, young possible to state nj
and statistical analysis can proceed. adulthood, middle age, old age). Yet, most dis or ratios (e.g., IIIIlIl
crete variables cannot be conceptualized as con years of formal scl
Continuous and Discrete Variables. Vari tinuous variables. For example, sex, religion, and years).
ables can be thought of as being either contin marital status cannot be conceptualized as con In most pra.:Q
uous or discrete. Continuous variables have an tinuous; however, related constructs can be con between inten~l.
infinite number ofvalues or attributes that flow ceptualized as continuous (e.g., femininity, ference. The
along a continuum. The values can be divided degree of religiousness, commitment to a mari measures can be
into many smaller increments; in mathematical tal relationship, etc.). in temperature
theory, there is an infinite number of incre The level of measurement limits the statisti ally a doublingoft1
ments. Examples of continuous variables in cal measures that can be used. A wide range of I
clude temperature, age, income, crime rate, and powerful statistical procedures are available for
amount of schooling. Discrete variables have a the higher levels of measurement, but the types
relatively fixed set of separate values or variable of statistics that can be used with the lowest lev TABLE 5.2
attributes. Instead of a smooth continuum of els are very limited.
values, discrete variables contain distinct cate There is a practical reason to conceptualize
gories. Examples of discrete variables include and measure variables at higher levels of mea
gender (male or female), religion (Protestant, surement. You can collapse higher levels of mea
Nominal
Catholic, Jew, Muslim, atheist), and marital surement to lower levels, but the reverse is not
status (never married single, married, di true. In other words, it is possible to measure a Ordinal
vorced or separated, widowed). Whether a construct very precisely, gather very specific in Interval
variable is continuous or discrete affects its formation, and then ignore some of the preci
Ratio
level of measurement. sion. But it is not possible to measure a construct
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 123
with less precision or with less specific informa numbers double, because zero degrees is not the
tion and then make it more precise later. absence of all heat.
Discrete variables are nominal and ordinal,
Distinguishing among the FourLevels. The four whereas continuous variables can be measured
levels from lowest to greatest or highest preci at the interval or ratio level. A ratio-level mea
sion are nominal, ordinal, interval, and ratio. sure can be turned into an interval, ordinal, or
Each level gives a different type of information nominal level. The interval level can always be
(see Table 5.2). Nominal measures indicate only turned into an ordinal or nominal level, but the
that there is a difference among categories (e.g., process does not work in the opposite way!
religion: Protestant, Catholic, Jew, Muslim; In general, use at least five ordinal categories
racial heritage: African, Asian, Caucasian, His and obtain many observations. This is because
panic, other). Ordinal measures indicate a dif the distortion created by collapsing a continu
ference, plus the categories can be ordered or ous construct into a smaller number of ordered
ranked (e.g., letter grades: A, B, C, D, F; opinion categories is minimized as the number of cate
measures: Strongly Agree, Agree, Disagree, gories and the number of observations increase.
Strongly Disagree). Interval measures everything The ratio level of measurement is rarely
the first two do, plus it can specifythe amount of used in the social sciences. For most purposes, it
distance between categories (e.g., Fahrenheit or is indistinguishable from interval measurement.
Celsius temperature: 5°,45°, 90°; IQ scores: 95, The only difference is that ratio measurement
110,125). Arbitrary zeroes may be used in inter has a "true zero." This can be confusing because
val measures; they are just there to help keep some measures, like temperature, have zeroes
score. Ratio measures do everything all the other that are not true zeroes. The temperature can be
levels do, plus there is a true zero, which makes it zero, or below zero, but zero is an arbitrary
possible to state relations in terms of proportion number when it is assigned to temperature. This
or ratios (e.g., money income: $10, $100, $500; can be illustrated by comparing zero degrees
years of formal schooling: 1 year, 10 years, 13 Celsius with zero degrees Fahrenheit-they are
years). different temperatures. In addition, doubling
In most practical situations, the distinction the degrees in one system does not double the
between interval and ratio levelsmakes little dif degrees in the other. Likewise, it does not make
ference. The arbitrary zeroes of some interval sense to say that it is "twice as warm," as is pos
measures can be confusing. For example, a rise sible with ratio measurement, if the temperature
in temperature from 30 to 60 degrees is not re rises from 2 to 4 degrees, from 15 to 30 degrees,
ally a doubling of the temperature, although the or from 40 to 80 degrees. Another common ex-
Nominal Yes
Ordinal Yes Yes
Interval Yes Yes Yes
Ratio Yes Yes Yes Yes
124 PART ONE / FOUNDATIONS
ample of arhitrary-not true-zeroes occurs Keep two things in mind. First, virtually
when measuring attitudes where numbers are every social phenomenon can be measured.
assigned to statements (c.g., -1 = disagree, 0 = Some constructs can be measured directly and
no opinion, +1 = agree). True zeroes exist for produce precise numerical values (e.g., family
variables such as income, age, or years of educa income). Other constructs require the use of
tion. Examples of the four levelsof measurement surrogates or proxies that indirectly measure a _,ill..........
are shown in Table 5.3. variable and may not be as precise (e.g., predis
position to commit a crime). Second, a lot can
be learned from measures used by other re
Specialized Measures: Scales
searchers. You are fortunate to have the work of
and Indexes thousands of researchers to draw on. It is not al
• 5
Researchers have created thousands of different ways necessary to start from scratch. You can usc
'Ii. -
scales and indexes to measure social variables.
For example, scales and indexes have been de
veloped to measure the degree of formalization
in bureaucratic organizations, the prestige of oc
a past scale or index, or you can modify it for
your own purposes.
cupations, the adjustment of people to a mar index and scale confusing because they are often
riage, the intensity of group interaction, the level used interchangeably. One researcher's scale is
of social activity in a community, the degree to another's index. Both produce ordinal- or inter
which a state's sexual assault laws reflect feminist val-level measures of a variable. To add to the
values, and the level of socioeconomic develop confusion, scale and index techniques can be
ment of a nation. I cannot discuss the thousands combined in one measure. Scales and indexes
of scales and indexes. Instead, I will focus on give a researcher more information about vari
principles of scale and index construction and ables and make it possible to assess the quality of
explore some major types. measurement. Scales and indexes increase relia
---------------------.
Religion (nominal) Different religious denominations (Jewish, Catholic, Lutheran, Baptist) are not
ranked, just different (unless one belief is conceptualized as closer to heaven).
Attendance (ordinal) "How often do you attend religious services? (0) Never, (1) less than once a
year, (3) several times a year, (4) about once a month, (5) two or three times
a week, or (8) several times a week?" This might have been measured at a ratio
level if the exact number of times a person attended was asked instead.
IQ Score (interval) Most intelligence tests are organized with 100 as average, middle, or normal.
Scores higher or lower indicate distance from the average. Someone with a
score of 11 5 has somewhat above average measured intelligence for people
who took the test, while 90 is slightly below. Scores of below 65 or above 140
are rare.
Age (ratio) Age is measured by years of age. There is a true zero (birth). Note that a 40
year-old has lived twice as long as a 20-year-old.
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 1ZS
der. Feminist ideology is a highly abstract and inflation, is created by totaling the cost of buying
general construct. It includes specific beliefs and a list of goods and services (e.g., food, rent, and
attitudes toward social, economic, political, fam utilities) and comparing the total to the cost of
ily, and sexual relations. The ideology's five be buying the same list in the previous year. The
lief areas are parts of the single general construct. consumer price index has been used by the U.S.
The parts are mutually reinforcing and together Bureau of Labor Statistics since 1919; wage in
form a system of beliefs about the dignity, creases, union contracts, and social security pay
strength, and power of women. ments are based on it. An index is a combination
If feminist ideology is unidimensional, then of items into a single numerical score. Various
-"!ail_III
--_ .
there is a unified belief system that varies from components or subparts of a construct are each
_cw'r.'4
very antifeminist to very profeminist. We can test measured, then combined into one measure. ~ j
the validity of the measure that includes multiple There are many types of indexes. For exam
indicators that tap the construct's subparts. If ple, if you take an exam with 25 questions, the
one belief area (e.g., sexual relations) is consis total number of questions correct is a kind of in a..:,. • -• ..
tently distinct from the other areas in empirical
tests, then we question its unidimensionality.
dex. It is a composite measure in which each
question measures a small piece of knowledge,
-.a'"
...........
It is easy to become confused: A specific and all the questions scored correct or incorrect
measure can be an indicator of a unidimensional are totaled to produce a single measure.
construct in one situation and indicate a part of Indexes measure the most desirable place to -
a different construct in another situation. This is live (based on unemployment, commuting time,
-
..
...
possible because constructs can be used at dif crime rate, recreation opportunities, weather,
ferent levels of abstraction. and so on), the degree of crime (based on com -=. ..
For example, a person's attitude toward bining the occurrence of different specific
gender equality with regard to pay is more spe crimes), the mental health of a person (based on
cificand less abstract than feminist ideology (i.e., the person's adjustment in various areas oflife),
beliefs about gender relations throughout soci and the like.
ety). An attitude toward equal pay can be both a One way to demonstrate that indexes are
unidimensional construct in its own right and a not very complicated is to use one. Answer yes or
subpart of the more general and abstract unidi no to the seven questions that follow on the
mensional construct, ideology toward gender characteristics of an occupation. Base your an
relations. swers on your thoughts regarding the following
four occupations: long-distance truck driver,
medical doctor, accountant, telephone operator.
Score each answer 1 for yes and 0 for no.
INDEX CONSTRUCTION
1. Does it pay a good salary?
The Purpose
2. Is the job secure from layoffs or unemploy
You hear about indexes all the time. For example, ment?
U.S. newspapers report the Federal Bureau of In 3. Is the work interesting and challenging?
vestigation (FBI) crime index and the consumer 4. Are its working conditions (e.g., hours,
price index (CPI). The FBI index is the sum of safety, time on the road) good?
police reports on seven so-called index crimes 5. Are there opportunities for career advance
(criminal homicide, aggravated assault, forcible ment and promotion?
rape, robbery, burglary, larceny of $50 or more, 6. Is it prestigious or looked up to by others?
and auto theft). It began with the Uniform Crime 7. Does it permit self-direction and the free
Report in 1930. The CPI, which is a measure of dom to make decisions?
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 127
Total the seven answers for each of the four are threatened whenever data for some cases are
occupations. Which had the highest and which missing. There are four ways to attempt to re
had the lowest score? The seven questions are solve the problem, but none fully solve it.
my operational definition of the construct good For example, I construct an index of the de
occupation. Each question represents a subpart gree of societal development in 1975 for 50 na
of my theoretical definition. A different theoret tions. The index contains four items: life
ical definition would result in different ques expectancy, percentage of homes with indoor
tions, perhaps more than seven. plumbing, percentage ofpopulation that is liter
Creating indexes is so easy that it is impor ate, and number of telephones per 100 people. I
tant to be careful that every item in the index has locate a source of United Nations statistics for
face validity. Items without face validity should my information. The values for Belgium are 68 +
be excluded. Each part of the construct should 87 + 97 + 28; for Turkey, the scores are 55 + 36
be measured with at least one indicator. Of + 49 + 3; for Finland, however, I discover that
course, it is better to measure the parts of a con literacy data are unavailable. I check other
struct with multiple indicators. sources of information, but none has the data
because they were not collected.
Weighting
Rates and Standardization
An important issue in index construction is
whether to weight items. Unless it is otherwise You have heard of crime rates, rates of popula
stated, assume that an index is unweighted. Like tion growth, and the unemployment rate. Some
wise, unless you have a good theoretical reason indexes and single-indicator measures are ex
for assigning different weights, use equal pressed as rates. Rates involve standardizing the
weights. An unweighted index gives each item value of an item to make comparisons possible.
equal weight. It involves adding up the items The items in an index frequently need to be stan
without modification, as if each were multiplied dardized before they can be combined.
by 1 (or -1 for items that are negative). Standardization involves selecting a base
In a weighted index, a researcher values or and dividing a raw measure by the base. For ex
weights some items more than others. The size ample, City A had 10 murders and City B had 30
of weights can come from theoretical assump murders in the same year. In order to compare
tions, the theoretical definition, or a statistical murders in the two cities, the raw number
technique such as factor analysis. Weighting of murders needs to be standardized by the city
changes the theoretical definition of the con population. If the cities are the same size, City B
struct. is more dangerous. But City B may be safer if it is
Weighting can produce different index much larger. For example, if City A has 100,000
scores, but in most cases, weighted and un people and City B has 600,000, then the murder
weighted indexes yield similar results. Re rate per 100,000 is 10 for City A and 5 for City B.
searchers are concerned with the relationship Standardization makes it possible to com
between variables, and weighted and un pare different units on a common base. The
weighted indexes usually give similar results for process of standardization, also called norming,
the relationships between variables. removes the effect of relevant but different char
acteristics in order to make the important differ
ences visible. For example, there are two classes
Missing Data
of students. An art class has 12 smokers and a bi
Missing data can be a serious problem when ology class has 22 smokers. A researcher can
constructing an index. Validity and reliability compare the rate or incidence of smokers by
128 PART ONE / FOUNDATIONS
standardizing the number of smokers by the size When combining several items into an index, it --
of the classes. The art class has 32 students and is best to standardize items on a common base . Be'
the biology class has 143 students. One method (see Box 5.3). 5.3
of standardization that you already know is the
use of percentages, whereby measures are stan Sports fans in a.r
u
dardized to a common base of 100. In terms of SCALES "winning· at 1tIr 201
percentages, it is easy to see that the art class has most gold -.:i& I
more than twice the rate of smokers (37.5 per The Purpose sto~ft·_
Unemployment rate ==
Number of
unemployed people
Total number of
people working
struct. For example, a researcher believes that
there is a single ideological dimension that un
derlies people's judgments about specific poli
cies (e.g., housing, education, foreign affairs,
2
3
4
1
....
\
I.6A
0iIIII
l
Sports fans in the United States were jubilant about ulations and wealth. The results show that the Ba
"winning" at the 2000 Olympics by carrying off the hamas, with less than 300,000 citizens (smallerthan
most gold medals. However, because they failed to a medium-sized U.S. city), proportionately won the
standardize, the "win" is an illusion. Of course, the most gold. Adjusted for its population size or wealth,
world's richest nation with the third largest popula the United States is not even near the top; it appears
tion does well in one-on-one competition among all to be the leader only because of its great size and
nations. To see what reallyhappened, one must stan wealth. Sports fans in the United States can perpet
dardize on a base of the population or wealth. Stan uate the illusion of being at the top only if they ig
dardization yields a more accurate picture by nore the comparative advantage of the United
adjusting the results as ifthe nations had equal pop States.
TOP TEN GOLD MEDAL WINNING COUNTRIES AT THE 2000 OLYMPICS IN SYDNEY
Note: 'Population is gold medals per 10 million people and GDP is gold medals per $1 0 billion;
Figure 5.4 is an example of a "feeling ther measure attitudes toward candidates, social
mometer" scale that is used to find out how peo groups, and issues.
ple feel about various groups in society (e.g., the
National Organization of Women, the Ku Klux
Commonly Used Scales
Klan, labor unions, physicians, etc.). This type
of measure has been used by political scientists "'tLikert Scale. You have probably used Likert
in the National Election Study since 1964 to scales; they are widely used and very common in
130
FIGURE 5.4
PART ONE / FOUNDATIONS
"Feeling Thermometer"
Graphic Rating Scale
..
forth. Keep the number of choices to eight or
nine at most. More distinctions than that are
probably not meaningful, and people will be
come confused. The choices should be evenly
100 Very Warm
balanced (e.g., "strongly agree," "agree" with The Rosenberg !i
90 "strongly disagree," "disagree"). Aliinall,lam~
Researchers have debated about whether to 1. Almosta~
80 offer a neutral category (e.g., "don't know," "un 2. Often true 1
decided," "no opinion") in addition to the di 3. Sometimes til
70
rectional categories (e.g., "disagree," "agree"). 4. Seldom true j
60 A neutral category implies an odd number of 5. Never true
10
tudes among college roommates (see Box 5.5). ~
As part of a larger survey, they asked four ques Brand ~
0 Very Cold tions about group inequality. The answer to each " J
question was a seven-point Likert scale with
x
y
choices from Strongly Disagree to Strongly
Agree. They created the index by adding the an
swers for each student to create scores that WOrkGroup~
ranged from 4 to 28. Notice that they worded
My supervisor: i
survey research. They were developed in the question number four in a reverse direction from
1930s by Rensis Likert to provide an ordinal-level the other questions. The reason for switching di
measure of a person's attitude. Likert scales usu rections in this way is to avoid the problem ofthe
Lets members
ally ask people to indicate whether they agree or response set. The response set, also called response
disagree with a statement. Other modifications style and response bias, is the tendency of some Is friendly and
are possible; people might be asked whether they people to answer a large number of items in the Treats all unit
approve or disapprove, or whether they believe same way (usually agreeing) out of laziness or a
something is "almost always true." Box 5.4 pre psychological predisposition. For example, if
sents several examples of Likert scales. items are worded so that saying "strongly agree"
Likert scales need a minimum of two cate always indicates self-esteem, we would not know
gories, such as "agree" and "disagree." Using whether a person who always strongly agreed had ated with . J
only two choices creates a crude measure and high self-esteem or simply had a tendency to indexuses~
forces distinctions into only two categories. It is agree with questions. The person might be an reliability. lhe~'
usually better to use four to eight categories. A swering "strongly agree" out of habit or a ten measure several
researcher can combine or collapse categories af dency to agree. Researchers word statements in 100 .IIIlpI'O¥l5
ter the data are collected, but data collected with alternative directions, so that anyone who agrees 5C~Tes give ~ ~
crude categories cannot be made more precise all the time appears to answer inconsistently or to ot a person's ~
later.
You can increase the number of categories
at the end of a scale by adding "strongly agree,"
"somewhat agree," "very strongly agree," and so
have a contradictory opinion.
Researchers often combine many Likert
scaled attitude indicators into an index. The
scale and indexes have properties that are associ-
from
-5troo~'
-str~-
1O:S
500'S opinion ~
. ~
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 131
x
y
ated with improving reliability and validity. An Instead of scoring Likert items, as in the pre
index uses multiple indicators, which improves vious example, the scores -2, -1, + 1, +2 could
reliability. The use of multiple indicators that be used. This scoring has an advantage in that a
measure several aspects of a construct or opin zero implies neutrality or complete ambiguity,
ion improves content validity. Finally, the index whereas a high negative number means an atti
scores give a more precise quantitative measure tude that opposes the opinion represented by a
of a person's opinion. For example, each per high positive number.
son's opinion can be measured with a number The numbers assigned to the response cate
from 10 to 40, instead of in four categories: gories are arbitrary. Remember that the use of a
"strongly agree," "agree," "disagree," and zero does not give the scale or index a ratio level
"strongly disagree." of measurement. Likert scale measures are at the
132 PART ONE I FOUNDATIONS
ordinal level of measurement because responses cator measurement is possible. The scale has two
indicate a ranking only. Instead of 1 to 4 or -2 limitations: Different combinations of several
to +2, the numbers 100, 70, 50, and 5 would scale items can result in the same overall score or
have worked. Also, do not be fooled into think result, and the response set is a potential danger.
ing that the distances between the ordinal cate
gories are intervals just because numbers are Bogardus Social Distance Scale. The Bogardus
assigned. Although the number system has nice social distance scale measures the social distance
mathematical properties, the numbers are used separating ethnic or other groups from each
for convenience only. The fundamental mea other. It is used with one group to determine
surement is only ordinal. how much distance it feels toward a target or
The simplicity and ease of use of the Likert "out-group."
scale is its real strength. When several items are The scale has a simple logic. People respond
combined, more comprehensive multiple indi to a series of ordered statements; those that are
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 133
most threatening or most socially distant are at cially distant items will refuse the socially closer
one end, and those that might be least threaten items (see Box 5.6).
ing or socially intimate are at the other end. The Researchers use the scale in several ways. For
logic of the scale assumes that a person who re example, people are given a series of statements:
fuses contact or is uncomfortable with the so- People from Group X are entering your country,
In 1993, Kleg and Yamamoto (1 998) replicated the gories, but they were worded slightly differently (see
original 1925 study by Emory Bogardus that first below). Third, both studies had seven categories
used the social distance scale. The original study had (called anchor points) printed left to right at the top.
I
1 1 0 subjects from the Pacific Coast. Participants in In the Bogardus original it said: "According to my first
cluded 107 White Americans of non-Jewish Euro feeling reactions I would willingly admit members of
pean ancestry, 1 Jewish White, 1 Chinese, and 1 each race (as a class, and not the best I have known,
Japanese (about 70 percent were female). In their nor the worst members) to one or more of the clas
1993 replication, Kleg and Yamamoto selected 1 35 sifications under which I have placed a cross (x)." In
middle school teachers from an affluent school dis the 1993 replication it said: "Social distance means
trict in a Colorado metropolitan area. There were the degree that individuals desire to associate with
1 1 9 non-Jewish Whites, 7 Jewish Whites, 6 African others. This scale relates to a special form of social
Americans, 1 American Indian, 1 Asian, and 1 un distance known as person to group distance. You are
known (65 percent were female). There were three given a list of groups. Across from each group there
minor deviations from the 1925 study. First, the are boxes identified by the labels at the top. Place an
original Bogardus respondents were given a list of "x" in the boxes that indicate the degree of associa
39 groups. Those in the replication had a list of 36 tion you would desire to have with each group. Give
groups. The two lists shared 24 groups in common. your first reaction." The main finding was that al
Three target groups were renamed: Negroes in though the average social distance declined a great
1925 versus African Americans in 1993; Syrians ver deal over over 68 years, the ranking of the 25
sus Arabs; and German-Jews and Russian-Jews vs. groups changed very little (see below).
Jews. Second, both studies contained seven cate-
Instructions
social distance ~
dent or a depera
A researchc
see how distane
Results versus another.
ethnic groups. iI
tor-patient disIl
associates (200t
English 1.27 1 1.17 2 ported differem
Scottish 1.69 2 1.22 6 with different
Irish 1.93 3 1.14 1 would be willin
French 2.04 4 1.20 4 with arthritis, UI
Dutch 2.12 5 1.25 9 tion. Fewer than
Swedish 2.44 6 1.21 5 being a friend tel
tion. The social
Danish 2.48 7 1.23 7
way to determis
Norwegian 2.67 8 1.25 8
toward a social !
German 2.89 9 1.27 10
tations. First, an
Spanish 3.28 10 1.29 11 egories to Ci!- speci
Italian 3.98 11 1.19 3 Second, it ~not.
Hindu 4.35 12 1.95 23 how a responde!
Polish 4.57 13 1.30 12 groups unless th
Russian 4.57 14 1.33 13 lar social distaoo
Native American 4.65 15 1.44 16 same time. Of Cl
Jewish 4.83* 16 1.42 15 pletes the scale a
Greek 4.89 17 1.38 14 havior in specific
Arab 5.00* 18 2.21 24
Mexican 5.02 19 1.56 18 f Semantic DiffO!
provides an indi
Black American 5.10* 20 1.55 17
feels about a COl
Chinese 5.28 21 1.68 20
The technique III
Japanese 5.30 22 1.62 19 ward something I
Korean 5.55 23 1.72 21 cause people COIII
Turk 5.80 24 1.77 22 adjectives in spoI
Grand Mean 3.82 1.43 cause most adieci
light/dark, ha;d 51
"Slight change in name of group. po site adjectives
scale. The Semai
connotations as5lII
evaluated and pm
are in your town, work at your place of employ feel uncomfortable with the relationship. People The Semantic
ment, live in your neighborhood, become your may be asked to respond to all statements, or many purposes. I
personal friends, and marry your brother or sis they may keep reading statements until they are how consumers •
ter. People are asked whether they feel comfort not comfortable with a relationship. There is no advisers use it it
able "ith the statement or if the contact is set number of statements required; the number about a candidate
acceptable. It is also possible to ask whether they usually ranges from five to nine. The measure of
CHAPTER 5 / QUALITATIVE AND QUANTITATIVE MEASUREMENT 135
social distance can be used as either an indepen to determine how a client perceives himself or
dent or a dependent variable. herself (see Box 5.7).
A researcher can use the Bogardus scale to To use the Semantic Differential, a re
see how distant people feel from one out-group searcher presents subjects with a list of paired
versus another. In addition to studying racial opposite adjectives with a continuum of 7 to 11
ethnic groups, it has been used to examine doc points between them. The subjects mark the
tor-patient distance. For example, Gordon and spot on the continuum between the adjectives
associates (2004) found that college students re that expresses their feelings. The adjectives can
ported different social distance toward people be very diverse and should be well mixed (e.g.,
with different disabilities. Over 95 percent positive items should not be located mostly on
would be willing to be a friend with someone either the right or the left side). Studies of a wide
with arthritis, cancer, diabetes, or a heart condi variety of adjectives in English found that they
tion. Fewer than 70 percent would ever consider fall into three major classes of meaning: evalua
being a friend to someone with mental retarda tion (good-bad), potency (strong-weak), and ac
tion. The social distance scale is a convenient tivity (active-passive). Of the three classes of
way to determine how close a respondent feels meaning, evaluation is usually the most signifi
toward a social group. It has two potentiallimi cant. The analysis of results is difficult, and a re
tations. First, a researcher needs to tailor the cat searcher needs to use statistical procedures to
egories to a specific out-group and social setting. analyze a subject's feelings toward the concept.
Second, it is not easy for a researcher to compare Results from a Semantic Differential tell a
how a respondent feels toward several different researcher how one person perceives different
groups unless the respondent completes a simi concepts or how different people view the same
lar social distance scale for all out-groups at the concept. For example, political analysts might
same time. Of course, how a respondent com discover that young voters perceive their candi
pletes the scale and the respondent's actual be date as traditional, weak, and slow, and as
havior in specific social situations may differ. halfway between good and bad. Elderly voters
perceive the candidate as leaning toward strong,
f Semantic Differential. Semantic Differential fast, and good, and as halfway between tradi
provides an indirect measure of how a person tional and modern.
feels about a concept, object, or other person.
The technique measures subjective feelings to f'<;uttman Scaling. Guttman scaling, or cumu
ward something by using adjectives. This is be lative scaling, differs from the previous scales or
cause people communicate evaluations through indexes in that researchers use it to evaluate data
adjectives in spoken and written language. Be after they are collected. This means that re
cause most adjectives have polar opposites (e.g., searchers must design a study with the Guttman
light/dark, hard/soft, slow/fast), it uses polar op scaling technique in mind.
posite adjectives to create a rating measure or Guttman scaling begins with measuring a
scale. The Semantic Differential captures the set of indicators or items. These can be ques
connotations associated with whatever is being tionnaire items, votes, or observed characteris
evaluated and provides an indirect measure of it. tics. Guttman scaling measures many different
The Semantic Differential has been used for phenomena (e.g., patterns of crime or drug use,
many purposes. In marketing research, it tells characteristics of societies or organizations, vot
how consumers feel about a product; political ing or political participation, psychological dis
advisers use it to discover what voters think orders). The indicators are usually measured in a
about a candidate or issue; and therapists use it simple yes/no or present/absent fashion. From 3
to 20 indicators can be used. The researcher se
136 PART ONE I FOUNDATIONS
. -
As part of her undergraduate thesis, Daina Hawkes restaurant, clothing store, or grocery store;
studied attitudes toward women with tattoos using boyfriend or not; average grades or failing grades.
the semantic differential (Hawkes, Senn, and Thorn, They used a semantic differential with 22 adjective
2004). The researchers had 268 students at a pairs. They also had participants complete two
medium-sized Canadian university complete a se scales: Feminist and Women's Movement scale and
mantic differential form in response to several sce Neosexism scale. The semantic differential terms
narios about a 22-year-old woman college student were selected to indicate three factors: evaluative,
with a tattoo. They had five scenarios in which they activity, and potency (strong/weak). Based on sta
varied the size of the tattoo (small versus large) and tistical analysis three adjectives were dropped. The
whether or not it was visible, and one with no details 19 items used are listed below. Among other find
about the tattoo. The authors also varied features of ings, the authors found that there were more nega
the senario: weight problem or not; part-time job at tive feelings toward a woman with a visible tattoo.
Good Bad'
Beautiful Ugly
Clean Dirty
Kind Cruel'
Rich Poor'
Honest Dishonest'
Pleasant Unpleasant'
Successful Unsuccessful
Reputable Disreputable
Safe Dangerous
Gentle Violent'
Feminine Masculine
Weak Powerful'
Passive Active'
Cautious Rash'
Soft Hard
Weak Strong
Mild Intense
Delicate Rugged'
lects items on the belief that there is a logical re Once a set of items is measured, the re
lationship among them. He or she then places searcher considers all possible combinations of
the results into a Guttman scale and determines responses for the items. For example, three items
whether the items form a pattern that corre are measured: whether a child knows her age,
sponds to the relationship. (See Box 5.8 for an her telephone number, and three local elected
example of a study using Guttman scaling.) political officials. The little girl may know her
Crozat (1998) examined public responses to various strations), but not all who accepted modest forms
forms of political protest. He looked at survey data accepted the more intense forms. In addition to
on the public's acceptance of forms of protest in showing the usefulness of the Guttman scale, Crozat
Great Britain, Germany, Italy, Netherlands, and the also found that people in different nations saw
United States in 1974 and 1 990. He found that the protest similarily and the degree of Guttman scala
pattern of the public's acceptance formed a Guttman bility increased over time. Thus, the pattern of ac
scale. Those who accepted more intense forms of ceptance of protest activities was Guttman "scalable"
protest (e.g., strikes and sit-ins) almost always ac in both time periods, but it more closely followed the
cepted more modest forms (e.g., petitions or demon- Guttman pattern in 1990 than 1974.
FORM OF PROTEST
Guttman Patterns
N N N N N
Y N N N N
Y Y N N N
Y Y Y N N
Y Y Y Y I\J
y y y y y
Other Patterns (examples only)
N y N Y N
Y N Y Y N
Y N Y N N
N Y Y N N
Y N N Y Y
138 PART ONE / FOUNDATIONS
age but no other answer, or all three, or only her tern. Alternative statistics to measure scalability As you are poroI
age and telephone number. In fact, for three have also been suggested. search involves doiI
items there are eight possible combinations of a study. Serious III
answers or patterns of responses, from not one phase can do i
knowing any through knowing all three. There is sults, even if the G
CONCLUSION
a mathematical way to compute the number of project were coodg
combinations (e.g., 2 3 ) , but you can write down In this chapter, you learned about the principles
all the combinations of yes or no for three ques and processes of measurement in quantitative
tions and see the eight possibilities. and qualitative research. All researchers concep Key Terms
The logical relationship among items in tualize-or refine and clarify their ideas into
Guttman scaling is hierarchical. Most people or conceptual definitions. All researchers opera Bogardus Social Dill
cases have or agree to lower-order items. The tionalize-or develop a set of techniques or conceptual cIe6wjrW
smaller number of cases that have the higher-or processes that will link their conceptual defini conceptual h~
der items also have the lower-order ones, but not tions to empirical reality. Qualitative and quan conceptualintie
vice versa. In other words, the higher-order titative researchers differ in how they approach concurrent ~ .
items build on the lower ones. The lower-order these processes, however. The quantitative re content validity <
items are necessary for the appearance of the searcher takes a more deductive path, whereas continuous ,ai' " ~
higher-order items. the qualitative researcher takes a more inductive criterion v~
An application of Guttman scaling, known path. The goal remains the same: to establish un discrete variables I
as scalogram analysis, lets a researcher test ambiguous links between a reseacher's abstract empirical h~
whether a hierarchical relationship exists among ideas and empirical data. exhaustive ~
the items. For example, it is easier for a child to You also learned about the principles of re external v~ ~
know her age than her telephone number, and to liability and validity. Reliability refers to the de face validitv
know her telephone number than the names of pendability or consistency of a measure; validity Guttman~
political leaders. The items are called scalable, or refers to its truthfulness, or how well a construct index
capable of forming a Guttman scale, if a hierar and data for it fit together. Quantitative and internal~
chical pattern exists. qualitative styles of research significantly diverge interval-level
The patterns of responses can be divided in how they understand these principles. None IevelsOfJDe2!il1Rl1lllllll
into two groups: scaled and errors (or nonscal theless, both quantitative and qualitative re Likert scale
able). The scaled patterns for the child's knowl searchers try to measure in a consistent way, and
edge example would be as follows: not knowing both seek a tight fit between the abstract ideas
any item, knowing only age, knowing only age they use to understand social world and what
plus phone number, knowing all three. Other occurs in the actual, empirical social world. In
combinations of answers (e.g., knowing the po addition, you saw how quantitative researchers
liticalleaders but not her age) are possible but apply the principles of measurement when they
are nonscalable. If a hierarchical relationship ex create indexes and scales, and you read about
ists among the items, then most answers fit into some major scales they use.
the scalable patterns. Beyond the core ideas of reliability and va
The strength or degree to which items can lidity, good measurement requires that you cre
be scaled is measured with statistics that mea ate clear definitions for concepts, use multiple
sure whether the responses can be reproduced indicators, and, as appropriate, weigh and stan
based on a hierarchical pattern. Most range from dardize the data. These principles hold across all
zero to 100 percent. A score of zero indicates a fields of study (e.g., family, criminology, in
random pattern, or no hierarchical pattern. A equality, race relations, etc.) and across the
score of 100 percent indicates that all responses many research techniques (e.g., experiments,
to the answer fit the hierarchical or scaled pat surveys, etc.).
CHAPTER 5 ! QUALITATIVE AND QUANTITATIVE MEASUREMENT 139
Ia:tiaa ...
Introduction
ca~~....
Pk
~
p'
.
.-1I:DI1iiIIII.
!'
Win" SIIIIIIf. .L4
Nonprobability Sampling
Haphazard, Accidental, or Convenience Sampling
Quota Sampling
Purposive or Judgmental Sampling
Snowball Sampling
Deviant Case Sampling
Sequential Sampling
Probability Sampling
Populations, Elements, and Sampling Frames
Why Random?
Types of Probability Samples
Hidden Populations
How Large Should a Sample Be?
Drawing Inferences
Conclusion
140
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 141
---------~
TAB L E 6.1 Types of Nonprobability
Samples
..
a researcher may draw a sample of 2,000; the
data from those 2,000 are equal for most pur
poses to the data from all 20 million. The sec
Haphazard Get any cases in any manner
ond purpose of probability sampling is that is convenient.
accuracy. The results of a well-designed, care
fully executed probability sample will produce Quota Get a preset number of cases in
each of several predetermined
results that are equally if not more accurate than
categories that will reflect the
trying to reach every single person in the whole
diversity of the population,
population. A census is usually an attempt to
using haphazard methods.
count everyone. In 2000, the U.S. Census Bu
reau tried to count everyone in the nation, but it Purposive Get all possible cases that fit
would have been more accurate if it used very particular criteria, using various
specialized statistical sampling. methods.
Qualitative researchers focus less on a sam Snowball Get cases using referrals from
ple's representativeness or on detailed techniques one or a few cases, and then
for drawing a probability sample. Instead, they referrals from those cases, and
focus on how the sample or small collection of so forth.
cases, units, or activities illuminates key features Deviant Case Get cases that substantially
of sociallife.The purpose ofsampling is to collect differ from the dominant
cases, events, or actions that clarify and deepen pattern (a special type of
understanding. Qualitative researchers' concern purposive sample).
is to find cases that will enhance what the re Sequential Get cases until there is no
searchers learn about the processes of social life additional information or new
in a specific context. For this reason, qualitative characteristics (often used with
researchers tend to collect a second type of sam other sampling methods).
pling: nonprobability sampling.
142 PART ONE / FOUNDATIONS
shows a variety of nonprobability sampling in each category. Thus, the number of people in
techniques. various categories of the sample is fixed. For ex FIGURE 6.1
ample, a researcher decides to select 5 males and
5 females under age 30, 10 males and 10 females
Haphazard, Accidental, or
aged 30 to 60, and 5 males and 5 females over age
Convenience Sampling
60 for a 40-person sample. It is difficult to repre
Haphazard sampling can produce ineffective, sent all population characteristics accurately (see
highly unrepresentative samples and is not rec Figure 6.1).
ommended. When a researcher haphazardly se Quota sampling is an improvement because
lects cases that are convenient, he or she can the researcher can ensure that some differences
easily get a sample that seriously misrepresents are in the sample. In haphazard sampling, all
the population. Such samples are cheap and those interviewed might be of the same age, sex,
quick; however, the systematic errors that easily or race. But once the quota sampler fixesthe cat
occur make them worse than no sample at all. egories and number of cases in each category, he
The person-on-the-street interview conducted or she uses haphazard sampling. For example, Of 32 a
by television programs is an example of a hap the researcher interviews the first five males un
hazard sample. Television interviewers go out on der age 30 he or she encounters, even if all five
the street with camera and microphone to talk to just walked out of the campaign headquarters of
a few people who are convenient to interview. a political candidate. Not only is misrepresenta
The people walking past a television studio in tion possible because haphazard sampling is
the middle of the day do not represent everyone used within the categories, but nothing prevents
(e.g., homemakers, people in rural areas, etc.). the researcher from selecting people who "act
Likewise, television interviewers often select friendly" or who want to be interviewed.
people who look "normal" to them and avoid A case from the history of sampling illus
people who are unattractive, poor, very old, or trates the limitations of quota sampling. George
inarticulate. Gallup's American Institute of Public Opinion,
Another example of a haphazard sample is using quota sampling, successfully predicted the
that of a newspaper that asks readers to clip a outcomes of the 1936, 1940, and 1944 U.S. pres
questionnaire from the paper and mail it in. Not idential elections. But in 1948, Gallup predicted Purposive 5aJlII
everyone reads the newspaper, has an interest in the wrong candidate. The incorrect prediction situations. First, a
the topic, or will take the time to cut out the had several causes (e.g., many voters were unde unique cases that i1III
questionnaire and mail it. Some people will, and cided, interviewing stopped early), but a major example, a researda
the number who do so may seem large (e.g., reason was that the quota categories did not ac sisto study ma~
5,000), but the sample cannot be used to gener curately represent all geographical areas and all or she selects a speI
alize accurately to the population. Such haphaz people who actually cast a vote. zine to study becais
ard samples may have entertainment value, but Second, a Tesl
they can give a distorted view and seriously mis sampling to select
Purposive or Judgmental Sampling
represent the population. reach, specialized pj
Purposive samplingis used in situations in which lations later in t:h.iIl
an expert uses judgment in selecting cases with a researcher wants til
Quota Sampling
specific purpose in mind. It is inappropriate if it possible to list all I
Quotasampling is an improvement over haphaz is used to pick the "average housewife" or the domly from the Ii
ard sampling. In quota sampling, a researcher "typical school." With purposive sampling, the subjective informa
first identifies relevant categories of people (e.g., researcher never knows whether the cases se prostitutes solicit.
male and female; or under age 30, ages 30 to 60, lected represent the population. It is often used prostitutes associaa
over age 60, etc.), then decides how many to get in exploratory research or in field research. lice who work on ..
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 143
--------------------.
FIGURE 6.1 Quota Sampling
o 0 0
•
• ~9o • w°9°~ ol t
iiti~ ~i' il
• • • • Q•
I i 01
Of 32 adults and children in the street scene, select 10 for the sample:
•0
g II• 0
••
V
4 Adult Males
~
0
Q
"
4 Adult Females
Purposive sampling is appropriate in three etc.) to identify a "sample" of prostitutes for in
situations. First, a researcher uses it to select clusion in the research project. The researcher
unique cases that are especially informative. For uses many different methods to identify the
example, a researcher wants to use content analy cases, because his or her goal is to locate as many
sis to study magazines to find cultural themes. He cases as possible.
or she selects a specific popular women's maga Another situation for purposive sampling
zine to study because it is trend setting. occurs when a researcher wants to identify par
Second, a researcher may use purposive ticular types of cases for in-depth investigation.
sampling to select members of a difficult-to The purpose is less to generalize to a larger pop
reach, specializedpopulation (see Hidden Popu ulation than it is to gain a deeper understanding
lations later in this chapter). For example, the of types. For example, Gamson (1992) used pur
researcher wants to study prostitutes. It is im posive sampling in a focus group study of what
possible to list all prostitutes and sample ran working-class people think about politics.
domly from the list. Instead, he or she uses (Chapter 11 discusses focus groups.) Gamson
subjective information (e.g., locations where wanted a total of 188 working-class people to
prostitutes solicit, social groups with whom participate in one of37 focus groups. He sought
prostitutes associate, etc.) and experts (e.g., po respondents who had not completed collegebut
lice who work on vice units, other prostitutes, who were diverse in terms of age, ethnicity, reli
144 PART ONE / FOUNDATIONS
gion, interest in politics, and type of occupation. case, and the lines represent friendship or other poverty rate and
He recruited people from 35 neighborhoods in linkages (see Figure 6.2). percent I s.eaioIl
the Boston area by going to festivals, picnics, Researchers also use snowball sampling in nonpublic SCITi.:r
fairs, and flea markets and posting notices on combination with purposive sampling as in the books, the 1nti:Dllll.
many public bulletin boards. In addition to ex case of Kissane (2003) in a descriptive field re ingdown~~
plaining the study, he paid the respondents well search study oflow-income women in Philadel tified 50 DOD~iI
so as to attract people who would not tradition phia. The U.S. policy to provide aid and services observed dw 3.
ally participate in a study. to low-income people changed in 1996 to in
crease assistance (e.g., food pantries, domestic
violence shelters, drug rehabilitation services,
Snowball Sampling
clothing distribution centers) deliveredby non
Snowball sampling (also called network, chain re public as opposed to government/public agen
ferral, or reputationai sampling) is a method for cies. As frequently occurs, the policy change was
identifying and sampling (or selecting) the cases made without a study of its consequences in
in a network. It is based on an analogy to a snow advance. Noone knew whether the affected low
ball, which begins small but becomes larger as it income people would use the assistance pro
is rolled on wet snow and picks up additional vided by nonpublic agencies as much as that
snow. Snowball sampling is a multistage tech provided by public agencies. One year after the
nique. It begins with one or a few people or cases new policy, Kissane studied whether low-in
and spreads out on the basis oflinks to the initial come women were equally likely to use nonpub
cases. lie aid. She focused on the Kensington area of
One use ofsnowball sampling is to sample a Philadelphia. It had a high (over 30 percent)
network. Social researchers are often interested
in an interconnected network of people or orga
nizations. The network could be scientists
around the world investigating the same prob
lem, the elites of a medium-sized city, the mem
----------.,
poverty rate and was a predominately White (85 previous research suggested that a majority of
percent) section of the city. First, she identified dropouts come from families that have low in
nonpublic service providers by using telephone come, are single parent or unstable, have been
books, the Internet, referral literature, and walk geographically mobile, and are racial minorities.
ing down every street of the area until she iden The family environment is one in which parents
tified 50 nonpublic social service providers. She and/or siblings have low education or are them
observed that a previous study found low-in selves dropouts. In addition, dropouts are often
come women in the area distrusted outsiders engaged in illegal behavior and have a criminal
and intellectuals. Her snowball sample began record prior to dropping out. A researcher using
asking service providers for the names of a few deviant case sampling would seek majority
low-income women in the area. She then asked group dropouts who have no record of illegalac
those women to refer her to others in a similar tivities and who are from stable two-parent,
situation, and asked those respondents to refer upper-middle-income families who are geo
her to still others. She identified 20 low-income graphically stable and well educated.
women aged 21 to 50, most who had received
public assistance. She conducted in-depth,
Sequential Sampling
open-ended interviews about their awareness
and experience with nonpublic agencies. She Sequential sampling is similar to purposive sam
learned that the women were less likely to get pling with one difference. In purposive sam
nonpublic than public assistance. Compared to pling, the researcher tries to find as many
public agencies, the women were less aware of relevant cases as possible, until time, financial
nonpublic agencies. Nonpublic agencies created resources, or his or her energy is exhausted. The
more social stigma, generated greater adminis goal is to get every possible case. In sequential
trative hassles, were in worse locations, and in sampling, a researcher continues to gather cases
volved more scheduling difficulties because of until the amount of new information or diver
limited hours. sity of cases is filled. In economic terms, infor
mation is gathered until the marginal utility, or
incremental benefit for additional cases, levels
Deviant Case Sampling
off or drops significantly. It requires that a re
A researcher uses deviant case sampling (also searcher continuously evaluate all the collected
called extreme case sampling) when he or she cases. For example, a researcher locates and
seeks casesthat differ from the dominant pattern plans in-depth interviews with 60 widows over
or that differ from the predominant characteris 70 years old who have been living without a
tics of other cases.Similar to purposive sampling, spouse for 10 or more years. Depending on the
a researcher uses a variety of techniques to locate researcher's purposes, getting an additional 20
cases with specific characteristics. Deviant case widows whose life experiences, social back
sampling differs from purposive sampling in that grounds, and worldviews differ little from the
the goal is to locate a collection of unusual, dif first 60 may be unnecessary.
ferent, or peculiar cases that are not representa
tive of the whole. The deviant cases are selected
because they are unusual, and a researcher hopes
PROBABILITY SAMPLING
to learn more about the social life by considering
cases that fall outside the general pattern or in A specialized vocabulary or jargon has devel
cluding what is beyond the main flow of events. oped around terms used in probability sam
For example, a researcher is interested in pling. Before examining probability sampling, it
studying high school dropouts. Let us say that is important to review its language.
146 PART ONE / FOUNDATIONS
Populations, Elements, and commercials, etc.) and geographical and time many types of sampI
Sampling Frames boundaries. tories, tax records, •
A researcher begins with an idea ofthe pop so on. Listing the
A researcher draws a sample from a larger pool
ulation (e.g., all people in a city) but defines it sounds simple. It is4
of cases, or elements. A sampling element is the
more precisely. The term target population refers may be no good 1ist4
unit of analysis or case in a population. It can be to the specific pool of cases that he or she wants
a person, a group, an organization, a writt~n A good samplil
to study. The ratio ofthe size ofthe sample to the sampling. A mismi
document or symbolic message, or even a social
size of the target population is the sampling ratio. frame and the rona
action (e.g., an arrest, a divorce, or a kiss) that
For example, the population has 50,000 people, can be a major 50IIl
is being measured. The large pool is the and a researcher draws a sample of 150 from it. match between ~I
population, which has an important role in sam
Thus, the sampling ratio is 150/50,000 = 0.003, definitions of a ,dIiiI
pling. Sometimes, the term universe is used in or 0.3 percent. If the population is 500 and the ment, so a miSIWIII
terchangeably with population. To define the researcher samples 100, then the sampling ratio
population, a researcher specifies the unit being frame and the pops
is 100/500 = 0.20, or 20 percent. pling. Researchers b
sampled, the geographical location, and the tem
A population is an abstract concept. How For example, you .~
poral boundaries of populations. Consider the
can population be an abstract concept, when in a region of the C.
examples of populations in Box 6.1. All the
there are a given number of people at a certain get a list of everyors
examples include the elements to be sampled
time? Except for specific small populations, one some people do 00
(e.g., people, businesses, hospital admissions,
can never truly freeze a population to measure it. the lists of those ""~
For example, in a city at any given moment, regularly, quickly glI
some people are dying, some are boarding or income tax records.. 1
getting off airplanes, and some are in cars dri some people cheat I
ving across city boundaries. The researcher must no income and do 1
decide exactly who to count. Should he or she died or have not ~
count a city resident who happens to be on vaca others have entered,
1. All persons aged 1 6 or older living in Singapore
tion when the time is fixed? What about the time taxes were due
on December 2, 1999, who were not incarcer tourist staying at a hotel in the city when the ries, but they are 00
ated in prison, asylums, and similar institutions time is fixed? Should he or she count adults, chil are not listed in a tell
dren, people in jails, those in hospitals? A popu ple have unlisted nl
2. All business establishments employing more
lation, even the population of all people over the cently moved. With
than 1 00 persons in Ontario Province, Canada,
that operated in the month of July 2005
age of 18 in the city limits of Milwaukee, Wis of all students enroU
consin, at 12:01 A.M. on March 1,2006, is an ab frames are almost all
3. All admissions to public or private hospitals in
stract concept. It exists in the mind but is
the state of New Jersey between August 1,
impossible to pinpoint concretely.
1988, and July 31,1993
Because a population is an abstract concept,
4. All television commercials aired between 7:00 except for small specialized populations (e.g., all FIGURE 6.3 AI
A.M. and 11 :00 P.M. Eastern Standard Time on
the students in a classroom), a researcher needs
three major U.S. networks between November 1
to estimate the population. As an abstract con
and November 25, 2006
cept, the population needs an operational defin
5. All currently practicing physicians in Australia ition. This process is similar to developing
who received medical degrees between January operational definitions for constructs that are
1, 1960, and the present
measured.
6. All African American male heroin addicts in the A researcher operationalizes a population
Vancouver, British Columbia, or Seattle, Wash by developing a specific list that closely approxi
ington, metropolitan areas during 2003 mates all the elements in the population. This list
is a sampling frame. He or she can choose from
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 147
many types of sampling frames: telephone direc frame can include some of those outside the tar
tories, tax records, driver's license records, and get population (e.g., a telephone directory that
so on. Listing the elements in a population lists people who have moved away) or might
sounds simple. It is often difficult because there omit some of those inside it (e.g., those without
may be no good list of elements in a population. telephones).
A good sampling frame is crucial to good Any characteristic of a population (e.g., the
sampling. A mismatch between the sampling percentage of city residents who smoke ciga
frame and the conceptually defined population rettes, the average height of all women over the
can be a major source of error. Just as a mis age of 21, the percent of people who believe in
match between the theoretical and operational UFOs) is a population parameter. It is the true
definitions ofa variable creates invalid measure characteristic of the population. Parameters are
ment, so a mismatch between the sampling determined when all elements in a population
frame and the population causes invalid sam are measured. The parameter is never known
pling. Researchers try to minimize mismatches. with absolute accuracy for large populations
For example, you would like to sample all people (e.g., an entire nation), so researchers must esti
in a region of the United States, so you decide to mate it on the basis of samples. They use infor
get a list of everyone with a driver's license. But mation from the sample, called a statistic, to
some people do not have driver's licenses, and estimate population parameters (see Figure 6.3).
the lists of those with licenses, even if updated A famous case in the history of sampling il
regularly, quickly go out of date. Next, you try lustrates the limitations of the technique. The
income tax records. But not everyone pays taxes; Literary Digest, a major U.S. magazine, sent
some people cheat and do not pay, others have postcards to people before the 1920, 1924, 1928,
no income and do not have to file, some have and 1932 U.S. presidential elections. The maga
died or have not begun to pay taxes, and still zine took the names for the sample from auto
others have entered or left the area since the last mobile registrations and telephone directories
time taxes were due. You try telephone directo the sampling frame. People returned the post
ries, but they are not much better; some people cards indicating whom they would vote for. The
are not listed in a telephone directory, some peo magazine correctly predicted all four election
ple have unlisted numbers, and others have re outcomes. The magazine's success with predic
cently moved. With a few exceptions (e.g., a list tions was well known, and in 1936, it increased
of all students enrolled at a university), sampling the sample to 10 million. The magazine pre
frames are almost alwaysinaccurate. A sampling dicted a huge victory for Alf Landon over
----------------------9..
FIG U R E 6.3 A Model of the Logic of Sampling
What You Population
Would Like to
Talk About
Sampling Process
148 PART ONE / FOUNDATIONS
Franklin D. Roosevelt. But the Literary Digest sample. For example, if conducting a telephone research methods bQ
was wrong; Franklin D. Roosevelt won by a survey, the researcher needs to try to reach the era ted by a pure ~
landslide. specific sampled person, by calling back four or number has an equ.JIj
The prediction was wrong for several rea five times, to get an accurate random sample. any position. Comps
sons, but the most important were mistakes in Random samples are most likely to yield a duce lists of randorm
sampling. Although the magazine sampled a sample that truly represents the population. In You may ask, OIl
large number of people, its sampling frame did addition, random sampling lets a researcher sta the sampling frame"
not accurately represent the target population tistically calculate the relationship between the sampling frame or.
(i.e., all voters). It excluded people without tele sample and the population-that is, the size of common answer is d
phones or automobiles, a sizable percentage of the sampling error. A nonstatistical definition of stricted random sa.
the population in 1936, during the worst of the the sampling error is the deviation between sam with replacemen~
Great Depression of the 1930s. The frame ex ple results and a population parameter due to after sampling it
simple random s.mJI
so,
cluded as much as 65 percent of the population random processes.
and a segment of the voting population (lower Random sampling is based on a great deal of the researcher ignoRIl
income) that tended to favor Roosevelt. The sophisticated mathematics. This chapter focuses into the sample.
magazine had been accurate in earlier elections on the fundamentals of how sampling works, the The logic of ~
because people with higher and lower incomes difference between good and bad samples, how illustrated with an ~
did not differ in how they voted. Also, during to draw a sample, and basic principles of sam pling marbles from a
earlier elections, before the Depression, more pling in social research. This does not mean that 5,000 marbles, so.lIlill1
lower-income people could afford to have tele random sampling is unimportant. It is essential 5,000 marbles are rot
phones and automobiles. to first master the fundamentals. If you plan to meter I want to ~
You can learn two important lessons from pursue a career using quantitative research, you marbles in it. I ~
the Literary Digest mistake. First, the sampling should get more statistical background than close my eyes, shab
frame is cruciaL Second, the size of a sample is space permits here. and repeat the pr~
less important than whether or not it accurately a random sample ofl
represents the population. A representative sam ber of red marbles i:oI
Types of Probability Samples
ple of2,500 can give more accurate predications percentage of red ~
about the U.S. population than a nonrepresenta Simple Random. The simple random sample is population. This is ~
tive sample of 1 million or 10 million. both the easiest random sample to understand 5,000 marbles. ~f:, ill
and the one on which other types are modeled. red marbles.
In simple random sampling, a researcher devel Does this me:.m I
Why Random?
ops an accurate sampling frame, selects elements meter is 48 percent nl
The area of applied mathematics called proba from the sampling frame according to a mathe cause of random ~
bility theory relies on random processes. The matically random procedure, then locates the might be off. I can ~
word random has a special meaning in mathe exact element that was selected for inclusion in the 100 marbles ~
matics. It refers to a process that generates a the sample. bles, and drawing .a ~
mathematically random result; that is, the selec After numbering all elements in a sampling 100 marbles. On the.l.
tion process operates in a truly random method frame, a researcher uses a list of random num 49 white marbles ~
(i.e., no pattern), and a researcher can calculate bers to decide which elements to select. He or problem. \\b:ich is
the probability of outcomes. In a true random she needs as many random numbers as there are random sampling
process, each element has an equal probability elements to be sampled; for example, for a sam from the same
of being selected. ple of 100, 100 random numbers are needed. sultsr I repeat the
Probability samples that rely on random The researcher can get random numbers from a I have drawn 130
processes require more work than nonrandom random-number table, a table of numbers chosen bles each (see Box
ones. A researcher must identify specific sam in a mathematically random way. Random might empty the w
pling elements (e.g., person) to include in the number tables are available in most statistics and Kant to see what is
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 149
research methods books. The numbers are gen 130 different samples reveal a clear pattern. The
erated by a pure random process so that any most common mix of red and white marbles is
number has an equal probability of appearing in 50/50. Samples that are close to that split are
any position. Computer programs can also pro more frequent than those with more uneven
duce lists of random numbers. splits. The population parameter appears to be
You may ask, Once I select an element from 50 percent white and 50 percent red marbles.
the sampling frame, do I then return it to the Mathematical proofs and empirical tests
sampling frame or do I keep it separate? The demonstrate that the pattern found in Box 6.2
common answer is that it is not returned. Unre always appears. The set of many random sam
stricted random sampling is random sampling ples is my samplingdistribution. It is a distribu
with replacement-that is, replacing an element tion of different samples that shows the
after sampling it so it can be selected again. In frequency of different sample outcomes from
simple random sampling without replacement, many separate random samples. The pattern will
the researcher ignores elements already selected appear if the sample size is 1,000 instead of 100;
into the sample. if there are 10 colors of marbles instead of 2; if
The logic of simple random sampling can be the population has 100 marbles or 10 million
illustrated with an elementary example-sam marbles instead of 5,000; and if the population is
pling marbles from a jar. I have a large jar full of people, automobiles, or colleges instead of mar
5,000 marbles, some red and some white. The bles. In fact, the pattern will become clearer as
5,000 marbles are my population, and the para more and more independent random samples
meter I want to estimate is the percentage of red are drawn from the population.
marbles in it. I randomly select 100 marbles (I The pattern in the sampling distribution
close my eyes, shake the jar, pick one marble, suggests that over many separate samples, the
and repeat the procedure 99 times). I now have true population parameter (i.e., the 50/50 split in
a random sample of marbles. I count the num the preceding example) is more common than
ber of red marbles in my sample to estimate the any other result. Some samples deviate from the
percentage of red versus white marbles in the population parameter, but they are less com
population. This is a lot easier than counting all mon. When many different random samples are
5,000 marbles. My sample has 52 white and 48 plotted as in the graph in Box 6.2, then the sam
red marbles. pling distribution looks like a normal or bell
Does this mean that the population para shaped curve. Such a curve is theoretically
meter is 48 percent red marbles? Maybe not. Be important and is used throughout statistics.
cause of random chance, my specific sample The central limit theorem from mathematics
might be off. I can check my results by dumping tells us that as the number of different random
the 100 marbles back in the jar, mixing the mar samples in a sampling distribution increases to
bles, and drawing a second random sample of ward infinity, the pattern of samples and the
100 marbles. On the second try, my sample has population parameter become more predictable.
49 white marbles and 51 red ones. Now I have a With a huge number of random samples, the
problem. Which is correct? How good is this sampling distribution forms a normal curve, and
random sampling business if different samples the midpoint of the curve approaches the popu
from the same population can yield different re lation parameter as the number of samples
sults? I repeat the procedure over and over until increases.
I have drawn 130 different samples of 100 mar Perhaps you want only one sample because
bles each (see Box 6.2 for results). Most people you do not have the time or energy to draw
might empty the jar and count all 5,000, but I many different samples. You are not alone. A
want to see what is going on. The results of my researcher rarely draws many samples. He or she
1SO PART ONE / FOUNDATIONS
42 58 1
43 57 1
45 55 2
46 54 4
47 53 8
48 52 12 Number of red and white marbles that were
49 51 21 randomly drawn from a jar of 5,000 marbles
50 50 31 with 100 drawn each time, repeated 1 30
51 49 20 times for 1 30 independent random samples.
52 48 13
53 47 9
54 46 5
55 45 2
57 43 1
Total 130
Number of Samples
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
variation in it
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 predict specific
Number of RedMarbles in a Sample -ith a great ~
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 151
usually draws only one random sample, but the Systematic Sampling. Systematic sampling is
central limit theorem lets him or her generalize simple random sampling with a shortcut for
from one sample to the population. The theorem random selection. Again, the first step is to num
is about many samples, but lets the researcher ber each element in the sampling frame. Instead
calculate the probability of a particular sample of using a list of random numbers, a researcher
being off from the population parameter. calculates a sampling interval, and the interval
Random sampling does not guarantee that becomes his or her quasi-random selection
every random sample perfectly represents the method. The sampling interval (i.e., 1 in k,
population. Instead, it means that most random where k is some number) tells the researcher
samples will be close to the population most of how to select elements from a sampling frame
the time, and that one can calculate the proba by skipping elements in the frame before select
bility of a particular sample being inaccurate. A ing one for the sample.
researcher estimates the chance that a particular For instance, I want to sample 300 names
sample is off or unrepresentative (i.e., the size of from 900. After a random starting point, I select
the sampling error) by using information from every third name of the 900 to get a sample of300.
the sample to estimate the sampling distribu My sampling interval is 3. Sampling intervals are
tion. He or she combines this information with easy to compute. I need the sample size and the
knowledge of the central limit theorem to con population size (or sampling frame size as a best
struct confidence intervals. estimate). You can think of the sampling interval
The confidence interval is a relatively simple as the inverse of the sampling ratio. The sampling
but powerful idea. When television or newspa ratio for 300 names out of900 is 300/900 = .333 =
per polls are reported, you may hear about 33.3 percent. The sampling interval is 900/300 = 3.
something called the margin of error being plus In most cases, a simple random sample and
or minus 2 percentage points. This is a version of a systematic sample yield virtually equivalent re
confidence intervals. A confidence interval is a sults. One important situation in which system
range around a specific point used to estimate a atic sampling cannot be substituted for simple
population parameter. A range is used because random sampling occurs when the elements in a
the statistics of random processes do not let a re sample are organized in some kind of cycle or
searcher predict an exact point, but they let the pattern. For example, a researcher's sampling
researcher say with a high level of confidence frame is organized by married couples with the
(e.g., 95 percent) that the true population para male first and the female second (see Table 6.2).
meter lies within a certain range. Such a pattern gives the researcher an unrepre
The calculations for sampling errors or con sentative sample if systematic sampling is used.
fidence intervals are beyond the level of this dis His or her systematic sample can be nonrepre
cussion, but they are based on the idea of the sentative and include only wives because of how
sampling distribution that lets a researcher cal the cases are organized. When his or her sample
culate the sampling error and confidence inter frame is organized as couples, even-numbered
val. For example, I cannot say, "There are sampling intervals result in samples with all hus
precisely 2,500 red marbles in the jar based on a bands or all wives.
random sample." However, I can say, "I am 95 Table 6.3 illustrates simple random sam
percent certain that the population parameter pling and systematic sampling. Notice that dif
lies between 2,450 and 2,550." I can combine ferent names were drawn in each sample. For
characteristics of the sample (e.g., its size, the example, H. Adams appears in both samples, but
variation in it) with the central limit theorem to C. Droullard is only in the simple random sam
predict specific ranges around the parameter ple. This is because it is rare for any two random
with a great deal of confidence. samples to be identical.
152 PART ONE / FOUNDATIONS
---------.......,
TAB L E 6. 2 Problems with Systematic
Sampling of Cyclical Data
...
The sampling frame contains 20 males and
20 females (gender is in parenthesis after each
name). The simple random sample yielded 3
TABLE 6.3 C~
---------------------..
----------------------l~
TAB LE 6. 3 Continued
dam or systematic sampling. In stratified sam supplemental information about strata is not al
each stratum, rather than letting random In general, stratified sampling produces
processes control it. This guarantees representa samples that are more representative of the pop
tiveness or fixes the proportion of different ulation than simple random sampling if the stra
strata within a sample. Of course, the necessary tum information is accurate. A simple example
illustrates why this is so. Imagine a population later chapter) oversampled African Americans.
that is 51 percent female and 49 percent male; A random sample of the U.S. population yielded
the population parameter is a sex ratio of 51 to 191 Blacks. Davis and Smith conducted a sepa
49. With stratified sampling, a researcher draws rate sample of African Americans to increase the
random samples among females and among total number of Blacksto 544. The 191 Blackre
males so that the sample contains a 51 to 49 per spondents are about 13 percent of the random
cent sex ratio. If the researcher had used simple sample, roughly equal to the percentage of
random sampling, it would be possible for a ran Blacks in the U.S. population. The 544 Blacksare
dom sample to be off from the true sex ratio in 30 percent of the disproportionate sample. The
the population. Thus, he or she makes fewer er researcher who wants to use the entire sample
rors representing the population and has a must adjust it to reduce the number of sampled
smaller sampling error with stratified sampling. African Americans before generalizing to the SuIf~
Researchers use stratified sampling when a U.S. population. Disproportionate sampling
..,..~~
stratum of interest is a small percentage of a helps the researcher who wants to focus on is
population and random processes could miss sues most relevant to a subpopulation. In this ~~
the stratum by chance. For example, a researcher case, he or she can more accurately generalize to
draws a sample of 200 from 20,000 college stu African Americans using the 544 respondents
~ tftilllioii'"
dents. He or she gets information from the col than using a sample of only 191. The larger sam
lege registrar indicating that 2 percent of the ple is more likely to reflect the full diversity of ~
20,000 students, or 400, are divorced women the African American subpopulation. Clets
with children under the age of 5. This group is ~fII;l/I
important to include in the sample. There would Cluster Sampling. Cluster sampling addresses
be 4 such students (2 percent of200) in a repre two problems: Researchers lack a good sampling ~stI
sentative sample, but the researcher could miss frame for a dispersed population and the cost to
them by chance in one simple random sample. reach a sampled element is very high. For exam
With stratified sampling, he or she obtains a list ple, there is no single list of all automobile me Iow~-xl.
of the 400 such students from the registrar and chanics in North America. Even if a researcher ~T.iE!H '}L.
randomly selects 4 from it. This guarantees that got an accurate sampling frame, it would cost T~~..-
the sample represents the population with re too much to reach the sampled mechanics who
gard to the important strata (see Box 6.3). are geographically spread out. Instead of using a
In special situations, a researcher may want single sampling frame, researchers use a sam
the proportion of a stratum in a sample to differ pling design that involves multiple stages and
from its true proportion in the population. For clusters.
example, the population contains 0.5 percent
Aleuts, but the researcher wants to examine
Aleuts in particular. He or she oversamples so
that Aleuts make up 10 percent of the sample.
With this type of disproportionate stratified
A cluster is a unit that contains final sam
pling elements but can be treated temporarily as
a sampling element itself.A researcher first sam
ples clusters, each of which contains elements,
then draws a second sample from within the
.._.-r" ......
45t~
-..·.CSJ8Qlr
".
¥' i
, ..
sample, the researcher cannot generalize directly clusters selected in the first stage of sampling. In ia. . . or R:ldI. .1II
from the sample to the population without spe other words, the researcher randomly samples _-\R:kada
cial adjustments. clusters, then randomly samples elements from -dRaa-s."II&-.•
In some situations, a researcher wants the
proportion of a stratum or subgroup to differ
within the selected clusters. This has a big prac
tical advantage. He or she can create a good sam ,,_,5
1 . . . . . . . ' -. . . .
. -
from its true proportion in the population. For pling frame of clusters, even if it is impossible to . . .... ~d.-II:I;.l"
example, Davis and Smith (1992) reported that create one for sampling elements. Once the re . .ftc Mi'-
the 1987 General Social Survey (explained in a searcher gets a sample of clusters, creating a --'for'c i 5
.
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 155
Administrators 15 2.88 1 3 -2
Staff physicians 25 4.81 2 5 -3
Intern physicians 25 4.81 6 5 +1
Registered nurses 100 19.23 22 19 +3
Nurse assistants 100 19.23 21 19 +2
Medical technicians 75 14.42 9 14 +5
Orderlies 50 9.62 8 10 -2
Clerks 75 14.42 5 14 +1
Maintenance staff 30 5.77 3 6 -3
Cleaning staff 25 4.81 3 5 -2
- -- - -
Total 520 100.00 100 100
Note: Traditionally, N symbolizes the number in the population and n represents the number in the sample.
The simple random sample overrepresents nurses, nursing assistants, and medical technicians, but underrepresents
administrators, staff physicians, maintenance staff, and cleaning staff. The stratified sample gives an accurate representation
sampling frame for elements within each cluster individuals from Mapleville. First, he or she ran
becomes more manageable. A second advantage domly samples city blocks, then households
for geographically dispersed populations is that within blocks, then individuals within house
elements within each cluster are physically closer holds (see Box 6.4). Although there is no accurate
to one another. This may produce a savings in list of all residents of Mapleville, there is an accu
locating or reaching each element. rate list of blocks in the city. After selecting a ran
A researcher draws several samples in stages dom sample of blocks, the researcher counts all
in cluster sampling. In a three-stage sample, stage households on the selected blocks to create a
1 is random sampling of big clusters; stage 2 is sample frame for each block. He or she then uses
random sampling of small clusters within each the list of households to draw a random sample
selected big cluster; and the last stage is sampling at the stage of sampling households. Finally, the
of elements from within the sampled small clus researcher chooses a specific individual within
ters. For example, a researcher wants a sample of each sampled household.
156 PART ONE I FOUNDATIONS
Cluster sampIi
than simple r~
curate. Each stage ..
sampling errors. Th
Goal: Draw a random sample of 240 people in Mapleville. sample has more 51
Step 1: Mapleville has 55 districts. Randomly select 6 districts. stage random samP
1 2 3* 4 5 6 7 8 9 10 11 12 13 14 15* 16 17 18 1920 21 22 23 2425 26 A researcher 1
27* 282930 31* 323334353637383940* 4142434445464748 must decide the II
4950 51 525354* 55 number of element!
* = Randomly selected. ample, in a two-stall
Step 2: Divide the selected districts into blocks. Each district contains 20 blocks. Randomly select 4 blocks ple from Mapleei
from the district. randomly select 1]
Example of District 3 (selected in step 1): ments from each. a
and select 120 ekmi
1 234* 5 678910* 11 12 13* 14 15 16 17* 18 1920
The general aJ1S'lo-~
* = Randomly selected.
clusters is better. ~
Step 3: Divide blocks into households. Randomlyselect households. clusters (e.g., peopI
Example of Block 4 of District 3 (selected in step 2): tend to be similar •
Block 4 contains a mix of single-family homes, duplexes, and four-unit apartment buildings. It is bounded by the same block ~
Oak Street, RiverRoad,South Avenue, and GreenviewDrive. There are 45 householdson the block. Randomly on different blocks)
select 1 a households from the 45. many similar e1emll
would be less repR:lll
1 #1 Oak Street 16 31 *
17*
tion. For example.:
2 #3 Oak Street #154 River Road 32'
two blocks with rei
3* #5 Oak Street 18 # 1 56 River Road 33
4 19* #158 River Road 34 # 1 56 Greenview Drive
draw 120 people fnl
5 20* 35* representative than]
6 21 #13 South Avenue 36 city blocks and .1 inII
7 #7 Oak Street 22 37 When a researd
8 23 #11 South Avenue 38 ographical area andl
9* #150 River Road 24 #9 South Avenue 39 #158 Greenview Drive cluster sampling •
10* 25 #7 South Avenue 40 costs. As usual, tbeIll
11 26 # 5 South Avenue 41 racy and cost.
12 27 #3 South Avenue 42 For example. A
13 #152 River Road 28 #1 South Avenue 43 #160 Greenview Drive each plan to visit •
14 29* 44 sample of 1,500 ~
15 30 #152 Greenview Drive 45 ulation of all colle!!lli
Alan obtains an aal
* = Randomly selected.
students and uses si
Step 4: Select a respondent within each household. travels to 1,000 ditl
Summary of cluster sampling: one or two stu~
random sample of
1 person randomly selected per household
1 a households randomly selected per block
3,000 colleges, .
4 blocks randomly selected per district 500 students from ~
6 districts randomly selected in the city dom sample of ~
1 X 10 X 4 6 = 240 people in sample and selects 5 studell
erage $250 per loIj
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 157
Cluster sampling is usually less expensive $250,000, Ricardo's is $750, and Barbara's is
than simple random sampling, but it is less ac $75,000. Alan's sample is highly accurate, but
curate. Each stage in cluster sampling introduces Barbara's is only slightly less accurate for one
sampling errors. This means a multistage cluster third the cost. Ricardo's sample is the cheapest,
sample has more sampling errors than a one but it is not representative at all.
stage random sample.
A researcher who uses cluster sampling Probability Proportionate to Size (PPS). There
must decide the number of clusters and the are two methods of cluster sampling. The
number of elements within each cluster. For ex method just described is proportionate or un
ample, in a two-stage cluster sample of240 peo weighted cluster sampling. It is proportionate
ple from Mapleville, the researcher could because the size of each cluster (or number of
randomly select 120 clusters and select 2 ele elements at each stage) is the same. The more
ments from each, or randomly select 2 clusters common situation is for the cluster groups to
and select 120 elements in each. Which is best? be of different sizes. When this is the case, the
The general answer is that a design with more researcher must adjust the probability or sam
clusters isbetter. This is because elements within pling ratio at various stages in sampling (see
clusters (e.g., people living on the same block) Box6.5).
tend to be similar to each other (e.g., people on The foregoing cluster sampling example
the same block tend to be more alike than those with Alan, Barbara, and Ricardo illustrates the
on different blocks). If few clusters are chosen, problem with unweighted cluster sampling. Bar
many similar elements could be selected, which bara drew a simple random sample of 300 col
would be less representative of the total popula leges from a list of all 3,000 colleges, but she
tion. For example, the researcher could select made a mistake-unless every college has an
two blocks with relatively wealthy people and identical number of students. Her method gave
draw 120 people from each. This would be less each college an equal chance of being selected
representative than a sample with 120 different a 300/3,000 or 10 percent chance. But colleges
city blocks and 2 individuals chosen from each. have different numbers of students, so each stu
When a researcher samples from a large ge dent does not have an equal chance to end up in
ographical area and must travel to each element, her sample.
cluster sampling significantly reduces travel Barbara listed every college and sampled
costs. As usual, there is a tradeoff between accu from the list. A large university with 40,000 stu
racy and cost. dents and a small college with 400 students had
For example, Alan, Ricardo, and Barbara an equal chance of being selected. But if she
each plan to visit and personally interview a chose the large university, the chance of a given
sample of 1,500 students who represent the pop student at that college being selected was 5 in
ulation of all college students in North America. 40,000 (5/40,000 = 0.0125 percent), whereas a
Alan obtains an accurate sampling frame of all student at the small collegehad a 5 in 400 (5/400
students and uses simple random sampling. He = 1.25 percent) chance of being selected. The
travels to 1,000 different locations to interview small-college student was 100 times more likely
one or two students at each. Ricardo draws a to be in her sample. The total probability of be
random sample of three collegesfrom a list of all ing selected for a student from the large univer
3,000 colleges, then visits the three and selects sity was 0.125 percent (10 X 0.0125), while it
500 students from each. Barbara draws a ran was 12.5 percent (10 X 1.25) for the small
dom sample of 300 colleges. She visits the 300 college student. Barbara violated a principle of
and selects 5 students at each. If travel costs av random sampling-that each element has an
erage $250 per location, Alan's travel bill is equal chance to be selected into the sample.
158 PART ONE / FOUNDATIONS
Sampling has many terms for the different parts of county or SMSA. This gave them a sample of 84 Vaquera and Kao (2
samples or types of samples. A complex sample illus counties or SMSAs. tion among adolesa
trates how researchers use them. Look at the 1980 For the second stage, the researchers identified were either from the
sample for the best-known national U.S.survey in so city blocks, census tracts, or the rural equivalent in Their data were from
ciology, the General Social Survey. each county or SMSA. Each sampling element (e.g., adolescent health I
The population is defined as all resident adults city block) had a minimum of 50 housing units. In or through 12 in 80
(1 8 years or older) in the U.S. for the universe of all der to get an accurate count of the number of hous schools. There were
Americans. The target population consists of all Eng ing units for some counties, a researcher counted schools. After the s
lish-speaking adults who livein households, excluding addresses in the field. The researchers selected 6 or mately 200 studenl
those living in institutional settings such as college more blocks within each county or SMSA using PPS from within those sci
dormitories, nursing homes, or military quarters. The to yield 562 blocks. the school, and stud
researchers estimated that 97.3 percent of all resi In the third stage, the researchers used the the school. Because t
dent adults lived in households and that 97 percent household as a sampling element. They randomly se size, ranging from 11
of the household population spoke sufficient English lected households from the addresses in the block. thors adjusted usinf;
to be interviewed. After selecting an address, an interviewer contacted size (PPS). They fau
The researchers used a complex multistage prob the household and chose an eligible respondent from dents had a relatiom
ability sample that is both a cluster sample and a it. The interviewer looked at a selection table for pos site sex in the Pf"et
stratified sample. First, they created a national sible respondents and interviewed a type of respon Blacks were more iii
sampling frame of all U.S.counties, independent cities, dent (e.g., second oldest) based on the table. In ships (90 percent) (]
and Standard Metropolitan Statistical Areas (SMSAs), total, 1,934 people were contacted for interviews ics (70 percent). Th
a Census Bureau designation for larger cities and sur and 75.9 percent of interviews were completed. This mixed-race couples d
rounding areas. Each sampling element at this first gave a final sample size of 1,468. We can calculate affection, but the inti
level had about 4,000 households. They divided the sampling ratio by dividing 1,468 by the total to do so in public thl
these elements into strata. The strata were the four number of adults living in households, which was
major geographic regions as defined by the Census about 150 million, which is 0.01 percent. To check
Bureau, divided into metropolitan and nonmetropol the representativeness of their sample, the re
itan areas. They then sampled from each strata using searchers also compared characteristics of the sam Three kinds ot
probability proportionate to size (PPS) random selec ple to census results (see Davis and Smith, 1992: sampling frame iu
tion, based on the number of housing units in each 31-44). without telephooa
moved, and prop
Those without phQ
ucated, and transa
If Barbara uses probability proportionate to with 40,000 students will be 100 times more phone interview stI
size (PPS) and samples correctly, then each final likely to be selected than one with 400 students. general public ",itb
sampling element or student will have an equal (See Box 6.6 for another example.) cent in advanced •
probability of being selected. She does this by percentage of the PI
adjusting the chances ofselecting a college in the Random-Digit Dialing. Random-digit dialing creased, the perces
first stage of sampling. She must give large col (RDD) is a special sampling technique used in has also grown. Se"Il
leges with more students a greater chance of be research projects in which the general public is listed numbers: rc
ing selected and small colleges a smaller chance. interviewed by telephone. It differs from the tra lection agencies; tt.
She adjusts the probability of selecting a college ditional method of sampling for telephone in want privacy and ,
on the basis of the proportion of all students in terviews because a published telephone directory salespeople, and pII
the population who attend it. Thus, a college is not the sampling frame. eas, the percentag
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 159
uations each person may have a separate phone government service agencies identified as vic pleted an initial i
number or more than one phone number. This tims. Law-enforcement estimates depend on the ceived more than
means that after a working residential phone is specific level of enforcement efforts and are most times this yieldes
reached, a second stage of sampling is necessary, likely to identify a small percent of the most vis other times more:
within household sampling, to select the person ible and serious cases. Similar difficulties exist ferral coupons ~
to be interviewed. with nongovernment service agencies that pro young man heardi
Box 6.5 presents an example of how the vide aid to victims. Thus, during the first 10 parlor and called 1
many sampling terms and ideas can be used to months of 2004, Norwegian police detected 42 He (participant U
gether in a specific real-life situation. sex trafficking victims. This is subset of all possi user and in his ill
ble trafficking victims. For this population other drug users.l
Tyldum and Brunovskis suggested using a cap pants (parti~
Hidden Populations
In contrast to sampling the general population
ture-recapture method borrowed from biology.
In capture-recapture, a percentage of the same
about one rn:::J
refer anyone
or visible and accessible people, sampling hidden cases will reappear across multiple attempts to four new people. ~
populations (i.e., people who engage in con "capture" cases (with a release after past cap 148) referred ~
cealed activities) is a recurrent issue in the ture). This percentage recaptured allows re ferred four D~
studies of deviant or stigmatized behavior. It il searchers to estimate the size of the total person, and tha
lustrates the creative application of sampling population. A third population is that of mi referred four
principles, mixing qualitative and quantitative grants who have returned to their country of ori took place in .
styles of research and often using nonprobability gin. By surveying returnees and estimating the duced 249 users
techniques. Examples of hidden populations in proportion of them who are former trafficking mine between
clude illegal drug users, prostitutes, homosexu victims, researchers have another way to esti You are now
als, people with HIV/AIDS, homeless people, mate the size of the hidden population. types of probalul1l
and others. Draus and associates (2005) described their supplementarr
Tyldum and Brunovskis (2005) described sampling a hidden population in a field research PPS, within- .
ways to measure the hidden population of study of illicit drug users in four rural Ohio may be app'TOtlrial.
women and children victims ofsex trafficking in counties. They used respondent-driven sam how researchers
Norway. They suggested using multiple sam pling (RDS), which is a version ofsnowball sam probabilin" sanlJPli.
pling approaches and thinking of in terms of pling and appropriate when members of a as hidden 'P<Jt:-1iI
several overlapping populations in which vic hidden population are likely to maintain contact mining a sampje
tims are a subset. One population is all working with one another. This type of sampling begins
prostitutes. By telephoning all identifiable escort by identifying an eligible case or participant.
and massage services, then calculating response This person, called a "seed," is given referral
rates and the number of women per phone, the coupons to distribute among other eligible peo
authors estimated that 600 female prostitutes ple who engage in the same activity. For each
worked in the Oslo metro area in October 2003. successful referral, the "seed" receives some
Based on number of months most women work money. This process is repeated with several
in prostitution and their turnover rate each year, waves of new recuits until the a point of satura
they estimated that 1,100 different women work tion (see Sequential Sampling earlier in this
as prostitutes in Oslo in a year. Of these, about chapter). In the study by Draus and associates,
80 percent of them are of non-Norwegian ori each interviewed drug-using participant was
gin. Victims ofsex trafficking are a subset among paid $50 for an initial two-hour interview and
the roughly 800 non-Norwegians who work as $35 for an hour-long follow-up interview. The
prostitutes who are being exploited by others participants received three referral coupons at
and working involuntary. A second population the end of the initial interview and got $10 for
is the women law-enforcement officials or non each eligible participant they referred who com
CHAPTER 6 I QUALITATIVE AND QUANTITATIVE SAMPLING 161
Types of Probability
TABLE 6.4
times this yielded no new participants, but at Samples
other times more than the three people with re
ferral coupons were recruited. In one case, a
young man heard about the study at a local tatoo
parlor and called the study office in July 2003. Simple Random Create a sampling frame for all
cases, then select cases using
He (participant 157) had been a powder cocaine
a purely random process (e.g.,
user and in his interview said he knew many
random-number table or
other drug users. He referred two new partici
computer program).
pants (participants 161 and 146) who came in
about one month later. Participant 161 did not Stratified Create a sampling frame for
refer anyone new, but participant 146 referred each of several categories of
four new people, and two of the four (154 and cases, draw a random sample
148) referred still others. Participant 154 re from each category, then
combine the several samples.
ferred four new people and 146referred one new
person, and that one person, (participant 158) Systematic Create a sampling frame,
referred four others. This sampling process that calculate the sampling interval
took place in different geographic locations pro 1/k, choose a random starting
duced 249 users of cocaine or methanmpheta place, then take every l/k
mine between June 2002 and February 2004. case.
You are now familiar with several major Cluster Create a sampling frame for
types of probability samples (see Table 6.4) and larger cluster units, draw a
supplementary techniques used with them (e.g., random sample of the cluster
PPS, within-household, RDD, and RDS) that units, create a sampling frame
may be appropriate. In addition, you have seen for cases within each selected
how researchers combine nonprobability and cluster unit, then draw a
probability sampling for special situations, such random sample of cases, and
as hidden populations. Next, we turn to deter so forth.
mining a sample size for probability samples.
by the statistical method and because it gives equal, larger samples are needed if one wants
sample sizes close to those of the statistical high accuracy, if the population has a great deal FIGURE 6.4
method. Rules of thumb are not arbitrary but of variability or heterogeneity, or if one wants to
are based on past experience with samples that examine many variables in the data analysis si
have met the requirements of the statistical multaneously. Smaller samples are sufficient
method. when less accuracy is acceptable, when the pop
One principle of sample sizes is, the smaller ulation is homogeneous, or when only a few
the population, the bigger the sampling ratio has variables are examined at a time.
to be for an accurate sample (i.e., one with a high The analysis of data on subgroups also af
pro bability of yielding the same results as the en fects a researcher's decision about sample size. If
tire population). Larger populations permit the researcher wants to analyze subgroups in the
smaller sampling ratios for equally good sam population, he or she needs a larger sample. For
ples. This is because as the population size example, I want to analyze four variables for
grows, the returns in accuracy for sample size males between the ages of 30 and 40 years old. If
shrink. this sample is of the general public, then only a
For small populations (under 1,000), a re small proportion (e.g., 10 percent) of sample
searcher needs a large sampling ratio (about 30 cases will be males in that age group. A rule of
percent). For example, a sample size of about thumb is to have about 50 cases for each sub
300 is required for a high degree of accuracy. For group to be analyzed. Thus, if I want to analyze
moderately large populations (10,000), a smaller a group that is only 10 percent of the popula
sampling ratio (about 10 percent) is needed to tion, then I should have lOX 50 or 500 cases in
be equally accurate, or a sample size of around the sample to be sure I get enough for the sub
1,000. For large populations (over 150,000), group analysis.
smaller sampling ratios (1 percent) are possible,
and samples of about 1,500 can be very accurate.
Drawing Inferences
To sample from very large populations (over 10
million), one can achieve accuracy using tiny A researcher samples so he or she can draw in
sampling ratios (0.025 percent) or samples of ferences from the sample to the population. In
about 2,500. The size of the population ceases to fact, a subfield of statistical data analysis that
be relevant once the sampling ratio is very small, concerns drawing accurate inferences is called
and samples of about 2,500 are as accurate for inferential statistics. The researcher directly ob
........-....
populations of 200 million as for 10 million. serves variables using units in the sample. The
These are approximate sizes, and practicallimi sample stands for or represents the population.
tations (e.g., cost) also playa role in a re Researchers are not interested in samples in
searcher's decision. themselves; they want to infer to the population.
'p'
A related principle is that for small samples, Thus, a gap exists between what the researcher
small increases in sample size produce big gains concretely has (a sample) and what is of real in
in accuracy. Equal increases in sample size pro terest (a population) (see Figure 6.4).
duce more of an increase in accuracy for small In the last chapter, you saw how the logic of
than for large samples. measurement could be stated in terms of a gap
A researcher's decision about the best sam between abstract constructs and concrete indica
ple size depends on three things: (1) the degree tors. Measures of concrete, observable data are
of accuracy required, (2) the degree of variability approximations for abstract constructs. Re
or diversity in the population, and (3) the num searchers use the approximations to estimate
ber of different variables examined simultane what is of real interest (i.e., constructs and causal
ously in data analysis. Everything else being laws). Conceptualization and operationalization
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 163
--------------------..
FIG U R E 6. 4 Model of the Logic of Sampling and of Measurement
A Model of the Logic of Sampling
WhatYou Population
Would Like to
Talk About
Sampling Process
What You
Sample
Would Like to
Talk About
bridge the gap in measurement just as the use of erage annual family income. He or she has two
sampling frames, the sampling process, and in samples. Sample 1 gives a confidence interval of
ference bridge the gap in sampling. $30,000 to $36,000 around the estimated popu
Researchers put the logic of sampling and lation parameter of $33,000 for an 80 percent
the logic of measurement together by directly level of confidence. For a 95 percent level of con
observing measures of constructs and empirical fidence, the range is $23,000 to $43,000. A sam
relationships in samples (see Figure 6.4). They ple with a smaller sampling error (because it is
infer or generalize from what they can observe larger or is more homogeneous) might give a
empirically in samples to the abstract causal laws $30,000 to $36,000 range for a 95 percent confi
and constructs in the population. dencelevel. Key Terms
Validity and sampling error have similar
functions, as can be illustrated by the analogy central limit
between the logic of sampling and the logic of clusteT~
CONCLUSION
measurement-that is, between what is ob confidence ir dI:I....
served and what is discussed. In measurement, a In this chapter, you learned about sampling. deviant case
researcher wants valid indicators of constructs Sampling is widely used in social research. You haphazard sanlpliill
that is, concrete indicators that accurately repre learned about types of sampling that are not hidden pop1obt....
sent abstract constructs. In sampling, he or she based on random processes. Only some are ac inferential staI:i5b~
wants samples that have little sampling error ceptable, and their use depends on special cir nonrandom
concrete collections of cases that accurately rep cumstances. In general, probability sampling is parameter
resent unseen and abstract populations. A valid preferred by quantitative researchers because it population
measure deviates little from the construct it rep produces a sample that represents the popula
resents. A sample with little sampling error per tion and enables the researcher to use powerful
mits estimates that deviate little from population statistical techniques. In addition to simple ran
parameters. dom sampling, you learned about systematic,
Researchers try to reduce sampling errors. stratified, and cluster sampling. Although this
The calculation of the sampling error is not pre book does not cover the statistical theory used in
sented here, but it is based on two factors: the random sampling, from the discussion of sam
sample size and the amount of diversity in the pling error, the central limit theorem, and sam
sample. Everything else being equal, the larger ple size, it should be clear that random sampling
the sample size, the smaller the sampling error. produces more accurate and precise sampling.
Likewise, the greater the homogeneity (or the Before moving on to the next chapter, it
less the diversity) in a sample, the smaller its may be useful to restate a fundamental principle
sampling error. of social research: Do not compartmentalize the
Sampling error is also related to confidence steps of the research process; rather, learn to see
intervals. Iftwo samples are identical except that the interconnections between the steps. Re
one is larger, the one with more cases will have a search design, measurement, sampling, and spe
smaller sampling error and narrower confidence cific research techniques are interdependent.
intervals. Likewise, if two samples are identical Unfortunately, the constraints of presenting in
except that the cases in one are more similar to formation in a textbook necessitate presenting
each other, the one with greater homogeneity the parts separately, in sequence. In practice, re
will have a smaller sampling error and narrower searchers think about data collection when they
confidence intervals. A narrow confidence inter design research and develop measures for vari
val means more precise estimates of the popula ables. Likewise, sampling issues influence re
tion parameter for a given level of confidence. search design, measurement of variables, and
For example, a researcher wants to estimate av data collection strategies. As you will see in fu-
CHAPTER 6 / QUALITATIVE AND QUANTITATIVE SAMPLING 165
ture chapters, good social research depends on probability proportionate to size (PPS)
simultaneously controlling quality at several dif purposive sampling
ferent steps-research design, conceptualiza quota sampling
tion' measurement, sampling, and data random-digit dialing (RDD)
collection and handling. The researcher who random-number table
makes major errors at anyone stage may make random sample
an entire research project worthless. sample
sampling distribution
sampling element
Key Terms sampling error
sampling frame
central limit theorem sampling interval
cluster sampling sampling ratio
confidence intervals sequential sampling
deviant case sampling simple random sampling
haphazard sampling snowball sampling
hidden populations sociogram
inferential statistics statistic
nonrandom sample stratified sampling
parameter systematic sampling
population target population
Survey Research
Introduction
Research Questions Appropriate for a Survey
The Logic of Survey Research
What Is a Survey?
Steps in Conducting a Survey
Constructing the Questionnaire
Principles of Good Question Writing
Aiding Respondent Recall
Types of Questions and Response Categories
Open versus Closed Questions
Wording Issues
Questionnaire Design Issues
Types of Surveys: Advantages and Disadvantages
Mail and Self-Administered Questionnaires
Web Surveys Itesearda 0-.
Telephone Interviews tiara SanE,
Face-to-Face Interviews ~rc:st:adl
Interviewing ~1D
The Role of the Interviewer ~.aIal
opinions. dIIIrxII
Stages of an Interview
behroot.
Training Interviewers
~~
Interviewer Bias tioosmu.
Computer-Assisted Telephone Interviewing
The Ethical Survey
Conclusion
_. ..
lbn-~mulllll!lll:l'"
1D<psinns
166
CHAPTER 7 / SURVEY RESEARCH 167
he or she actually does. This is illustrated by Survey research is often called correlational.
Pager and Quillian (2005), who compared tele Survey researchers use questions as control vari FIGURE 7.1 s
phone survey responses from Milwaukee-area ables to approximate the rigorous test for causal 5
employers about their willingness to hire ex-of ity that experimenters achieve with their
fenders of different races with an "audit." In the physical control over temporal order and alter Step 1:
audit, a trained pair of young males with specific native explanations. • Develop hfI
characteristics applied for 350 job openings in • Decideoo~
December 2001. Employers agreed to hire 34 (mail. inten
Steps in Conducting a Survey
percent of White and 14 percent of Black appli • Write StnIIl!
cants. The applicants had identical job experi The survey researcher follows a deductive ap • Decide on ~
ence and credentials and no criminal records. proach. He or she begins with a theoretical or • Design Iaya
The same employers agreed to hire 17 percent of applied research problem and ends with empir
Whites and 5 percent of Blackswith identical job ical measurement and data analysis. Once a
experience and credentials but also with a crim researcher decides that the survey is an appro
Step 2:
inal record for illegal drug use. The employers priate method, basic steps in a research project
• Plan how..
were telephoned a few months later. Pager and can be divided into the substeps outlined in
• Pilot test SUI
Quillian found in the telephone survey far more Figure 7.1.
employers expressed a willingness to hire an In the first phase, the researcher develops an
ex-offender (67 percent) and there were no dif instrument-a survey questionnaire or inter
ferences in the offender's race. Also, certain em view schedule-that he or she uses to measure Step 3:
ployers said they were more willing to hire an variables. Respondents read the questions them • Decide on III
ex-offender, but in the audit all employers acted selves and mark answers on a questionnaire. An • Get safllJJiItl
the same. The authors said, "Survey responses interviewschedule is a set of questions read to the • Decide on 51
have very little connection to the actual behav respondent by an interviewer, who also records • Select~
iors exhibited by these employers" (2005:367). responses. To simplify the discussion, I will use
only the term questionnaires.
A survey researcher conceptualizes and op
Step 4:
THE lOGIC OF SURVEY erationalizes variables as questions. He or she
• Locate feSIJlII
RESEARCH writes and rewrites questions for clarity and Conduct inlIl!
completeness, and organizes questions on the • Carefully relIIi
What Is a Survey?
questionnaire based on the research question,
Survey researchers sample many respondents the respondents, and the type of survey. (The
who answer the same questions. They measure types of surveys are discussed later.)
many variables,test multiple hypotheses, and in When preparing a questionnaire, the re Step 5:
fer temporal order from questions about past
behavior, experiences, or characteristics. For ex
ample, years of schooling or a respondent's race
searcher thinks ahead to how he or she will
record and organize data for analysis. He or she
pilot-tests the questionnaire with a small set of • Perfonn_
aI.
• Enter data ill
• Recheck
I
are prior to current attitudes. An association respondents similar to those in the final survey.
among variables is measured with statistical If interviewers are used, the researcher trains
techniques. Survey researchers think of alterna them with the questionnaire. He or she asks re Step 6:
tive explanations when planning a survey, spondents in the pilot-test whether the questions
• Describe InIIIl
measure variables that represent alternative ex were clear and explores their interpretations to in resean:h ~
planations (i.e., control variables), then statisti see ifhis or her intended meaning was clear. The • Present findil
cally examine their effectsto rule out alternative researcher also draws the sample during this critique and~
explanations. phase.
----------
FIGURE 7.1 Steps in the Process of
Survey Research
..
CHAPTER 7 / SURVEY RESEARCH 169
Step 2:
t tionnaire and are thanked, he or she organizes
the data and prepares them for statisticalanalysis.
Survey research can be complex and expen
0 Plan how to record data.
sive and it can involve coordinating many peo
· Pilot test survey instrument. ple and steps. The administration of survey
research requires organization and accurate
Step 3:
1
record keeping. The researcher keeps track of
each respondent, questionnaire, and inter
viewer. For example, he or she gives each sam
·
0
Decide on target population.
Get sampling frame.
pled respondent an identification number,
which also appears on the questionnaire. He or
0 Decide on sample size. she then checks completed questionnaires
0 Select sample. against a list of sampled respondents. Next, the
Step 4:
t
researcher reviewsresponses on individual ques
tionnaires, stores original questionnaires, and
transfers information from questionnaires to a
0 Locate respondents. format for statistical analysis. Meticulous book
0 Conduct interviews. keeping and labeling are essential. Otherwise,
0 Carefully record data. the researcher may find that valuable data and
effort are lost through sloppiness.
Step 5:
1
CONSTRUCTING THE
0 Enter data into computers. QUESTIONNAIRE
0 Recheck all data.
0 Perform statistical analysis on data. Principles of Good Question Writing
Step 6:
1
A good questionnaire forms an integrated
whole. The researcher weaves questions together
so they flow smoothly. He or she includes intro
• Describe methods and findings ductory remarks and instructions for clarifica
in research report. tion and measures each variable with one or
· Present findings to others for
critique and evaluation.
more survey questions.
Three principles for effective survey ques
tions are: Keep it clear, keep it simple, and keep
170 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
the respondent's perspective in mind. Good sur For the general public, this is the language used
vey questions give the researcher valid and reli on television or in the newspaper (about an
able measures. They also help respondents feel eighth-grade reading vocabulary). Survey re
that they understand the question and that their searchers have learned that some respondents
answers are meaningful. Questions that do not may not understand basic terminology. Here are three ~
mesh with a respondent's viewpoint or that re enced professional
spondents find confusing are not good mea 2. Avoid ambiguity, confusion, and vagueness. original wording after
sures. A survey researcher must exercise extra Ambiguity and vagueness plague most question percent of respordei~
care if the respondents are heterogeneous or writers. A researcher might make implicit as
sumptions without thinking of the respondents. Original Quemc. l
come from different life situations than his or
her own. For example, the question, "What is your in Do you exercise Of" ~
Researchers face a dilemma. They want each come?" could mean weekly, monthly, or annual;
family or personal; before taxes or after taxes; for
sports regularly? l
respondent to hear exactly the same questions,
What is the average ~
but will the questions be equally clear, relevant, this year or last year; from salary or from all
and meaningful to all respondents? If respon sources. The confusion causes inconsistencies in
days each week you ~
dents have diverse backgrounds and frames of how different respondents assign meaning to
reference, the exact same wording may not have and answer the question. The researcher who
wants before-tax annual family income for last
the same meaning. Yet, tailoring question word
ing to each respondent makes comparisons al year must explicitly ask for it.! [Following question ad
Another source of ambiguity is the use ofin What is the number af:~
most impossible. A researcher would not know in a typical day? j
whether the wording of the question or the dif definite words or response categories. For exam
ple, an answer to the question, "Do you jog 1
ferences in respondents accounted for different
regularly? Yes _ _ No _ _," hinges on the j
answers.
meaning of the word regularly. Some respon j
Question writing is more of an art than a I
science. It takes skill, practice, patience, and cre dents may define regularly as every day, others as
once a week. To reduce respondent confusion j
ativity. The principles of question writing are il
and get more information, be specific-ask j
lustrated in the following 12 things to avoid Exercise question ('I. ~
whether a person jogs "about once a day," "a few
when writing survey questions. The list does not
include every possible error, only the more fre times a week," "once a week," and so on. (See Butter question ('1.1
quent problems. Box 7.1 on improving questions.) Egg question (% sarW1
3. Avoid emotional language. Words have im ~
1. Avoid jargon, slang, and abbreviations. Jar plicit connotative as well as explicit denotative
i
Source: Adapted from F~
gon and technical terms come in many forms. meanings. Words with strong emotional conno
Plumbers talk about snakes, lawyers about a con tations can color how respondents hear and an
tract of uberrima fides, psychologists about the swer survey questions.
Oedipus complex. Slang is a kind ofjargon within Use neutral language. Avoid words with
a subculture-for example, the homeless talk emotional "baggage," because respondents may or status. Issues link£IlI
about a snowbird and skiers about a hotdog. Also react to the emotionally laden words rather than status can color hOWl
avoid abbreviations. NATO usually means to the issue. For example, the question, "What swer survey quest:ioal
North Atlantic Treaty Organization, but for a re do you think about a policy to pay murderous ment with a PT~
spondent, it might mean something else (Na terrorists who threaten to steal the freedoms of Respondents may . .
tional Auto Tourist Organization, Native peace-loving people?" is full of emotional words feelings toward the pi
Alaskan Trade Orbit, or North African Tea Of (m urderous, freedoms, steal, and peace). addressing the issue. •
fice). Avoid slang and jargon unless a specialized doctors say that cigad
population is being surveyed. Target the vocab 4. Avoid prestige bias. Titles or positions in so ease for those near a I
ulary and grammar to the respondents sampled. ciety (e.g., president, expert, etc.) carry prestige fects people who "..
CHAPTER 7 I SURVEY RESEARCH 171
Here are three survey questions written by experi inadequate answers (e.g., don't know). As you can
enced professional researchers. They revised the see, question wording is an art that may improve with
original wording after a pilot test revealed that 1 5 practice, patience, and pilot testing.
percent of respondents asked for clarification or gave
Do you exercise or play What counts as exercise? Do you do any sports or hobbies,
sports regularly? physical activities, or exercise,
including walking, on a regular basis?
What is the average number of Does margarine count The next question is just about
days each week you have butter? as butter? butter-not including margarine.
How many days a week do you have
butter?
[Following question on eggs] How many eggs is a serving? On days when you eat eggs, how
What is the number of servings What is a typical day? many eggs do you usually have?
in a typical day?
or status. Issues linked to people with high social Likewise, a question such as, "Do you support
status can color how respondents hear and an the president's policy regarding Kosovo?" will be
swer survey questions. Avoid associating a state answered by respondents who have never heard
ment with a prestigious person or group. of Kosovo on the basis of their view of the
Respondents may answer on the basis of their president.
feelings toward the person or group rather than
addressing the issue. For example, saying, "Most 5. Avoid double-barreled questions. Make each
doctors say that cigarette smoke causes lung dis question about one and only one topic. A
ease for those near a smoker. Do you agree?" af double-barreled question consists of two or more
fects people who want to agree with doctors. questions joined together. It makes a respon
172 PART TWO! CONDUCTING QUANTITATIVE RESEARCH
dent's answer ambiguous. For example, if asked, "Should the mayor spend even more tax money hours later or cl
"Does this company have pension and health in trying to keep the streets in top shape?" leads re day?" Answers to a
surance benefits?" a respondent at a company spondents to disagree, whereas "Should the very reliable, but
with health insurance benefits only might an mayor fix the pot-holed and dangerous streets tration.
swer either yes or no. The response has an am in our city?" is loaded for agreement.
biguous meaning, and the researcher cannot be
8. A void asking questions that are beyond re
certain of the respondent's intention. A re
spondents' capabilities. Asking something that
searcher who wants to ask about the joint occur
few respondents know frustrates respondents
rence of two things-for example, a company
and produces poor-quality responses. Respon
with both health insurance and pension bene
dents cannot always recall past details and may
fits-should ask two separate questions.
not know specific factual information. For ex
6. Do not confuse beliefs with reality. Do not ample, asking an adult, "How did you feel about
confuse what a respondent believes with what your brother when you were 6 years old?" is
you, the researcher, measures. A respondent probably worthless. Asking respondents to make
may think that a relationship exists between two a choice about something they know nothing
variables but this is not an empirical measure about (e.g., a technical issue in foreign affairs or
ment ofvariables in a relationship. For example, an internal policy of an organization) may result
a researcher wants to find out if students rate in an answer, but one that is unreliable and
teachers higher who tell many jokes in class. The meaningless. When many respondents are un
two variables are "teacher tells jokes" and "rating likely to know about an issue, use a full-filter 11. A void doulM
the teacher." The wrong way to approach the is question form (to be discussed). ordinary language
sue is to ask students, "Do you rate a teacher Phrase questions in the terms in which re and confusing. For"
higher if the teacher tells many jokes?" This mea spondents think. For example, few respondents logically means tluI
sures whether or not students believe that they will be able to answer, "How many gallons of job, but the second
rate teachers based on joke telling; it does not gasoline did you buy last year for your car?" Yet, for emphasis. Sur.:h
measure the empirical relationship. The correct respondents may be able to answer a question more subtle fonm
way is to ask two separate empirically based about gasoline purchases for a typical week, also confusing. ~
questions: "How do you rate this teacher?" and which the researcher can multiply by 52 to esti asked to agree or
"How many jokes does the teacher tell in class?" mate annual purchases.? example, respo
Then the researcher can examine answers to the statement, "S~
9. Avoid false premises. Do not begin a ques take a compreheD!l;iq::4j
two questions to determine if there is an associ
tion with a premise with which respondents may ically stating a
ation between them. People's beliefs about a re
not agree, then ask about choices regarding it. li..;agree with not
lationship among variables are distinct from an
Respondents who disagree with the premise will
actual empirical relationship.
be frustrated and not know how to answer. For 12. Avoid OW'rMf..llII
7. Avoid leading questions. Make respondents example, the question, "The post office is open caegories. Make
feel that all responses are legitimate. Do not let too many hours. Do you want it to open four mutually exclusise,
them become aware of an answer that the re hours later or close four hours earlier each day?" _\lurually exdusoe
searcher wants. A leading (or loaded) question is leaves those who either oppose the premise or ~ries do not oqn.1I
one that leads the respondent to choose one re oppose both alternatives without a meaningful :hat are n umerial
sponse over another by its wording. There are choice. ~30lcanbe~
many kinds of leading questions. For example, A better question explicitly asks the respon ~29). The arD~""
the question, "You don't smoke, do you?" leads dent to assume a premise is true, then asks for a ~ofoY~
respondents to state that they do not smoke. preference. For example, "Assuming the post of .:DFle, ~Are ~"'OU
Loaded questions can be stated to get either fice has to cut back its operating hours, which ihere things ~ ~
positive or negative answers. For example, would you find more convenient, opening four :De2Il5 that e'\-crT
CHAPTER 7 I SURVEY RESEARCH 173
hours later or closing four hours earlier each place to go. For example, asking respondents,
day?"Answers to a hypothetical situation are not "Are you working or unemployed?" leaves out
very reliable, but being explicit will reduce frus respondents who are not working but do not
tration. consider themselves unemployed (e.g., full-time
homemakers, people on vacation, students, peo
10. Avoid asking about intentions in the distant ple with disabilities, retired people, etc.). A re
future. Avoid asking people about what they searcher first thinks about what he or she wants
might do under hypothetical circumstances far to measure and then considers the circum
in the future. Responses are poor predictors of stances of respondents. For example, when ask
behavior removed far from their current situa ing about a respondent's employment, does the
tion or far in the future. Questions such as, researcher want information on the primary job
"Suppose a new grocery store opened down the or on all jobs? On full-time work only or both
road in three years. Would you shop at it?" are full- and part-time work? On jobs for pay only
usually a waste of time. It is better to ask about or on unpaid or volunteer jobs as well?
current or recent attitudes and behavior. In gen Keep response categories balanced. A case of
eral, respondents answer specific, concrete ques unbalanced choices is the question, "What kind
tions that relate to their experiences more of job is the mayor doing: outstanding, excellent,
reliably than they do those about abstractions very good, or satisfactory?" Another type of un
that are beyond their immediate experiences. balanced question omits information-for ex
11. Avoid double negatives. Double negatives in ample, "Which of the five candidates running
ordinary language are grammatically incorrect for mayor do you favor: Eugene Oswego or one
and confusing. For example, "I ain't got no job" of the others?" Researchers can balance re
logically means that the respondent does have a sponses by offering bipolar opposites. It is easy
job, but the second negative is used in this way to see that the terms honesty and dishonesty have
for emphasis. Such blatant errors are rare, but different meanings and connotations. Asking re
more subtle forms of the double negative are spondents to rate whether a mayor is highly,
also confusing. They arise when respondents are somewhat, or not very honest is not the same as
asked to agree or disagree with a statement. For asking them to rate the mayor's level of
example, respondents who disagree with the dishonesty. Unless there is a specific purpose for
statement, "Students should not be required to doing otherwise, it is better to offer respondents
take a comprehensive exam to graduate" are log equal polar opposites at each end of a contin
ically stating a double negative because they uum.' For example, "Do you think the mayor is:
disagree with not doing something. very honest, somewhat honest, neither honest
nor dishonest, somewhat dishonest, or very dis
12. Avoid overlapping or unbalanced response honest?" (see Table 7.1).
categories. Make response categories or choices
mutually exclusive, exhaustive, and balanced.
Aiding Respondent Recall
Mutually exclusive means that response cate
gories do not overlap. Overlapping categories Recalling events accurately takes more time and
that are numerical ranges (e.g., 5-10, 10-20, effort than the five seconds that respondents
20-30) can be easily corrected (e.g., 5-9,10-19, have to answer survey questions. Also, one's
20-29). The ambiguous verbal choice is another ability to recall accurately declines over time.
type of overlapping response category-for ex Studies in hospitalization and crime victimiza
ample, "Are you satisfied with your job or are tion show that although most respondents can
there things you don't like about it?" Exhaustive recall significant events that occurred in the past
means that every respondent has a choice-a several weeks, half are inaccurate a year later.
174 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
--------------------.
TABLE 7.1 Summary of Survey Question Writing Pitfalls
Survey resear
is less trustworthv
affected by man);
curring simultan
significance of an
1. Jargon, slang, abbreviations Did you drown in brew until you were Last night, about how much beer did conditions (questX.
totally blasted last night? you drink?
style), and the r'estJlJllllll
2. Vagueness Do you eat out often? In a typical week, about how many nal consistency.
meals do you eat away from home, at a The complexirr
restaurant, cafeteria, or other eating
not mean that 5
establishment?
about past events;
3. Emotional language "The respected Grace Commission doc How important is it to you that Con questions and int
4. Prestige bias uments that a staggering $350 BILLION gress adopt measures to reduce gov
searchers should p
of our tax dollars are being completely ernment waste?
cial instructions and
wasted through poor procurement prac Very Important
tices, bad management, sloppy book
should also provide
Somewhat Important such as a fixed time
keeping, 'defective' contract
Neither Important or Unimportant Rather than ask, ~
management, personnel abuses and
other wasteful practices. Is cutting pork Somewhat Unimportant sporting event last
barrel spending and eliminating govern Not Important At All want to know how
ment waste a top priority for you r: tended last winter.
5. Double-barreled questions Do you support or oppose raising so Do you support or oppose raising so Think back to D
cial security benefits and increased cial security benefits? sporting events for
spending for the military? Do you support or oppose increasing December? Now, think.
spending on the military? attend any sporting
6. Beliefs as real Do you think more educated people What is your education level? Do you
smoke less? smoke cigarettes? Types of Questions
7. Leading questions Did you do your patriotic duty and Did you vote in last month's mayoral Categories
vote in the last election for mayor? election? Threatening (bra__
8. Issues beyond respondent Two years ago, how many hours did In the past two weeks, about how many sometimes ask aboul
capabilities you watch TV every month? hours do you think you watched TV on that respondents ~
a typical day? sentation of self, such
9. False premises When did you stop beating your Have you ever slapped, punched, or hit behavior, drug or
girl/boyfriend? your girl/boyfriend? problems, or devm.
10. Distant future intentions After you graduate from college, get a Do you have definite plans to put some may be reluctant to
job, and are settled, will you invest a money into the stock market within the answer completely
lot of money in the stock market? coming two months? searchers who wish to
11 . Double negatives Do you disagree with those who do not There is a proposal to build a new city do so with great care
want to build a new city swimming swimming pool. Do you agree or dis about the results! (see
pool? agree with the proposal?
Threatening q
1 2. Unbalanced responses Did you find the service at our hotel to Please rate the service at our hotel: sue of self-presentation
be, Outstanding, Excellent, Superior, or Outstanding, Very Good, Adequate, or
spondents often trv to
Good? Poor.
of themselves to others..
embarrassed, or afraid
'Actual question taken from a mailquestionnaire that was sent to me in May 1998 by the National Republican Congres
sional Committee. It is also a double-barreled question.
or find it emotionalh
own actions honestlv,
CHAPTER 7 / SURVEY RESEARCH 175
--------------------~
TAB L E 7.3 Over- and Underreporting Behavior on Surveys
.. questions can be ~
dents do not wilDli
may measure o~
about factual infun
have inaccurate iacl
Some simple II
Low Threat/Normative the number of ~
not always ilIlS"'-~
Registered to vote +15 +17 +12
some households
Voted in primary +39 +31 +36 boyfriend who
who left after an DI
kit.
Have own library card +19 +21 +18
or the uncle who ..
I
High Threat money-may be •
Bankruptcy -32 -29 -32 household, but be ~
permanent resi~
Drunk driving -47 -46 -54
herself to live tberd
Source: Adapted from Bradburnand Sudman (1 980:8)0 Others have i
oppose foreign DII
is based on enr~
want honest answers. They can phrase the ques money to charity, having a good marriage, lov cost of the p~
tion in an "enhanced way" to provide a context ing their children, and so forth. For example, would prefer to spcl
that makes it easier for respondents to give hon one study found that one-third of people who an amount much ~
est answers. For example, the following en reported in a survey that they gave money to a spent. I
hanced question was asked of heterosexual local charity really did not. Because a norm says A researcher ~
males: "In past surveys, many men have re that one should vote in elections, many report
questionsareat3
ported that at some point in their lives they have voting when they did not. In the United States, Little is gained ii
had some type ofsexual experience with another those under the greatest pressure to vote (i.e., not answer the
male. This could have happened before adoles
cence, during adolescence, or as an adult. Have
highly educated, politically partisan, highly reli
gious people who had been contacted by an or
can be worded
able saying ~ ~
sO'
you ever had sex with a male at some point in
your life?" In contrast, a standard form of the
ganization that urged them to vote) are the
people most likely to overreport voting.
example, "How
heard about. . . .
or' .
question would have asked, "Have you ever had Questionnaire writers try to reduce social l
sex with another male?"
Also, by embedding a threatening response
desirability bias by phrasing questions in ways
that make norm violation appear less objection
SkiPOT~
avoid asking
within more serious activities, it may be made to able and that presents a wider range of behavior respondent. Yet.
seem less deviant. For example, respondents as acceptable. They can also offer multiple re specific r·espODlclem.
might hesitate to admit shoplifting if it is asked sponse categories that give respondents "face two- (or more;
first, but after being asked about armed robbery saving" alternatives. first part of the
or burglary, they may admit to shoplifting be two different
cause it appears less serious. Knowledge Questions. Studies suggest that a ceives, Con,ti°il¥lIi3IllII
large majority of the public cannot correctly an dents for whom a
Socially Desirable Questions. Social desirabil swer elementary geography questions or identify Sometimes thrr
ity bias occurs when respondents distort answers important political documents (e.g., the Decla 00ns.. On thebasis
to make their reports conform to social norms. ration of Independence). Researchers some non. the r~OD"
People tend to overreport being cultured (i.e., times want to find out whether respondents instructed to ~
reading, attending high-culture events), giving know about an issue or topics, but knowledge .;uestions.
CHAPTER 7 / SURVEY RESEARCH 177
(e.g., age), open questions may be best. In addi question responses from the answers given to
tion, sensitive topics (e.g., sexual behavior, the open questions.
liquor consumption) may be more accurately Researchers writing closed questions have to
measured with closed questions. make many decisions. How many response
The disadvantages of a question form can be choices should be given? Should they offer a
reduced by mixing open-ended and closed-ended middle or neutral choice? What should be the
questions in a questionnaire. Mixing them also order of responses? What types of response
offers a change of pace and helps interviewers es choices? How will the direction of a response be
tablish rapport. Periodic probes (i.e., follow-up measured?
questions by interviewers) with closed-ended Answers to these questions are not easy. For
questions can reveal a respondent's reasoning. example, two response choices are too few, but
Having interviewers periodically use probes more than five response choices are rarelyeffec
to ask about a respondent's thinking is a way to tive. Researchers want to measure meaningful
check whether respondents are understanding distinctions and not collapse them. More spe
the questions as the researcher intended. How cific responses yield more information, but too
ever, probes are not substitutes for writing clear many specifics create confusion. For example,
questions or creating a framework of under rephrasing the question, "Are you satisfied with
standing for the respondent. Unless carefully your dentist?" (which has a yes/no answer) to
stated, probes might shape the respondent's an "How satisfied are you with your dentist: very
swers or force answers when a respondent does satisfied, somewhat satisfied, somewhat dissatis
not have an opinion or information. Yet, flexible fied, or not satisfied at all?" gives the researcher
or conversational interviewing in which inter more information and a respondent more
viewers use many probes can improve accuracy choices.
on questions about complex issues on which re
spondents do not clearly understand basic terms No nattitudes and the Middle Positions.
or about which they have difficulty expressing Survey researchers debate whether to include
their thoughts. For example, to the question, choices for neutral, middle, and nonattitudes
"Did you do any work for money last week?" a (e.g., "not sure," "don't know," or "no opin
respondent might hesitate then reply, "Yes." An ion")." Two types of errors can be made: accept
interviewer probes, "Could you tell me exactly ing a middle choice or "no attitude" response
what work you did?" The respondent may reply, when respondents hold a nonneutral opinion, or
"On Tuesday and Wednesday, I spent a couple forcing respondents to choose a position on an
hours helping my buddy John move into his new issue when they have no opinion about it.
apartment. For that he gave me $40, but I didn't Many fear that respondents will choose
have any other job or get paid for doing anything nonattitude choices to evade making a choice.
else." If the researcher's intention was only to get Yet, it is usually best to offer a nonattitude
reports of regular employment, the probe re choice, because people will express opinions on
vealed a misunderstanding. Researchers also use fictitious issues, objects, and events. Byoffering a
partiallyopenquestions (i.e., a set of fixed choices nonattitude (middle or no opinion) choice, re
with a final open choice of "other"), which al searchers identify those holding middle posi
lows respondents to offer an answer that the re tions or those without opinions.
searcher did not include. The issue of non attitudes can be approached
Open-ended questions are especially valu by distinguishing among three kinds of attitude
able in early or exploratory stages of research. questions: standard-format, quasi-filter, and full
For large-scale surveys, researchers use open filter questions (see Box 7.3). The standard-for
questions in pilot-tests, then develop closed- mat question does not offer a "don't know"
180 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
ment? "The Russian leaders are basically trying to get along with America."
men are better suited,
Quasi-Filter or both are equally SIll
respondents are more I
Here is a statement about an other country: "The Russian leaders are basically trying
ment, whereas for-cell
to get along with America." Do you agree, disagree, or have no opinion on that?
courage thought ~
Full Filter bias-a tendency of.
Here is a statement about an other country. Not everyone has an opinion on this. If you
and not really deode, I
do not have an opinion, just say so. Here's the statement: "The Russian leaders are ba Researchers ~
sically trying to get along with America." Do you have an opinion on that? If yes, do you gives respondents oil RlI
agree or disagree? ternative. For e~
whether they suppoIUlj
Example of Results from Different Question Forms ergy conservation. The
Standard Form (%) Quasi-Filter (%) Full Filter (%) spondents heard, -Dol
you oppose it becausej
Agree 48.2 27.7 22.9 to enforce?" instead. ali
Disagree 38.2 29.5 20.9 or oppose the law~- J
No opinion 13.6' 42.8 56.3 It is better to ~
among alternatives ~
items along an ~
'Volunteered dents can rate se\-edl.
Source: Adapted from Schuman and Presser (1981 :116-125). Standard format is from Fall place them in a hi~
1978; quasi- and full-filter are from February 1977.
Wording Issues
Survey researChers~
first, discussed earlier,
choice; a respondent must volunteer it. A quasi quasi -filter or full-filter questions. Filtered ques lary and grammar to
filter question offers respondents a "don't know" tions do not eliminate all answers to nonexistent second issue involves
alternative. A full-filter question is a special type of issues, but they reduce the problem. phrases. This is tri "
contingency question. It first asks if respondents to know in advance ..
have an opinion, then asks for the opinion of Agree/Disagree, Rankings or Ratings? Survey affects responses. l
those who state that they do have an opinion. researchers who measure values and attitudes
The well-dos
Many respondents will answer a question if have debated two issues about the responses of forbid and not allow
a "no opinion" choice is missing, but they will fered." Should questionnaire items make a state wording differences.
choose "don't know" when it is offered, or say ment and ask respondents whether they agree or
meaning, but manS"
that they do not have an opinion if asked. Such disagree with it, or should it offer respondents "not allow" some' "
respondents are called floaters because they specific alternatives? Should the questionnaire general, less well ed
"float" from giving a response to not knowing. include a set of items and ask respondents to rate influenced by minor
Their responses are affected by minor wording them (e.g., approve, disapprove), or should it Certain words ~
changes, so researchers screen them out using give them a list of items and force them to rank- tional reaction, and rd
CHAPTER 7 / SURVEY RESEARCH 181
order items (e.g., from most favored to least ning to learn of them. For example, Smith
favored)? (1987) found large differences (e.g., twice as
It is best to offer respondents explicit alter much support) in U.S. survey responses de
natives. For example, instead of asking, "Do you pending on whether a question asked about
agree or disagree with the statement, 'Men are spending "to help the poor" or "for welfare."
better suited to ... .' " instead ask, "Do you think He suggested that the word welfare has such
men are better suited, women are better suited, strong negative connotations for Americans
or both are equally suited?" Less well educated (lazy people, wasteful and expensive programs,
respondents are more likely to agree with a state etc.) that it is best to avoid it.
ment, whereas forced-choice alternatives en Many respondents are confused by words or
courage thought and avoid the response set their connotations. For example, respondents
bias-a tendency of some respondents to agree were asked whether they thought television news
and not really decide. was impartial. Researchers later learned that
Researchers create bias if question wording large numbers of respondents had ignored the
gives respondents a reason for choosing one al word irnpartial-a term the middle-class, edu
ternative. For example, respondents were asked cated researchers assumed everyone would
whether they supported or opposed a law on en know. Less than half the respondents had inter
ergy conservation. The results changed when re preted the word as intended with its proper
spondents heard, "Do you support the law or do meaning. Over one- fourth ignored it or had no
you oppose it because the law would be difficult idea of its meaning. Others gave it unusual
to enforce?" instead of simply, "Do you support meanings, and one-tenth thought it was directly
or oppose the law?" opposite to its true meaning. Researchers need
It is better to ask respondents to choose to be cautious, because some wording effects
among alternatives by ranking instead of rating (e.g., the difference between forbid and not al
items along an imaginary continuum. Respon low) remain the same for decades, while other
dents can rate several items equally high, but will effects may appear."
place them in a hierarchy if asked to rank them.f
Questionnaire Design Issues
Wording Issues
Length ofSurvey or Questionnaire. How long
Survey researchers face two wording issues. The should a questionnaire be or an interview last?
first, discussed earlier, is to use simple vocabu Researchers prefer long questionnaires or inter
lary and grammar to minimize confusion. The views because they are more cost effective. The
second issue involves effects of specific words or cost for extra questions-once a respondent has
phrases. This is trickier because it is not possible been sampled, has been contacted, and has com
to know in advance whether a word or phrase pleted other questions-is small. There is no ab
affects responses. solute proper length. The length depends on the
The well-documented difference between survey format (to be discussed) and on the re
forbid and not allow illustrates the problem of spondent's characteristics. A 5-minute tele
wording differences. Both terms have the same phone interview is rarely a problem and may be
meaning, but many more people are willing to extended to 20 minutes. A few researchers
"not allow" something than to "forbid" it. In stretched this to beyond 30 minutes. Mail ques
general, less well educated respondents are most tionnaires are more variable. A short (3- or 4
influenced by minor wording differences. page) questionnaire is appropriate for the
Certain words seem to trigger an emo general population. Some researchers have had
tional reaction, and researchers are just begin success with questionnaires as long as 10 pages
182 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
(about 100 items) with the general public, but not when the question is by itself or before a
responses drop significantly for longer question question about fetus defects. A classic example
naires. For highly educated respondents and a of order effects is presented in Box 7.4.
salient topic, using questionnaires of 15 pages Respondents may not perceive each issue of
Question 1
may be possible. Face-to-face interviews lasting a survey as isolated and separate. They respond
an hour are not uncommon. In special situa to survey questions based on the set ofissues and "Do you think that the I
tions, face-to-face interviews as long as three to their order of presentation in a questionnaire. porters from other- CDIIIlI
five hours have been conducted. Previous questions can influence later ones in
two ways: through their content (i.e., the issue)
'Ow, as they ':~ j
Question 2
Question Order or Sequence. A survey re and through the respondent's response. For ex
"Doyou thinka
searcher faces three question sequence issues:
organization of the overall questionnaire, ques
ample, a student respondent is asked, "Do you
support or favor an educational contribution for
reporters come in ~ 1
tion order effects, and context effects. students?" Answers vary depending on the topic j
of the preceding question. If it comes after, -J
Organization of Questionnaire. In general, you "How much tuition does the average U.S. stu I
should sequence questions to minimize the dis
comfort and confusion of respondents. A ques
tionnaire has opening, middle, and ending
dent pay?" respondents interpret "contribution"
to mean support for what students will pay. If it
comes after "How much does the Swedish gov
--1
Heard First
#1
questions. After an introduction explaining the ernment pay to students?" respondents interpret #2
survey, it is best to make opening questions it to mean a contribution that the government
pleasant, interesting, and easy to answer to help will pay. Responses can be also influenced by
The context created tJy
a respondent feel comfortable about the ques previous answers, because a respondent having second question.
tionnaire. Avoid asking many boring back already answered one part will assume no over
ground questions or threatening questions first. lap. For example, a respondent is asked, "How is Source: Adapted frOf'1
Organize questions into common topics. Mix your wife?" The next question is, "How is your
ing questions on different topics causes confu family?" Most respondents will assume that the
sion. Orient respondents by placing questions second question means family members other
on the same topic together and introduce the than the wife because they already gave an an
section with a short introductory statement (e.g., swer about the wife.!!
"Now I would like to ask you questions about
housing"). Make question topics flow smoothly Context Effects. Researchers found powerful
and logically, and organize them to assist re context effects in surveys. As a practical matter,
spondents' memory or comfort levels. Do not two things can be done regarding context effects.
end with highly threatening questions, and al Use a funnel sequence of questions-that is, ask
ways end with a "thank you." more general questions before specific ones (e.g.,
ask about health in general before asking about
Order Effects. Researchers are concerned that specific diseases). Or, divide the number of re
the order in which they present questions may spondents in half and give half of the questions
influence respondent answers. Such "order ef in one order and the other half in the alternative
fects" appear to be strongest for people who lack order, then examine the results to see whether
strong views, for less educated respondents, and question order mattered. If question order ef
for older respondents or those with memory fects are found, which order tells you what the
10ss.1° For example, support for an unmarried respondents really think? The answer is that you
woman having an abortion rises if the question cannot know for sure.
is preceded by a question about abortion being For example, a few years ago, a class of my
acceptable when a fetus has serious defects, but students conducted a telephone survey on two
CHAPTER 7 / SURVEY RESEARCH 183
Question 1
"Do you think that the United States should let Communist newspaper re
porters from other countries come in here and send back to their papers the
news as they see it?"
Question 2
"Do you think a Communist country like Russia should let American newspaper
reporters come in and send back to America the news as they see it?"
#1 54% 75%
#2 64% 82%
The context created by answering the first question affects the answer to the
second question.
topics: concern about crime and attitudes to less important issue. Bycontrast, after they were
ward a new anti-drunk-driving law. A random asked about drunk driving and thought about
half of the respondents heard questions about drunk driving as a crime, they may have ex
the drunk-driving law first; the other half heard pressed less concern about crime in general.
about crime first. I examined the results to see Respondents answer all questions based on
whether there was any context effect-a differ a context of preceding questions and the inter
ence by topic order. I found that respondents view setting. A researcher needs to remember
who were asked about the drunk-driving law that the more ambiguous a question's meaning,
first expressed less fear about crime than did the stronger the context effects, because re
those who were asked about crime first. Like spondents will draw on the context to interpret
wise, they were more supportive of the drunk and understand the question. Previous ques
driving law than were those who first heard tions on the same topic and heard just before a
about crime. The first topic created a context question can have a large context effect. For ex
within which respondents answered questions ample, Sudman and associates (1996:90-91)
on the second topic. After they were asked about contrasted three ways of asking how much a re
crime in general and thought about violent spondent followed politics. When they asked
crime, drunk driving may have appeared to be a the question alone, about 21 percent of respon
184 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
dents said they followed politics "now and a different style from the questions (e.g., in a dif Question Format S.
then" or "hardly at all." When they asked the ferent color or font or in all capitals) to distin a format for ques:tioal
question after asking what the respondent's guish them. This is so an interviewer can easily spondents circle respl
elected representative recently did, the percent distinguish between questions for respondents dots, or put an X in a
age who said they did not follow nearly dou and instructions intended for the interviewer make responses lJJlBI
bled, going to 39 percent. The knowledge alone. ets to be checked aId
question about the representative made many Layout is crucial for mail questionnaires be
respondents feel that they did not really know cause there is no friendly interviewer to interact
much. When a question about the amount of with the respondent. Instead, the question
"public relations work" the elected representa naire's appearance persuades the respondent. In
tive provided to the area came between the two mail surveys, include a polite, professional cover
questions, 29 percent of respondents said they letter on letterhead stationery, identifying the re
Example of Horizonal
did not follow politics. This question gave re searcher and offering a telephone number for
spondents an excuse for not knowing the first questions. Details matter. Respondents will be Doyou think it istoo ~
question-they could blame their representa turned off if they receive a bulky brown enve o Too Easy = T.
tive for their ignorance. The context of a ques lope with bulk postage addressed to Occupant 1
tion can make a difference and researchers need or if the questionnaire does not fit into the re Do you think it istoo eiil
to be aware of it at all times. turn envelope. Always end with "Thank you for o TooEasy l
your participation." Interviewers and question :::J Too Difficult .!
Format and Layout. There are two format or naires should leave respondents with a positive
layout issues: the overall physical layout of the
questionnaire and the format of questions and
responses.
feeling about the survey and a sense that their
participation is appreciated.
Question design matters. One study of col
lege students asked how many hours they stud
Example Ofa_,
:::J About Right 1
1
Questionnaire Layout. Layout is important, ied per day. Some students saw five answer
whether a questionnaire is for an interviewer or choices ranging from 0.5 hour to more than 2.5
for the respondent. Questionnaires should be hours; others saw five answer choices ranging
clear, neat, and easy to follow. Give each ques from less than 2.5 hours to more than 4.5 hours.
tion a number and put identifying information Of students who saw the first set, 77 percent said
(e.g., name of organization) on questionnaires. they studied under 2.5 hours versus 31 percent
Never cramp questions together or create a con of those receiving the second set. When the mail
fusing appearance. A few cents saved in postage questionnaire and telephone interview were
or printing will ultimately cost more in terms of compared, 58 percent of students hearing the
Examples of Some
lower validity due to a lower response rate or of first set said under 2.5 hours, but there was no
confusion of interviewers and respondents. change among those hearing the second set. Excellent, Good. Fa...
Make a cover sheet or face sheet for each inter More than differences in response categories Approve/Disapprove
view, for administrative use. Put the time and were involved, because when students were Favor/Oppose
date of interview, the interviewer, the respon asked about hours of television watching per day Strongly Agree, Agree..
dent identification number, and the inter with similar response categories and then the al Too Much, Too little.
viewer's comments and observations on it. A ternative response categories made no differ Better, Worse, About
professional appearance with high-quality ence. What can we learn from this? Respondents Regularly, Often.
graphics, space between questions, and good lay without clear answers tend to rely on question
Always, MostofTrme.
out improves accuracy and completeness and naire response categories for guidance and more
helps the questionnaire flow. anonymous answering formats tend to yield 'wIore Likely, Less L.ikefy.
Give interviewers or respondents instruc more honest responses (see Dillman 2000:32-39 Very Interested, I
tions on the questionnaire. Print instructions in for more details).
CHAPTER 7 I SURVEY RESEARCH 185
Question Format. Survey researchers decide on usually clearest. Also, listing responses down a
a format for questions and responses. Should re page rather than across makes them easier to see
spondents circle responses, check boxes, fill in (see Box 7.5). As mentioned before, use arrows
dots, or put an X in a blank? The principle is to and instructions for contingency questions. Vi
make responses unambiguous. Boxes or brack sual aids are also helpful. For example, hand out
ets to be checked and numbers to be circled are thermometer-like drawings to respondents
Do you think it is too easy or too difficult to get a divorce, or is it about right?
o Too Easy
o Too Difficult
o About Right
when asking about how warm or cool they feel and statements (i.e., making eye contact, ex
toward someone. A matrix question (or grid pressing sincerity, explaining the sampling or
question) is a compact way to present a series of survey, emphasizing importance of the inter
questions using the same response categories. It view, clarifying promises of confidentiality, etc.).
saves space and makes it easier for the respon Survey researchers can also use alternative in 1. Address the qui
dent or interviewer to note answers for the same terviewers (i.e., different demographic charac not "Occupant."
response categories. teristics, age, race, gender, or ethnicity), use 2. Include a <:areIuI
alternative interview methods (i.e., phone ver lettemead stA1tiII
Nonresponse. The failure to get a valid response sus face to face), or accept alternative respon cooperation. ~
from every sampled respondent weakens a sur dents in a household. the purpose of
vey. Have you ever refused to answer a survey? A critical area of nonresponse or refusal to searcher's name
In addition to research surveys, people are asked participate occurs with the initial contact be 3. Always indude a I
to respond to many requests from charities, tween an interviewer and a respondent. A face envelope.
marketing firms, candidate polls, and so forth. to-face or telephone interview must overcome 4. The quest:iofyyi
Charities and marketing firms get low response resistance and reassure respondents. layout and rea5iII
rates, whereas government organizations get Research on the use of incentives found that
5. The questiornl
much higher cooperation rates. Nonresponse prepaid incentives appear to increase respon
printed and easy
can be a major problem for survey research be dent cooperation in all types of surveys. They do
6. Send two~
cause if a high proportion of the sampled re not appear to have negative effects on survey
notres~
spondents do not respond, researchers may not composition or future participation.
one week after- !
be able to generalize results, especially if those There is a huge literature on ways to in
seconda~1iIl
who do not respond differ from those who re crease response rates for mail questionnaires
again and offer- ..
spond. (see Box 7.6),13 Heberlein and Baumgartner
7. Do not send ~
Public cooperation in survey research has (1978, 1981) reported 71 factors affecting mail
day periods, ;
declined over the past 20 to 30 years across many questionnaire response rates.
countries, with the Netherlands having the high 8. Do not put
est refusal rate, and with refusal rates as high as leave a btra
30 percent in the United States.V There is both a TYPES OF SURVEYS: ADVANTAGES general commelll1l
growing group of "hard core" refusing people AND DISADVANTAGES 9. Sponsors ~
and a general decline in participation because mate (e.g..
Mail and Self-Administered large firms. ere..)
many people feel there are too many surveys.
Questionnaires 10. Include a ~
Other reasons for refusal include a fear of crime
and strangers, a more hectic life-style, a loss of Advantages. Researchers can give question possible,
privacy, and a rising distrust of authority or gov naires directly to respondents or mail them to
ernment. The misuse of the survey to sell prod respondents who read instructions and ques
ucts or persuade people, poorly designed tions, then record their answers. This type of
questionnaires, and inadequate explanations of survey is by far the cheapest, and it can be con
surveys to respondents also increase refusals for ducted by a single researcher. A researcher can
legitimate surveys. send questionnaires to a wide geographical area.
Survey researchers can improve eligibility The respondent can complete the questionnaire
rates by careful respondent screening, better when it is convenient and can check personal
sample-frame definition, and multilingual inter records if necessary. Mail questionnaires offer
viewers. They can decrease refusals by sending anonymity and avoid interviewer bias. They can
letters in advance of an interview, offering to be effective, and response rates may be high for
reschedule interviews, using small incentives an educated target population that has a strong
(i.e., small gifts), adjusting interviewer behavior interest in the topic or the survey organization.
CHAPTER 7 / SURVEY RESEARCH 187
--------------------~.
TAB L E 7.4 Types of Surveys and Their Features
Administrative Issues
Research Control
Sources of Bias
..
. . sriooNMC.
Respondent's reading skill Yes Yes No No _-\ third
...,.".. ••• wile
~~
...-r3Dd
*-s~
CHAPTER 7 / SURVEY RESEARCH 189
Advantages. Web-based surveys over the In still learning what is most effective for web sur
ternet or bye-mail are very fast and inexpensive. veys. It is best to provide screen-by-screen ques
They allow flexibledesign and can use visual im tions and make an entire question visible on the
ages, or even audio or video in some Internet screen at one time in a consistent format with
versions. Despite great flexibility, the basic prin drop-down boxes for answer choices. It is best to
ciples for question writing and for paper ques include a progress indicator (as motivation),
tionnaire design generally apply. such as a clock or waving hand. Visual appear
ance of a screen, such as the range of colors and
Disadvantages. Web surveys have three areas fonts, should be kept simple for easy readability
of concern: coverage, privacy and verification, and consistency. Be sure to provide very clear in
and design issues. The first concern involves structions for all computer actions (e.g., use of
sampling and unequal Internet access or use. drop-down screens) where they are needed and
Despite high coverage rates, older, less-educated, include "click here" instructions. Also, make it
lower-income, and more rural people are less easy for respondents to move back and forth
likely to have good Internet access. In addition, across questions. Researchers using web surverys
many people have multiple e-mail addresses, need to avoid technical glitches at the imple
which limits using them for sampling purposes. mentation stage by repeated pretesting, having a
Self-selection is a potential problem with web dedicated server, and obtaining sufficient broad
surveys. For example, a marketing department band to handle high demand.
could get very distorted results ofthe population
of new car buyers. Perhaps half of the new car
Telephone Interviews
buyers for a model are over age 55, but 75 per
cent of respondents to a web survey are under Advantages. The telephone interview is a pop
age 32 and only 8 percent are over age 55. Not ular survey method because about 95 percent of
only would the results be distorted by age but the population can be reached by telephone. An
the relatively small percentage of over-55 re interviewer calls a respondent (usually at home),
spondents may not be representative of all over asks questions, and records answers. Researchers
55 potential new car buyers (e.g., they may be sample respondents from lists, telephone direc
higher income or more educated). tories, or random digit dialing, and can quickly
A second concern is protecting respondent reach many people across long distances. A staff
privacy and confidentiality. Researchers should of interviewers can interview 1,500 respondents
encrypt collected data, only use secure websites across a nation within a few days and, with sev
and erase nonessential respondent identification eral callbacks, response rates can reach 90 per
or linking information on a daily or weeklybasis. cent. Although this method is more expensive
They should develop a system of respondent than a mail questionnaire, the telephone inter
verification to ensure that only the sampled re view is a flexible method with most of the
spondent participates and does so only once. strengths efface-to-face interviews but for about
This may involve a system such as giving each half the cost. Interviewers control the sequence
respondent a unique PIN number to access the ofquestions and can use some probes. A specific
questionnaire. respondent is chosen and is likely to answer all
A third concern involves the complexity of the questions alone. The researcher knows when
questionnaire design. Researchers need to check the questions were answered and can use con
and verify the compatibility ofvarious web soft tingency questions effectively, especially with
ware and hardware combinations for respon computer-assisted telephone interviewing (CATI)
dents using different computers. Researchers are (to be discussed).
190 PART TWO! CONDUCTING QUANTITATIVE RESEARCH
Disadvantages. Higher cost and limited inter roles, norms, and expectations. The interview is
view length are among the disadvantages of tele a short-term, secondary social interaction be TABLE 7.5
phone interviews. In addition, respondents tween two strangers with the explicit purpose of
without telephones are impossible to reach, and one person's obtaining specific information
the call may come at an inconvenient time. The from the other. The social roles are those of the
use of an interviewer reduces anonymity and in interviewer and the interviewee or respondent.
troduces potential interviewer bias. Open-ended Information is obtained in a structured conver
questions are difficult to use, and questions re sation in which the interviewer asks prearranged
quiring visual aids are impossible. Interviewers questions and records answers, and the respon
can only note serious disruptions (e.g., back dent answers. It differs in several ways from or
ground noise) and respondent tone of voice dinary conversation (see Table 7.5).
(e.g., anger or flippancy) or hesitancy. An important problem for interviewers is
that many respondents are unfamilar with the
survey respondents' role. As a result, they substi
Face-to-Face Interviews
tute another role that may affect their responses.
Advantages. Face-to-face interviews have the Some believe the interview is an intimate con
highest response rates and permit the longest versation or thearpy session, some see it as a bu
questionnaires. Interviewers also can observe the reaucratic exercise in completing forms, some
surroundings and can use nonverbal communi view it as a citizen referendum on policy choices,
cation and visual aids. Well-trained interviewers some view it as a testing situation, and some
can ask all types of questions, can ask complex consider it as a form of deceit in which inter
questions, and can use extensive probes. viewers are trying to trick or entrap respondents.
Even in a well-designed, professional survey, fol
Disadvantages. High cost is the biggest disad low-up research found that only about half the
vantage of face-to- face interviews. The training, respondents understand questions exactly as in
travel, supervision, and personnel costs for in tended by researchers. Respondents reinter
terviews can be high. Interviewer bias is also preted questions to make them applicable to
greatest in face-to-face interviews. The appear their ideosynactic, personal situations or to
ance, tone of voice, question wording, and so make them easy to answer. 16
forth of the interviewer may affect the respon The role of interviewers is difficult. They ob
dent. In addition, interviewer supervision is less tain cooperation and build rapport, yet remain
than for telephone interviews, which supervisors neutral and objective. They encroach on the re
monitor by listening in. 14 spondents' time and privacy for information
:i
that may not directly benefit the respondents.
They try to reduce embarrassment, fear, and 1
suspicion so that respondents feel comfortable
INTERVIEWING revealing information. They may explain the na litely redirects
ture of survey research or give hints about social such questions are
The Role of the Interviewer
roles in an interview. Good interviewers moni if a respondent
Interviews to gather information occur in many tor the pace and direction of the social interac interviewer rna. ~
settings. Survey research interviewing is a special tion as well as the content of answers and the ested in what you 111
ized kind of interviewing. As with most inter behavior of respondents. matter." Likewise. i
viewing,its goal is to obtain accurate information Survey interviewers are nonjudgmental and shocking answer i lIII
from another person.l'' do not reveal their opinions, verbally or nonver times for beating IIIJl
The survey interview is a social relationship. bally (e.g., by a look of shock). If a respondent ing her with cigardl
Like other social relationships, it involves social asks for an interviewer's opinion, he or she po- not show shock, SUI!
CHAPTER 7 / SURVEY RESEARCH 191
-----------------------I1t'
1. Questions and answers from each participant 1. Interviewer asks and respondent answers most
are relatively equally balanced. of the time.
2. There is an open exchange of feelings and 2. Only the respondent reveals feelings and
opinions. opinions.
3. Judgments are stated and attempts made to 3. Interviewer is nonjudgmental and does not try
persuade the other of a particular points of to change respondent's opinions or beliefs.
view. 4. Interviewer tries to obtain direct answers to
4. A person can reveal deep inner feelings to gain specific questions.
sympathy or as a therapeutic release. 5. Interviewer avoids making ritual responses that
5. Ritual responses are common (e.g., "Uh huh," influence a respondent and also seeks genuine
shaking head, "How are you?" "Fine"). answers, not ritual responses.
6. The participants exchange information and 6. Respondent provides almost all information.
correct the factual errors that they are aware Interviewer does not correct a respondent's
of. factual errors.
7. Topics rise and fall and either person can 7. Interviewer controls the topic, direction, and
introduce new topics. The focus can shift pace. He or she keeps the respondent "on task,"
directions or digress to less relevant issues. and irrelevant diversions are contained.
8. The emotional tone can shift from humor, to 8. Interviewer attempts to maintain a consistently
joy, to affection, to sadness, to anger, and so warm but serious and objective tone
on. throughout.
9. People can evade or ignore questions and give 9. Respondent should not evade questions and
flippant or noncommittal answers. should give truthful, thoughtful answers.
Source: Adapted from Gorden (1980:19-25) and Sudman and Bradburn (1983:5-10).
litely redirects the respondent and indicates that the answer in a matter-of-fact manner. He or she
such questions are inappropriate. For example, helps respondents feel that they can give any
if a respondent asks, "What do you think?" the truthful answer.
interviewer may answer, "Here, we are inter You might ask, "If the survey interviewer
ested in what you think; what I think doesn't must be neutral and objective, why not use a ro
matter." Likewise, if the respondent gives a bot or machine?" Machine interviewing has not
shocking answer (e.g., "I was arrested three been successful because it lacks the human
times for beating my infant daughter and burn warmth, sense oftrust, and rapport that an in
ing her with cigarettes"), the interviewer does terviewer creates. An interviewer helps define
not show shock, surprise, or disdain but treats the situation and ensures that respondents have
192 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
the information sought, understand what is ex she'll do when she gets older. Heart surgery is too
pected, give relevant answers, are motivated to risky for her and it costs so much. She'll have to
cooperate, and give serious answers. learn to live with it." If the interviewer writes,
Interviewers do more than interview re "concerned about daughter's health," much is
spondents. Face-to-face interviewers spend only lost. Interviewer Questiorr:
about 35 percent of their time interviewing. The interviewer knows how and when to
About 40 percent is spent in locating the correct use probes. A probe is a neutral request to clarify
respondent, 15 percent in traveling, and 10 per an ambiguous answer, to complete an incom
cent in studying survey materials and dealing plete answer, or to obtain a relevant response.
with administrative and recording details. I? Interviewers recognize an irrelevant or inaccu
rate answer and use probes as needed.l'' There
are many types of probes. A three- to five-second
Stages of an Interview
pause is often effective. Nonverbal communica
The interview proceeds through stages, begin tion (e.g., tilt of head, raised eyebrows, or eye
ning with an introduction and entry. The inter contact) also works well. The interviewer can re
viewer gets in the door, shows authorization, peat the question or repeat the reply and then
and reassures and secures cooperation from the pause. She or he can ask a neutral question, such
respondent. He or she is prepared for reactions as, "Any other reasons?" "Can you tell me more
such as, "How did you pick me?" "What good about that?" "How do you mean?" "Could you
will this do?" "I don't know about this," "What's explain more for me?" (see Box 7.7).
this about, anyway?" The interviewer can ex The last stage is the exit, when the inter Interviewer Q~
plain why the specific respondent is interviewed viewer thanks the respondent and leaves. He or the death penalty.
and not a substitute. she then goes to a quiet, private place to edit the posed to it?
The main part of the interview consists of questionnaire and record other details such as (Favor) 1 _ 2
asking questions and recording answers. The in the date, time, and place of the interview; a Respondent AnsMr.
terviewer uses the exact wording on the ques thumbnail sketch of the respondent and inter should get death t.t
tionnaire-no added or omitted words and no view situation; the respondent's attitude (e.g.,
rephrasing. He or she asks all applicable ques serious, angry, or laughing); and any unusual
tions in order, without returning to or skipping circumstances (e.g.,"Telephone rang at question
questions unless the directions specify this. He 27 and respondent talked for four minutes be
or she goes at a comfortable pace and gives fore the interview started again"). He or she
nondirective feedback to maintain interest. notes anything disruptive that happened during
In addition to asking questions, the inter the interview (e.g., "Teenage son entered room,
viewer accurately records answers. This is easy sat at opposite end, turned on television with the
for closed-ended questions, where interviewers volume loud, and watched a music video"). The
just mark the correct box. For open-ended ques interviewer also records personal feelings and
tions, the interviewer's job is more difficult. He anything that was suspected (e.g., "Respondent
or she listens carefully,must have legiblewriting, became nervous and fidgeted when questioned
and must record what is said verbatim without about his marriage").
correcting grammar or slang. More important,
the interviewer never summarizes or para curate, mature,
Training Interviewers
phrases. This causes a loss of information or dis ~t, stable, . . .
torts answers. For example, the respondent says, A large-scale survey requires hiring multiple in nonthreat~
"I'm really concerned about my daughter's heart terviewers. Few people appreciate the difficulty with manv dj"iImI-'f
problem. She's only 10years old and already she of the interviewer's job. A professional-quality poise and tact..
has trouble climbing stairs. I don't know what interview requires the careful selection of inter- ~-ers'ph~
CHAPTER 7 ! SURVEY RESEARCH 193
the death penalty, where 1 is strongly infavorof the death penalty. and 7 is strongly op
posed to it?
(Favor) 1 _ 2 3 4 5 6 7 (Oppose)
Respondent Answer: About a 4. I think that all murderers. rapists. and violent criminals
should get death, but I don't favor it for minor crimes like stealing a car.
viewers and extensive training. As with anyem guages spoken, and even the sound of their
ployment situation, adequate pay and good voice.
supervision are important for consistent high Professional interviewers will receive a two
quality performance. Unfortunately, profes week training course. It includes lectures and
sional interviewing has not always paid well or reading, observation of expert interviewers,
provided regular employment. In the past, inter mock interviews in the office and in the field that
viewers were largely drawn from a pool of mid are recorded and critiqued, many practice inter
dle-aged women willing to accept irregular views, and role-playing. The interviewers learn
part-time work. about survey research and the role of the inter
Good interviewers are pleasant, honest, ac viewer. They become familiar with the question
curate, mature, responsible, moderately intelli naire and the purpose of questions, although not
gent, stable, and motivated. They have a with the answers expected.
nonthreatening appearance, have experience The importance of carefully selecting and
with many different types ofpeople, and possess training interviewers was evident during the
poise and tact. Researchers may consider inter 2004 U.S. presidential election. Exit polls are
viewers' physical appearance, age, race, sex, lan quick, very short surveys conducted outside a
194 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
polling place for people immediately after they not vary from what they would be if asked by
voted. On Election Day of 2004 exit polls any other interviewer.
showed candidate John Kerry well ahead, but Survey researchers know that interviewer
after final votes were counted he lost to his op expectations can create significant bias. Inter
ponent, George W. Bush. A major cause of the viewers who expect difficult interviews have
mistake was that the research organization, paid them, and those who expect certain answers are
$10 million by six major news organizations to more likely to get them (see Box 7.8). Proper in
conduct the exit polls, had hired many young in terviewer behavior and exact question reading
experienced interviewers and gave them only may be difficult, but the issue is larger.
minimal training. Younger voters tended to sup The social setting in which the interview oc
port John Kerry, whereas older voters tended to curs can affect answers, including the presence
support George Bush. The young inexperienced of other people. For example, students answer
interviewers were less successful in gaining co differently depending on whether they are asked
operation from older voters and felt more com questions at home or at school. In general, sur
fortable handing the questionnaire to someone vey researchers do not want others present be
of a similar age. As a result, exit poll participants cause they may affect respondent answers. It
Interviewer
did not reflect the composition of all voters and may not always make a difference, however, es
poll results showed greater support for Kerry pecially if the others are small children.r" Looked Jewish . . .
than actually existed among all voters.l? An interviewer's visible characteristics, in
Looked Jewish 0IIIj
Although interviewers largely work alone, cluding race and gender, often affect interviews
researchers use an interviewer supervisor in and respondent answers, especially for questions Non-Jewish apJ:JeiI ..
large-scale surveys with several interviewers. Su about issues related to race or gender. For exam Non-Jewish apJ:JeilI 1I
pervisors are familiar with the area, assist with ple, African American and Hispanic American non-Jewi!Sll--5DI__
problems, oversee the interviewers, and ensure respondents express different policy positions
that work is completed on time. For telephone on race- or ethnic-related issues depending on
interviewing, this includes helping with calls, the apparent race or ethnicity of the interviewer.
checking when interviewers arrive and leave, and This occurs even with telephone interviews
monitoring interview calls. In face-to-face inter when a respondent has clues about the inter
views, supervisors check to find out whether the viewer's race or ethnicity. In general, interview
interview actually took place. This means calling ers of the same ethnic-racial group get more
back or sending a confirmation postcard to a accurate answers.I! Gender also affects inter
sample of respondents. They can also check the views both in terms of obvious issues, such as and Illicropboc.
response rate and incomplete questionnaires to sexual behavior, as well as support for gender tions from a CDI",.
see whether interviewers are obtaining coopera related collective action or gender equality.22 if'Ondent wile is
tion, and they may reinterview a small subsam
ple, analyze answers, or observe interviews to see
Survey researchers need to note the race and
gender of both interviewers and respondents.
'ria the~ II"
ilol-er, the COI. . .rt
whether interviewers are accurately asking ques ~~1een.
tions and recording answers. Computer-Assisted Telephone
Compusee
Interviewing ~ 10-rc:nil~"
~Italsoet.......
Interviewer Bias ~ infonnom..
Advances in computer technology and lower
Survey researchers proscribe interviewer behav computer prices have enabled professional sur a.uP£()C~
ior to reduce bias. This goes beyond reading vey research organizations to install computer ~tina....1
each question exactly as worded. Ideally, the ac assisted telephone interviewing(CATI) systems.P ~~ol
tions of a particular interviewer will not affect With CATI, the interviewer sits in front of a ~tOr~. . .
bow it respondent answers, and responses will computer and makes calls. Wearing a headset
CHAPTER 7 / SURVEY RESEARCH 195
Note: Racialstereotypes held by respondents can affect how they respond in interviews.
Source: Adapted from Hyman (1975:115, 163).
and microphone, the interviewer reads the ques for a specific respondent; interviewers do not
tions from a computer screen for the specific re have to turn pages looking for the next question.
spondent who is called, then enters the answer In addition, the computer can check an answer
via the keyboard. Once he or she enters an an immediately after the interviewer enters it. For
swer, the computer shows the next question on example, if an interviewer enters an answer that
the screen. is impossible or clearly an error (e.g., an H
Computer-assisted telephone interviewing instead of an M for "Male"), the computer will
speeds interviewing and reduces interviewer er request another answer. Innovations with com
rors. It also eliminates the separate step of enter puters and web surveys also help to gather data
ing information into a computer and speeds on sensitive issue (see Box 7.9).
data processing. Of course, CATI requires an in Several companies have developed software
vestment in computer equipment and some programs for personal computers that help re
knowledge of computers. The CATI system is searchers develop questionnaires and analyze
valuable for contingency questions because the survey data. They provide guides for writing
computer can show the questions appropriate questions, recording responses, analyzing data,
196 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
process of asking many people the same ques Respondents who are highly loyal to one brand of
tions and examining their answers. traditional carbonated sodas can answer the ques
Survey researchers try to minimize errors, tion easily.Other respondents must implicitly ad
but survey data often contain them. Errors in dress the following questions to answer the
surveys can compound each other. For example, question as it was asked: (a) What time period is
involved-the past month, the past year, the last
errors can arise in sampling frames, from nonre
10 years? (b) What conditions count-at home,
sponse, from question wording or order, and
at restaurants, at sporting events? (c) Buying for
from interviewer bias. Do not let the existence of oneself alone or for other family members? (d)
errors discourage you from using the survey, What is a "soft drink"? Do lemonade, iced tea,
however. Instead, learn to be very careful when mineral water, or fruit juices count? (e) Does
designing survey research and cautious about "usually" mean a brand purchased as 51 percent
generalizing from the results of surveys. or more of all soft drink purchases, or the brand
purchased more frequently than any other? Re
spondents rarely stop and ask for clarification;
they make assumptions about what the researcher
Key Terms
means.
2. See Dykema and Schaeffer (2000) and Sudman
closed-ended question and colleagues (1996:197-226).
computer-assisted telephone interviewing 3. SeeOstrom and Gannon (1996).
(CAT!) 4. See Bradburn (1983), Bradburn and Sudman
context effect (1980), and Sudman and Bradburn (1983) on
contingency question threatening or sensitive questions. Backstrom and
cover sheet Hursh-Cesar (1981 :219) and Warwick and
double-barreled question Lininger (1975:150-151) provide useful sugges
floaters tions as well.
full-filter question 5. On how "Who knows who lives here?" can be
complicated, see Martin (1999) and Tourangeau
funnel sequence
et al. (1997).
interview schedule
6. For a discussion of the "don't know," "no opin
matrix question ion," and middle positions in response categories,
open-ended question see Backstrom and Hursh-Cesar (1981:148-149),
order effects Bishop (1987), Bradburn and Sudman (1988:
partially open question 154), Brody (1986), Converse and Presser
prestige bias (1986:35-37), Duncan and Stenbeck (1988), and
probe Sudman and Bradburn (1983:140-141).
quasi-filter question 7. The disagree/agree versus specificalternatives de
response set bate can be found in Bradburn and Sudman
social desirability bias (1988:149-151), Converse and Presser (1986:38
39), and Schuman and Presser (1981:179-223).
standard-format question
8. The ranking versus ratings issue is discussedin Al
threatening questions
win and Krosnick (1985) and Krosnick and Alwin
wording effects (1988). Also see Backstrom and Hursh-Cesar
(1981:132-134) and Sudman and Bradburn
(1983:156-165) for formats of asking rating and
Endnotes ranking questions.
9. See Foddy (1993) and Presser (1990).
1. Sudman and Bradburn (1983:39) suggested that 10. Studies by Krosnick (1992) and Narayan and
even simple questions (e.g., "What brand of soft Krosnick (1996) show that education reduces re
drink do you usually buy?") can cause problems. sponse-order (primacy or recency) effects, but
CHAPTER 7 / SURVEY RESEARCH 199
Knauper (1999) found that age is strongly associ 17. From Moser and Kalton (1972:273).
ated with response-order effects. 18. The use of probes is discussed in Backstrom and
11. This example comes from Strack (1992). Hursh-Cesar (1981:266--273), Gorden (1980:368
12. For a discussion, see Couper, Singer et al. (1998), 390), and Hyman (1975:236--241).
de Heer (1999), Keeter et al. (2000), Sudman and 19. Report by Jacques Steinberg (2005). "Study Cites
Bradburn (1983: 11), and "Surveys Proliferate, but Human Failings in Election Day Poll System,"
Answers Dwindle," New York Times, October 5, New York Times (January 20,2005).
1990, p. 1. Smith (1995) and Sudman (1976:114 20. See Bradburn and Sudman (1980), Pollner and
116) also discuss refusal rates. Adams (1997), and Zane and Matsoukas (1979).
13. Bailey (1987:153-168), Church (1993), Dillman 21. The race or ethnicity of interviewers is discussed
(1978, 1983), Fox and colleagues (1988), Goyder in Anderson and colleagues (1988), Bradburn
(1982), Heberlein and Baumgartner (1978,1981), (1983), Cotter and colleagues (1982), Davis
Hubbard and Little (1988), Jones (1979), and (1997), Finkel and colleagues (1991), Gorden
Willimack and colleagues (1995) discuss increas (1980:168-172), Reese and colleagues (1986),
ing return rates in surveys Schaffer (1980), Schuman and Converse (1971),
14. For a comparison among types of surveys, see and Weeks and Moore (1981).
Backstrom and Hursh-Cesar (1981 :16-23), Brad 22. See Catania and associates (1996) and Kane and
burn and Sudman (1988:94-110), Dillman (1978: MacAulay (1993).
39-78), Fowler (1984:61-73), and Frey (1983:27 23. CAT! is discussed in Bailey (1987:201-202), Brad
55). burn and Sudman (1988:100-101), Frey (1983:
15. For more on survey research interviewing, see 24-25, 143-149), Groves and Kahn (1979:226),
Brenner and colleagues (1985), Cannell and Kahn Groves and Mathiowetz (1984), and Karweit and
(1968), Converse and Schuman (1974), Dijkstra Meyers (1983).
and van der Zouwen (1982), Foddy (1993), Gor 24. On reporting survey results in the media, see
den (1980), Hyman (1975), and Moser and Channels (1993) and MacKeun (1984).
Kalton (1972:270-302). 25. See Singer (1988).
16. See Turner and Martin (1984:262-269,282). 26. From Turner and Martin (1984:62).
INTltODUC.
~"W'ft
~-
-
.
•
at a r- - -:zri:a:
Experimental Research
theOlba- ..
~Ki- !iI
JonM 1154
iug. mil
rondnaapalil
Introduction
Research Questions Appropriate for an Experiment
Random Assignment
Why Randomly Assign?
How to Randomly Assign
Matching versus Random Assignment
Experimental Design logic
The Language of Experiments
Types of Design
Design Notation
Internal and External Validity
The Logic of Internal Validity
Threats to Internal Validity
External Validity and Field Experiments
Practical Considerations
Planning and Pilot-Tests
Instructions to Subjects
Postexperiment Interview
Results of Experimental Research: Making Comparisons
A Word on Ethics
Conclusion
200
CHAPTER 8 / EXPERIMENTAL RESEARCH 201
thousands to attend college and prevent others mental method. For example, a researcher may
from attending to discover who later earns more wish to study attitudes toward people in wheel
income. He or she cannot induce couples to chairs. An experimenter might ask people to re
have either many children or a single child so he spond (e.g., Would you hire this person? How
or she can examine how leadership skills develop comfortable would you be if this person asked
in children. He or she cannot compel people to you for a date?) to photos of some people in
join or quit organizations then see whether they wheelchairs and some people not in wheelchairs.
vote. Experimenters are highly creative in simu A survey researcher might ask people their opin
lating such interventions or conditions, but they ions about people in wheelchairs. The field re
cannot manipulate many of the variables of in searcher might observe people's reactions to
terest to fit the pure experimental logic. someone in a wheelchair, or the researcher him
The experiment is usually best for issues self or herself might be in wheelchair and care
that have a narrow scope or scale. This strength fully note the reactions of others.
allows experimenters to assemble and "run"
many experiments with limited resources in a
short period. Some carefully designed experi
RANDOM ASSIGNMENT
ments require assembling only 50 or 60 volun
teers and can be completed in one or two Social researchers frequently want to compare.
months. In general, the experiment is better For example, a researcher has two groups of 15
suited for micro-level (e.g., individual or small students and wants to compare the groups on
group phenomena) than for macro-level con the basis of a key difference between them (e.g.,
cerns or questions. Experiments can rarely a course that one group completed). Or a re
address questions that require looking at condi searcher has five groups of customers and wants
tions across an entire society or across decades. to compare the groups on the basis of one char
The experiment also limits one's ability to gen acteristic (e.g., geographic location). The cliche,
eralize to larger settings (see External Validity "Compare apples to apples, don't compare ap
and Field Experiments later in this chapter). ples to oranges," is not about fruit; it is about
Experiments encourage researchers to iso comparisons. It means that a valid comparison
late and target the impact that arises from one or depends on comparing things that are funda
a few causal variables. This strength in demon mentally alike. Random assignment facilitates
strating causal effects is a limitation in situations comparison in experiments by creating similar
where a researcher tries to examine numerous groups.
variables simultaneously. The experiment is When making comparisons, researchers
rarely appropriate for research questions or is want to compare casesthat do not differ with re
sues that require a researcher to examine the im gard to variables that offer alternative explana
pact of dozens of diverse variables all together. tions. For example, a researcher compares two
Although the accumulated knowledge from groups of students to determine the impact of
many individual experiments, each focused on completing a course. In order to be compared,
one or two variables, advances understanding, the two groups must be similar in most respects
the approach of expermimental research differs except for taking the course. If the group that
from doing research on a highly complex situa completed the course is also older than the
tion in which one examines how dozens of vari group that did not, for example, the researcher - . ~.-1111111
ables operate simultaneously. cannot determine whether completing the
5
, F
&
Often, researchers study closely related top course or being older accounts for differences , fill
___ 1I11al.
ics using either an experimental or a nonexperi between the groups. _.~
CHAPTER 8 / EXPERIMENTAL RESEARCH 203
males in the first group are only children; one is have all these I
EXPERIMENTAL DESIGN LOGIC
from a divorced family, one from an intact fam parts plus otbes
ily. One is tall, slender, and Jewish; the other is The Language of Experiments cussed here, InlI
short, heavy, and Methodist. In order to match
groups, does the researcher have to find a tall Experimental research has its own language or
set of terms and concepts. You already encoun 1. Treatment I
Jewish male only child from a divorced home 2. Dependent
and a short Methodist male only child from an tered the basic ideas: random assignment and
independent and dependent variables. In exper 3. Pretest
intact home? The tall, slender, Jewish male only 4. Posttest
imental research, the cases or people used in re
child is 22 years old and is studying to become a 5. Experimers
search projects and on whom variables are
physician. The short, heavy Methodist male is 20 6. Control gJ1
measured are called the subjects.
years old and wants to be an accountant. Does 7. Random ail
the researcher also need to match the age and ca
reer aspirations of the two males? True matching Parts ofthe Experiment. We can divide the ex In most eq
soon becomes an impossible task. periment into seven parts. Not all experiments situation or enI
-----~-----------------
FIGURE 8.1 Random Assignment and Random Sampling
Random Sampling
.. then modifies iI
or manipulatiol
fies. The term a
physician admi
the physician inI
logical conditio
dent variable 01
Population (Sampling Frame) variables. In earl
researcher d~
Random
ment or indicas
Process
applied it to a pi
searchers "meal
creating a cood
the independes
anxiety"; the If''lll
Random Assignment stead of asking I
experimenters •
Step 1: Begin with a collection of subjects. fear or a low-fell
dependent varil
so that some sui
feel little.
Researchers
treatments. SoIlli
Step 2: Devise a method to randomize that is purely mechanical (e.g., flip a coin).
groups of subioj
Step 3: Assign subjects with "Heads" to one group and ''Tails'' to the other group. can be as compl
tions with elabo
~ settings, or conti
have all these parts, and some have all seven cial behaviors, attitudes, feelings, or beliefs of
parts plus others. The following seven, to be dis subjects that change in response to a treatment.
cussed here, make up a true experiment: Dependent variables can be measured by paper
and-pencil indicators, observation, interviews,
1. Treatment or independent variable or physiological responses (e.g., heartbeat or
2. Dependent variable sweating palms).
3. Pretest Frequently, a researcher measures the de
4. Posttest pendent variable more than once during an ex
5. Experimental group periment. The pretest is the measurement of the
6. Control group dependent variable prior to introduction of the
7. Random assignment treatment. The posttest is the measurement of
the dependent variable after the treatment has
In most experiments, a researcher creates a been introduced into the experimental situation.
situation or enters into an ongoing situation, Experimental researchers often divide sub
then modifies it. The treatment (or the stimulus jects into two or more groups for purposes of
or manipulation) is what the researcher modi comparison. A simple experiment has two
fies. The term comes from medicine, in which a groups, only one of which receives the treat
physician administers a treatment to patients; ment. The experimentalgroup is the group that
the physician intervenes in a physical or psycho receives the treatment or in which the treatment
logical condition to change it. It is the indepen is present. The group that does not receive the
dent variable or a combination of independent treatment is called the control group. When the
variables. In earlier examples of measurement, a independent variable takes on many different
researcher developed a measurement instru values, more than one experimental group is
ment or indicator (e.g., a survey question), then used.
applied it to a person or case. In experiments, re We can review the variables in the three ex
searchers "measure" independent variables by periments used as examples in previous chap
creating a condition or situation. For example, ters. In Chapter 2 you read about an experiment
the independent variable is "degree of fear or by Brase and Richmond (2004) about doctor
anxiety"; the levelsare high fear and low fear. In patient interactions and perceptions. After ran
stead of asking subjects whether they are fearful, dom assignment, subjects saw same- and oppo
experimenters put subjects into either a high site-gender models identified as being medical
fear or a low-fear situation. They measure the in doctors but who wore either informal or for
dependent variable by manipulating conditions mal/traditional attire (independent variable).
so that some subjects feel a lot of fear and others The experimenters then measured the subjects'
feel little. judgments about trust in the physican and the
Researchers go to great lengths to create physician's abilities (dependent variable). In
treatments. Some are as minor as givingdifferent Goar and Sell's (2005) experiment about mixed
groups of subjects different instructions. Others race task groups described in Chapter 4, ran
can be as complex as putting subjects into situa domly assigned three-person groups were told
tions with elaborate equipment, staged physical they were either to a complete complex task re
settings, or contrived social situations to manip quiring diverse skills or not (independent vari
ulate what the subjects see or feel. Researchers able). The experimenters measured the time it
want the treatment to have an impact and pro took the group to complete a task and involve
duce specific reactions, feelings, or behaviors. ment by group members of different races (de
Dependent variables or outcomes in experi pendent variable). In the study on college
mental research are the physical conditions, so women with tattoos discussed in Chapter 5 by
206 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
Experimental researchers use deception to for experiments you conduct or create your own
control the experimental setting. Deception oc variations.
curs when the researcher intentionally misleads The designs are illustrated with a simple ex
subjects through written or verbal instructions, ample. A researcher wants to learn whether wait
the actions of others, or aspects of the setting. It staff (waiters and waitresses) receive more in tips
may involve the use of confederates or stooges if they first introduce themselves by first name
people who pretend to be other subjects or and return to ask "Is everything fine?" 8 to 10
bystanders but who actually work for the re minutes after delivering the food. The dependent
searcher and deliberately mislead subjects. variable is the size of the tip received. The study
Through deception, the researcher tries to con occurs in two identical restaurants on different
trol what the subjects see and hear and what they sides of a town that have had the same types of
believe is occurring. For example, a researcher's customers and average the same amount in tips.
instructions falsely lead subjects to believe that
they are participating in a study about group co ClassicalExperimental Design. All designs are
operation. In fact, the experiment is about variations of the classical experimental design, the
male/female verbal interaction, and what sub type of design discussed so far, which has ran
jects say is being secretly tape recorded. Decep dom assignment, a pretest and a posttest, an ex
tion lets the researcher control the subjects' perimental group, and a control group.
definition of the situation. It prevents them from Example. The experimenter gives 40 newly
altering their cross-sex verbal behavior because hired wait staff an identical two-hour training
they are unaware of the true research topic. By session and instructs them to follow a script in
focusing their attention on a false topic, the re which they are not to introduce themselves by
searcher induces the unaware subjects to act first name and not to return during the meal to
"naturally." For realistic deception, researchers check on the customers. They are next randomly
may invent falsetreatments and dependent vari divided into two equal groups of 20 and sent to
able measures to keep subjects unaware of the the two restaurants to begin employment. The
true ones. The use of deception in experiments experimenter records the amount in tips for
raises ethical issues (to be discussed). all subjects for one month (pretest score). Next,
the experimenter "retrains" the 20 subjects at
Types of Design restaurant 1 (experimental group). The experi
menter instructs them henceforth to introduce
Researchers combine parts of an experiment themselves to customers by first name and to
(e.g., pretests, control groups, etc.) together into check on the customers, asking, "Is everything
an experimental design. For example, some de fine?" 8 to 10 minutes after delivering the food
signs lack pretests, some do not have control (treatment). The group at restaurant 2 (control
groups, and others have many experimental group) is "retained" to continue without an in
groups. Certain widely used standard designs troduction or checking during the meal. Over
have names. the second month, the amount of tips for both
You should learn the standard designs for groups is recorded (posttest score).
two reasons. First, in research reports, re
searchers give the name of a standard design in Preexperimental Designs. Some designs lack
stead of describing it. When reading reports, you random assignment and are compromises or
will be able to understand the design of the ex shortcuts. These preexperimental designs are used
periment if you know the standard designs. Sec in situations where it is difficult to use the classi
ond, the standard designs illustrate common cal design. They have weaknesses that make in
ways to combine design parts. You can use them ferring a causal relationship more difficult.
208 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
One-Shot Case Study Design. Also called the Static Group Comparison. Also called the
one-group posttest-only design, the one-shot posttest-only nonequivalent group design, static TABLE 8.1 I
case study design has only one group, a treat group comparison has two groups, a posttest, and
ment, and a posttest. Because there is only one treatment. It lacks random assignment and a
. ....
~
group, there is no random assignment. pretest. A weakness is that any posttest outcome '"'.;;
difference between the groups could be due to
Example. The experimenter takes a group of 40 group differences prior to the experiment in Classical !
newly hired wait staff and gives all a two-hour stead of to the treatment. One-Shot~
training session in which they are instructed to One-~~
introduce themselves to customers by first name Example. The experimenter gives 40 newly
and to check on the customers, asking, "Is every hired wait staff an identical two-hour training Static~<1
thing fine?" 8 to 10 minutes after delivering the session and instructs them to follow a script in Two-~~
food (treatment). All subjects begin employ which they are not to introduce themselves by Time Series ~
ment, and the experimenter records the amount first name and not to return during the meal to
in tips for all subjects for one month (posttest check on the customers. They can choose one of j
score). the two restaurants to work at, so long as each l
restaurant ends up with 20 people. Allbegin em
One-Group Pretest-Posttest Design. This design ployment. After one month, the experimenter
has one group, a pretest, a treatment, and a "retrains" the 20 subjects at restaurant 1 (exper
posttest. It lacks a control group and random as imental group). The experimenter instructs
signment. them henceforth to introduce themselves to cus
tomers by first name and to check on the cus
Example. The experimenter takes a group of 40 tomers, asking, "Is everything fine?" 8 to 10
newly hired wait staff and gives all a two-hour minutes after delivering the food (treatment).
training session. They are instructed to follow a The group at restaurant 2 (control group) is "re
script in which they are not to introduce them tained" to continue without an introduction or
selvesby first name and not to return during the checking during the meal. Over the second
meal to check on the customers. All begin em month, the amount of tips for both groups is
ployment, and the experimenter records the recorded (posttest score).
amount in tips for all subjects for one month
(pretest score). Next, the experimenter "re Quasi-Experimental and Special Designs.
trains" all 40 subjects (experimental group). The These designs, like the classical design, make
experimenter instructs the subjects henceforth identifying a causal relationship more certain
to introduce themselves to customers by first than do preexperimental designs. Quasi-experi
name and to check on the customers, asking, "Is mental designs help researchers test for causal re
everything fine?" 8 to 10 minutes after delivering lationships in a variety of situations where the
the food (treatment). Over the second month, classical design is difficult or inappropriate. They
the amount of tips is recorded (posttest score). are called quasibecause they are variations of the
This is an improvement over the one-shot classical experimental design. Some have ran
case study because the researcher measures the domization but lack a pretest, some use more
dependent variable both before and after the than two groups, and others substitute many ob
treatment. But it lacks a control group. The re servations of one group over time for a control
searcher cannot know whether something other group. In general, the researcher has less control
than the treatment occurred between the pretest over the independent variable than in the classi
and the posttest to cause the outcome. cal design (see Table 8.1).
CHAPTER 8 I EXPERIMENTAL RESEARCH 209
----------------------1..
TAB L E 8.1 A Comparison of the Classical Experimental Design with Other Major Designs
Two-Group Posttest-Only Design. This is iden to act the same toward all customers. The results
tical to the static group comparison, with one showed that higher tips came from customers
exception: The groups are randomly assigned. It who received the message about upcoming
has all the parts of the classical design except a specials.
pretest. The random assignment reduces the
chance that the groups differed before the treat Interrupted Time Series. In an interrupted time
ment, but without a pretest, a researcher cannot series design, a researcher uses one group and
be as certain that the groups began the same on makes multiple pretest measures before and af
the dependent variable. ter the treatment. For example, after remaining
In a study using a two-group posttest-only level for many years, in 1995, cigarette taxes
design with random assignment, Rind and jumped 35 percent. Taxes remained relatively
Strohmetz (1999) examined messages about a constant for the next 10 years. The hypothesis is
upcoming special written on the back of cus that increases in taxes lower cigarette consump
tomers' checks. The subjects were 81 dining par tion. A researcher plots the rate of cigarette con
ties eating at an upscale restaurant in New Jersey. sumption for 1985 through 2005. The researcher
The treatment was whether a female server notes that cigarette consumption was level dur
wrote a message about an upcoming restaurant ing the 10 years prior to the new taxes, then
special on the back of a check and the dependent dropped in 1995 and stayed about the same for
variable was the size of tips. The server with two the next 10 years.
years' experience was given a randomly shuffled
stack of cards, half of which said No Message Equivalent Time Series. An equivalent time se
and half of which said Message. Just before she ries is another one-group design that extends
gave a customer his or her check, she randomly over a time period. Instead of one treatment, it
pulled a card from her pocket. If it said Message, has a pretest, then a treatment and posttest, then
she wrote about an upcoming special on the treatment and posttest, then treatment and
back of the customer's check. If it said No Mes posttest, and so on. For example, people who
sage, she wrote nothing. The experimenters drive motorcycles were not required to wear hel
recorded the amount of the tip and the number mets before 1985,when a law was passed requir
of people at the table. They instructed the server ing helmets. In 1991, the law was repealed
210 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
because of pressure from motorcycle clubs. The four groups. For example, a mental health
helmet law was reinstated in 2003. The re worker wants to determine whether a new train FIGURE 8.2
searcher's hypothesis is that wearing protective ing method improves clients' coping skills. The
helmets lowers the number ofhead injury deaths worker measures coping skills with a 20-minute 3.25 r--
E 3.00 1---
in motorcycle accidents. The researcher plots test of reactions to stressful events. Because the ti 2.75 1---
head injury death rates in motorcycle accidents clients might learn coping skills from taking the ;;: 2.50 1---
over time. The rate was very high prior to 1985, test itself, a Solomon four-group design is used. 8-
«l 2.25 1---
dropped sharply between 1985 and 1991, then The mental health worker randomly divides a: 2.00
returned to pre-1985 levels between 1991 and clients into four groups. Two groups receive the ~ 1.75 1---
2003, then dropped again from 2003 to the pretest; one of them gets the new training
iii 1.50 1---
1.25 1---
present. method and the other gets the old method. An 1.00 L...-_ _
---------------------.
FIG U R E 8. 2 Blame, Resistance, and Schema: Interaction Effect
3.25 -------------------------
E 3.00
~ 2.75
:> 2.50 -------'----.::--------------
.. .....- - - - -
Q)
g. 2.25 -------------="'-~.
a:
~
2.00
1.75
-------------_._-'""'""-...-.~ '-"-"-~--
~"
~------=
.
~ 150
•.......""""""................
1.25
1.00
Fights Submits
Victim Tries to Fight off the Rapist (Resistance)
........- Sex schema
••••• Power schema
Interaction effects are illustrated in Figure ing of the crime of rape (i.e., the rape schema
8.2, which uses data from a study by Ong and held by each subject). The researchers found that
Ward (1999). As part of a study of 128 female two rape schemas caused subjects to interpret
undergraduates at the National University of victim resistance in opposite ways for the pur
Singapore, Ong and Ward measured which of pose of assigning responsibility for the crime.
two major ways subjects understood the crime Researchers discuss factorial design in a
of rape. Some of the women primarily under shorthand way.A "two by three factorial design"
stood it as sex and due to the male sex drive (sex is written 2 X 3. It means that there are two
schema); others understood it as primarily an treatments, with two categories in one and three
act of male power and domination of a woman categories in the other. A 2 X 3 X 3 design
(power schema). The researchers asked the sub means that there are three independent vari
jects to read a realistic scenario about the rape of ables, one with two categories and two with
a college student at their university. One ran three categories each.
domly selected group of subjects read a scenario The previously discussed experiment by
in which the victim tried to fight off the rapist. In Hawkes, Seen, and Thorn (2004) on tattoos
the other set, she passively submitted. The re among college women used a 3 X 2 X 2 X 2 X 2
searchers next asked the subjects to evaluate the factorial design. The full study considered four
degree to which the rape victim was at blame or independent variables, one with three categories,
responsible for the rape. the rest having two categories, and it had three
Results showed that the women who held measures of the dependent variable. The depen
the sex schema (and who also tended to embrace dent variable measures included a Semantic
traditionalist gender role beliefs) more strongly Differential measure (which contained three di
blamed the victim when she resisted. Blame de mensions). In addition, experimenters had
creased if she submitted. The women who held a subjects complete a Neosexism measure (an 11
power schema (and who also tended to be non item, 5-point Likert Scale statements summed
traditionalists) were less likely to blame the vic into an index) and a measure of Feminism and
tim if she fought. They blamed her more if she Women's Movement Support (a 10-item, 5
passively submitted. Thus, the subjects' re point Likert Scale summed into an index). The
sponses to the victim's act of resisting the attack experimenters manipulated two independent
varied by, or interacted with, their understand- variables in the descriptions of the tattoo read by
212 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
subjects: (1) whether the woman had no tattoo, ment, independent variable; R == random assign
a tattoo smaller than a Canadian $1 coin, or ment. The Os are numbered with subscripts TABLE '.2
larger than a $1 coin; and (2) the tattoo's visib
lity as always hidden versus always hidden. Two
independent variables were not manipulated but
from left to right based on time order. Pretests
are 01> posttests 02' When the independent
variable has more than two levels, the Xs are
......~ ~
C1assicaJ~
were preexisting characteristics of researcher numbered with subscripts to distinguish among
subjects, (3) whether the subject him/herselfhad them. Symbols are in time order from left to I
a tattoo or not, and (4) the subject's gender. The right. The R is first, followed by the pretest, the Preexpet-""'" ~
study included 268 subjects, 122 males and 146 treatment, and then the posttest. Symbols are
females, of them 43 (or 16 percent) had a tattoo. arranged in rows, with each row representing a One-shot Q5r ~
The study results showed that subjects group of subjects. For example, an experiment One-groep ~
viewed college women without a tattoo more with three groups has an R (if random assign Static~~
positivity and female subjects were more posi ment is used), followed by three rows of Os and ~
~
tive toward a college woman having a tattoo Xs. The rows are on top of each other because
than male subjects. There was also a significant the pretests, treatment, and posttest occur in
effect for visibility, with more favorable attitudes each group at about the same time. Table 8.2 TWo-~~
for a nonvisible tattoo. Generally, subjects who gives the notation for many standard experi
had tattoos themselves were more favorable to mental designs. Interrupted 1.-r
ward the woman having a tattoo. Size of tattoo Equivalent n.r
had little effect. Men and women with a tattoo LatinSCjUiR
were more favorable, regardless of tattoo size,
while those without a tattoo were negative. In INTERNAL AND EXTERNAL
addition, gender made no difference toward size VALIDITY
of tattoo. The experiment had many specific
The Logic of Internal Validity
findings for each combination of the five inde
pendent variables. One specific finding was that Internal validity means the ability to eliminate al
female subjects who had a tattoo themselves ternative explanations of the dependent variable.
were least likely to react negatively to a larger tat Variables, other than the treatment, that affect
too. Results from the attitude measures suggest the dependent variable are threats to internal va
that "the tattooed woman may be seen by some lidity. They threaten the researcher's ability to say
Factorial~
as flaunting her freedom from gender norms or that the treatment was the true causal factor pro
as threatening women's traditional place in soci ducing change in the dependent variable. Thus,
ety" (Hawkes, Seen, and Thorn 2004:603). the logic of internal validity is to rule out vari
ables other than the treatment by controlling ex
perimental conditions and through experimental
Design Notation
designs. Next, we examine major threats to inter
Experiments can be designed in many ways. nal validity.
Design notation is a shorthand system for sym dom~gDD"~
bolizing the parts of experimental design. Once experime:obI
Threats to Internal Validity
you learn design notation, you will find it easier th~~dq::~'"
to think about and compare designs. For exam The following are nine common threats to inter experiment a.
ple, design notation expresses a complex, para nal validity. 1 mentgIOOF
graph-long description of the parts of an ...i to are k'JOtiIMllI
experiment in five or six symbols arranged in Selection Bias. Selection bias is the threat that ...-hereas the
two lines. It uses the following symbols:
observation of dependent variable; X == treat
°
== research participants will not form equivalent
groups. It is a problem in designs without ran-
cians, chess
ample is an
CHAPTER 8 I EXPERIMENTAL RESEARCH 213
----------------~.,
TAB L E 8.2 Summary of Experimental Designs with Notation
R~g
Classical experimental design X 0
0
Preexperimental Designs
One-shot case study X 0
One-group pretest-posttest 0 X 0
Static group comparison X 0
0
Quasi-Experimental Designs
Two-group posttest only R -c:::::. X 0
0
Interrupted time series 0 0 0 0 X 000
Equivalent time series 0 X 0 X OXOXO
~~
Latin square designs Xa 0 Xb 0 XcO
Xb 0 Xa 0 XcO
Xc 0 Xb 0 XaO
Xa 0 Xc 0 XbO
Xb 0 Xc 0 XaO
0 Xc 0 Xa 0 XbO
R~g
Solomon four-group design X 0
0
X 0
0
~X1X1
Factorial designs Z1 0
R Z2 0
X2 Z1 0
X2 Z2 0
dam assignment. It occurs when subjects in one to dodge heavy traffic. All subjects assigned to
experimental group have a characteristic that af one group come from rural areas, and all sub
fects the dependent variable. For example, in an jects in the other grew up in large cities. An ex
experiment on physical aggressiveness,the treat amination of pretest scores helps a researcher
ment group unintentionally contains subjects detect this threat, because no group differences
who are football, rugby, and hockey players, are expected.
whereas the control group is made up of musi
cians, chess players, and painters. Another ex History. This is the threat that an event unre
ample is an experiment on the ability of people lated to the treatment will occur during the ex
214 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
periment and influence the dependent variable. effects occur, a researcher cannot say that the
History effects are more likelyin experiments that treatment alone has affected the dependent
continue over a long time period. For example, variable.
halfway through a two-week experiment to eval
uate subjects' attitudes toward space travel, a Instrumentation. This threat is related to reli
spacecraft explodes on the launch pad, killing ability. It occurs when the instrument or depen
the astronauts. The history effect can occur in dent variable measure changes during the
the cigarette tax example discussed earlier (see experiment. For example, in a weight-loss ex
the discussion of interrupted time-series de periment, the springs on the scaleweaken during
sign). If a public antismoking campaign or re the experiment, giving lower readings in the
duced cigarette advertising also began in 1989, it posttest. Another example might have occurred
would be hard to say that higher taxes caused in an experiment by Bond and Anderson (1987)
less smoking. on the reluctance to transmit bad news. The ex
perimenters asked subjects to tell another per
Maturation. This is the threat that some bio son the results of an intelligence test and varied
logical, psychological, or emotional process the test results to be either well above or well be
within the subjects and separate from the treat low average. The dependent variable was the
ment will change over time. Maturation is more length of time it took to tell the test taker the re
common in experiments over long time periods. sults. Some subjects were told that the session
For example, during an experiment on reason was being Videotaped. During the experiment,
ing ability, subjects become bored and sleepy the video equipment failed to work for one sub
and, as a result, score lower. Another example is ject. If it had failed to work for more than one
an experiment on the stylesof children's play be subject or had worked for only part of the ses
tween grades 1 and 6. Play styles are affected by sion, the experiment would have had instru
physical, emotional, and maturation changes mentation problems. (By the way, subjects took
that occur as the children grow older, instead of longer to deliver bad news only if they thought
or in addition to the effects of a treatment. De they were doing so publicly-that is, being
signs with a pretest and control group help re videotaped.)
searchers determine whether maturation or
history effects are present, because both experi Mortality. Mortality, or attrition, arises when
mental and control groups will show similar some subjects do not continue throughout the
changes over time. experiment. Although the word mortality means
death, it does not necessarily mean that subjects
Testing. Sometimes, the pretest measure itself have died. If a subset of subjects leaves partway
affects an experiment. This testing effect threat through an experiment, a researcher cannot
ens internal validity because more than the treat know whether the results would have been dif
ment alone affects the dependent variable. The ferent had the subjects stayed. For example, a re
Solomon four-group design helps a researcher searcher begins a weight-loss program with 50
detect testing effects. For example, a researcher subjects. At the end of the program, 30 remain,
gives students an examination on the first day of each of whom lost 5 pounds with no side effects.
class. The course is the treatment. He or she tests The 20 who left could have differed from the 30
learning by giving the same exam on the last day who stayed, changing the results. Maybe the
of class. If subjects remember the pretest ques program was effective for those who left, and
tions and this affectswhat they learned (i.e., paid they withdrew after losing 25 pounds. Or per
attention to) or how they answered questions on haps the program made subjects sick and forced
the posttest, a testing effect is present. If testing them to quit. Researchers should notice and re
CHAPTER 8 / EXPERIMENTAL RESEARCH 215
port the number of subjects in each group dur treatment. Researchers avoid it by isolating
ing pretests and posttests to detect this threat to groups or having subjects promise not to reveal
internal validity. anything to others who will become subjects.
For example, subjects participate in a day-long
Statistical Regression. Statistical regression is experiment on a new way to memorize words.
not easy to grasp intuitively. It is a problem of During a break, treatment-group subjects tell
extreme values or a tendency for random errors those in the control group about the new way to
to move group results toward the average. It can memorize, which control-group subjects then
occur in two ways. use. A researcher needs outside information,
One situation arises when subjects are un such as postexperiment interviews, with subjects
usual with regard to the dependent variable. Be to detect this threat.
cause they begin as unusual or extreme, subjects
are unlikely to respond further in the same di Experimenter Expectancy. Although it is not
rection. For example, a researcher wants to see always considered a traditional internal validity
whether violent films make people act violently. problem, the experimenter's behavior, too, can
He or she chooses a group of violent criminals threaten causal logic,"A researcher may threaten
from a high-security prison, givesthem a pretest, internal validity, not by purposefully unethical
shows violent films, then administers a posttest. behavior but by indirectly communicating
To the researcher's shock, the prisoners are experimenter expectancy to subjects. Researchers
slightly less violent after the film, whereas a con may be highly committed to the hypothesis and
trol group of prisoners who did not see the film indirectly communicate desired findings to the
are slightly more violent than before. Because subjects. For example, a researcher studies the
the violent criminals began at an extreme, it is effects of memorization training on student
unlikely that a treatment could make them more learning ability, and also sees the grade tran
violent; by random chance alone, they appear scripts of subjects. The researcher believes that
less extreme when measured a second time.e students with higher grades tend to do better at
A second situation involves a problem with the training and will learn more. Through eye
the measurement instrument. If many research contact, tone ofvoice, pauses, and other nonver
participants score very high (at the ceiling) or bal communication, the researcher uncon
very low (at the floor) on a variable, random sciously trains the students with higher grades
chance alone will produce a change between the more intensely; the researcher's nonverbal be
pretest and the posttest. For example, a re havior is the opposite for students with lower
searcher gives 80 subjects a test, and 75 get per grades.
fect scores. He or she then gives a treatment to Here is a way to detect experimenter ex
raise scores. Because so many subjects already pectancy. A researcher hires assistants and
had perfect scores, random errors will reduce the teaches them experimental techniques. The as
group average because those who got perfect sistants train subjects and test their learning abil
scores can randomly move in only one direc ity. The researcher gives the assistants fake
tion-to get some answers wrong. An examina transcripts and records showing that subjects in
tion of scores on pretests will help researchers one group are honor students and the others are
detect this threat to internal validity. failing, although in fact the subjects are identical.
Experimenter expectancy is present if the fake
Diffusion of Treatment or Contamination. honor students, as a group, do much better than
Diffusion of treatment is the threat that research the fake failing students.
participants in different groups will communi The double-blind experiment is designed to
cate with each other and learn about the other's control researcher expectancy. In it, people who
216 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
have direct contact with subjects do not know contains the new drug. Only another person
the details of the hypothesis or the treatment. It who does not deal with subjects directly knows
is double blind because both the subjects and which colored pill contains the drug and it is he
those in contact with them are blind to details of or she who examines the results.
the experiment (see Figure 8.3). For example, a
researcher wants to see if a new drug is effective.
Using pills of three colors-green, yellow, and
External Validity and Field
........
pink-the researcher puts the new drug in the
yellow pill, puts an old drug in the pink one, and
makes the green pill a placebo-a false treatment
that appears to be real (e.g., a sugar pill without
Experiments
Even if an experimenter eliminates all concerns
about internal validity, external validity remains
a potential problem. External validity is the abil
......
. . .-.GAl....
~a . .1S4II
6r .........
any physical effects). Assistants who givethe pills
and record the effects do not know which color
ity to generalize experimental findings to events
and settings outside the experiment itself. If a .A
---------------------.
FIG U R E 8.3 Double-Blind Experiments: An Illustration of Single-Blind, or Ordinary,
and Double-Blind Experiments
Single-Blind Experiment
o
Experimenter ~
7\~ EilMbfI__..
e e e e e e e e e e e e e e e e
\l~\l~\l~\l~\ly\ly\l~Q~
Subjects Who Are Blind to True Hypothesis
~
~a
,.,
.. ...
..-at.. ~lIIIiiI
...... a-I. . .
~
. ._ .
wairsa.a
~uIaI
o Double-Blind Experiment
pia.:r.a ......
Experimenter G'li e Assistant Who Is Blind ...·Nermll.
IT ~ to Details of Treatment pia.:riDthe
~~~a
7\~
Wwieas •
~tb6rr
"'raa.~
. .datn~
e e e e e e e e e e e e e e e e -* 00 a suII~.
\l~\l~\l~\l~\l~\l~\l~\ly 1ra1.=
study lacks external validity, its findings hold person encounters people who contradict the
true only in experiments, making them useless stereotype, especiallyif the others are respected.
to both basic and applied science. They used both a laboratory experiment (with a
two-group, posttest-only design) and a field ex
Reactivity. Research participants might react periment. Past studies focused on out-group
differently in an experiment than they would in stereotypes, but the authors wanted to examine
real life because they know they are in a study; the hypothesis for an in-group, women. In the
this is called reactivity. The Hawthorne effect is a laboratory experiment, experimenters randomly
specific kind of reactivity.? The name comes assigned female subjects to view either (1) a set
from a series of experiments by Elton Mayo at photographs and biographies of 16 famous
the Hawthorne, Illinois, plant of Westinghouse women leaders or (2) photos and descriptions of
Electric during the 1920sand 1930s.Researchers 16 flowers. The experimenters used deception
modified many aspects of working conditions and told subjects the study was about testing
(e.g., lighting, time for breaks, etc.) and mea memory. The dependent variable was attitudes
sured productivity. They discovered that pro and beliefs about women and was measured
ductivity rose after each modification, no matter with a Implicit Association Test (rAT). The re
what it was. This curious result occurred because sults showed that subjects associated gendered
the workers did not respond to the treatment first names (e.g., John vs. Emily) with leadership
but to the additional attention they received or follower traits (e.g., assertive and sympa
from being part of the experiment and knowing thetic). A high IAT score indicated that a sub
that they were being watched. Later research ject viewed women more than men as having
questioned whether this occurred, but the name leadership more than supportive traits. The re
is used for an effect from the attention of re searchers also used a scale on beliefs about
searchers. A related effect is the effect of some women. They found support for the hypothesis
thing new, which may wear off over time. that exposure to famous women in leadership
positions increased IAT scores, compared to ex
Field Experiments. So far, this chapter has fo posure to neutral information about flowers.
cused on experiments conducted under the con The field experiment had a pretest and a posttest
trolled conditions of a laboratory. Experiments but no random assignment. Subjects were fe
are also conducted in real-life or field settings males who attended two colleges in the same
where a researcher has less control over the ex town. One was a coeducational college and the
perimental conditions. The amount of control other had all female students. Subjects were re
varies on a continuum. At one end is the highly cruited from first-year classes at the beginning
controlled laboratory experiment, which takes of the academic year and completed the IAT
place in a specialized setting or laboratory; at the measure, the beliefs about women scale, and a
opposite end is the field experiment, which takes general campus questionnaire. The experi
place in the "field"-in natural settings such as a menters documented that the all-female college
subway car, a liquor store, or a public sidewalk. had more females in administrative and faculty
Subjects in field experiments are usually un leadership positions. Pretest IAT scores were
aware that they are involved in an experiment very similar, with subjects from coeducational
and react in a natural way. For example, re college having slightly lower scores. This helped
searchers have had a confederate fake a heart at the experimenters to check for possible selection
tack on a subway car to see how the bystanders bias. Subjects were contacted one year later and
react.P asked to complete the same measures as pre
Dasgupta and Asgari (2004) tested the hy sented in the posttest. Experimenters watched
pothesis that stereotypical beliefsweaken when a very carefully for experimental mortality since
218 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
some students stopped attending college or did imental mortality, since some students stopped
not complete later surveys. The IAT scores for attending college,left college dormitories, or did
subjects at the coeducational college declined not complete the later surveys. They tested the
(i.e., they were lesslikelyto see females as having hypotheses that students who were randomly as
leadership traits), whereas the IAT scores for signed to live with an out-group member (the
subjects at the all-female college greatly in independent variable) developed less prejudicial
creased. In addition, the experimenters found attitudes toward members of that out-group.
that the more female teachers a student had at They found that compared to pretest measures,
either college,the higher the posttest IAT scores, prejudicial attitudes declined as predicted by the
and this was especiallythe case for math and sci contact hypothesis with one exception. Appar
ences courses. Thus, exposure to women in lead ently having an Asian American roommate
ership positions caused the IAT scores to worked in the opposite way and actually in
increase, whereas the absence of such exposure, creased prejudice, especially among the White
if anything, lowered the scores. students.
Von Larr and colleagues (2005) used a field Experimenter control relates to internal and
experiment to test the well-known contact hy external validity. Laboratory experiments tend
pothesis that says intergroup contact reduces to have greater internal validity but lower exter
racial-ethnic prejudice as people replace their nal validity; that is, they are logically tighter and
...
stereotypes with personal experience, although better controlled, but lessgeneralizable.Field ex
this happens so long as the contact involves peo periments tend to have greater external validity
ple of equal status pursuing common goals in a but lower internal validity; that is, they are more
cooperative setting and is approved byauthori generalizable but less controlled. Quasi-experi
ties. In addition, informal contact in which peo
ple get to know about out-group members as
acquaintances also reduces out-group prejudice.
The experiment took place at UCLA, where the
student body is very racially and ethnically di
mental designs are common in field experi
ments. Table 8.3 summarizes threats to internal
and external validity. ._-.....
. .JiiIIII. . . . .
5 •
...........
.
racial-ethnic groups. These were the experi Diffusion of treatment
ment's pretest and multiple posttest measures.
Experimenters watched very carefully for exper
Experimenter expectancy ."pIIIl...
;
CHAPTER 8 / EXPERIMENTAL RESEARCH 219
---------------------.
Enrique's Natalie's
Slim Clinic Nutrition Center
Pretest Pasttest Pretest Pasttest
Experimental 190 (30) 140 (29) Experimental 190 (30) 188 (29)
Control group 189 (30) 189 (30) Control group 192 (29) 190 (28)
Carl's Calorie
Counters
Pretest Pastiest
Experimental 160 (30) 152 (29)
Control group 191 (29) 189 (29)
pears that the treatment was not effective. the maturation, history, or diffusion of treat
Pauline's Pounds Off also avoided selection bias ment effects may have occurred. Thus, the treat
and experimental mortality problems. People in ment at Enrique's Slim Clinic appears to be the
her experimental group lost 32 pounds, but so most effectiveone. See Box 8.2 for a practical ap
did those in the control group. This suggests that plication of comparing experimental results.
...._$_..
...-.sa
Occasionally, a "natural" experiment ;s possible due in the New Orleans area. When public officials
............
a fa
to public policy changes or a government interven planned to shift enforcement priorities, Scribner and
tion, and researchers are able to measure, partici Cohen (2001) examined its impact. They had sev
pate, and learn from it and conduct a field experiment eral people who clearly looked under 1 8 years old
with high external validity. This occurred in New Or attempt to purchase alcoholic beverages illegally
leans, Lousiana. Until the mid-1990s, laws on selling (the law required being at least 21 years of age) at
liquor to underage customers were barely enforced 143 randomly selected liquor outlets between No
in New Orleans. If caught, the offending liquor re vember 1 995 and January 1 996 (Time 0). The per
tailer met privately with the liquor commission and centage who could buy liquor illegallywas the pretest
paid a small fine. Enforcing liquor laws was low prior
ity for state and local government, so only three en
forcement officers monitored 5,000 alcohol outlets
measure. After assessing the rate of illegal sales, the
dependent variable, the police issued citations to 51
of the sales outlets, the primary independent variable .. "
- -llil
-.dI ...
CHAPTER 8 / EXPERIMENTAl RESEARCH 221
or treatment. About the same time, government offi March to April 1996 (Time 1) and again in Novem
cials initiated a media campaign urging better law ber 1996 to January 1997 (Time 2), during which
compliance. There were two posttest measures, first in the experimenters checked the 143 outlets.
The results allow us to compare rates of illegal the citation campaign, but were unable to do so.
selling activity before and after citations plus media Since outlets that did not receive the treatment (i.e.,
campaign (pretest and posttest measures) and to a citation for law violation) probably learned about it
compare outlets that received citations (experimental from others in the same business, a form of diffusion
group) with those that did not receive citations and of the treatment could be operating. Third, the re
only had media exposure (control group). We see that searchers report that they began with 155 outlets,
the citations and campaign did not stop the illegal but studied only 143 because 12 outlets went out
activity, but it had some effect. The impact was of business during the study. The authors noted that
greater on outlets that experienced direct punish none of the outlets that stopped selling alcohol
ment. In addition, by adding a later follow-up (Time closed due to new law enforcement, but if those out
2), we see how the law-enforcement impact slowly lets that received citations had more problems and
decayed over time. As frequently happens in a nat were more likely to go out of business, it suggests
ural experiment, internal validity is threatened: First, experimental mortality. The experimenters did not
the pretest measure shows a difference in the two mention any external events in New Orleans that
sets of outlets. with outlets that received the treat happened during the time of the study (e.g., a publi
ment showing higher rates of illegal behavior; this is cized event such as underage drinker dying of alcohol
potential selection bias. Second, the media campaign poisoning from overdrinking). Researchers need to
occurred for alloutlets, so the treatment is really a ci be aware of potential external events when a study
tation plus the media campaign. The authors noted continues for a long time and consider possible
that they had intended to compare the New Orleans history effects.
area with another area with neither the media nor
searchers must be very careful if they place re dence for causality. In general, experiments tend
search participants in physical danger or in em to be easier to replicate, less expensive, and less
barrassing or anxiety-inducing situations. They time consuming than the other techniques. Ex
must painstakingly monitor events and control perimental research also has limitations. First,
what occurs. some questions cannot be addressed using ex
Deception is common in social experi perimental methods because control and exper
ments, but it involves misleading or lying to imental manipulation are impossible. Another n_ J.
subjects. Such dishonesty is not condoned un limitation is that experiments usually test one or
conditionally and is acceptable only as the a few hypotheses at a time. This fragments
means to achieve a goal that cannot be achieved knowledge and makes it necessary to synthesize
otherwise. Even for a worthy goal, deception can results across many research reports. External
be used only with restrictions. The amount and validity is another potential problem because E
type of deception should not go beyond what is
minimally necessary, and research participants
many experiments rely on small nonrandom
samples of college students.P
..• .
Dj
....
should be debriefed. You learned how a careful examination and
comparison of results can alert you to potential
problems in research design. Finally, you saw
some practical and ethical considerations in ex
CONCLUSION
periments.
In this chapter, you learned about random as
signment and the methods of experimental re
search. Random assignment is an effective way
In the next chapters, you will examine other
research techniques. The logic of the nonexperi
mental methods differs from that of the experi n
e,
e ..
to create two (or more) groups that can be ment. Experimenters focus narrowly on a few
treated as equivalent and hence compared. In hypotheses. They usually have one or two inde
general, experimental research provides precise pendent variables, a single dependent variable, a
and relatively unambiguous evidence for a few small groups of subjects, and an indepen
causal relationship. It follows the positivist ap dent variable that the researcher induces. By
proach, produces quantitative results that can be contrast, other social researchers test many
analyzed with statistics, and is often used in eval hypotheses at once. For example, survey re
uation research (see Box 8.2). searchers measure a large number of indepen
This chapter also examined the parts of an dent and dependent variables and use a larger
experiment and how they can be combined to number of randomly sampled subjects. Their in
produce different experimental designs. In addi dependent variables are usually preexisting con
tion to the classical experimental design, you ditions in research participants.
learned about preexperimental and quasi-exper
imental designs. You also learned how to express
them using design notation. Key Terms
You learned that internal validity-the in
ternallogical rigor of an experiment-is a key classical experimental design
idea in experimental research. Threats to inter control group
nal validity are possible alternative explanations debrief
to the treatment. You also learned about exter deception
nal validity and how field experiments maximize demand characteristics
external validity. design notation
The real strength of experimental research is diffusion of treatment
its control and logical rigor in establishing evi double-blind experiment
CHAPTER 8 / EXPERIMENTAL RESEARCH 223
liquor consumption by 40 to 60 percent (Rathje those who had psychological problems in latter
and Murphy, 1992:71). Researchers have studied life versus those who did not. (Also see Box 9.2.) CIMISIIa:I
the listening habits of drivers by checking what GIll RR":a.I_.
Recording and Documentation
stations their radios are tuned to when cars are
repaired. They have measured interest in differ
ent exhibits by noting worn tiles on the floor in
Creating nonreactive measures follows the logic
of quantitative measurement. A researcher first
-...as.-nr
UJIbti. .
iIOre.
different parts of a museum. They have studied conceptualizes a construct, then links the con OIl door
differences in graffiti in male versus female high struct to nonreactive empirical evidence, which
school restrooms to show gender differences in is its measure. The operational definition of the
themes. Some have examined high school year variable includes how the researcher systemati
books to compare the high school activities of cally notes and records observations.
gests behavior.
cared, The
Example: A researcher watches students to see
Because nonreactive measures indicate a thor of this book, wrote words and drew dia
construct indirectly, the researcher needs to rule grams to communicate research methods con
out reasons for the observation other than the tent to you, the student. The way the book was
construct of interest. For example, a researcher written and the way you read it are without any
wants to measure customer walking traffic in a knowledge or intention of its ever being content
store. The researcher's measure is dirt and wear analyzed.
on floor tiles. He or she first clarifies what the Content analysis lets a researcher reveal the
customer traffic means (e.g., Is the floor a path content (i.e., messages, meanings, etc.) in a
to another department? Does it indicate a good source of communication (i.e., a book, article,
location for a visual display?) Next, he or she sys movie, etc.). It lets him or her probe into and
tematically measures dirt or wear on the tiles, discover content in a different way from the or
compares it to that in other locations, and dinary way of reading a book or watching a tele
records results on a regular basis (e.g., every vision program.
month). Finally, the researcher rules out other With content analysis, a researcher can
reasons for the observations (e.g., the floor tile is compare content across many texts and analyze
oflower quality and wears faster, or the location it with quantitative techniques (e.g., charts and
is near an outside entrance). tables). In addition, he or she can reveal aspects
of the text's content that are difficult to see. For
example, you might watch television commer
cials and feel that non-Whites rarely appear in
CONTENT ANALYSIS commercials for expensive consumer goods
(e.g., luxury cars, furs, jewelry, perfume, etc.).
What Is Content Analysis?
Content analysis can document-in objective,
Content analysis is a technique for gathering and quantitative terms-whether your vague feel
analyzing the content of text. The content refers ings based on unsystematic observation are
to words, meanings, pictures, symbols, ideas, true. It yields repeatable, precise results about
themes, or any message that can be communi the text.
cated. The text is anything written, visual, or Content analysis involves random sam
spoken that serves as a medium for communica pling, precise measurement, and operational de
tion. It includes books, newspaper and magazine finitions for abstract constructs. Coding turns
articles, advertisements, speeches, official docu aspects of content that represent variables into
ments, films and videotapes, musical lyrics, pho numbers. After a content analysis researcher
tographs, articles of clothing, and works of art. gathers the data, he or she analyzes them with
The content analysis researcher uses objec statistics in the same way that an experimenter
tive and systematic counting and recording pro or survey researcher would.
cedures to produce a quantitative description of
the symbolic content in a text. 2 There are also
Topics Appropriate for
qualitative or interpretive versions of content
Content Analysis
analysis, but in this chapter the emphasis is on
quantitative data about a text's content. Researchers have used content analysis for many
Content analysis is nonreactive because the purposes: to study themes in popular songs and
process of placing words, messages, or symbols religious symbols in hymns, trends in the topics
in a text to communicate to a reader or receiver that newspapers cover and the ideological tone
occurs without influence from the researcher of newspaper editorials, sex-role stereotypes in
who analyzes its content. For example, I, as au textbooks or feature films, how often people of
228 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
different races appear in television commercials system is a set of instructions or rules on how to given week? 11
and programs, answers to open-ended survey systematically observe and record content from are they, or •
questions, enemy propaganda during wartime, text. A researcher tailors it to the specific type of theyap'~ ~
the covers of popular magazines, personality text or communication medium being studied ~
characteristics from suicide notes, themes in ad (e.g., television drama, novels, photos in maga Direction.
vertising messages, gender differences in conver zine advertisements, etc.). The coding system messages in
sations, and so on. also depends on the researcher's unit of analysis. (e.g., posi
Generalizations that researchers make on For example, in the study by Lauzen and Dozier posed). FOI"
the basis of content analysis are limited to the (2005) on gender stereotypes in the most popu ways an
cultural communication itself. Content analysis lar U.S. films in 2002 (discussed in Chapter 4), Some are
cannot determine the truthfulness of an asser the authors developed a coding system based on ate) and
tion or evaluate the aesthetic qualities of litera prior studies ofprime-time television shows and selfish).
ture. It reveals the content in text but cannot film.
interpret the content's significance. Researchers
should examine the text directly. Units. The unit of analysis can vary a great deal
Content analysis is useful for three types of in content analysis. It can be a word, a phrase, a
research problems. First, it is helpful for prob theme, a plot, a newspaper article, a character,
lems involving a large volume of text. A re and so forth. In addition to units of analysis, re
searcher can measure large amounts oftext (e.g., searchers use other units in content analysis that
years of newspaper articles) with sampling and mayor may not be the same as units of analysis:
multiple coders. Second, it is helpful when a recording units, context units, and enumeration
topic must be studied "at a distance." For exam units. There are few differences among them, and
ple, content analysis can be used to study histor they are easily confused, but each has a distinct
ical documents, the writings of someone who role. In simple projects, all three are the same.
has died, or broadcasts in a hostile foreign coun
try. Finally, content analysis can reveal messages What Is Measured? Measurement in content
in a text that are difficult to see with casual ob analysis uses structured observation: systematic,
servation. The creator of the text or those who careful observation based on written rules. The
read it may not be aware of all its themes, biases, rules explain how to categorize and classify ob
or characteristics. For example, authors of servations. As with other measurement, cate
preschool picture books may not consciously gories should be mutually exclusive and
intend to portray children in traditional stereo exhaustive. Written rules make replication pos
typed sex roles, but a high degree of sex stereo sible and improve reliability. Although re
typing has been revealed through content searchers begin with preliminary coding rules,
analysis.' they often conduct a pilot study and refine cod
ing on the basis of it.
Coding systems identify four characteristics
Measurement and Coding
of text content: frequency, direction, intensity,
General Issues. Careful measurement is cru and space. A researcher measures from one to all
cial in content analysis because a researcher four characteristics in a content analysis research
converts diffuse and murky symbolic communi project.
cation into precise, objective, quantitative data.
He or she carefully designs and documents pro Frequency. Frequency simply means counting
cedures for coding to make replication possible. whether or not something occurs and, if it oc
The researcher operationalizes constructs in curs, how often. For example, how many elderly
content analysis with a coding system. A coding people appear on a television program within a
CHAPTER 9 / NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 229
given week? What percentage of all characters words or phrases, and puts the text into a form
are they, or in what percentage of programs do that computers can read."
they appear? Manifest coding is highly reliable because
the phrase or word either is or is not present.
Direction. Direction is noting the direction of Unfortunately, manifest coding does not take
messages in the content along some continuum the connotations of words or phrases into ac
(e.g., positive or negative, supporting or op count. The same word can take on different
posed). For example, a researcher devises a list of meanings depending on the context. The possi
ways an elderly television character can act. bility that there are multiple meanings of a word
Some are positive (e.g., friendly, wise, consider limits the measurement validity of manifest
ate) and some are negative (e.g., nasty, dull, coding.
selfish). For example, I read a book with a red cover
that is a real red herring. Unfortunately, its pub
Intensity. Intensity is the strength or power of a lisher drowned in red ink because the editor
message in a direction. For example, the charac could not deal with the red tape that occurs when
teristic of forgetfulness can be minor (e.g., not a book is red hot. The book has a story about a
remembering to take your keys when leaving red fire truck that stops at red lights only after the
home, taking time to recall the name ofsomeone leaves turn red. There is also a group of Reds who
you have not seen in years) or major (e.g., not carry red flags to the little red schoolhouse. They
remembering your name, not recognizing your are opposed by red-blooded redneeks who eat red
children). meat and honor the red, white, and blue. The
main character is a red-nosed matador who fights
Space. A researcher can record the size of a text red foxes, not bulls, with his red cape. Red-lipped
message or the amount of space or volume allo little Red Riding Hood is also in the book. She
cated to it. Space in written text is measured by develops red eyes and becomes red-faced after
counting words, sentences, paragraphs, or space eating a lot of red peppers in the red light district.
on a page (e.g., square inches). For video or au She is given a red backside by her angry mother,
dio text, space can be measured by the amount a redhead.
of time allocated. For example, a TV character In the study of gender stereotypes in films in
may be present for a few seconds or continu 2002, Lauzen and Dozier (2005) largely used
ously in every scene of a two-hour program. manifest coding. Coders coded each character in
a film as male or female, the estimated age of
each character in one of 7 categories, the occu
Coding, Validity, and Reliability
pation of each character, and whether a charac
Manifest Coding. Coding the visible, surface ter was formally appointed to provide guidance
content in a text is called manifest coding. For ex or direction in a group or informally emgered in
ample, a researcher counts the number of times such a function.
a phrase or word (e.g., red) appears in written
text, or whether a specific action (e.g., a kiss) ap Latent Coding. A researcher using latent cod
pears in a photograph or video scene. The cod ing (also called semantic analysis) looks for the
ing system lists terms or actions that are then underlying, implicit meaning in the content of a
located in text. A researcher can use a computer text. For example, a researcher reads an entire
program to search for words or phrases in text paragraph and decides whether it contains erotic
and have a computer do the counting work. To themes or a romantic mood. The researcher's
do this, he or she learns about the computer pro coding system has general rules to guide his or
gram, develops a comprehensive list of relevant her interpretation of the text and for determin
230 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
ing whether particular themes or moods are cent of .80 or better is generally required, al
present. though .70 may be acceptable for exploratory re
Latent coding tends to be less reliable than search. When the coding process stretches over a
manifest coding. It depends on a coder's knowl considerable time period (e.g., more than three
edge oflanguage and social meaning.f Training, months), the researcher also checks reliability by
practice, and written rules improve reliability, having each coder independently code samples
but still it is difficult to consistently identify of text that were previously coded. He or she
themes, moods, and the like. Yet, the validity of then checks to see whether the coding is stable or
latent coding can exceed that of manifest coding changing. For example, six hours of television
because people communicate meaning in many episodes are coded in April and coded again in
implicit ways that depend on context, not just in July without the coders looking at their original
specific words. coding decisions. Large deviations in coding ne
A researcher can use both manifest and la cessitate retraining and coding the text a second
tent coding. If the two approaches agree, the fi time.
nal result is strengthened; if they disagree, the In the study of the 100 most popular U.S.
researcher may want to reexamine the opera films of2002 by Lauzen and Dozier (2005), three
tional and theoretical definitions. graduate students worked as coders. During an
initial training period they studied the coding
Intercoder Reliability. Content analysis often system and variable definitions. Next, the coders
involves coding information from a very large practiced by coding independent of one another
number of units. A research project might in several films that were not in the study then
volve observing the content in dozens of books, comparing and discussing results. For coding of
hundreds of hours of television programming, study films, 10 percent of all films were double
or thousands of newspaper articles. In addition coded to calculate intercoder reliability mea
to coding the information personally, a re sures. Intercorder reliability measures were cal
searcher may hire assistants to help with the cod culated for each variable. For the gender of the
ing. He or she teaches coders the coding system major character in the film it was .99, for occu
and trains them to fill out a recording sheet. pation of the chaeters it was .91, and for the age
Coders should understand the variables, follow of characters it was .88.
the coding system, and ask about ambiguities. A
researcher records all decisions he or she makes Content Analysis with Visual Material. Using
about how to treat a new specific coding situa content analysis to study visual "text," such as
tion after coding begins so that he or she can be photographs, paintings, statues, buildings,
consistent. clothing, and videos and film, is difficult. It com
A researcher who uses several coders must municates messages or emotional content indi
always check for consistency across coders. He rectly through images, symbols, and metaphors.
or she does this by asking coders to code the Moreover, visual images often contain mixed
same text independently and then checking for messages at multiple levels of meaning.
consistency across coders. The researcher mea To conduct content analysis on visual text,
sures intercoder reliability with a statistical coef the researcher must "read" the meaning(s)
ficient that tells the degree of consistency among within visual text. He or she must interpret signs
coders. The coefficient is always reported with and discover the meanings attached to symbolic
the results of content analysis research. There are images. Such "reading" is not mechanical (i.e.,
several intercoder reliability measures that range image X always means G); it depends heavily on
from 0 to 1, with 1.0 signifying perfect agree the cultural context because the meaning of an
ment among coders. An interreliability coeffi image is culture bound. For example, a red light
CHAPTER 9 / NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 231
does not inevitably mean "stop"; it means "stop" and dress) in the photographs and the recurrent
only in cultures where people have given it that use of major symbols, such as the Statute of Lib
meaning. People construct cultural meanings erty or the U.S. flag, communicated messages.
that they attach to symbolic images, and the Chavez argued that magazine covers are a
meanings can change over time. Some meanings site, or location, where cultural meaning is cre
are clearer and more firmly attached to symbols ated. Visual images on magazine covers have
and images than others. multiple levels of meaning, and viewers con
Most people share a common meaning for struct specific meanings as they read the image
key symbols of the dominant culture, but some and use their cultural knowledge. Collectively,
people may read a symbol differently. For exam the covers convey a world view and express mes
ple, one group of people may "read" a national sages about a nation and its people. For example,
flag to mean patriotism, duty to nation, and a magazine cover that displayed the icon of the
honor of tradition. For others, the same flag Statute of Liberty as strong and full of compas
evokes fear, and they read it to indicate govern sion (message:welcome immigrants) was altered
ment oppression, abuse of power, and military to have strong Asian facial features (message:
aggression. A researcher pursuing the content Asian immigrants distorted the national culture
analysis of images needs to be aware of divergent and altered the nation's racial make-up), or
readings of symbols for people in different holding a large stop sign (message: go away im
situations or who may have diverse beliefs and migrants). Chavez (2001: 44) observed that "im
experiences. ages on magazines both refer to, and in the
Sociopolitical groups may invent or con process, help to structure and construct contem
struct new symbols with attached meanings porary 'American' identity." (SeeBox 9.3 for an
(e.g., a pink triangle came to mean gay pride). other content analysis example.)
They may wrestle for control of the meaning of
major existing symbols. For example, some peo
How to Conduct Content Analysis
ple want to assign a Christian religious meaning
Research
to the Christmas tree; others want it to represent
a celebration of tradition and family values with Question Formulation. As in most research,
out specific religious content; others see its ori content analysis researchers begin with a re
gins as an anti-Christian pagan symbol; and still search question. When the question involves
others want it to mean a festive holiday season variables that are messages or symbols, content
for commercial reasons. Because images have analysis may be appropriate. For example, I
symbolic content with complex, multilayer want to study how newspapers cover a political
meaning, researchers often combine qualitative campaign. My construct "coverage" includes the
judgments about the images with quantitative amount of coverage, the prominence of the cov
data in content analysis. erage, and whether the coverage favors one can
For example, Chavez (2001) conducted a didate over another. I could survey people about
content analysis of the covers of major U.S. mag what they think of the newspaper coverage, but a
azines that dealt with the issue of immigration better strategy is to examine the newspapers di
into the United States. Looking at the covers of rectly using content analysis.
10 magazines from the mid-1970s to the mid
1990s, he classified the covers as having one of Units ofAnalysis. A researcher decides on the
three major messages: affirmative, alarmist, or units of analysis (i.e., the amount of text that is
neutral or balanced. Beyond his classification assigned a code). For example, for a political
and identifying trends in messages, he noted campaign, each issue (or day) of a newspaper is
how the mix of people (i.e., race, gender, age, the unit of analysis.
232 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
Sampling.
pIing in cont
population
ample, the po
Two studies that examined race-ethnicity and ad In another study, Mastro and Atkin (2002) ex sentences, all
vertising in the United States illustrate how content amined whether alcohol advertising to promote types of do
analysis is conducted. Mastro and Stern (2003) brands and make drinking appear glamorous influ Likewise, it co
wanted to see whether television advertising repre enced high school students who are too young to uation, scene,
sents major racial-ethnic groups proportionate to drink legally. They looked at alcohol signs and bill
vision progra
their presence in U.S. society. They examined a one boards in a Mexican-American Chicago neighbor
For example, I
week random sample of prime-time television pro hood. They first photographed all outdoor billboards
minorities are
gramming for six U.S. television networks (ABC, CBS, and signs concerning alcohol in the neighborhood
magazines.. My
NBC, Fox, UPN, and WB) drawn from a three-week over a two-day period in March 1999. After a period
population in
period in February 2001. Prime time was Monday of coder training, two female graduate students con
through Saturday 8:00 P.M. to 11 :00 P.M. EST and tent-analyzed the photographs, coding the following Time, Newsw
Sunday 7:00-1 1:00 P.M. Four undergraduate stu variables: product type, product name, number of portbetween 1
dents were trained as coders. They used two units of human models, and the race, age, gender of each three magazines
analysis: a commercial (excluding local commercials, model. More subjective-latent aspects of models and define pr
political advertisements, and trailers for upcoming coded included attractiveness, sexiness, stylishness, cle." For install
programs) and the first three speaking characters in friendliness, and activity level. In addition, placement cles? Is there a
a commercial. Variables included product type based of products and colors in the billboard were coded. an article? Is a
on a 30-product coding scheme, setting (e.g., work, Coders also classified an overall theme of the bill or two articles?
outdoors), relation to product (e.g., endorse, use, board as romance, individuality, relaxation, sports, Second, I
neither or both), job authority, family status, social adventure, or tradition. Next, a questionnaire was de find that the avi
authority, sexual gazing, and affective state (e.g., cry, veloped for students at a high school in the neigh articles and that
show anger, laugh). Other variables included respect borhood where 89 percent of the students were weeks per year.
shown for a character, character's age, and affability Mexican American. Students in grades 10, 1 1, and population con
(friendly or hostile). The study coded 2,880 com 1 2 were asked to volunteer to complete the survey 45 X 52 X 20= 1
mercials with 2,315 speaking characters, among across a three-day period and 123 completed it. a list of all the
whom 2,290 had a race-ethnicity identified. Data Questionnaire items asked about attention, expo ple size and des'
analysis found that AfricanAmerican characters were sure, recall, and brand exposure to the outdoor signs and time, I d~'idi
most often shown advertising financial services (19 and billboards as wellas drinking intention, approval
1,400 articles. Thu
percent) or food (17 percent), Asians were associ of underage drinking, and pro-drinking beliefs. Re
cent. I also chooses
ated with technology products (30 percent), and sults showed that a student's recall of billboard im
tematic sampling I
Latinos were shown selling soap (40 percent). In ages did not affect his or her drinking attitudes.
published cwlical
general, Whites were slightly overrepresented, Blacks However, brand exposure and accepting the themes
equally represented, but Asians, Latinos, and Native (e.g., an inte~-a1 00
in the billboards were associated with greater ap
Americans underrepresented. For example, Latinos proval of underage drinking. The general impact on same week each yc
are 12 percent of the population but had 1 percent the students was present but not strong. The au magazine are imp
of speaking parts, and were usually scantly clad thors suggested that the weak impact was because pIing. I stratify by II
young people with noticeable accents. The authors there were few Mexican American models and the 467 articles from e;
said that AfricanAmericans appear in commercials in models were older. Also, survey measures of family tides represent e3I
a way that approximates their proportion in the beliefs suggested that the influence of the student's stratify by year. Th
United States, but other racial minorities are under family and culture may have weakened the billboard's per magazine per ~
represented or limited to specific products. impact on pro-drinking attitudes. Finally, I draw
random-number tAl
the 23 sample art:idI
year. I develop a iii
CHAPTER 9 I NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 233
Sampling. Researchers often use random sam keep track of my sampling procedure. See Table
pling in content analysis. First, they define the 9.1 for a sampling frame worksheet in which
population and the sampling element. For ex 1,398sample articles are randomly selected from
ample, the population might be all words, all 140,401 articles.
sentences, all paragraphs, or all articles in certain
types of documents over a specified time period. Variables and Constructing Coding Categories.
Likewise, it could include each conversation, sit In my example, I am interested in the construct
uation, scene, or episode of certain types of tele of an African American or Hispanic American
vision programs over a specified time period. woman portrayed in a significant leadership
For example, I want to know how women and role. I must define "significant leadership role"
minorities are portrayed in U.S. weekly news in operational terms and express it as written
magazines. My unit of analysis is the article. My rules for classifying people named in an article.
population includes all articles published in For example, if an article discusses the achieve
Time, Newsweek, and U.S. News and World Re ments of someone who is now dead, does the
port between 1985and 2005. I first verify that the dead person have a significant role? What is a
three magazines were published in those years significant role-a local Girl Scout leader or a
and define precisely what is meant by an "arti corporate president?
cle." For instance, do film reviews count as arti I must also determine the race and sex of
cles? Is there a minimum size (two sentences) for people named in the articles. What if the race
an article? Is a multipart article counted as one and sex are not evident in the text or accompa
or two articles? nying photographs? How do I decide on the per
Second, I examine the three magazines and son's race and sex?
find that the average issue of each contains 45 Because I am interested in positive leader
articles and that the magazines are published 52 ship roles, my measure indicates whether the
weeks per year. With a 20-year time frame, my role was positive or negative. I can do this with
population contains over 140,000 articles (3 X either latent or manifest coding. With manifest
45 X 52 X 20 = 140,400). My sampling frame is coding, I create a list of adjectives and phrases. If
a list of all the articles. Next, I decide on the sam someone in a sampled article is referred to with
ple size and design. After looking at my budget one of the adjectives, then the direction is de
and time, I decide to limit the sample size to cided. For example, the terms brilliant and top
1,400 articles. Thus, the sampling ratio is 1 per performer are positive, whereas drugkingpin and
cent. I also choose a sampling design. I avoid sys uninspired are negative. For latent coding, I cre
tematic sampling because magazine issues are ate rules to guide judgments. For example, I clas
published cyclically according to the calendar sify stories about a diplomat resolving a difficult
(e.g., an interval of every 52nd issue results in the world crisis, a business executive unable to make
same week each year). Because issues from each a firm profitable, or a lawyer winning a case into
magazine are important, I use stratified sam positive or negative terms. (Relevant questions
pling. I stratify by magazine, sampling 1,400/3 = for coding each article are in Box 9.4.)
467 articles from each. I want to ensure that ar In addition to written rules for coding deci
ticles represent each of the 20 years, so I also sions, a content analysis researcher creates a
stratify by year. This results in about 23 articles recording sheet (also called a coding form or tally
per magazine per year. sheet) on which to record information (see Box
Finally, I draw the random sample using a 9.5). Each unit should have a separate recording
random-number table to select 23 numbers for sheet. The sheets do not have to be pieces of pa
the 23 sample articles for each magazine for each per; they can be 3" X 5" or 4" X 6" file cards,
year. I develop a sampling frame worksheet to or lines in a computer record or file. When a lot
234 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
--------------------~
TABLE 9.1 Excerpt from Sampling Frame Worksheet
..
Time January 1-7, 1985 pp.2-3 000001 No
Time p. 4, bottom 000002 No
Time p. 4, top 000003 Yes-1 0001
•
• Total~.
U.S. News December 25-31,2005 p.62 140401 Yes-23 1389
No..~"
·"Yes" means the number was chosen from a random number table. The number after the dash is a count of the number of Person 1 :
articles selected for a year. P~ 2 :
P~ 3 _
P~ 4
--
PIfiOn
Person
PfnOn
--
1. Characteristics of the article. What is the maga 3. Leadership roles. For each significant person in --
zine? What is the date of the article? How large the article, which ones have leadership roles?
PfnOn
is the article? What was its topic area? Where What is the field of leadership or profession of
used?
4. Positive or negative roles. For each leadership or
2. People inthe article. How many people are named professional role, rate how positively or nega at infomuIia.
in the article? Of these, how many are significant tively it is shown. For example, 5 = highly posi uniL~th-.
in the article? What is the race and sex of each tive, 4 = positive, 3 = neutral, 2 = negative, 1 = When pbllli.lI~
person named? highly negative, 0 = ambiguous. ~ "'-ud RqlDR4
lot-test, I find
CHAPTER 9 / NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 235
Blank Example
Professor Neuman, Sociology Department Coder:_ _
Minority/Majority Group Representation in Newsmagazines Project
ARTICLE#_ _ MAGAZINE:_ _ DATE:_ _ SIZE:_ _ col. in.
Total number of people named _ _ Number of Photos
ARTICLE # 0454 MAGAZINE: Time DATE: March 1-7,2005 SIZE: 14 col. in.
Total number of people named_5_ Number of Photos Q
No. people with significant roles:-± Article Topic: Foreign Affairs
Person 1 Race: White Gender: M Leader?: Y Field? Banking Rating:_5_
Person 2 Race: White Gender: M Leader?: N Field? Government Rating: 1i6..
Person 3 Race: Black Gender: F Leader?: Y Field? Civil Rights Rating:_2_
Person 4 Race: White Gender: F Leader?: Y Field? Government Rating:_O_
Person Race: Gender: Leader?: Field? Rating: _ _
Person - Race: Gender: Leader?: Field? Rating: _ _
Person - Race: Gender: Leader?: Field? Rating: _ _
Person Race: Gender: Leader?: Field? Rating: _ _
of information is recorded for each recording utes to read and code an article. This does not
unit, more than one sheet of paper can be used. include sampling or locating magazine articles.
When planning a project, researchers calculate With approximately 1,400 articles, that is 350
the work required. For example, during my pi hours of coding, not counting time to verify the
lot-test, I find that it takes an average of 15 min accuracy of coding. Because 350 hours is about
236 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
nine weeks of nonstop work at 40 hours a week, search question and variables in mind, and then
I should consider hiring assistants as coders. reassemble the information in new ways to ad
Each recording sheet has a place to record dress the research question.
the identification number of the unit and spaces It is difficult to specify topics that are appro
for information about each variable. I also put priate for existing statistics research because they
identifying information about the research pro are so varied. Any topic on which information
ject on the sheet in case I misplace it or it looks has been collected and is publicly available can
similar to other sheets I have. Finally, if I use be studied. In fact, existing statistics projects
multiple coders, the sheet reminds the coder to may not fit neatly into a deductive model of re
check intercoder reliability and, if necessary, search design. Rather, researchers creatively re
makes it possible to recode information for in organize the existing information into the
accurate coders. After completing all recording variables for a research question after first find
sheets and checking for accuracy, I can begin ing what data are available.
data analysis. Experiments are best for topics where the re
searcher controls a situation and manipulates an
independent variable. Survey research is best for
Inferences
topics where the researcher asks questions and
The inferences a researcher can or cannot make learns about reported attitudes or behavior. Con
on the basis of results is critical in content analy tent analysis is best for topics that involve the
Sociallindiiallllll
sis. Content analysis describes what is in the text. content of messages in cultural communication.
It cannot reveal the intentions of those who cre Existing statistics research is best for topics
ated the text or the effects that messages in the that involve information routinely collected by
text have on those who receive them. For exam large bureaucratic organizations. Public or pri
ple, content analysis shows that children's books vate organizations systematically gather many
contain sex stereotypes. That does not necessar types of information. Such information is gath
ily mean that children's beliefs or behaviors are ered for policy decisions or as a public service. It
influenced by the stereotypes; such an inference is rarely collected for purposes directly related to
requires a separate research project on how chil a specific research question. Thus, existing sta
dren's perceptions develop. tistics research is appropriate when a researcher
wants to test hypotheses involving variables that
are also in official reports of social, economic,
and political conditions. These include descrip
EXISTING STATISTICS/ tions of organizations or the people in them. Of
DOCUMENTS AND SECONDARY ten, such information is collected over long time
ANALYSIS periods. For example, existing statistics can be
used by a researcher who wants to see whether
Appropriate Topics
unemployment and crime rates are associated in
Many types of information about the social 150 cities across a 20-year period.
world have been collected and are available to Downey (2005) conducted an existing sta
the researcher. Some information is in the form tistics study on racial inequality (Black/White)
of statistical documents (books, reports, etc.) and living near a toxic pollution site in Detroit.
that contain numerical information. Other in He used census data on the population/housing
formation is in the form of published compila and manufacturing directories of manufactur
tions available in a library or on computerized ing facilities. He also identified highly polluting
records. In either case, the researcher can search industries and used the Environmental Protec
through collections of information with a re tion Agency's inventory of toxic chemicals. His
CHAPTER 9 I NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 237
unit of analysis was the census tract. Downey health and nutrition, public safety, education
tested competing models of environmental in and training, work, income, culture and leisure,
equality: (1) racist siting policy: toxic sites were social mobility, and public participation.
placed in Black residential areas, (2) economic A more specific example of a social indica
inequality: low-income people who are dispro tor is the FBI's uniform crime index. It indicates
portionately Black move into areas near toxic the amount of crime in U.S. society. Social indi
sites because they find low-cost housing there, cators can measure negative aspects of social life,
and (3) residential segregation: Whites move such as the infant mortality rate (the death rate
into specific areas and keep out non-Whites. He of infants during the first year of life) or alco
found greatest support for the residential segre holism, or they can indicate positive aspects,
gation model. Paradoxically, it meant that such as job satisfaction or the percentage of
Blackswere lesslikelythan Whites to live close to housing units with indoor plumbing. Social in
a toxic pollution site. This was because Whites dicators often involve implicit value judgments
had obtained housing near the factories where (e.g., which crimes are serious or what consti
they worked and kept Blacks from moving in tutes a good quality oflife).
but those factories were the major sources of
toxic pollution.
Locating Data
Locating Existing Statistics. The main sources
Social Indicators
of existing statistics are government or interna
During the 1960s,some social scientists, dissatis tional agencies and private sources. An enor
fied with the information available to decision mous volume and variety of information exists.
makers, spawned the "social indicators' move If you plan to conduct existing statistics re
ment" to develop indicators of social well-being. search, it is wise to discuss your interests with an
Many hoped that information about social well information professional-in this case, a refer
being could be combined with widely used indi ence librarian, who can point you in the direc
cators of economic performance (e.g., gross tion of possible sources.
national product) to better inform government Many existing documents are "free"-that
and other policymaking officials. Thus, re is, publicly available at libraries-but the time
searchers wanted to measure the quality of social and effort it takes to search for specific informa
life so that such information could influence tion can be substantial. Researchers who con
public policy.6 duct existing statistics research spend many
Today, there are many books, articles, and hours in libraries or on the Internet. After the
reports on social indicators, and even a scholarly information is located, it is recorded on cards,
journal, Social Indicators Research, devoted to graphs, or recording sheets for later analysis. Of
the creation and evaluation of social indicators. ten, it is already available in a format for com
The U.S. Census Bureau produced a report, puters to read. For example, instead of recording
Social Indicators, and the United Nations has voting data from books, a researcher could use a
many measures of social well-being in different social science data archive at the University of
nations. Michigan (to be discussed).
A social indicator is any measure of social There are so many sources that only a small
well-being used in policy. There are many spe sample of what is available is discussed here. The
cific indicators that are operationalizations of single-most valuable source of statistical infor
well-being. For example, social indicators have mation about the United States is the Statistical
been developed for the following areas: popula Abstract of the United States, which has been
tion, family, housing, social security and welfare, published annually (with a few exceptions) since
238 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
1878. The Statistical Abstract is available in all UNESCO Statistical Yearbook, and United Na
public libraries and on the Internet and can be tions Statistical Yearbook.
purchased from the U.S. Superintendent of In addition to government statistical docu
Documents. It is a selected compilation of the ments, there are dozens of other publications.
many official reports and statistical tables pro Many are produced for business purposes and
duced by U.S. government agencies. It contains can be obtained only for a high cost. They in
statistical information from hundreds of more clude information on consumer spending, the
detailed government reports. You may want to location of high-income neighborhoods, trends
examine more specific government documents. in the economy, and the like.9
(The detail of what is available in government Over a dozen publications list characteris
documents is mind boggling. For example, you tics of businesses or their executives. These are
can learn that there were two African American found in larger libraries. Three such publications
females over the age of 75 in Tucumcari City, are as follows:
New Mexico, in 1980.)
The Statistical Abstract has over 1,400 Dun and Bradstreet Principal Industrial Busi
charts, tables, and statistical lists from over 200 nesses is a guide to approximately 51,000
government and private agencies. It is hard to businesses in 135 countries with informa
grasp all that it contains until you skim through tion on sales,number of employees, officers,
the tables. A two-volume set summarizes similar and products.
information across many years; it is called Who Owns Whom comes in volumes for na
Historical Statistics of the U.S.: Colonial Times to
tions or regions (e.g., North America, the
1970.
United Kingdom, Ireland, and Australia). It
Most governments publish similar statisti lists parent companies, subsidiaries, and as
cal yearbooks. Australia's Bureau of Statistics
sociated companies.
produces Yearbook Australia, Statistics Canada
produces Canada Yearbook, New Zealand's De Standard and Poor's Register ofCorporations,
partment of Statistics publishes New Zealand Of Directors and Executives lists about 37,000 .' «
ficial Yearbook, and in the United Kingdom, the U.S. and Canadian companies. It has infor . .,4. . . . .
Central Statistics Office publishes Annual Ab mation on corporations, products, officers, ·a l
stract of Statistics. 7 Many nations publish books industries, and sales figures. ~D
with historical statistics, as well.
Locating government statistical documents Many biographical sources list famous peo
.'
is an art in itself.Some publications exist solelyto ple and provide background information on
assist the researcher. For example, the American them. These are useful when a researcher wants
Statistics Index: A Comprehensive Guide and to learn about the social background, career, or Ie
Index to the Statistical Publications of the U.S. other characteristics of famous individuals. The
Government and Statistics Sources: A Subject publications are compiled by companies that
Guide to Data on Industrial, Business, Social Edu send out questionnaires to people identified as
cation, Financial and Other Topicsfor the U.S. and "important" by some criteria. They are public
Internationally are two helpful guides for the sources of information, but they depend on the
cooperation and accuracy of individuals who are
United States.f The United Nations and interna
tional agenciessuch as the World Bank have their
own publications with statistical information for
selected.
Politics has its own specialized publications.
'~.si;.
Rilriu:dw -
various countries (e.g., literacy rates, percentage There are two basic types. One has biographical
&nIliIIi pa...l.I
of the labor force working in agriculture, birth information on contemporary politicians. The
rates)-for example, the Demographic Yearbook, other type has information on voting, laws en -d..... afl.
CHAPTER 9 / NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 239
acted, and the like. Here are three examples of Large-scale data collection is expensive and
political information publications for the United difficult. The cost and time required for a major
States: national survey that uses rigorous techniques are
prohibitive for most researchers. Fortunately,
Almanac of American Politics is a biannual the organization, preservation, and dissemina
publication that includes photographs and a tion of major survey data sets have improved.
short biography of U.S. government offi Today, there are archives of past surveys that are
cials. Committee appointments, voting open to researchers.
records, and similar information are pro The Inter-University Consortium for Polit
vided for members of Congress and leaders ical and Social Research (ICPSR) at the Univer
in the executive branch. sity of Michigan is the world's major archive of
America Votes: A Handbook of Contempo social science data. Over 17,000 survey research
rary American Election Statistics contains de and related sets of information are stored and
tailed voting information by county for made available to researchers at modest costs.
most statewide and national offices. Pri Other centers hold survey data in the United
mary election results are included down to States and other nations. 10
the county level. A widely used source of survey data for the
United States is the General Social Survey (GSS),
Vital Statistics on American Politics provides
which has been conducted annually in most
dozens of tables on political behavior, such
years by the National Opinion Research Center
as the campaign spending of every candi
at the University of Chicago. In recent years, it
date for Congress, their primary and final
has covered other nations as well. The data are
votes, ideological ratings by various political
made publicly availablefor secondary analysis at
organizations, and a summary of voter reg
a low cost (see Box 9.6).
istration regulations by state.
The General Social Survey (GSS) is the best-known and in 2000, it was intergroup relations and multi
set of survey data used by social researchers for sec culturalism.
ondary analysis. The mission of the GSS is "to make Interviewers collect the data through face-to
timely, high quality, scientifically relevant data avail face interviews. The NORC staff carefully selects in
able to the social science research community" terviewers and trains them in social science
(Davis and Smith, 1992:1). It is available in many methodology and survey interviewing. About 120
computer-readable formats and is widely accessible to 1 40 interviewers work on the GSS each year.
for a low cost. Neither datasets nor codebooks are About 90 percent are women, and most are middle
copyrighted. Users may copy or disseminate them aged. The NORC recruits bilingual and minority in
without obtaining permission. You can find results terviewers. Interviewers with respondents are race
using the GSS in over 2,000 research articles and matched with respondents. Interviews are typically
books. 90 minutes long and contain approximately 500
The National Opinion Research Center (NORC) questions. The response rate has been 71 to 79
has conducted the GSS almost every year since percent. The major reason for non response is a re
1972. A typical year's survey contains a random fusal to participate.
sample of about 1 ,500 adult U.s. residents. A team The International Social Survey Program conducts
of researchers selects some questions for inclusion, similar surveys in other nations. Beginning with the
and individual researchers can recommend questions. German ALLBUS and British Social Attitudes Survey,
They repeat some questions and topics each year, participation has grown to include 33 nations. The
include some on a four- to six-year cycle, and add goal is to conduct on a regular basis large-scale na
othertopics in specific years. For example, in 1998, tional general surveys in which some common ques
the special topic was job experiences and religion, tions are asked across cooperating nations.
A second danger is that the researcher does sion by quoting statistics in greater detail than
not understand the substantive topic. Because warranted and "overloading" the details. For ex
the data are easily accessible, researchers who ample, existing statistics report that the popula
know very little about a topic could make erro tion of Australia is 19,169,083, but it is better to
neous assumptions or false interpretations about say that it is a little over 19 million. One might
results. Before using any data, a researcher needs calculate the percentage of divorced people as
to be well informed about the topic. For exam 15.65495 in a secondary data analysis of the 2000
ple, if a researcher uses data on high school grad General Social Survey, but it is better to report
uation rates in Germany without understanding that about 15.7 percent of people are divorced.'!
the Germany secondary education system with
its distinct academic and vocational tracks, he or Units ofAnalysis and Variable Attributes. A
she may make serious errors in interpreting common problem in existing statistics is finding
results. the appropriate units of analysis. Many statistics
A third danger is that a researcher may are published for aggregates, not the individual.
quote statistics in great detail to give an impres For example, a table in a government document
sion of scientific rigor. This can lead to the has information (e.g., unemployment rate,
fallacy of misplaced concreteness, which occurs crime rate, etc.) for a state, but the unit of analy
when someone gives a false impression of preci sis for the research question is the individual
CHAPTER 9 I NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 241
I
miscounts by using scientific sampling and adjusting physician or hospital. Many work injuries, as de
the census. The politicians proved to be less con fined by the researcher, would not be in official
cerned about improving the scientific accuracyof the statistics. Another example occurs when a re
census than retaining traditional census methods
searcher defines people as unemployed if they
that would benefit their own political fortunes or help
would work if a good job were available, if they
their constituencies, because the government uses
have to work part time when they want full-time
census data to draw voting districts and allocate
publicfunds to areas. work, and if they have given up looking for
work. The official definition, however, includes
only those who are now actively seeking work
(full or part time) as unemployed. The official
statistics exclude those who stopped looking,
(e.g., "Are unemployed people more likely to who work part time out of necessity, or who do
commit property crimes?"). The potential for not look because they believe no work is avail
committing the ecological fallacy is very real in able. In both cases, the researcher's definition
this situation. It is less of a problem for sec differs from that in official statistics (see Box
ondary survey analysis because researchers can 9.8).
242 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
In most countries, the official unemployment rate An economic policy or labor market perspective
measures only the unemployed (see below) as a per says the rate should measure those ready to enter
cent of all working people. It would be 50 percent the labor market immediately. It defines nonworking
higher if two other categories of nonemployed peo people as a supply of high-quality labor, an input for
ple were added: involuntary part-time workers and use in the economy available to employers. By con
discouraged workers (see below). In some countries trast, a social policy or human resource perspective
(e.g., Sweden and United States), it would be nearly says the rate should measure those who are not cur
double ifit included these people. This does not con rently working to their fullest potential. The rate
sider other nonworking people, transitional self-em should represent people who are not or cannot fully
ployed, or the underemployed (see below). What a utilize their talents, skills, or time to the fullest. It de
country measures is a theoretical and conceptual de fines nonworking people as a social problem of indi
finition issue: What construct should an unemploy viduals unable to realize their capacity to be
ment rate measure and why measure it? productive, contributing members of society.
....r;;aL ..bled ~
cIirilkdl
.......
. • L
Another validity problem arises when offi bery arrests as a proxy. But the measure is not ..... iaIu.. The.
: ,~
This is necessary because the researcher cannot not always result in an arrest. .-..iIr,. 1ItIal poIi
A third validity problem arises because the
collect original data. For example, the researcher
wants to know how many people have been researcher lacks control over how information is _aRpOrUd...
data the fastest. The quality of U.S. statistics suf valuable. Missing information is especially a to show an
fered from being highly decentralized, having problem when researchers cover long time peri the ease of
fewer statisticians employed than any nation, ods. For instance, a researcher interested in the ask respo
and politically motivated cutbacks on the range number of work stoppages and strikes in the variables,
of data collected.l' United States can obtain data from the 1890s to in the text,
the present, except for a five-year period after
Missing Data. One problem that plagues re 1911 when the federal government did not col
searchers who use existing statistics and docu lect the data. (See Box 9.9 for an existing statis Ethical u.
ments is that of missing data. Sometimes, the tics example.) Ethical con.:al
data were collected but have been lost. More fre nonreactive ..
quently, the data were never collected. The deci studied are ~
sion to collect official information is made ethical~
within government agencies. The decision to ask ISSUES OF INFERENCE AND of using i.nioQ
questions on a survey whose data are later made THEORY TESTING Another etIl.iq
publicly available is made by a group of re social and ~
Inferences from Nonreactive Data
searchers. In both cases, those who decide what and value as5IIIl
to collect may not collect what another re A researcher's ability to infer causality or test a
searcher needs in order to address a research theory on the basis of nonreactive data is lim isCOllect~
~I
ing it.
question. Government agencies start or stop col ited. It is difficult to use unobtrusive measures to official and
lecting information for political, budgetary, or establish temporal order and eliminate alterna iects of ..
other reasons. For example, during the early tive explanations. In content analysis, a re 'of policy. B~ 1
1980s, cost-cutting measures by the U.S. federal searcher cannot generalize from the content to
PUblic~~
government stopped the collection of much in its effects on those who read the text, but can would be .
formation that social researchers had found only use the correlation logic of survey research valid,m~
collection of ~
tions f e.z., tIIj
~::~Putul
ofthe 1930s..
defined as
public a
An androgynous first name is one that can be for ei rare (about 3 percent) and that there has been a
ther a girl or boy without clearly marking the child's very slight historical trend toward androgyny, but percentage ~
gender. Some argue that the feminist movement de only in very recent years. In addition, parents give an t..s. schools ~
creased gender marking in a child's name as part of drogynous names to girls more than to boys, and since 1953. 3IIIIl
its broader societal influence to reduce gender dis gender segregation in naming is unstable (i.e., a name onlv. since the'i
tinctions and inequality. Others observe that gender tends to lose its androgynous meaning over time). was not salie:uI!
remains the single-most predominant feature of nam The authors noted that the way parents name chil Thecolkd
ing in most societies. Even when racial groups or so dren mimics a pattern of collective behavior found to newatten~
cial classes invent distinctive new first names, the operate in another research area: the racial segrega aboutap
gender distinctions are retained. tion of neighborhoods. Change in residence is un official~
Lieberson and colleagues (2000) examined ex equal among races with less movement by the came a biggerji
isting statistical data in the form of computerized dominant group; the less powerful group moves to onthen~
records from the birth certificates of 1 1 millionbirths occupy areas that the dominant group has aban whether~
of White children in the state of Illinois from 1 91 6 to doned; and integration is unstable, with new segre Political ~
1 989. They found that androgynous first names are gation reappearing after some time. sions about -II
Most official sa
CHAPTER 9 / NONREACTIVE RESEARCH AND SECONDARY ANALYSIS 245
interest. Although existing statistics and sec 4. Stone and Weber (1992) and Weber (1984,1985)
ondary data analysis are low-cost research tech summarized computerized content analysis tech
niques, the researcher lacks control over, and niques.
substantial knowledge of, the data collection 5. See Andren (1981:58-66) for a discussion of reli
process. This introduces a potential source of er ability. Coding categorization in content analysis
is discussed in Holsti (1969:94-126).
rors about which researchers need to be espe
6. A discussion of social indicators can be found in
ciallyvigilant and cautious.
Carley (1981). Also see Bauer (1966), Duncan
In the next chapter, we move from design (1984:233-235), Juster and Land (1981), Land
ing research projects and collecting data to ana (1992), and Rossi and Gilmartin (1980).
lyzing data. The analysis techniques apply to the 7. Many non-English yearbooks are also produced;
quantitative data you learned about in the previ for example, Statistiches Jahrbuch for the Federal
ous chapters. So far, you have seen how to move Republic of Germany, Annuaire Statistique de la
from a topic, to a research design and measures, France for France, Year Book Australia for Aus
to collecting data. Next, you will learn how to tralia, and Denmark's Statiskisk Ti Arsoversigt.
look at data and see what they can tell you about Japan produces an English version of its yearbook
a hypothesis or research question. called the Statistical HandbookofJapan.
8. Guides exist for the publications of various gov
ernments-for example, the Guide to British
Key Terms GovernmentPublications, Australian Official Pub
lications, and Irish Official Publications. Similar
accretion measures publications exist for most nations.
coding 9. See Churchill (1983:140-167) and Stewart (1984)
for lists of business information sources.
coding system
10. Other major U.S. archives of survey data include
content analysis the National Opinion Research Center, University
erosion measures of Chicago; the Survey Research Center, Univer
fallacy of misplaced concreteness sity of California-Berkeley; the Behavioral Sci
General Social Survey (GSS) ences Laboratory, University of Cincinnati; Data
latent coding and Program Library Service, University ofWis
manifest coding consin-Madison; the Roper Center, University of
nonreactive Connecticut-Storrs; and the Institute for Re
recording sheet search in Social Science, University of North Car
Statistical Abstract ofthe United States olina-Chapel Hill. Also see KiecoIt and Nathan
structured observation (1985) and Parcel (1992).
11. For a discussion of these issues, see Dale and col
text
leagues (1988:27-31), Maier (1991), and Parcel
unobtrusive measures
(1992). Horn (1993:138) gives a good discussion
with examples of the fallacy of misplaced con
Endnotes creteness.
12. See Stevenson (1996).
1. See Webb and colleagues (1981:7-11). 13. See The Economist, "The Good Statistics Guide"
2. For definitions of content analysis, see Holsti (September 11, 1993), "The Overlooked House
(1968:597), Krippendorff (1980:21-24), Markoff keeper" (February 5, 1994), and "Fewer Damned
and associates (1974:5-6), Stone and Weber Lies?" (March 30, 1996).
(1992), and Weber (1983,1984, 1985:81, note 1). 14. See Block and Burns (1986), Carr-Hill (1984),
3. Weitzman and colleagues (1972) is a classicin this Hindess (1973), Horn (1993), Maier (1991), and
type of research. Van den Berg and Van derVeer (1985).
VVeber(1984,1985)
intent analysis tech-
1
a discussion of reli
in content analysis Analysis of Quantitative
4-126).
:ors can be found in
ier (1966), Duncan
Data
Publications. Similar
[lations. Results with One Variable
,7) and Stewart (1984) Frequency Distributions
Statistical Significance
Conclusion
186), Carr-Hill (1984),
~3), Maier (1991), and
rVeer (1985).
247
248 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
easily read it, and without the codebook, it is then use a bar-code reader to transfer the in
worthless. It condenses answers to 50 survey formation into a computer.
questions for three respondents into three lines
or rows. The raw data for many research projects
Cleaning Data
look like this, except that there may be over
1,000 rows, and the lines may be over 100 Accuracy is extremely important when coding
columns long. For example, a IS-minute tele data. Errors made when coding or entering data
phone survey of 250 students produces a grid of into a computer threaten the validity of measures
data that is 250 rows by 240 columns. and cause misleading results. A researcher who
The codebook in Figure 10.1 says that the has a perfect sample, perfect measures, and no
first two numbers are identification numbers. errors in gathering data, but who makes errors in
Thus, the example data are for the first (01), sec the coding process or in entering data into a
ond (02), and third (03) respondents. Notice computer, can ruin a whole research project.
that researchers use zeroes as place holders to re After very careful coding, the researcher ver
duce confusion between 1 and 01. The Is are al ifies the accuracy of coding, or "cleans" the data.
ways in column 2; the lOs are in column 1. The He or she may code a 10 to 15 percent random
codebook says that column 5 contains the vari sample of the data a second time. If no coding
able "sex": Cases 1 and 2 are male and Case 3 is errors appear, the researcher proceeds; if he or
female. Column 4 tells us that Carlos inter she finds errors, the researcher rechecks all
viewed Cases 1 and 2, and Sophia Case 3. coding.
There are four ways to get raw quantitative When the data are in the computer, re
data into a computer: searchers verify coding in two ways. Possible code
cleaning (or wildcode checking) involves checking
1. Code sheet. Gather the information, then the categories of all variables for impossible
transfer it from the original source onto a codes. For example, respondent sex is coded 1 =:
grid format (code sheet). Next, type what is Male, 2 =: Female. Finding a 4 for a case in the
on the code sheet into a computer, line by field for the sex variable indicates a coding error.
line. A second method, contingency cleaning (or
2. Direct-entry method, including CATI. As in consistency checking), involves cross-classifying
formation is being collected, sit at a com two variables and looking for logically impossible
puter keyboard while listening to/ observing combinations. For example, education is cross
the information and enter the information, classified by occupation. If a respondent is
or have a respondent/subject enter the in recorded as never having passed the eighth grade
formation himself or herself. The computer and also is recorded as being a legitimate medical
must be preprogrammed to accept the doctor, the researcher checks for a coding error.
information. A researcher can modify data after they are
3. Optical scan. Gather the information, then in the computer. He or she may not use more re
enter it onto optical scan sheets (or have a fined categories than were used when collecting
respondent/subject enter the information) the original data, but may combine or group in
by filling in the correct "dots." Next, use an formation. For example, the researcher may
optical scanner or reader to transfer the in group ratio-level income data into five ordinal
formation into a computer. categories. Also, he or she can combine informa
4. Bar code. Gather the information and con tion from several indicators to create a new
vert it into different widths of bars that are variable or add the responses to several ques
associated with specific numerical values, tionnaire items into an index score.
CHAPTER 10 I ANALYSIS OF QUANTITATIVE DATA 251
--------------------~.
FIG U R E 1 0 . 2 Examples of Univariate Statistics
Males
Females
Under $5,000 25
$5,000 to $9,999 50
$10,000 to $15,999 100
$16,000 to $19,999 150
$20,000 to $29,999 50
$30,000 and over 25
Total 400
Frequency
50
•
45
40
• • • •
35 •
30 • •
25
20 •
• •
15
• • •
10
• •
5 ••
• • • etc.
I
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 etc.
---.
CHAPTER 10 / ANALYSIS OF QUANTITATIVE DATA
--------'--.
TAB LE 1 0 . 1 Measures of Central
mode the lowest. If most cases have higher
scores with a few extreme low scores, the mean
Tendency and levels of will be the lowest, the median in the middle, and
Measurement the mode the highest. In general, the median is
, Distribution
best for skewed distributions, although the mean
Percentage
is used in most other statistics (see Figure 10.3).
25'7'0
75'7'0
---------------------11'
Cases
Mean, Median, Mode
Skewed Distributions
~~
groups of six with a range of 35 years: 30 30 30 30 It is based on the mean and gives an "average
3065 and 20 45 46 48 50 55. distance" between all scores and the mean. Peo
Percentiles tell the score at a specific place ple rarely compute the standard deviation by
within the distribution. One percentile you al hand for more than a handful of cases because
ready learned is the median, the 50th percentile. computers and calculators can do it in seconds.
Sometimes the 25th and 75th percentiles or the Look at the calculation of the standard devi
10th and 90th percentiles are used to describe a ation in Figure lOA. If you add up the absolute
distribution. For example, the 25th percentile is difference between each score and the mean (i.e.,
the score at which 25 percent of the distribution subtract each score from the mean), you get
have either that score or a lower one. The com zero. This is because the mean is equally distant
putation of a percentile follows the same logic as from all scores. Also notice that the scores that
the median. IfI have 100 people and want to find differ the most from the mean have the largest
the 25th percentile. I rank the scores and count effect on the sum of squares and on the standard
up from the bottom until I reach number 25. If deviation.
the total is not 100, I simply adjust the distribu The standard deviation is used for compar
tion to a percentage basis. ison purposes. For example, the standard devia
Standard deviation is the most difficult to tion for the schooling of parents of children in
compute measure of dispersion; it is also the class A is 3.317 years; for class B, it is 0.812; and
most comprehensive and widely used. The range for class C, it is 6.239. The standard deviation
and percentile are for ordinal-, interval-, and ra tells a researcher that the parents of children in
tio-level data, but the standard deviation re class B are very similar, whereas those for class C
quires an interval or ratio level of measurement. are very different. In fact, in class B, the school
...
---, ----------------~
FIG U R E 1 0.4 The Standard Deviation
from the mean. Scores are in terms of their rela Kings College and Queens College. She learns
tive position within a distribution, not as ab that the colleges are similar and that both grade
solute values. on a 4.0 scale.Yet, the mean grade-point average
For example, Katy, a sales manager in firm at Kings College is 2.62 with a standard deviation
A, earns $50,000 per year, whereas Mike in firm of .50, whereas the mean grade-point average at
B earns $38,000 per year. Despite the absolute Queens College is 3.24 with a standard deviation
income differences between them, the managers of .40. The employer suspects that grades at
are paid equally relative to others in the same Queens College are inflated. Suzette from Kings
firm. Katy is paid more than two-thirds of other College has a grade-point average of 3.62, and
employees in her firm, and Mike is also paid Jorge from Queens College has a grade-point av
more than two-thirds of the employees in his erage of 3.64. Both students took the same
firm. courses. The employer wants to adjust the grades
Z-scores are easy to calculate from the mean for the grading practices of the two colleges (i.e.,
and standard deviation (see Box 10.1). For ex create standardized scores). She calculates z
ample, an employer interviews students from scores by subtracting each student's score from
Z-score = (Score - Mean)/Standard Deviation, Now, I label the values of the mean and add or
or in symbols: subtract standard deviations from it. One standard
deviation above the mean (+1) when the mean is 7
z =-
x-x and standard deviation is 2 years is just 7 + 2, or 9
years. For a - 2 z-score, I put 3 years. This is because
s it is 2 standard deviations, of 2 years each (or 4
where: X = score, X = mean, 8 = standard deviation years), lower than the Mean of 7. My diagram now
I usually rely on a simple conceptual diagram that looks like this:
does the same thing and that shows what z-scores
really do. Consider data on the ages of schoolchild 3 5 7 9 11 13 age in years
ren with a mean of 7 years and a standard deviation I I I I I I I
of 2 years. How do I compute the z-score of 5-year -3-2 -10 +1 +2 +3
old Miguel, or what if I know that Yashohda's z-score
is a +2 and I need to know her age in years? First, I It is easy to see that Miguel, who is 5 years old,
draw a little chart from - 3 to +3 with zero in the has a z-score of -1, whereas Yashohda's z-score of
middle. I will put the mean value at zero, because a z +2 corresponds to 11 years old. I can read from z
score of zero is the mean and z-scores measure dis score to age, or age to z-score, For fractions, such as
tance above or below it. I stop at 3 because virtually a z-score of -1.5, I just apply the same fraction to
all cases fall within 3 standard deviations of the mean age to get 4 years. Likewise, an age of 1 2 is a z-score
in most situations. The chart looks like this: of +2.5.
CHAPTER 10 / ANALYSIS OF QUANTITATIVE DATA 257
Ilege. She learns the mean, then dividing by the standard devia other variable. For example, Rita wants to know
i that both grade tion. For example, Suzette's z-score is 3.62 whether number of siblings is related to life ex
lde-point average 2.62 = 1.00/.50 = 2, whereas Jorge's z-score is pectancy. If the variables are independent, then
:andard deviation 3.64 - 3.24. = .40/.40 = 1. Thus, the employer people with many brothers and sisters have the
:-point average at learns that Suzette is two standard deviations same life expectancy as those who are only chil
tandard deviation above the mean in her college, whereas Jorge is dren. In other words, knowing how many broth
ts that grades at only one standard deviation above the mean for ers or sisters someone has tells Rita nothing
uzette from Kings his college. Although Suzette's absolute grade about the person's life expectancy.
-rage of 3.62, and point average is lower than Jorge's, relative to the Most researchers state hypotheses in terms
; a grade-point av students in each of their colleges Suzette's grades of a causal relationship or expected covariation;
:s took the same are much higher than Jorge's. if they use the null hypothesis, the hypothesis is
o adjust the grades that there is independence. It is used in formal
e two colleges (i.e., hypothesis testing and is frequently found in in
She calculates z ferential statistics (to be discussed).
udent's score from RESULTS WITH TWO VARIABLES Three techniques help researchers decide
whether a relationship exists between two vari
A Bivariate Relationship
ables: (1) a scattergram, or a graph or plot of the
Univariate statistics describe a single variable in relationship; (2) cross-tabulation, or a percent
isolation. Bivariate statistics are much more aged table; and (3) measures of association, or
valuable. They let a researcher consider two vari statistical measures that express the amount of
ables together and describe the relationship be covariation by a single number (e.g., correlation
tween variables. Even simple hypotheses require coefficient) .
two variables. Bivariate statistical analysis shows
.1 I I a relationship between variables-that is, things
+-1 +2 +3 Seeing the Relationship:
that appear together.
The Scattergram
Statistical relationships are based on two
the mean and add or
ideas: covariation and independence. Covaria What Is a Scattergram (or Scatterplot)? A
from it. One standard
tion means that things go together or are associ scattergram is a graph on which a researcher
I) when the mean is 7
ears is just 7 + 2, or 9
ated. To covary means to vary together; cases plots each case or observation, where each axis
) years. This is because with certain values on one variable are likely to represents the value of one variable. It is used for
of 2 years each (or 4 have certain values on the other one. For exam variables measured at the interval or ratio level,
of 7. My diagram now ple, people with higher values on the income rarely for ordinal variables, and never if either
variable are likely to have higher values on the variable is nominal. There is no fixed rule for
life expectancy variable. Likewise, those with which variable (independent or dependent) to
13 age in years lower incomes have lower life expectancy. This is place on the horizontal or vertical axis, but usu
~ usually stated in a shorthand way by saying that ally the independent variable (symbolized by the
+3 income and life expectancy are related to each letter X) goes on the horizontal axis and the de
other, or covary. We could also say that knowing pendent variable (symbolized by Y) on the verti
uel, who is 5 years old, one's income tells us one's probable life ex cal axis. The lowest value for each should be the
s Yashohda's z-score of pectancy, or that life expectancy depends on in lower left corner and the highest value should be
s old. I can read from z come. at the top or to the right.
reo For fractions, such as Independence is the opposite of covariation.
ply the same fraction to It means there is no association or no relation How to Construct a Scattergram. Begin with
an age of 1 2 is a z-score ship between variables. If two variables are inde the range of the two variables. Draw an axis with
pendent, cases with certain values on one the values of each variable marked and write
variable do not have any particular value on the numbers on each axis (graph paper is helpful).
258 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
Next, label each axis with the variable name and Form. Relationships can take three forms: in
put a title at the top. dependence, linear, and curvilinear. Inde
You are now ready for the data. For each pendence or no relationship is the easiest to see.
case, find the value of each variable and mark the It looks like a random scatter with no pattern, or
graph at a place corresponding to the two values. a straight line that is exactly parallel to the hori
For example, a researcher makes a scattergram zontal or vertical axis. A linear relationship
of years of schooling by number of children. He means that a straight line can be visualized in the
or she looks at the first case to see years of middle of a maze of cases running from one cor
schooling (e.g., 12) and at the number of chil ner to another. A curvilinear relationship means
dren (e.g., 3). Then he or she goes to the place on that the center of a maze of caseswould form a U
the graph where 12 for the "schooling" variable curve, right side up or upside down, or an S
and 3 for the "number of children" variable in curve.
tersect and puts a dot for the case.
The scattergram in Figure 10.5 is a plot of Direction. Linear relationships can have a pos
data for 33 women. It shows a negative relation itive or negative direction. The plot of a positive
ship between the years of education the woman relationship looks like a diagonal line from the
completed and the number of children she gave lower left to the upper right. Higher values on X
birth to. tend to go with higher values on Y, and vice
versa. The income and life expectancy example
What Can You Learn from the Scattergram? described a positive linear relationship.
A researcher can see three aspects of a bivariate A negative relationship looks like a line from
relationship in a scattergram: form, direction, the upper left to the lower right. It means that
and precision. higher values on one variable go with lower val
-----------------.
5.00
<,
..
s:::
III
:2 <,.
r-.r-,
:E 4.00
0
..
iii
-
:J
3.00
-..
Cll
z
0
III
.0
E
:J
2.00
- - - ~
-
Z
~
r-.
1.00 II
0.00
8.00 10.00 12.00 14.00 16.00 18.00 20.00 22.00
Years of Formal Education
CHAPTER 10 / ANALYSIS OF QUANTITATIVE DATA 259
: three forms: in ues on the other. For example, people with more Figure 10.6 is a raw count or frequency
irvilinear. Inde education are less likely to have been arrested. If table. Its cells contain a count of the cases. It is
the easiest to see. we look at a scattergram of data on a group of easy to make, but interpreting a raw count table
rith no pattern, or males where years of schooling (X axis) are plot is difficult because the rows or columns can have
rrallel to the hori ted by number of arrests (Y axis), we see that different totals, and what is of real interest is the
near relationship most cases (or men) with many arrests are in the relative size of cells compared to others.
,evisualized in the lower right, because most of them completed Researchers convert raw count tables into
ing from one cor few years of school. Most cases with few arrests percentaged tables to see bivariate relationships.
elationship means are in the upper left because most have had more There are three ways to percentage a table: by
eswould form a U schooling. The imaginary line for the relation row, by column, and for the total. The first two
le down, or an S ship can have a shallow or a steep slope. More are often used and show relationships.
advanced statistics provide precise numerical Is it best to percentage by row or column?
measures of the line's slope. Either can be appropriate. Let us first review the
ps can have a pos mechanics of percentaging a table. When calcu
leplot of a positive Precision. Bivariate relationships differ in their lating column percentages, compute the per
degree of precision. Precision is the amount of centage each cell is of the column total. This
onal line from the
spread in the points on the graph. A high level of includes the total column or marginal for the
-ligher values on X
es on Y, and vice precision occurs when the points hug the line column variable. For example, the first column
rpectancy example that summarizes the relationship. A low leveloc total is 26 (there are 26 people under age 30),
curs when the points are widely spread around and the first cell of that column is 20 (there are
ationship.
oks like a line from the line. Researchers can "eyeball" a highly pre 20 people under age 30 who agree). The per
ight. It means that cise relationship. They can also use advanced centage is 20/26 = 0.769 or 76.9 percent. Or, for
: go with lower val statistics to measure the precision of a relation the first number in the marginal, 37/101 = 0.366
ship in a way that is analogous to the standard = 36.6 percent (see Table 10.2). Except for
deviation for univariate statistics. rounding, the total should equal 100 percent.
Computing row percentages is similar.
Compute the percentage of each cell as a per
Bivariate Tables
centage of the row total. For example, using the
What Is a Bivariate Table? The bivariate con same cell with 20 in it, we now want to know
tingency table is widely used. It presents the what percentage it is of the row total of 37, or
same information as a scattergram in a more 20/37 = 0.541 = 54.1 percent. Percentaging by
condensed form. The data can be measured at row or column gives different percentages for a
any level of measurement, although interval and cell unless the marginals are the same.
ratio data must be grouped if there are many dif The row and column percentages let a re
ferent values. The table is based on cross-tabula searcher address different questions. The row
tion; that is, the cases are organized in the table percentage table answers the question. Among
on the basis of two variables at the same time. those who hold an attitude, what percentage
A contingency table is formed by cross-tabu come from each age group? It says of respon
lating two or more variables. It is contingent be dents who agree, 54.1 percent are in the under
cause the cases in each category of a variable get 30 age group. The column percentage table
distributed into each category of a second (or addresses the question: Among those in each age
additional) variable. The table distributes cases group, what percentage hold different attitudes?
into the categories of multiple variables at the It says that among those who are under 30, 76.9
same time and shows how the cases, by category percent agree. From the row percentages, a re
)0
of one variable, are "contingent upon" the cate searcher learns that a little over half ofthose who
gories of other variables. agree are under 30 years old, whereas from col
260 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
FIG U R E 1 0.6
-------.,
Age Group by Attitude about Changing the Drinking
Age, Raw Count Table
Agree 20 10 4 3 37
No opinion 3 (d) 10 10 2 25
Disagree 3 ~ 21 lUi ---.3..9.
Total (c)
-----"-
26 25 +
- - - - - - - - - - - - - - - ,..-------_._._
35 15 101
Missing cases (f) = 8. (e)
umn percentages, the researcher learns that Unfortunately, there is no "industry stan
among the under-30 people, over three-quarters dard" for putting independent and dependent
agree. One way of percentaging tells about peo variables in a percentage table as row or column,
ple who have specific attitudes; the other tells or for percentage by row and column. A major
about people in specific age groups. ity of researchers place the independent variable
A researcher's hypothesis may imply look as the column and percentage by column, but a
ing at row percentages or the column percent large minority put the independent variable as
ages. When beginning, calculate percentages the row and percentage by row.
each way and practice interpreting, or figuring
out, what each says. For example, my hypothesis Reading a Percentaged Table. Once you un
is that age affects attitude, so column percent derstand how a table is made, reading it and fig
ages are most helpful. However, if my interest uring out what it saysare much easier. To read a
was in describing the age make-up of groups of table, first look at the title, the variable labels,
people with different attitudes, then row per and any background information. Next, look at
centages are appropriate. the direction in which percentages have been
CHAPTER 10 / ANALYSIS OF QUANTITATIVE DATA 261
computed-in rows or columns. Notice that the centages are computed. A rule of thumb is to
percentaged tables in Table 10.2 have the same compare across rows if the table is percentaged
title. This is because the same variables are used. down (i.e., by column) and to compare up and
It would have helped to note how the data were down in columns if the table is percentaged
percentaged in the title, but this is rarely done. across (i.e., by row).
Sometimes, researchers present abbreviated ta For example, in row-percentaged Table
bles and omit the 100 percent total or the mar 10.2, compare columns or age groups. Most of
ginals, which adds to the confusion. It is best to those who agree are in the youngest group, with
include all the parts of a table and clear labels. the proportion declining as age increases. Most
Researchers read percentaged tables to no-opinion people are in the middle-age groups,
make comparisons. Comparisons are made in whereas those who disagree are older, especially
the opposite direction from that in which per- in the 46-to-60 group. When reading column-
TAB L E 1 0.2 Age Group by Attitude about Changing the Drinking Age. •
Percentaged Tables
Column-Percentaged Table
) 40 100
) 20 100 MORE THAN TWO VARIABLES
2 8 100 Agree 26.2 (37)
Statistical Control
5 5 100 No opinion 44.5 (25)
Showing an association or relationship between
Disagree 61.9 (39)
two variables is not sufficient to say that an in
Missing cases ~ 8 dependent variable causes a dependent variable.
ichooling In addition to temporal order and association, a
researcher must eliminate alternative explana
tions-explanations that can make the hypothe
gest that the mean age of those who disagree is sized relationship spurious. Experimental
much higher than for those who agree or have researchers do this by choosing a research design
5 5 100 no opinion. that physically controls potential alternative ex
2 8 100 planations for results (i.e., that threaten internal
validity).
~O 20 100 Measures of Association
In nonexperimental research, a researcher
.0 40 100 A measure of association is a single number that
controls for alternative explanations with statis
expresses the strength, and often the direction,
tics. He or she measures possible alternative ex
. of a relationship. It condenses information
planations with controlvariables, then examines
about a bivariate relationship into a single
the control variables with multivariate tables and
nships. A positive re number. statistics that help him or her decide whether a
te variable increases, There are many measures of association. bivariate relationship is spurious. They also
e relationship means The correct one depends on the level of mea show the relative size of the effect ofmultiple in
;,the other decreases. surement. Many measures are called by letters of dependent variables on a dependent variable.
the Greek alphabet. Lambda, gamma, tau, chi A researcher controls for alternative expla
. Percentages. Re (squared), and rho are commonly used mea nations in multivariate (more than two vari
ition in another kind sures. The emphasis here is on interpreting the ables) analysis by introducing a third (or
:asure of central ten measures, not on their calculation. In order to sometimes a fourth or fifth) variable. For exam
.stead of percentages. understand each measure, you will need to com ple, a bivariate table shows that taller teenagers
e is nominal or ordi plete a beginning statistics course. like sports more than shorter ones do. But the
I. at the interval or ra If there is a strong association or relation bivariate relationship between height and atti
nilar measure) of the ship, then few errors are made predicting a sec tude toward sports may be spurious because
5 presented for each ond variable on the basis of knowledge of the teenage males are taller than females, and males
ordinal variable. All first, or the proportion of errors reduced is large. tend to like sports more than females. To test
ordinal or nominal A large number of correct guesses suggests that whether the relationship is actually due to sex, a
le mean is calculated the measure of association is a nonzero number researcher must control for gender; in other
ile category from the if an association exists between the variables. words, effects of sex are statistically removed.
Table 10.5 describes five commonly used bivari Once this is done, a researcher can see whether
lean age of people in ate measures of association. Notice that most the bivariate relationship between height and at
ries. The results sug- range from -1 to + 1, with negative numbers in titude toward sports remains.
264 PART TWO / CONDUCTING QUANTITATIVE RESEARCH
Lambda is used for nominal-level data. It is based be used only for data measured at the interval or
on a reduction in errors based on the mode and ratio level. Rho is used for the mean and standard
ranges between 0 (independence) and 1.0 deviation of the variables and tells how far cases
(perfect prediction or the strongest possible are from a relationship (or regression) line in a
relationship). scatterplot. Rho ranges from -1 .0 to + 1.0, with 0
Gamma is used for ordinal-level data. It is based on meaning no association. Ifthe value of rho is
comparing pairs of variable categories and seeing squared, sometimes called R-squared, it has a
whether a case has the same rank on each. Gamma unique proportion reduction in error meaning. R
ranges from -1.0 to + 1.0, with 0 meaning no squared tells how the percentage in one variable
association. (e.g., the dependent) is accounted for, or
explained by, the other variable (e.g., the
Tau is also used for ordinal-level data. It is based independent). Rho measures linear relationships
on a different approach than gamma and takes only. It cannot measure nonlinear or curvilnear
care of a few problems that can occur with gamma. relationships. For example, a rho of zero can
Actually, there are several statistics named tau (it indicate either no relationship or a curvilinear
is a popular Greek letter), and the one here is relationship.
Kendall's tau. Kendall's tau ranges from -1.0 to
+ 1.0, with 0 meaning no association. Chi-squared has two different uses. It can be used
as a measure of association in descriptive statistics
Rho is also called Pearson's product moment like the others listed here, or in inferential
correlation coefficient (named after the famous statistics. Inferential statistics are brieflydescribed
statistician Karl Pearson and based on a product next. As a measure of association, chi-squared can
moment statistical procedure). It is the most be used for nominal and ordinal data. It has an
commonly used measure of correlation, the upper limit of infinityand a lower limit of zero,
correlation statistic people mean if they use the meaning no association.
term correlation without identifying it further. It can
A researcher controls for a third variable by persists. This means that tall males and tall fe
seeing whether the bivariate relationship persists males both like sports more than short males
within categories of the control variable. For ex and short females do. In other words, the con
ample, a researcher controls for sex, and the re trol variable has no effect.When this is so, the bi
lationship between height and sports attitude variate relationship is not spurious.
CHAPTER 10 / ANALYSIS OF QUANTITATIVE DATA 265
- - - - I. . If the bivariate relationship weakens or dis in the partial. Thus, it is possible to break apart a
appears after the control variable is considered, bivariate table to form partials, or combine the
it means that tall males are no more likely than partials to restore the initial bivariate table.
t the interval or short males to like sports, and tall females are no Trivariate tables have three limitations.
an and standard more likely to like sports than short females. It First, they are difficult to interpret if a control
s howfar cases indicates that the initial bivariate relationship is variable has more than four categories. Second,
sion) line in a spurious and suggests that the third variable, sex, control variables can be at any level of measure
.0 to +1 .0, with 0 and not height, is the true cause of differences in ment, but interval or ratio control variables
ue of rho is attitudes toward sports. must be grouped (i.e., converted to an ordinal
ared, it has a Statistical control is a key idea in advanced level), and how cases are grouped can affect the
rror meaning. R statistical techniques. A measure of association interpretation of effects. Finally, the total num
~ in one variable like the correlation coefficient only suggests a re ber of cases is a limiting factor because the cases
-d for, or lationship. Until a researcher considers control are divided among cells in partials. The number
~e.g., the variables, the bivariate relationship could be of cells in the partials equals the number of cells
.ar relationships spurious. Researchers are cautious in interpret in the bivariate relationship multiplied by the
r or curvilnear ing bivariate relationships until they have con number of categories in the control variable. For
of zero can sidered control variables. example, a control variable has three categories,
, a curvilinear and a bivariate table has 12 cells, so the partials
have 3 X 12 = 36 cells. An average of five cases
The Elaboration Model of
es. It can be used per cell is recommended, so the researcher will
escriptive statistics Percentaged Tables
need 5 X 36 = 180 cases at minimum.
inferential Constructing Trivariate Tables. In order to For three variables, three bivariate tables are
'e briefly described meet all the conditions needed for causality, re logically possible. In the example, the combina
n, chi-squared can searchers want to "control for" or see whether tions are (1) gender by attitude, (2) age group by
data. It has an an alternative explanation explains away a causal attitude, and (3) gender by age group. The par
er limit of zero, relationship. If an alternative explanation ex tials are set up on the basis of the initial bivariate
plains a relationship, then the bivariate relation relationship. The independent variable in each
ship is spurious. Alternative explanations are is "age group" and the dependent variable is "at
operationalized as third variables, which are titude." "Gender" is the control variable. Thus,
called control variables because they control for the trivariate table consists of a pair of partials,
alternative explanation. each showing the age/attitude relationship for a
One way to take such third variables into given gender.
consideration and see whether they influence the A researcher's theory suggests the hypothe
bivariate relationship is to statistically introduce sis in the initial bivariate relationship; it also tells
control variables using trivariate or three-vari him or her which variables provide alternative
able tables. Trivariate tables differ slightly from explanations (i.e., the control variables). Thus,
bivariate tables; they consist ofmultiple bivariate the choice of the control variable is based on
tables. theory.
A trivariate table has a bivariate table of the The elaboration paradigm is a system for
independent and dependent variable for each reading percentaged trivariate tables.:' It de
category of the control variable. These new ta scribes the pattern that emerges when a control
1 males and tall fe bles are called partials. The number of partials variable is introduced. Five terms describe how
e than short males depends on the number of categories in the con the partial tables compare to the initial bivariate
ier words, the con trol variable. Partial tables look like bivariate ta table, or how the original bivariate relationship
len this is so, the bi bles, but they use a subset ofthe cases. Only cases changes after the control variable is considered.
urious. with a specific value on the control variable are The examples of patterns presented here show
266 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
strong cases. More advanced statistics are impact on abortion attitude. The control vari
needed when the differences are not as obvious. able is an intervening variable, which helps you
The replication pattern is the easiest to un interpret the meaning of the complete relation
derstand. It is when the partials replicate or re ship.
produce the same relationship that existed in the The explanation pattern looks the same as in
bivariate table before considering the control terpretation. The difference is the temporal order
variable. It means that the control variable has of the control variable. In this pattern, a control
no effect. variable comes before the independent variable
The specification pattern is the next easiest in the initial bivariate relationship. For example,
pattern. It occurs when one partial replicates the the original relationship is between religious up
initial bivariate relationship but other partials do bringing and abortion attitude, but now gender
not. For example, you find a strong (negative) is the control variable. Gender comes before reli
bivariate relationship between automobile acci gious upbringing because one's sex is fixed at
dents and college grades. You control for gender birth. The explanation pattern changes how a re
and discover that the relationship holds only for searcher explains the results. It implies that the
males (i.e., the strong negative relationship was initial bivariate relationship is spurious.
in the partial for males, but not for females). The suppressor variable pattern occurs when
This is specification because a researcher can the bivariate tables suggest independence but a
specify the category of the control variable in relationship appears in one or both of the par
which the initial relationship persists. tials. For example, religious upbringing and
The control variable has a large impact in abortion attitude are independent in a bivariate
both the interpretation and explanation pat table. Once the control variable "region of the
terns. In both, the bivariate table shows a rela country" is introduced, religious upbringing is
tionship that disappears in the partials. In other associated with abortion attitude in the partial
words, the relationship appears to be indepen tables. The control variable is a suppressor vari
dence in the partials. The two patterns cannot be able because it suppressed the true relationship.
distinguished by looking at the tables alone. The The true relationship appears in the partials.
difference between them depends on the loca (See Table 10.6 for a summary of the elabora
tion of the control variable in the causal order of tion paradigm.)
variables. Theoretically, a control variable can be
in one of two places, either between the original
Multiple Regression Analysis
independent and dependent variables (i.e., the
control variable is intervening), or before the Multiple regression is a statistical technique
original independent variable. whose calculation is beyond the level in this
The interpretation pattern describes the situ book. Although it is quickly computed by the
ation in which the control variable intervenes appropriate statistics software, a background in
between the original independent and depen statistics is needed to prevent making errors in
dent variables. For example, you examine a rela its calculation and interpretation. It requires in
tionship between religious upbringing and terval- or ratio-level data. It is discussed here for
abortion attitude. Political ideology is a control two reasons. First, it controls for many alterna
variable. You reason that religious upbringing tive explanations and variables simultaneously
affects current political ideology and abortion (it is rarely possible to use more than one control
attitude. You theorize that political ideology is variable at a time using percentaged tables). Sec
logically prior to an attitude about a specific is ond, it is widely used in sociology, and you are
sue, like abortion. Thus, religious upbringing likely to encounter it when reading research re
causes political ideology, which in turn has an ports or articles.
CHAPTER 10 I ANALYSIS OF QUANTITATIVE DATA 267
he control vari
which helps you
--------------------.
TAB L E 1 0.6 Summary of the Elaboration Paradigm
.mplete relation
re temporal order
Specification Bivariate relationship is only seen in one of the partial tables.
iattern, a control Interpretation Bivariate relationship weakens greatly or disappears in the partial tables (control
lip. For example, Explanation Bivariate relationship weakens greatly or disappears in the partial tables (control
, but now gender Suppressor variable No bivariate relationship; relationship appears in partial tables only.
TAB L E 1 0 . 8
CHAPTER 10 I ANALYSIS OF QUANTITATIVE DATA
deology Index
I
Univariate Frequency distribution, measure Describe one variable.
of central tendency, standard
deviation, z-score
Bivariate Correlation,percentage table, Describe a relationship or the
chi-square association between two variables.
-.19 MuItivariate Elaboration paradigm, multiple Describe relationships among several
.01 regression variables, or see howseveral
-.44 independent variables have an effect
___________________m m_.__variable.
on a dependent l ~ _
.23
-.39
tion, and decide whether differences in results Statistical Significance
(e.g., between the mean scores of two groups)
are big enough to indicate that a relationship re Statistical significance means that results are not
ally exists. Inferential statistics use probability likely to be due to chance factors. It indicates the
theory to test hypotheses formally, permit infer probability of finding a relationship in the sam
dex. The multiple
income and reli ences from a sample to a population, and test ple when there is none in the population. Be
effects, education whether descriptive results are likely to be due cause probability samples involve a random
~ age no effect. All to random factors or to a real relationship. process, it is always possible that sample results
1gether have a 38 This section explains the basic ideas of in will differ from a population parameter. A re
~ a person's politi ferential statistics but does not deal with inferen searcher wants to estimate the odds that sample
The example sug tial statistics in any detail. This area is more results are due to a true population parameter
requent religious complex than descriptive statistics and requires a or to chance factors of random sampling. Statis
.esidence are posi background in statistics. tical significance uses probability theory and
ervative opinions, Inferential statistics rely on principles specific statistical tests to tell a researcher
ition is associated from probability sampling, where a researcher whether the results (e.g., an association, a differ
rpact of income is uses a random process (e.g., a random number ence between two means, a regression coeffi
he impact of living table) to select cases from the entire popula cient) are produced by random error in random
ve been examining tion. Inferential statistics are a precise way to sampling.
ole 10.8); next, we talk about how confident a researcher can be Statistical significance only tells what is
-ntial statics. when inferring from the results in a sample to likely. It cannot prove anything with absolute
the population. certainty. It states that particular outcomes are
You have already encountered inferential more or less probable. Statistical significance is
statistics if you have read or heard about "statis not the same as practical, substantive, or theoret
ICS tical significance" or results "significant at the ical significance. Results can be statistically sig
.05 level." Researchers use them to conduct var nificant but theoretically meaningless or trivial.
II Statistics
ious statistical tests (e.g., a r-test or an F-test). For example, two variables can have a statistically
do more than de Statistical significance is also used in formal hy significant association due to coincidence, with
hypotheses, know pothesis testing, which is a precise way to decide no logical connection between them (e.g.,length
I true in a popula- whether to accept or to reject a null hypothesis." of fingernails and ability to speak French).
270 PART TWO I CONDUCTING QUANTITATIVE RESEARCH
Almost every social researcher who needs to calcu The most widely used program for statistics in the
late many statistics does so with a computer pro social sciences in SPSS. Its advantages are that social
gram, often using a basic spreadsheet program, researchers used it extensively for over three
such as Excel. Unfortunately, spreadsheets are de decades, it includes many ways to manipulate quan
signed for accounting and bookkeeping functions. titative data, and it contains most statistical mea
They include statistics, but are clumsy and limited sures. A disadvantage is that it can take a long time
for that purpose. There are many computer pro to learn because of its many options and complex
grams designed for calculating general statistics. statistics. Also, it is expensive to purchase unless the
The marketplace can be confusing to a beginner, user gets an inexpensive, "stripped down" student
for products evolve rapidly with changing computer version included with a textbook or workbook.
technology. As computer technology makes using a statistics
In recent years, the software has become less de program easier, the danger increases that some peo
manding for a user. The most popular programs in ple will use the programs, but not understand statis
the social sciences are Minitab, Microcase, and SPSS tics or what the programs are doing. They can easily
(Statistical Package for the Social Sciences). Others violate basic assumptions required by a statistical
include SAS (Statistical Analysis System), STATIS procedure, use the statistics improperly, and pro
TICAby StratSoft, and Strata. Many began as simple, duce results that are pure nonsense but that look
low-cost programs for research purposes. very technically sophisticated.
create confusion, so the number 1 is usually the 4. In formal hypothesis testing, researchers test the
lowest value. null hypothesis. They usually want to reject the null
2. There are other statistics to measure a special kind because rejection of the null indirectly supports
of mean for ordinal data and for other special sit the alternative hypothesis to the null, the one they
uations, which are beyond the level of discussion deduced from theory as a tentative explanation.
in this book.
3. For a discussion of the elaboration paradigm and
its history, see Babbie (1998:393-401) and Rosen
berg (1968).
ing, researchers test the
ly want to reject the null
aull indirectly supports
to the null, the one they
tentative explanation.
Field Research
Introduction
Research Questions Appropriate for Field Research
The Logic of Field Research
What Is Field Research?
Building Rapport
Maintaining Relations
Taking Notes
Data Quality
Informants
Interview Context
Confidentiality
Conclusion
275
276 PART THREE / CONDUCTING QUALITATIVE RESEARCH
et of techniques to
s various methods
1researcher is a re
who has ingenuity
A field researcher does the following: 1. Prepare oneself, read the literature, and
or his feet while in
defocus.
1. Observes ordinary events and everyday activi 2. Select a field site and gain access to it.
1 naturalism, which ties as they happen in natural settings, in addi
phenomena (e.g., 3. Enter the field and establish social relations with
tion to any unusual occurrences
Naturalism involves members.
2. Becomes directly involved with the people being
natural settings, not 4. Adopt a social role, learn the ropes, and get
studied and personally experiences the process
~archer-created set along with members.
of daily social life in the field setting
e field and outside 5. Watch, listen, and collect quality data.
3. Acquires an insider's point of view while main
laboratory, or class- 6. Begin to analyze data and to generate and eval
taining the analytic perspective or distance of an
outsider uate working hypotheses.
is to examine social 7. Focus on specific aspects of the setting and use
4. Uses a variety of techniques and social skills in a
ple perspectives in theoretical sampling.
flexible manner as the situation demands
. she wants to get in 8. Conduct field interviews with member infor
. members and then 5. Produces data in the form of extensive written
notes, as well as diagrams, maps, or pictures to mants.
earch viewpoint. To
provide very detailed descriptions 9. Disengage and physically leave the setting.
hes perspectives and
I1tiple points of view 6. Sees events holistically (e.g., as a whole unit, not 10. Complete the analyses and write the research
in pieces) and individually in their social context report.
r conducted by a sin 7. Understands and develops empathy for mem
bers in a field setting, and does not just record Note: There is no fixed percentage of time needed for each
iall teams have been
step. For a rough approximation, Junker (1960:12)
'he researcher is di "cold" objective facts
suggested that, once in the field, the researcher should
t of the social world 8. Notices both explicit (recognized, conscious. expect to spend approximately one-sixth of his or her time
nal characteristics are spoken) and tacit (less recognized, implicit, un observing, one-third recording data, one-third of the time
esearcher's direct in spoken) aspects of culture analyzing data, and one-sixth reporting results. Also see
en has an emotional 9. Observes ongoing social processes without up Denzin (1989:176) for eight steps of field research.
be fun and exciting, setting, disrupting, or imposing an outside point
s personal life, physi of view
dl-being. More than 10. Copes with high levels of personal stress, uncer
ch, it reshapes friend basis of their value for providing information. In
tainty, ethical dilemmas, and ambiguity
entity, and personal the beginning, the researcher expects little con
trol over data and little focus. Once socialized to
the setting, however, he or she focuses the in
quiry and asserts control over the data.
:h Project Flexibility. Field researchers rarely follow
rolvement mean that fixed steps. In fact, flexibility is a key advantage Getting Organized in the Beginning. Human
tured than quantita of field research, which lets a researcher shift di and personal factors can playa role in any re
, it essential for a re rection and follow leads. Good field researchers search project, but they are crucial in field re
zed and prepared for recognize and seize opportunities, "play it by search. Field projects often begin with chance
rat the steps of a pro
ear," and rapidly adjust to fluid social situations. occurrences or a personal interest. Field re
termined but serve as A field researcher does not begin with a set searchers can begin with their own experiences,
r road map (see Box of methods to apply or explicit hypotheses to such as working at a job, having a hobby, or be
test. Rather, he or she chooses techniques on the ing a patient or an activist.
280 PART THREE / CONDUCTING QUALITATIVE RESEARCH
Field researchers use the skills of careful may be personally transformed by the field ex
looking and listening, short-term memory, and perience. Some adopt new values, interests, and
regular writing. Before entering the field, a new moral commitments, or change their religion or
researcher practices observing the ordinary de political ideology.>
tails of situations and writing them down. At
tention to details and short-term memory can
improve with practice. Likewise, keeping a daily
CHOOSING A SITE AND GAINING
diary or personal journal is good practice for
ACCESS
writing field notes.
As with all social research, reading the Although a field research project does not pro
scholarly literature helps the researcher learn ceed by fixed steps, some common concerns
concepts, potential pitfalls, data collection meth arise in the early stages. These include selecting a
ods, and techniques for resolving conflicts. In site and gaining access to the site, entering the
addition, a field researcher finds diaries, novels, field, learning the ropes, and developing rapport
journalistic accounts, and autobiographies use with members in the field.
ful for gaining familiarity and preparing emo
tionally for the field.
Selecting a Site and Entering
Field research begins with a general topic,
not specific hypotheses. A researcher does not Where to Observe. Field researchers talk about
get locked into any initial misconceptions. He or doing research on a setting, or field site, but this
she needs to be well informed but open to dis term is misleading. A site is the context in which
covering new ideas. Finding the right questions events or activities occur, a socially defined terri
to ask about the field takes time. tory with shifting boundaries. A social group
A researcher first empties his or her mind of may interact across several physical sites. For ex
preconceptions. The researcher should move ample, a college football team may interact on
outside his or her comfortable social niche to ex the playing field, in the locker room, in a dormi
perience as much as possible in the field without tory, at a training camp, or at a local hangout.
betraying a primary commitment to being a re The team's field site includes all five locations.
searcher. The field site and research question are
Another preparation for field research is bound together, but choosing a site is not the
self-knowledge. A field researcher needs to know same as focusing on a case for study. A case is a
himself or herself and reflect on personal experi social relationship or activity; it can extend be
ences. He or she can expect anxiety, self-doubt, yond the boundaries of the site and have links to
frustration, and uncertainty in the field. Espe other social settings. A researcher selects a site,
cially in the beginning, the researcher may feel then identifies cases to examine within it-for
that he or she is collecting the wrong data and example, how football team members relate to
may suffer emotional turmoil, isolation, and authority figures.
confusion. He or she often feels doubly mar Selecting a field site is an important deci
ginal: an outsider in the field setting and also dis sion, and researchers take notes on the site selec
tant from friends, family, and other researchers.' tion processes. Three factors are relevant when
The relevance of a researcher's emotional make choosing a field research site: richness of data,
up, personal biography, and cultural experiences unfamiliarity, and suitability.P Some sites are
makes it important to be aware of his or her per more likely than others to provide rich data.
sonal commitments and inner conflicts (see Box Sites that present a web of social relations, a va
11.3). Fieldwork can have a strong impact on a riety of activities, and diverse events over time
researcher's identity and outlook. Researchers provide richer, more interesting data. Beginning
CHAPTER 11 / FIELD RESEARCH 281
Eliasoph (1 998) conducted field research on several shirts and jeans, women in curly perms and tiered
groups in a California community to understand how flounces of lace or denim skirts, or jeans, and belts with
Americans avoid political expression. One was a so their names embroidered in glitter on the back.
) GAINING cial club. Eliasoph describes herself as an "urban, bi (1998:92)
coastal, bespectacled, Jewish, Ph.D. candidate from a
long line of communists, atheists, liberals, book-read Eliasoph introduced herself as a student. During
Iject does not pro
ers, ideologues, and arguers" (p. 270). The social her two years of research, she endured smoke-filled
ommon concerns
club's world was very foreign to her. The social club, rooms as well as expensive beer and bottled-water
include selecting a
the Buffalos, centered on country and western music prices; attended a wedding and many dance lessons;
e site, entering the
at a bar, the Silverado Club. She describes it: and participated in countless conversations and
:levelopingrapport heard many abusive sexist/racist jokes. She listened,
The Silverado huddled on a vast, rutted parking lot on asked questions, observed, and took notes in the
whatwas oncewetlands and nowwas a truck stop, a mile bathroom. When she returned home after hours
ring and a half from Amargo's [town name) nuclear battle with club members, it was to a university crowd who
ship station. Occasional gulleys of salt water cattails had little understanding of the world she was study
;earchers talk about pokedthrough the wide flat miles of pavedmalls andgas ing. For them, witty conversation was central and
or field site, but this stations. Giant four-wheeled-drive vehicles filled the being bored was to be avoided. The club members
he context in which parking lot, making my miniature Honda looklike a toy. used more nonverbal than verbal communication
iciallydefined terri ... Inside the windowless Silverado, initial blinding dark and being bored, or sitting and doing nothing, was
ies. A social group nessgave way to a huge Confederate flag pinned up be just fine. The research forced Eliasoph to reexamine
hysical sites. For ex hindthe bandstand, the standardcollection of neon beer her own views and tastes, which she had taken for
un may interact on signs and beer mirrors, men in cowboys hats, cowboys granted.
T room, in a dormi
at a local hangout.
s all five locations.
earch question are field researchers should choose an unfamiliar and regulations in institutions (e.g., public
ing a site is not the setting. It is easier to see cultural events and so schools, hospitals, prisons, etc.) restrict access.
for study. A case is a cial relations in a new site. When "casing" possi In addition, institutional review boards may
ty; it can extend be ble field sites, one must consider such practical limit field research on ethical grounds.
.ite and have links to issues as the researcher's time and skills, serious
archer selects a site, conflicts among people in the site, the re Level of Involvement. Field roles can be
mine within it-for searcher's personal characteristics and feelings, arranged on a continuum by the degree of de
1 members relate to and access to parts of a site. tachment or involvement a researcher has with
A researcher's ascriptive characteristics (e.g., members. At one extreme is a detached outsider;
an important deci age, gender, race) can limit access. Physical ac at the other extreme is an intimately involved
otes on the site selec cess to a site can be an issue. Sites are on a con insider.
rs are relevant when tinuum, with open and public areas (e.g., public The field researcher's level of involvement
ite: richness of data, restaurants, airport waiting areas, etc.) at one depends on negotiations with members,
ity.6 Some sites are end and closed and private settings (e.g., private specifics of the field setting, the researcher's
) provide rich data. firms, clubs, activities in a person's home, etc.) at personal comfort, and the particular role
.ocial relations, a va the other. A researcher may find that he or she is adopted in the field. Many move from outsider
rse events over time not welcome or not allowed on the site, or there to insider levels with more time in the field.
sting data. Beginning are legal and political barriers to access. Laws Each level has its advantages and disadvantages.
282 PART THREE / CONDUCTING QUALITATIVE RESEARCH
Different field researchers advocate different findings might provide evidence for someone to
levels of involvement. criticize them.
Roles at the outsider end of the continuum Dealing with gatekeepers is a recurrent issue
reduce the time needed for acceptance, make as a researcher enters new levels or areas. In ad
overrapport less an issue, and can sometimes dition, a gatekeeper can shape the direction of
help members open up. They facilitate detach research. In some sites, gatekeeper approval cre
ment and protect the researcher's self-identity. A ates a stigma that inhibits the cooperation of
researcher feels marginal. Although there is less members. For example, prisoners may not be
risk of "going native," he or she is also less likely cooperative if they know that the prison warden
to know an insider's experience and misinter gave approval to the researcher.
pretation is more likely. To really understand so
cial meaning for those being studied, the field
Strategy for Entering
researcher must participate 'm \ne se\tmg, as o\n
ers do. Entering a field site requires having a flexible
By contrast, roles at the insider end of the strategy or plan of action, negotiating access and
continuum facilitate empathy and sharing of a relations with members, and deciding how
member's experience. The goal of fully experi much to disclose about the research to field
encing the intimate social world of a member is members or gatekeepers.
achieved. Nevertheless, a lack of distance from,
too much sympathy for, or overinvolvement Planning. Entering and gaining access to a
with members is likely. A researcher's reports field site is a process that depends on common
may be questioned, data gathering is difficult, sense judgment and social skills. Field sites usu
there can be a dramatic impact on the re ally have different levels or areas, and entry is an
searcher's self, and the distance needed for issue for each. Entry is more analogous to peel
analysis may be hard to attain. ing the layers of an onion than to opening a
door. Moreover, bargains and promises of entry
Gatekeepers. A gatekeeper is someone with the may not remain stable over time. A researcher
formal or informal authority to control access to needs fallback plans or may have to return later
a site? It can be the thug on the corner, an ad for renegotiation. Because the specific focus of
ministrator of a hospital, or the owner of a busi research may not emerge until later in the re
ness. Informal public areas (e.g., sidewalks, search process or may change, it is best to avoid
public waiting rooms, etc.) rarely have gatekeep being locked into specifics by gatekeepers.
ers; formal organizations have authorities from
whom permission must be obtained. Negotiation. Social relations are negotiated
Field researchers expect to negotiate with and formed throughout the process of field
gatekeepers and bargain for access. The gate work." Negotiation occurs with each new mem
keepers may not appreciate the need for concep ber until a stable relationship develops to gain
tual distance or ethical balance. The researcher access, develop trust, obtain information, and
must set nonnegotiable limits to protect research reduce hostile reactions. The researcher expects
integrity. If there are many restrictions initially, a to negotiate and explain what he or she is doing
researcher can often reopen negotiations later, over and over in the field (see Normalizing So
and gatekeepers may forget their initial demands cial Research later in the chapter).
as trust develops. It is ethically and politically as Deviant groups and elites often require spe
tute to call on gatekeepers. Researchers do not cial negotiations for gaining access. To gain ac
expect them to listen to research concerns or cess to deviant subcultures, field researchers have
care about the findings, except insofar as these used contacts from the researcher's private life,
CHAPTER 11 / FIELD RESEARCH 283
ence for someone to gone to social welfare or law-enforcement agen who we are-the type of person we are or would
cies where the deviants are processed, advertised like to be-through our physical appearance,
rsis a recurrent issue for volunteers, offered a service (e.g., counseling) what we say, and how we act. The presentation
evels or areas. In ad in exchange for access, or gone to a location of self sends a symbolic message. It may be, "I'm
ape the direction of where deviants hang out and joined a group. a serious, hard-working student," "I'm a warm
keeper approval ere and caring person," "I'm a cool jock," or ''I'm a
the cooperation of Disclosure. A researcher must decide how rebel and party animal." Many selves are possi
'isoners may not be much to reveal about himself or herself and the ble, and presentations of selves can differ de
It the prison warden research project. Disclosing one's personal life, pending on the occasion.
her. hobbies, interests, and background can build A field researcher is conscious ofthe presen
trust and close relationships, but the researcher tation of self in the field. For example, how
will also lose privacy, and he or she needs to en should he or she dress in the field? The best
sure that the focus remains on events in the field. guide is to respect both oneself and those being
~es having a flexible A researcher also decides how much to dis studied. Do not overdress so as to offend or
egotiating access and close about the research project. Disclosure stand out, but copying the dress of those being
and deciding how ranges on a continuum from fully covert re studied is not always necessary. A professor who
he research to field search, in which no one in the field is aware that studies street people does not have to dress or
research is taking place, to the opposite end, act like one; dressing and acting informally is
where everyone knows the specifics of the re sufficient. Likewise, more formal dress and pro
gaining access to a search project. The degree and timing of disclo fessional demeanor are required when studying
epends on common sure depends on a researcher's judgment and corporate executives or top officials.
;kills. Field sites usu particulars in the setting. Disclosure may unfold A researcher must be aware that self-presen
areas, and entry is an over time as the researcher feels more secure. tation will influence field relations to some de
~e analogous to peel Researchers disclose the project to gatekeep gree. It is difficult to present a highly deceptive
1 than to opening a ers and others unless there is a good reason for front or to present oneself in a way that deviates
nd promises of entry not doing so, such as the presence of gatekeepers sharply from the person one is ordinarily.
:r time. A researcher who would seriously limit or inhibit research for
r have to return later illegitimate reasons (e.g., to hide graft or corrup Researcher as Instrument. The researcher is
the specific focus of tion). Even in these cases, a researcher may dis the instrument for measuring field data. This has
until later in the re close his or her identity as a researcher, but may two implications. First, it puts pressure on the
ige, it is best to avoid pose as one who seems submissive, harmless, researcher to be alert and sensitive to what hap
by gatekeepers. and interested in nonthreatening issues. pens in the field and to be disciplined about
recording data. Second, it has personal conse
lions are negotiated quences. Fieldwork involves social relationships
learning the Ropes
he process of field and personal feelings. Field researchers are flex
with each new mem After a field site is selected and access obtained, ible about what to include as data and admit
hip develops to gain researchers must learn the ropes, develop rap their own subjective insights and feelings. Per
in information, and port with members, adopt a role in the setting, sonal, subjective experiences are part of field
re researcher expects and maintain social relations. Before con data. They are valuable both in themselves and
tat he or she is doing fronting such issues, the researcher should ask: for interpreting events in the field. Instead of try
see Normalizing So How will I present myself? What does it mean ing to be objective and eliminate personal reac
apter). for me to be a "measurement instrument"? How tions, field researchers treat their feelings toward
tes often require spe can I assume an "attitude of strangeness"? field events as data.
g access. To gain ac Field research can heighten a researcher's
field researchers have Presentation ofSelf. People explicitly and im awareness of personal feelings. For example, a
earcher's private life, plicitly present themselves to others. We display researcher may not be fully aware of personal
284 PART THREE / CONDUCTING QUALITATIVE RESEARCH
feelings about nudity until he or she is in a nud Strangeness also encourages a researcher to
ist colony, or about personal possessions until he reconsider his or her own social world. Immer
or she is in a setting where others "borrow" sion in a different setting breaks old habits of
many items. The researcher's own surprise, in thought and action. He or she finds reflection
dignation, or questioning then may become an and introspection easier and more intense when
opportunity for reflection and insight. encountering the unfamiliar, whether it is a dif
ferent culture or a familiar culture seen through
An Attitude ofStrangeness. It is hard to recog a stranger's eyes.
nize what we are very close to. The everyday
world we inhabit is filled with thousands of de
Building Rapport
tails. If we paid attention to everything all the
time, we would suffer from severe information A field researcher builds rapport by getting along
overload. We manage by ignoring much of what with members in the field. He or she forges a
is around us and by engaging in habitual think friendly relationship, shares the same language,
ing. Unfortunately, we fail to see the familiar as and laughs and cries with members. This is a
distinctive, and assume that others experience step toward obtaining an understanding of
reality just as we do. We tend to treat our own members and moving beyond understanding to
way of living as natural or normal. empathy-that is, seeing and feeling events from
Field research in familiar surroundings is another's perspective.
difficult because of a tendency to be blinded by It is not always easy to build rapport. The
the familiar. By studying other cultures, re social world is not all in harmony, with warm,
searchers encounter dramatically different as friendly people. A setting may contain fear, ten
sumptions about what is important and how sion, and conflict. Members may be unpleasant,
things are done. This confrontation of cultures, untrustworthy, or untruthful; they may do
or culture shock, has two benefits: It makes it things that disturb or disgust a researcher. An
easier to see cultural elements and it facilitates experienced researcher is prepared for a range of
self-discovery. Researchers adopt the attitude of events and relationships. He or she may find,
strangeness to gain these benefits. The attitude however, that it is impossible to penetrate a set
of strangeness means questioning and noticing ting or get really close to members. Settings
ordinary details or looking at the ordinary where cooperation, sympathy, and collaboration
through the eyes of a stranger. Strangeness helps are impossible require different techniques."
a researcher overcome the boredom of observ
ing ordinary details. It helps him or her see the Charm and Trust. A field researcher needs so
ordinary in a new way, one that reveals aspects of cial skills and personal charm to build rapport.
the setting of which members are not con Trust, friendly feelings, and being well liked fa
sciously aware. A field researcher adopts both a cilitate communication and help him or her to
stranger's and an insider's point of view. understand the inner feelings of others. There is
People rarely recognize customs they take no magical way to do this. Showing a genuine
for granted. For example, when someone gives concern for and interest in others, being honest,
us a gift, we say thank you and praise the gift. By and sharing feelings are good strategies, but they
contrast, gift-giving customs in many cultures are not foolproof. It depends on the specific set
include complaining that the gift is inadequate. ting and members.
The attitude of strangeness helps make the tacit Many factors affect trust and rapport-how
culture visible-for example, that gift givers ex a researcher presents himself or herself; the role
pect to hear "Thank you" and "The gift is nice," he or she chooses for the field; and the events
and become upset otherwise. that encourage, limit, or make it impossible to
CHAPTER 11 I FIELD RESEARCH 285
ges a researcher to achieve trust. Trust is not gained once and for the field. In addition, they learn the ropes and
.ial world. Imrner all. It is a developmental process built up over maintain relations with members.
eaks old habits of time through many social nuances (e.g., sharing
ne finds reflection of personal experiences, story telling, gestures,
Roles in the Field
nore intense when hints, facial expressions). It is constantly re-cre
whether it is a dif ated and seems easier to lose once it has been Preexisting versus Created Roles. At times, a
Iture seen through built up than to gain in the first place. researcher adopts an existing role. Some existing
Establishing trust is important, but it does roles provide access to all areas of the site, the
not ensure that all information will be re ability to observe and interact with all members,
vealed. It may be limited to specific areas. For the freedom to move around, and a way to bal
example, trust can be built up regarding finan ance the requirements of researcher and mem
ort by getting along cial matters but not to disclose intimate dating ber. At other times, a researcher creates a new
He or she forges a behavior. Trust may have to be created anew role or modifies an existing one. Duneier (1999),
the same language, in each area of inquiry; it requires constant in his four-year study of New York City street
nembers. This is a reaffirmation. vendors, assumed the role ofbrowser, customer,
understanding of and even magazine vendor.
d understanding to Understanding. Rapport helps field research
feeling events from ers understand members, but understanding is a Limits on the Role Chosen. The field roles
precondition for greater depth, not an end in it open to a researcher are affected by ascriptive
build rapport. The self. It slowly develops in the field as the re factors and physical appearance. He or she can
rmony, with warm, searcher overcomes an initial bewilderment with change some aspects of appearance, such as
iY contain fear, ten a new or unusual language and system of social dress or hairstyle, but not ascriptive features
may be unpleasant, meaning. Once he or she attains an understand such as age, race, gender, and attractiveness.
rful; they may do ing of the member's point of view, the next step Nevertheless, such factors can be important in
ist a researcher. An is to learn how to think and act within a mem gaining access and can restrict the available roles.
epared for a range of ber's perspective. This is empathy, or adopting For example, Gurney (1985) reported that being
Ie or she may find, another's perspective. Empathy does not neces a female in a male-dominated setting required
le to penetrate a set- sarily mean sympathy, agreement, or approval; it extra negotiations and "hassles." Nevertheless,
members. Settings means feeling things as another does. Rapport her gender provided insights and created situa
IY, and collaboration helps create understanding and ultimately em tions that would have been absent with a male
rent techniques.i' pathy, and the development of empathy facili researcher.
tates greater rapport. Since many roles are sex-typed, gender is an
researcher needs so important consideration. Female researchers of
rm to build rapport. ten have more difficulty when the setting is per
I being well liked fa ceived as dangerous or seamy and where males
RELATIONS IN THE FIELD
i help him or her to are in control (e.g., police work, fire fighting,
gs of others. There is You play many social roles in daily life-daugh etc.). They may be shunned or pushed into lim
. Showing a genuine ter/son, student, customer, sports fan-and iting gender stereotypes (e.g., "sweet kid," "mas
others, being honest, maintain social relations with others. You cot," "loud mouth," etc.).
Idstrategies, but they choose some roles and others are structured for New researchers face embarrassment, expe
is on the specific set you. Few have a choice but to play the role of son rience discomfort, and are overwhelmed by the
or daughter. Some roles are formal (e.g., bank details in the field. For example, in her study of
st and rapport-how teller, police chief, etc.), others are informal u.S. relocation camps for Japanese Americans
M or herself; the role (flirt, elder statesperson, buddy, etc.). You can during World War II, respected field researcher
field; and the events switch roles, play multiple roles, and playa role Rosalie Wax (I971) reported that she endured
rake it impossible to in a particular way. Field researchers play roles in the discomfort of 120-degree Fahrenheit tem
286 PART THREE / CONDUCTING QUAliTATIVE RESEARCH
peratures, filthy and dilapidated living condi Members who are cool at first may warm up
tions, dysentery, and mosquitoes. She felt iso later. Or they may put on a front ofinitial friend
lated, she cried a lot, and she gained 30 pounds 1iness, and their fears and suspicions surface only
from compulsive eating. After months in the later. A researcher is in a delicate position. Early
field, she thought she was a total failure; she was in a project, when not yet fully aware of every
distrusted by members and got into fights with thing about a field site, the researcher does not
the camp administration. form close relationships because circumstances
Maintaining a "marginal" status is stressful; may change. Yet, if he or she does develop close
it is difficult to be an outsider who is not fully in friends, they can become allies who will defend
volved, especially when studying settings full of the researcher's presence and help him or her
intense feelings (e.g., political campaigns, reli gain access.
gious conversions, etc.). The loneliness and iso A field researcher monitors how his or her
lation of fieldwork may combine with the desire actions or appearance affects members. For ex
to develop rapport and empathy to cause over ample, a physically attractive researcher who in
involvement. A researcher may "go native" and teracts with members of the opposite sex may
drop the professional researcher's role to be encounter crushes, flirting, and jealousy. He or
come a full member of the group being studied. she develops an awareness of these field relations
Or the researcher may feel guilt about learning and learns to manage them.
intimate details as members drop their guard, In addition to developing social relation
and may come to overidentify with members. ships, a field researcher must be able to break or
withdraw from relationships as well. Ties with
Normalizing Social Research. A field re one member may have to be broken in order to
searcher not only observes and investigates forge ties with others or to explore other aspects
members in the field but is observed and inves of the setting. As with the end of any friendly re
tigated by members as well. In overt field lationship, the emotional pain of social with
research, members are usually initially un drawal can affect both the researcher and the
comfortable with the presence of a researcher. member. The researcher must balance social
Most are unfamiliar with field research and fail sensitivity and the research goals.
to distinguish between sociologists, psycholo
gists, counselors, and social workers. They may Small Favors. Exchange relationships develop
see the researcher as an outside critic or spy, or in the field, in which small tokens or favors, in
as a savior or all-knowing expert. cluding deference and respect, are exchanged. A
An overt field researcher must normalize so researcher may gain acceptance by helping out
cial research-that is, help members redefine so in small ways. Exchange helps when access to
cial research from something unknown and sensitive issues is limited. A researcher may offer
threatening to something normal and predictable. small favors but not burden members by asking
He or she can help members manage research by for return favors. As the researcher and mem
presenting his or her own biography, explaining bers share experiences and see each other again,
field research a little at a time, appearing non members recall the favors and reciprocate by al
threatening, or accepting minor deviance in the lowing access.For example, Duneier (1999) used
setting (e.g., minor violations of official rules). the small favor of watching the tables of street
vendors when they had to leave for a short time,
such as to use the bathroom.
Maintaining Relations
Social Relations. With time, a field researcher Conflicts in the Field. Fights, conflict, and dis
develops and modifies social relationships. agreements can erupt in the field, or a researcher
CHAPTER 11 I FIELD RESEARCH 287
first may warm up may study groups with opposing positions. In senses, noticing what is seen, heard, smelled,
~ont of initial
friend such situations, the researcher will feel pressure tasted, or touched. The researcher becomes an in
.picions surface only to take sides and will be tested to see ifhe or she strument that absorbs all sources of information.
kate position. Early can be trusted. In such occasions, a researcher A field researcher carefully scrutinizes the
ully aware of every usually stays on the neutral sidelines and walks a physical setting to capture its atmosphere. He or
researcher does not tightrope between opposing sides. This is be she asks: What is the color of the floor, walls,
.ause circumstances cause once he or she becomes aligned with one ceiling? How large is the room? Where are the
e does develop close side, the researcher will cut off access to the windows and doors? How is the furniture
lies who will defend other side. In addition, he or she will see the sit arranged, and what is its condition (e.g., new or
nd help him or her uation from only one point of view. old and worn, dirty or clean)? What type of
lighting is there? Are there signs, paintings,
itors how his or her Appearing Interested. Field researchers main plants? What are the sounds or smells?
ts members. For ex tain an appearance of interest in the field. An ex Why bother with such details? You may
e researcher who in perienced researcher appears to be interested in have noticed that stores and restaurants often
Le opposite sex may and involved with field events by statements and plan lighting, colors, and piped-in music to cre
and jealousy. He or behaviors (e.g., facial expression, going for cof ate a certain atmosphere. Maybe you know that
f these field relations fee, organizing a party, etc.) even if he or she is used-car sales people spray a new-car scent into
not truly interested. This is because field rela cars or that shops in shopping malls intention
.ing social relation tions may be disrupted if the researcher appears ally send out the odor of freshly made cookies.
.t be able to break or to be bored or distracted. Putting up such a tem These subtle, unconscious signals influence hu
os as well. Ties with porary front of involvement is a common small man behavior.
e broken in order to deception in daily life and is part ofbeing polite. Observing in field research is often detailed,
xplore other aspects Of course, selective inattention (i.e., not tedious work. Instead ofthe quick flash, motiva
id of any friendly re staring or appearing not to notice) is also part of tion arises out of a deep curiosity about the de
pain of social with- acting polite. If a person makes a social mistake tails. Good field researchers are intrigued about
researcher and the (e.g., accidentally uses an incorrect word, passes details that reveal "what's going on here"
nust balance social gas, etc.), the polite thing to do is to ignore it. Se through careful listening and watching. Field re
goals. lective inattention is used in fieldwork, as well. It searchers believe that the core of social life is
gives an alert researcher an opportunity to learn communicated through the mundane, trival,
elationships develop by casually eavesdropping on conversations or everyday minutia. This is what people often
tokens or favors, in observing events not meant to be public. overlook, but field researchers need to learn how
.ct, are exchanged. A to notice.
ance by helping out In addition to physical surroundings, a field
elps when access to researcher observes people and their actions, not
OBSERVING AND COllECTING
researcher may offer ing each person's observable physical character
DATA
l members by asking istics: age, sex, race, and stature. People socially
searcher and mern This section looks at how to get good qualitative interact differently depending on whether an
,ee each other again, field data. Field data are what the researcher ex other person is 18,40, or 70 years old; male or fe
rid reciprocate by al periences and remembers, and what are male; White or non-White; short and frail or tall,
Duneier (1999) used recorded in field notes and become available for heavyset, and muscular. When noting such char
~ the tables of street systematic analysis. acteristics, the researcher is included. For exam
.ave for a short time, ple, an attitude of strangeness heightens
l. sensitivity to a group's racial composition. A re
Watching and Listening
searcher who ignores the racial composition of a
Its, conflict, and dis Observing. In the field, researchers payatten group of Whites in a multiracial society because
field, or a researcher tion, watch, and listen carefully. They use all the he or she too is White is being racially insensitive.
288 PART THREE I CONDUCTING QUALITATIVE RESEARCH
The researcher records such details because relevance of what he or she is observing until
something of significance might be revealed. It is later. This has two implications. First is the im
better to err by including everything than to ig portance of keen observation and excellent notes
nore potentially significant details. For example, at all times, even when "nothing seems to be
"The tall, White muscular 19-year-old male happening." Second is the importance of look
sprinted into the brightly lit room just as the ing back over time and learning to appreciate
short, overweight Black woman in her sixties wait time. Most field researchers say that they
eased into a battered chair" says much more spend a lot of time "waiting." Novice field re
than "One person entered, another sat down." searchers get frustrated with the amount of time
A field researcher notes aspects of physical they seem to "waste," either waiting for other
appearance such as neatness, dress, and hairstyle people or waiting for events to occur.
because they express messages that can affect so A field researcher needs must be attuned to
cial interactions. People spend a great deal of the rhythms of the setting, operate on other peo
time and money selecting clothes, styling and ple's schedules, and observe how events occur
combing hair, grooming with make-up, shaving, within their own flow of time. Wait time is not
ironing clothes, and using deodorant or per always wasted time. Wait time is time for reflec
fumes. These are part of their presentation of tion, for observing details, for developing social
self. Even people who do not groom, shave, or relations, for building rapport, and for becom
wear deodorant present themselves and send a ing a familiar sight to people in the field setting.
symbolic message by their appearance. Noone Wait time also displays that a researcher is com
dresses or looks "normal." Such a statement sug mitted and serious; perseverance is a significant
gests that a researcher is not seeing the social trait field researchers need to cultivate. The re
world through the eyes of a stranger or is insen searcher may be impatient to get in, get the re
sitive to social signals. search over, and get on with his or her "real life"
Behavior is also significant. A field re but for the people in the field site, this is real life.
searcher notices where people sit or stand, the The researcher should subordinate his or her
pace at which they walk, and their nonverbal personal wants to the demands of the field site.
communication. People express social informa
tion, feelings, and attitudes through nonverbal Listening. A field researcher listens carefully to
communication, including gestures, facial ex phrases, accents, and incorrect grammar, listen
pressions, and how one stands or sits (standing ing both to what is said and how it is said or what
stiffly,sitting in a slouched position, etc.). People was implied. For example, people often use
express relationships by how they position them phrases such as "you know" or "ofcourse" or "et
selves in a group and through eye contact. A re cetera." A field researcher knows the meaning
searcher may read the social communication of behind such phrases. He or she can try to hear
people by noting that they are standing close to everything, but listening is difficult when many
gether, looking relaxed, and making eye contact. conversations occur at once or when eavesdrop
A field researcher also notices the context in ping. Luckily, significant events and themes usu
which events occur: Who was present? Who just ally recur.
arrived or left the scene? Was the room hot and
stuffy? Such details may help the researcher as
Taking Notes
sign meaning and understand why an event oc
curred. If they are not noticed, the details are Most field research data are in the form of field
lost, as is a full understanding of the event. notes. Full field notes can contain maps, dia
Serendipity is important in field research. grams, photographs, interviews, tape recordings,
Many times, a field researcher does not know the videotapes, memos, artifacts or objects from the
CHAPTER 11 / FIELD RESEARCH 289
is observing until field, notes jotted in the field, and detailed notes in a public setting looks strange when furiously
ms. First is the im written away from the field. A field researcher ex writing. More important, when looking down
and excellent notes peets to fill many notebooks, or the equivalent in and writing, the researcher cannot see and hear
thing seems to be computer memory. He or she may spend more what is happening. The attention given to note
nportance of look time writing notes than being in the field. Some writing is taken from field observation where it
ning to appreciate researchers produce 40 single-spaced pages of belongs. The specificsetting determines whether
.hers say that they notes for three hours of observation. With prac any notes in the field can be taken. The re
.," Novice field re tice, even a new field researcher can produce sev searcher may be able to write, and members may
ne amount of time eral pages of notes for each hour in the field. expect it, or he or she may have to be secretive
. waiting for other Writing notes is often boring, tedious work (e.g., go to the restroom).
o occur. that requires self-discipline. The notes contain Jotted notes are written in the field. They are
must be attuned to extensive descriptive detail drawn from mem short, temporary memory triggers such as
erate on other peo ory. A researcher makes it a daily habit or com words, phrases, or drawings taken inconspicu
how events occur pulsion to write notes immediately after leaving ously, often scribbled on any convenient item
e. Wait time is not the field. The notes must be neat and organized (e.g., napkin, matchbook). They are incorpo
e is time for reflec because the researcher will return to them over rated into direet observation notes but are never
r developing social and over again. Once written, the notes are pri substituted for them.
rt, and for becom vate and valuable. A researcher treats them with
in the field setting. care and protects confidentiality. Field notes Direct Observation Notes. The basic source of
researcher is com may be of interest to hostile parties, blackmail field data are notes a researcher writes immedi
mce is a significant ers, or legal officials, so some researchers write ately after leaving the field, which he or she can
) cultivate. The re field notes in code. add to later. The notes should be ordered
) get in, get the re A researcher's state of mind, level of atten chronologically with the date, time, and place on
lis or her "real life" tion, and conditions in the field affect note tak each entry. They serve as a detailed description
site, this is real life. ing. He or she will usually begin with relatively of what the researcher heard and saw in con
rdinate his or her short one- to three-hour periods in the field be crete, specific terms. To the extent possible, they
cis of the field site. fore writing notes. are an exact recording of the particular words,
phrases, or actions.
r listens carefully to Types of Field Notes. Field researchers take A researcher's memory improves with prac
:t grammar, listen notes in many ways.!? The recommendations tice. A new researcher can soon remember exact
iw it is said or what here (also see Box 11.4) are suggestions. Full phrases from the field. Verbatim statements
people often use field notes have several types or levels.Fivelevels should be written with double quote marks to
r "of course" or "et will be described. It is usually best to keep all the distinguish them from paraphrases. Dialogue
nows the meaning notes for an observation period together and to accessories (nonverbal communication, props,
she can try to hear distinguish types of notes by separate pages. tone, speed, volume, gestures) should be
fficult when many Some researchers include inferences with direct recorded as well. A researcher records what was
,r when eavesdrop observations if they are set off by a visible device actually said and does not clean it up; notes in
ts and themes usu such as brackets or colored ink. The quantity of clude ungrammatical speech, slang, and mis
notes varies across types. For example, six hours statements (e.g., write, "Vh, I'm goin' home,
in the field might result in 1 page of jotted notes, Sal," not "I am going home, Sally").
40 pages of direct observation, 5 pages of re A researcher puts concrete details in notes,
searcher inference, and 2 pages total for method not summaries. For example, instead of, "We
n the form of field ological, theoretical, and personal notes. talked about sports," he or she writes, "Anthony
ontain maps, dia argued with Sam and Jason. He said that the
ITS, tape recordings, Jotted Notes. It is nearly impossible to take Cubs would win next week because they traded
rr objects from the good notes in the field. Even a known observer for a new shortstop, Chiappetta. He also said
290 PART THREE / CONDUCTING QUALITATIVE RESEARCH
1. Record notes as soon as possible after each pe 1 O. "Let your feelings flow" and write quickly with
riod in the field, and do not talk with others un out worrying about spelling or "wild ideas." As
til observations are recorded. sume that no one else will see the notes, but use
2. Begin the record of each field visit with a new pseudonyms.
page, with the date and time noted. 11. Never substitute tape recordings completely for
3. Use jotted notes only as a temporary memory field notes.
aid, with key words or terms, or the first and last 12. Include diagrams or maps of the setting, and
things said. outline your own movements and those of oth
4. Use wide margins to make it easy to add to ers during the period of observation.
notes at any time. Go back and add to the notes 1 3. Include the researcher's own words and behav
if you remember something later. ior in the notes. Also record emotional feelings
5. Plan to type notes and keep each level of notes and private thoughts in a separate section.
separate so it will be easy to go back to them 14. Avoid evaluative summarizing words. Instead of
later. "The sink looked disgusting," say, "The sink was
6. Record events in the order in which they oc rust-stained and looked as if it had not been
curred, and note how long they last (e.g., a 15 cleaned in a long time. Pieces of food and dirty
minute wait, a one-hour ride). dishes looked as if they had been piled in it for
several days."
7. Make notes as concrete, complete, and compre
hensible as possible. 15. Reread notes periodically and record ideas gen
erated by the rereading.
8. Use frequent paragraphs and quotation marks.
Exact recall of phrases is best, with double 1 6. Always make one or more backup copies, keep
quotes; use single quotes for paraphrasing. them in a locked location, and store the copies
in different places in case of fire.
9. Record small talk or routines that do not appear
to be significant at the time; they may become
important later.
that the team was better than the Mets, who he Researcher Inference Notes. A field researcher
thought had inferior infielders. He cited last listens to members in order to "climb into their
week's game where the Cubs won against Boston skin" or "walk in their shoes." This involves a
by 8 to 3." A researcher notes who was present, three-step process. The researcher listens with
what happened, where it occurred, when, and out applying analytical categories; he or sbr
under what circumstances. New researchers may compares what is heard to what was heard _
not take notes because "nothing important hap other times and to what others say; then the re
pened." An experienced researcher knows that searcher applies his or her own interpretation ID
events when "nothing happened" can reveal a infer or figure out what it means. In ordinary m
lot. For example, members may express feelings teraction, we do all three steps simultaneousse
and organize experience into folk categories and jump quickly to our own inferences. A fidII
even in trivial conversations. researcher learns to look and listen without ....
CHAPTER 11 / FIELD RESEARCH 291
ferring or imposing an interpretation. His or her having an affair with each other, or a brother
observations without inferences go into direct and sister whose father just died?
observation notes.
A researcher records inferences in a separate Analytic Notes. Researchers make many deci
vrite quickly with section that is keyed to direct observations. Peo sions about how to proceed while in the field.
,r "wild ideas." As ple never see social relationships, emotions, or Some acts are planned (e.g., to conduct an inter
the notes, but use meaning. They see specific physical actions and view, to observe a particular activity, etc.) and
hear words, then use background cultural others seem to occur almost out of thin air. Field
19S completely for knowledge, clues from the context, and what is researchers keep methodological ideas in ana
done or said to assign social meaning. For exam lytic notes to record their plans, tactics, ethical
r the setting, and ple, one does not see love or anger; one sees and and procedural decisions, and self-critiques of
and those of oth hears specific actions (red face, loud voice, wild tactics.
-vation. gestures, obscenities) and draw inferences from Theory emerges in field research during
words and behav them (the person is angry). data collection and is clarified when a researcher
emotional feelings People constantly infer social meaning on reviews field notes. Analytic notes have a run
arate section. the basis of what they see and hear, but not al ning account of a researcher's attempts to give
words. Instead of ways correctly. For example, my niece visited me meaning to field events. He or she thinks out
say, "The sink was and accompanied me to a store to buy a kite. The loud in the notes by suggesting links between
f it had not been clerk at the cash register smiled and asked her ideas, creating hypotheses, proposing conjec
.of food and dirty whether she and her "Daddy" (looking at me) tures, and developing new concepts .
oeen piled in it for were going to fly the kite that day. The clerk ob Analytic memos are part of the theoretical
served our interaction, then inferred a notes. They are systematic digressions into the
Irecord ideas gen father/daughter, not an uncle/niece relationship. ory, where a researcher elaborates on ideas in
She saw and heard a male adult and a female depth, expands on ideas while still in the field,
ckup copies, keep child, but she inferred the social meaning and modifies or develops more complex theory
d store the copies incorrectly. by rereading and thinking about the memos.
Ire. A researcher keeps inferred meaning sepa
rate from direct observation because the mean Personal Notes. As discussed earlier, personal
ing of actions is not always self-evident. feelings and emotional reactions become part of
Sometimes, people try to deceive others. For ex the data and color what a researcher sees or
ample, an unrelated couple register at a motel as hears in the field. A researcher keeps a section of
Mr. and Mrs. Smith. More frequently, social be notes that is like a personal diary. He or she
havior is ambiguous or multiple meanings are records personal life events and feelings in it
. field researcher possible. For example, I see a White male and fe (''I'm tense today. I wonder if it's because of the
"climb into their male, both in their late twenties, get out of a car fight I had yesterday with Chris," ''I've got a
, This involves a and enter a restaurant together. They sit at a headache on this gloomy, overcast day").
:her listens with table, order a meal, and talk with serious expres Personal notes serve three functions: They
aries; he or she sions in hushed tones, sometimes leaning for provide an outlet for a researcher and a way to
rat was heard at ward to hear each other. As they get up to leave, cope with stress; they are a source of data about
say; then the re the woman, who has a sad facial expression and personal reactions; and they give him or her a
interpretation to appears ready to cry, is briefly hugged by the way to evaluate direct observation or inference
s. In ordinary in male. They then leave together. Did I witness a notes when the notes are later reread. For exam
s simultaneously couple breaking up, two friends discussing a ple, if the researcher was in a good mood during
nferences, A field third, two people trying to decide what to do be observations, it might color what he or she ob
isten without in cause they have discovered that their spouses are served (see Figure 11.2).
292 PART THREE I CONDUCTING QUALITATIVE RESEARCH
---------------------11
FIGURE 11.2 Types offield Notes
Direct Observation Inference Analytic Personal Journal
Maps and Diagrams. Field researchers often the field. They cannot be introduced into all field
make maps and draw diagrams or pictures of the sites, and can be used only after a researcher de
features of a field site. This serves two purposes: velops rapport. Recorders and videotapes provide
It helps a researcher organize events in the field a close approximation to what occurred and a
and it helps convey a field site to others. For ex permanent record that others can review. They
ample, a researcher observing a bar with 15 serve as "jotted notes" to help a researcher recall
stools may draw and number 15 circles to sim events and observe what is easy to miss. Never
plify recording (e.g., "Yosuke came in and sat on theless, these items can create disruption and an
stool 12; Phoebe was already on stool 10"). Field increased awareness of surveillance. Researchers
researchers find three types of maps helpful: spa who rely on them must address associated prob
tial, social, and temporal. The first helps orient lems (e.g.,ensure that batteries are fresh and there
the data; the latter two are preliminary forms of are enough blank tapes). Also, relistening to or
data analysis. A spatial map locates people, viewing tapes can be time consuming. For exam
equipment, and the like in terms of geographical ple, it may take over 100 hours to listen to 50
physical space to show where activities occur hours recorded in the field. Transcriptions oftape
(Figure 11.3A). A social map shows the number are expensive and not always accurate; they do
or variety of people and the arrangements not always convey subtle contextual meanings or
among them of power, influence, friendship, di mumbled words. Duneier (1999) had a tape
vision of labor, and so on (Figure 11.313). A recorder on all the time in his study of New York
temporal map shows the ebb and flow of people, City street vendors. He made others aware of the
goods, services, and communications, or sched machine and took reponsibility for what behav
ules (Figure ll.3e). iors he focused on, and he left the machine visible.
The taping may have created some distortion but
Machine Recordings to Supplement Memory. it also provided a record of everyday routines. He
Tape recorders and videotapes can be helpful also had a collaborator who took a large collection
supplements in field research. They never substi of photographs of his field site and informants,
tute for field notes or a researcher's presence in which helped him to see things differently.
--------------------.
---.
urnal
FIG U R Ell. 3
A Spatial Map
Types of Maps Used in Field Research
o Wong
O
Empty
Chair
Chalkboard
Sampson
jay.
o
00
De Onis Harris
Door
Window
B Social Map
oduced into all field
ter a researcher de
lvideotapes provide
hat occurred and a 11
~s can review. They
J a researcher recall
asy to miss. Never
e disruption and an 9
illance. Researchers
ess associated prob
Apartment A
s are fresh and there
;0, relistening to or
o
Friendship Pattern of 11 People
isuming, For exam
mrs to listen to 50 Males Females ApartmentB
anscriptions oftape
s accurate; they do
textual meanings or C Temporal Map
(1999) had a tape Day of Week, Buzz's Bar
,study of New York
Mon Tue Wed Thr Fri Sat
others aware of the Open 10:00
Old Old Old Old Skip Going
ity for what behav
Drunks Drunks Drunks Drunks Work or to
the machine visible.
Leave Fish
some distortion but Early
eryday routines. He 5:00
ok a large collection Football Neighbors Softball Young Loud Loners
ite and informants, Watchers and Team Crowd Music, and
~ differently. Bridge (All-Male Mixed No
Players Night) Crowd Dates
Close 1:00
294 PART THREE / CONDUCTING QUALITATIVE RESEARCH
Interview Notes. If a researcher conducts field saying that to please me? Is there anything that
interviews (to be discussed), he or she keeps the might limit his spontaneity?
interview notes separate. Field researchers take subjectivity and con
text into account as they evaluate credibility.
They know that a person's statements or actions
Data Quality
are affected by subjective perceptions. State
Reliability in Field Research. The reliability of ments are made from a particular point of view
field data addresses the question: Are researcher and colored by an individual's experiences. In
observations about a member or field event in stead of evaluating each statement to see if it is
ternallyand externally consistent? Internal con true, a field researcher finds statements useful in
sistency refers to whether the data are plausible themselves. Even inaccurate statements and ac
given all that is known about a person or event, tions can be revealing from a researcher's
eliminating common forms of human decep perspective.
tion. In other words, do the pieces fit together As mentioned before, actions and state
into a coherent picture? For example, are a ments are shaped by the context in which they
member's actions consistent over time and in appear. What is said in one setting may differ in
different social contexts? External consistency is other contexts. For example, when asked "Do
achieved by verifying or cross-checking observa you dance?" a member may say no in a public
tions with other, divergent sources of data. In setting full of excellent dancers, but yes in a
other words, does it all fit into the overall con semiprivate setting with few good dancers and
text? For example, can others verify what a re different music. It is not that the member islying
searcher observed about a person? Does other but that the answer is shaped by the context.
evidence confirm the researcher's observations? Duneier (1999) has warned us to avoid the
Reliability in field research also includes ethnographic fallacy. It occurs when a field re
what is not said or done, but is expected. Such searcher takes what he or she oberves at face
omissions can be significant but are difficult to value, does not question what people in a field
detect. For example, when observing a cashier site say, and focuses solely on the immediate
end her shift, a researcher notices that the concrete details of a field setting while ignoring
money in the cash drawer is not counted. He or larger social forces. Duneier noted that he tried
she may notice the omission only if other to avoid the fallacy by being aware of larger so
cashiers always count the money at the end of cial context and forces. Thus, he studied people
the shift. who took responsibilty for their own failures
Reliability in field research depends on a re (such as dropping out of school in the ninth
searcher's insight, awareness, suspicions, and grade) and blamed themselves. Duneier was
questions. He or she looks at members and fully aware from many other studies of the larger
events from different angles (legal, economic, forces (e.g., family situation, violence, poor
political, personal) and mentally asks questions: quality school, racial prejudice, joblessness) that
Where does the money come from for that? often contributed to their experience of failure. I I
What do those people do all day?
Field researchers depend on what members Validity in Field Research. Validity in field re
tell them. This makes the credibility of members search is the confidence placed in a researcher's
and their statements part of reliability. To check analysis and data as accurately representing the
member credibility, a researcher asks: Does the social world in the field. Replicability is not a cri
person have a reason to lie?Is she in a position to terion because field research is virtually impossi
know that? What are the person's values and ble to replicate. Essential aspects of the field
how might that shape what she says? Is he just change: The social events and context change,
CHAPTER 11 I FIELD RESEARCH 295
here anything that the members are different, the individual re
searcher differs, and so on. There are four kinds
-----------.
Field researchers sample people by focusing ended, informal, and long. Generally, they in
their attention on different kinds of people (old volve one or more people being present, occur in
timers and newcomers, old and young, males the field, and are informal and nondirective (i.e.,
and females, leaders and followers). As a re the respondent may take the interview in various
searcher identifies types of people, or people directions).
with opposing outlooks, he or she tries to inter A field interview involves a mutual sharing
act with and learn about all types. A field re of experiences. A researcher might share his or
searcher also samples various kinds of events, her background to build trust and encourage the
such as routine, special, and unanticipated. Rou informant to open up, but does not force an
tine events (e.g., opening up a store for business) swers or use leading questions. She or he en
happen every day and should not be considered courages and guides a process of mutual
unimportant simply because they are routine. discovery.
Special events (e.g., annual office party) are an In field interviews, members express them
nounced and planned in advance. They focus selves in the forms in which they normally speak,
member attention and reveal aspects of social think, and organize reality. A researcher retains
life not otherwise visible. Unanticipated events members' jokes and narrative stories in their
are those that just happen to occur while a re natural form and does not repackage them into
searcher is present (e.g., unsupervised workers a standardized format. The focus is on the mem
when the manager gets sick and cannot oversee bers' perspectives and experiences. In order to
workers at a store for a day). In this case, the re stay close to a member's experience, the re
searcher sees something unusual, unplanned, or searcher asks questions in terms of concrete ex
rare by chance. amples or situations-for example, "Could you
tell me things that led up to your quitting in
June?" instead of "Why did you quit your job?"
Field interviews can occur in a series over
THE FiElD RESEARCH INTERVIEW
time. A researcher begins by building rapport
So far, you have learned how field researchers and steering conversation away from evaluative
observe and take notes. They also interview or highly sensitive topics. He or she avoids prob
members, but field interviews differ from survey ing inner feelings until intimacy is established,
research interviews. This section introduces the and even then, the researcher expects apprehen
field interview. sion. After several meetings, he or she may be
able to probe more deeply into sensitive issues
and seek clarification of less sensitive issues. In
The Field Interview
later interviews, he or she may return to topics
Field researchers use unstructured, nondirective, and check past answers by restating them in a
in-depth interviews, which differ from formal nonjudgmental tone and asking for verifica
survey research interviews in many ways (see tion-for example, "The last time we talked, you
Table ILl). The field interview involves asking said that you started taking things from the store
questions, listening, expressing interest, and after they reduced your pay. Is that right?"
recording what was said. It is a joint production The field interview is closer to a friendly
of a researcher and a member. Members are ac conversation than the stimuluslresponse model
tive participants whose insights, feelings, and co found in a survey research interview. You are fa
operation are essential parts of a discussion miliar with a friendly conversation. It has its own
process that reveals subjective meanings. informal rules and the following elements: (1) a
Field research interviews go by many greeting ("Hi, it's good to see you again"); (2)
names: unstructured, depth, ethnographic, open the absence of an explicit goal or purpose (we
CHAPTER 11 / FIELD RESEARCH 297
Generally, they in don't say, "Let's now discuss what we did last motive race"); (9) a pause or brief silence when
19present, occur in weekend"); (3) avoidance of repetition (we neither person talks is acceptable; (10) a closing
i nondirective (i.e., don't say, "Could you clarify what you said (we don't say, "Let's end this conversation"; in
nterview in various about"); (4) question asking ("Did you see the stead, we give a verbal indicator before physically
race yesterday?"); (5) expressions of interest leaving: "I've got to get back to work now-see
:s a mutual sharing ("Really? I wish I could have been there!"); (6) ya tomorrow").
might share his or expressions of ignorance ("No, I missed it. What The field interview differs from a friendly
: and encourage the happened?"); (7) turn taking, so the encounter is conversation. It has an explicit purpose-to
does not force an balanced (one person does not always ask ques learn about the informant and setting. A re
oris. She or he en tions and the other only answer); (8) abbrevia searcher includes explanations or requests that
rocess of mutual tions ("I missed the Derby, but I'm going to the diverge from friendly conversations. For exam
Indy," not "I missed the Kentucky Derby horse ple, he or she may say, "I'd like to ask you about
bers express them race but I will go to the Indianapolis 500 auto- ..." or "Could you look at this and see if I've
ley normally speak,
l researcher retains
ive stories in their
~package them into
lCUS is on the mem
---------------------It'
iences. In order to
xperience, the re
rms of concrete ex 1 . It has a clear beginning and end. 1 . The beginning and end are not clear. The
o your quitting in 2. The same standard questions are asked of all 2. The questions and the order in which they
ou quit your job?" respondents in the same sequence. are asked are tailored to specific people and
.ur in a series over situations.
y building rapport 3. The interviewer appears neutral at all times. 3. The interviewer shows interest in responses,
~ expects apprehen
• he or she may be 5. It is almost always with one respondent alone. 5. It can occur in group setting or with others in
nto sensitive issues area, but varies.
sensitive issues. In 6. It has a professional tone and businesslike focus; 6. It is interspersed with jokes, asides, stories,
,ay return to topics diversions are ignored. diversions, and anecdotes, which are recorded.
~estating them in a
7. Closed-ended questions are common, with rare 7. Open-ended questions are common, and
Is that right?"
loser to a friendly 9. The social context in which the interview occurs 9. The social context of the interview is noted
us/response model is ignored and assumed to make little difference. and seen as important for interpreting the
ation. It has its own 1 O. The interviewer attempts to mold the framework 10. The interviewer adjusts to the member's
ring elements: (1) a communication pattern into a standard. norms and language usage.
~e you again"); (2)
ial or purpose (we Source: Adapted from Briggs (1 986), Denzin (1989), Douglas (1 985), Misher (1 986), Spradley (1979a).
298 PART THREE / CONDUCTING QUALITATIVE RESEARCH
written it down right?" The field interview is less adds structural questions until, in the middle
balanced. A higher proportion of questions stage after analysis has begun, they make up a
come from the researcher, who expresses more majority of the questions. Contrast questions
ignorance and interest. Also, it includes repeti begin to appear in the middle of a field research
tion, and a researcher asks the member to elabo study and increase until, by the end, they are
rate on unclear abbreviations. asked more than any other type. 13
Kissane (2003) used depth interviews in her A descriptive question is used to explore the
field study of low-income women in Philadel setting and learn about members. Descriptive
phia (discussed in Chapter 6). Interviews lasted questions can be about time and space-for ex
from 30 minutes to three hours. Kissane noted ample, "Where is the bathroom?" "When does
that she asked the women what services they the delivery truck arrive?" "What happened
used, and then named specific agencies. Often a Monday night?" They can also be about people
woman would then say she was aware of the and activities: "Who is sitting by the window?"
named agency. She asked the women to describe "What is your uncle like?" "What happens dur
their experiences with various agencies, when ing the initiation ceremony?" They can be about
they had used them or if they would use services objects: "When do you use a saber saw?""Which
of various agencies, and what other social ser tools do you carry with you on an emergency
vices they used. Open-ended interviewing al water leak job?" Questions asking for examples
lowed her to see the women's decision-making are descriptive questions-for example, "Could
process. you give me an example of a great date?" "What
were your experiences as a postal clerk?" De
scriptive questions may ask about hypothetical
Types of Questions in Field Interviews
situations: "If a student opened her book during
Many field researchers ask three types of ques the exam, how would you deal with it?" They
tions in a field interview: descriptive, structural, also ask members about the argot of the setting:
and contrast questions. All are asked concur "What do you call a deputy sheriff?" (The an
rently, but each type is more frequent at a differ swer is a "county Mountie.")
ent stage in the research process (see Figure A researcher introduces a structuralquestion
11.5). During the early stage, a researcher pri after spending time in the field and starting to
..
marily asks descriptive questions, then gradually analyze data. It begins after a researcher orga
nizes specific field events, situations, and con
versations into conceptual categories. For
---------~ example, a researcher's observations of a high
FIG U R Ell . 5 Types of Questions in way truck-stop restaurant revealed that the em
Field Research Interviews ployees informally classify customers who
patronize the truck stop. In a preliminary analy
sis, he or she creates a conceptual category of
kinds of customers and has members verify the
categories with structural questions. A common
way to pose a structural question is to ask the
Number of members whether a category includes elements
Questions in addition to those already identified-for ex
ample, "Are there any types of customers other
than regulars, greasers, pit stoppers, and long
haulers?" In addition, a researcher asks for con
Time in the Field firmation: "Is a greaser a type of customer that
CHAPTER 11 / FIELD RESEARCH 299
, in the middle you serve?" "Would a pit stopper ever eat a setting, using categories from the media or
they make up a three-course dinner?" education.
rtrast questions The contrastquestion builds on the analysis
: a field research already verified by structural questions. Contrast A field researcher may interview several
.e end, they are questions focus on similarities or differences be types of informants. Contrasting types of infor
13
tween elements in categories or between cate mants who provide useful perspectives include
d to explore the gories. The researcher asks members to verify rookies and old-timers, people in the center of
ers. Descriptive the similarities and differences: "You seem to events and those on the fringes of activity, peo
1space-for ex have a number of different kinds of customers ple who recently changed status (e.g., through
l?" "When does come in here. I've heard you call some cus promotion) and those who are static, frustrated
Vhat happened tomers 'regulars' and others 'pit stoppers.' How or needy people and happy or secure people, the
oe about people are a regular and a pit stopper alike?" or "Is the leader in charge and the subordinate who fol
y the window?" difference between a long hauler and a greaser lows. A field researcher expects mixed messages
at happens dur that the greaser doesn't tip?" or "Two types of when he or she interviews a range of informants.
ley can be about customers just stop to use the restroom-entire
ersaw?" "Which families and a lone male. Do you call both pit
Interview Context
1 an emergency stoppers?"
ng for examples Field researchers recognize that a conversation
xample, "Could in a private office may not occur in a crowded
at date?" "What Informants
lunchroom. Often, interviews take place in the
stal clerk?" De An informant or key actor in field research is a member's home environment so that he or she is
JUt hypothetical member with whom a field researcher develops a comfortable. This is not always best. If a member
her book during relationship and who tells about, or informs on, is preoccupied or there is no privacy, a re
I with it?" They the field. 14 Who makes a good informant? The searcher will move to another setting (e.g.,
ot of the setting: ideal informant has four characteristics: restaurant or university office).
eriffi" (The an The interview's meaning is shaped by its
1. The informant is totally familiar with the Gestalt or whole interaction of a researcher and
ruauralquestion culture and is in position to witness signifi a member in a specific context. For example, a
and starting to cant events. He or she lives and breathes the researcher notes nonverbal forms of communi
'esearcher orga culture and engages in routines in the set cation that add meaning, such as a shrug, a ges
itions, and con ting without thinking about them. ture, and so on.
::ategories. For 2. The individual is currently involved in the
tions of a high field. Ex-members who have reflected on
led that the em the field may provide useful insights, but the
:ustomers who LEAVING THE FIELD
longer they have been away from direct in
eliminary analy volvement, the more likely it is that they Work in the field can last for a few weeks to a
tual category of have reconstructed their recollections. dozen years. In either case, at some point work
nbers verify the 3. The person can spend time with the re in the field ends. Some researchers (e.g., Schatz
ons. A common searcher. Interviewing may take many man and Strauss, 1973) suggest that the end
on is to ask the hours, and some members are simply not comes naturally when theory building ceases or
eludes elements available for extensive interviewing. reaches a closure; others feel that fieldwork
ntified-for ex 4. Nonanalytic individuals make better infor could go on without end and that a firm decision
.ustomers other mants. A nonanalytic informant is familiar to cut off relations is needed.
Jpers, and long with and uses native folk theory or prag Experienced field researchers anticipate a
ier asks for con matic common sense. This is in contrast to process of disengaging and exiting the field. De
f customer that the analytic member, who preanalyzes the pending on the intensity of involvement and the
300 PART THREE I CONDUCTING QUALITATIVE RESEARCH
length of time in the field, the process can be dis ished until the process of disengagement and ex
ruptive or emotionally painful for both the re iting is complete.
searcher and the members. A researcher may
experience the emotional pain of breaking inti
mate friendships when leaving the field. He or
FOCUS GROUPS
she may feel guilty and depressed immediately
before and after leaving. He or she may find it The focus group is a special qualitative research
difficult to let go because of personal and emo technique in which people are informally "inter
tional entanglements. If the involvement in the viewed" in a group-discussion setting. IS Focus
field was intense and long, and the field site dif group research has grown over the past 20 years.
fered from his or her native culture, the re The procedure is that a researcher gathers to
searcher may need months of adjustment before gether 6 to 12 people in a room with a modera
feeling at home with his or her original cultural tor to discuss a few issues. Most focus groups last
surroundings. about 90 minutes. The moderator is trained to
Once a researcher decides to leave-be be nondirective and to facilitate free, open dis
cause the project reaches a natural end and little cussion by all group members (i.e., not let one
new is being learned, or because external factors person dominate the discussion). Group mem
force it to end (e.g., end of a job, gatekeepers or bers should be homogenous, but not include
der the researcher out, etc.)-he or she chooses close friends or relatives. In a typical study, a re
a method of exiting. The researcher can leave by searcher uses four to six separate groups. Focus
a quick exit (simply not return one day) or group topics might include public attitudes (e.g.,
slowly withdraw, reducing his or her involve race relations, workplace equality), personal be
ment over weeks. He or she also needs to decide haviors (e.g., dealing with AIDS), a new product
how to tell members and how much advance (e.g., breakfast cereal), a political candidate, or a
warning to give. number of other topics. Researchers often com
The exit process depends on the specific bine focus groups with quantitative research,
field setting and the relationships developed. In and the procedure has its own specific strengths
general, a researcher lets members know a short and weaknesses (see Box 1l.S).
period ahead of time. He or she fulfills any bar Several years ago, I conducted an applied
gains or commitments that were made and study on why parents and students chose to at
leaves with a clean slate. Sometimes, a ritual or tend a private high school. In addition to col
ceremony, such as a going-away party or shaking lecting quantitative survey data, I formed six
hands with everyone, helps signal the break for focus groups, each with 8 to 10 students from
members. Maintaining friendships with mem the high school. A trained college-student mod
bers is also possible and is preferred by feminist erator asked questions, elicited comments from
researchers. group members, and prevented one person from
A field researcher is aware that leaving af dominating discussions. The six groups were co
fects members. Some members may feel hurt or ed and contained members of either one grade
rejected because a close social relationship is level or two adjacent grades (e.g., freshmen and
ending. They may react by trying to pull a re sophomores). Students discussed their reasons
searcher back into the field and make him or her for attending the high school and whether spe
more of a member, or they may become angry cific factors were important. I tape- recorded the
and resentful. They may grow cool and distant discussions, which lasted about 4S minutes, then
because of an awareness that the researcher is re analyzed the tapes to understand what the stu
ally an outsider. In any case, fieldwork is not fin dents saw as important to their decisions. In ad
CHAPTER 11 / FIELD RESEARCH 301
mplications of say focus group 7. For more on gatekeepers and access, see Beck
e researcher is di go native (1970:11-29), Bogdan and Taylor (1975:30-32),
ing studied and is guilty knowledge and Wax (1971:367).
19. Doing field re internal consistency 8. Negotiation in the field is discussed in Gans
impact on the re jotted notes (1982), Johnson (1975:58-59, 76-77), and Schatz
llife, and sense of member validation man and Strauss (1973:22-23).
:research. Field re 9. See Douglas (1976), Emerson (1981:367-368),
naturalism
and Johnson (1975:124-129) on the question of
, but it is a way to normalize social research
whether the researcher should always be patient,
rld that otherwise structural question polite, and considerate.
10. For more on ways to record and organize field
ires a combination data, see Bogdan and Taylor (1975:60-73), Ham
I1g sense of self, the Endnotes mersley and Atkinson (1983:144-173), and Kirk
m incredible ability and Miller (1986: 49-59).
" tremendous pa 1. For studies of these sites or topics, see Neuman 11. See Duneier (1999:342-343) for detailed discus
thy for others, su (2000, 2003). On studies of children or schools, sion.
think very quickly see Corsaro (1994), Corsaro and Molinari (2000), 12. For more on validity in field research, see Briggs
ee subtle intercon Eder (1995), Eder and Kinney (1995), Kelle (1986:24), Bogdan and Taylor (1975), Douglas
Jr events, and a su (2000), and Merten (1999). On studies of home (1976), Emerson (1981:361-363), and Sanjek
less people, see Lankenau (1999), and on studies (1990).
elfin writing.
offemale strippers, see Wood (2000). 13. The types of questions are adapted from Spradley
t when a researcher
2. Ethnography is described in Agar (1986), Franke (1979a, 1979b).
le interacting in the (1983), Hammersley and Atkinson (1983), San 14. Field research informants are discussed in Dean
icro-level or small day (1983), and Spradley (1979a:3-12, 1979b:3 and associates (1969), Kemp and Ellen (1984),
n. It is less effective 16). Schatzman and Strauss (1973), Spradley
level processes and 3. For a general discussion of field research and nat (1979a:46-54), and Whyte (1982).
, useless for events uralism, see Adler and Adler (1994), Georges and 15. For a discussion of focus groups, see Bischoping
t past or processes Jones (1980), Holy (1984), and Pearsall (1970). and Dykema (1999), Churchill (1983:179-184),
listorical-compara For discussions of contrasting types of field re Krueger (1988), Labaw (1980:54-58), and Mor
he next chapter, is search, see Clammer (1984), Gonor (1977), Hol gan (1996).
ng these types of stein and Gubrium (1994), Morse (1994), 16. See Lofland and Lofland (1995:26, 63, 75,
Schwandt (1994), and Strauss and Corbin (1994). 168-177), Miles and Huberman (1994:288-297),
4. See Lofland (1976:13-23) and Shaffir and col and Punch (1986).
leagues (1980:18-20) on feeling marginal. 17. Fetterman (1989) discusses the idea of guilty
5. See Adler and Adler (1987:67-78). knowledge.
6. See Hammersley and Atkinson (1983:42-45) and
Lofland and Lofland (1995:16-30).
Historical-Comparative
Research
Introduction
Research Questions Appropriate for Historical-Comparative
Research
The Logic of Historical-Comparative Research
The Logic of Historical-Comparative Research and Quantitative
Research
The Logic of Historical-Comparative Research and Interpretive
Research
A Distinct Historical-Comparative Approach
Steps in a Historical-Comparative Research Project
Conceptualizing the Object of Inquiry
Locating Evidence
Evaluating Quality of Evidence
Organizing Evidence
Synthesizing
Writing a Report
Data and Evidence in Historical Context
Types of Historical Evidence
Research with Secondary Sources
Research with Primary Sources
Comparative Research
Types of Comparative Research
The Units Being Compared
Data in Cross-Cultural Research
Equivalence in Historical-Comparative Research
The Importance of Equivalence
Types of Equivalence
Ethics
Conclusion
304
CHAPTER 12 I HISTORICAL-COMPARATIVE RESEARCH 305
omparative method
:omparative
In Women of the Klan, Kathleen Blee (1 991) noted To locate survivors 60 years after the Klan was
ve Research that, prior to her research, no one had studied the active, Blee had to be persistent and ingenious. She
'Tical-Comparative estimated 500,000 women in the largest racist, mailed a notice about her research to every local
he confusion is that right-wing movement in the United States. She sug newspaper, church bulletin, advertising supplement,
.ly oriented and in gested that this may have been due to an assumption historical society, and public library in Indiana. She
tatively oriented re that women were apolitical and passive. Her sixyears obtained 3 written recollections, 3 unrecorded in
'comparative issues. of research into the unknown members of a secret terviews, and 15 recorded interviews.Most of her in
the idea that there is society over 60 years ago shows the ingenuity formants were over age 80. They recalled the Klan as
needed in historical-sociological research. an important part of their lives. Bleeverified parts of
y measure variables,
Blee focused on the state of Indiana, where as their memories through newspaper and other docu
iantitative data, and
many as 32 percent of White Protestant women mentaryevidence.
er generalizable laws
were members of the Ku Klux Klan at its peak in the Membership in the Klan remains controversial. In
societies. They see
1 920s.ln addition to reviewing published studies on the interviews, Bleedid not reveal her opinions about
: between quantita the Klan, her documentary investigation included the Klan. Although she was tested, Blee remained
.torical-cornparative newspapers, pamphlets, and unpublished reports. neutral and did not denounce the Klan. She stated,
She conducted library research on primary and sec "My own background in Indiana (where I lived from
ramines social life in ondary materials at over half a dozen college, gov primary school through college) and white skin led
ion-that of the re ernment, and historical libraries. The historical informants to assume-lacking spoken evidence to
.ative research can be photographs, sketches, and maps in the book give the contrary-that I shared their worldview" (p. 5).
ensions: Is the focus readers a feel for the topic. She did not find Klan women brutal, ignorant, and
In, a small set of na Finding information was difficult. Blee did not have full of hatred. Blee got an unexpected response to a
the focus on a single access to membership lists. She identified Klan question on why the women had joined the Klan.
'oss many years, or a women by piecing together a few surviving rosters, Most were puzzled by the question. To them it
ialysis based primar locating newspaper obituaries that identified women needed no explanation-it was just "a way of grow
:ative data? as Klan members, scrutinizing public notices or anti ing up" and "to get together and enjoy."
Klan documents for the names of Klan women, and
interviewingsurviving women of the Klan.
Comparative
ve Research
.torical-comparative ative approach for using a small number of approach and rejects causal statements, system
rs from the positivist cases. They believe that historical-comparative atic concepts, or abstract theoretical models. In
'erne interpretive ap research is inadequate because it rarely produces the extremist interpretive approach, each social
itive researchers who probabilistic causal generalizations that they setting is unique and comparisons are impossi
.tive data may depart take as indicating a "true" (i.e., positivist) ble.
Their research is an SCIence.
1 limited number of Like interpretive field researchers, H-C re A Distinct Historical
ling and context are searchers focus on culture, try to see through the
Comparative Approach
m one nation, can be eyes of those being studied, reconstruct the lives
es can elaborate his ofthe people studied, and examine particular in The distinct historical-comparative research
y concrete historical dividuals or groups. An extremist interpretive method avoids the excesses of the positivist and
position says that an empathic understanding of extreme interpretive approaches. It combines a
e positivist approach the people being studied is the sole goal of social sensitivity to specific historical or cultural con
e historical-compar- research. It takes a strict, idiographic, descriptive texts with theoretical generalization. The logic
308 PART THREE / CONDUCTING QUALITATIVE RESEARCH
and goals of H -C research are closer to those of research. Both involve interpretation, which in
. field research than to positivist approaches. The troduces the interpreter's location in time, place,
following discussion describes similarities be and worldview. Historical-comparative research
tween H -C research and field research, and six does not try to produce a single, unequivocal set
more unique features of historical-comparative of objective facts. Rather, it is a confrontation of
research (see Table 12.l). old with new or of different worldviews. It rec
ognizes that a researcher's reading of historical
Similarities to Field Research. First, both H -C or comparative evidence is influenced by an
research and field research recognize that the re awareness of the past and by living in the
searcher's point ofview is an unavoidable part of present.
----------------------....1..
etation, which in Second, both field and H-C research exam searcher has a unique perspective and assembles
tion in time, place, ine a great diversity of data. In both, the re a unique body of evidence. Instead, researchers
nparative research searcher becomes immersed in data to gain an offer plausible accounts and limited generaliza
le, unequivocal set empathic understanding of events and people. tions.
a confrontation of Both capture subjective feelings and note how
vorldviews. It rec everyday, ordinary activities signify important Unique Features of Historical-Comparative
ading of historical social meaning. Research. Despite its many similarities to field
influenced by an The researcher inquires, selects, and focuses research, some important differences distinguish
by living in the on specific aspects of social life from the vast ar H-C research. Research on the past and on an
ray of events, actions, symbols, and words. An alien culture share much in common, and what
H-C researcher organizes data and focuses at they share distinguishes them from other
tention on the basis of evolving concepts. He or approaches.
she examines rituals and symbols that dramatize First, H-C research usually relies on limited
Qualitative culture (e.g., parades, clothing, placement of ob and indirect evidence. Direct observation or in
jects, etc.) and investigates the motives, reasons, volvement by a researcher is often impossible.
and justifications for behaviors. An H-C researcher reconstructs what occurred
Third, both field and H-C researchers use from the evidence, but cannot have absolute
groundedtheory. Theory usually emerges during confidence in the reconstruction. Historical evi
h process
the process of data collection. dence depends on the survival of data from the
Next, in both field and H-C research the re past, usually in the form of documents (e.g., let
ize variables searcher's meaning system frequently differs ters and newspapers). The researcher is limited
from that of the people he or she studies, but he to what has not been destroyed and what leaves
:npared with or she tries to penetrate and understand their a trace, record, or other evidence behind.
point of view. Once the life, language, and per Historical-comparative researchers must
spective of the people being studied have been also interpret the evidence. Different people
mastered, the researcher "translates" it for others looking at the same evidence often ascribe dif
.ehavior
who read his or her report. ferent meanings to it, so a researcher must re
Fifth, both field and H -C researchers focus flect on evidence. An understanding of it based
ontext-free laws on process and sequence. They see the passage of on a first glance is rarely possible. To do this, a
time and process as essential to how people con researcher becomes immersed in and absorbs
struct social reality. This is related to how both details about a context. For example, a re
are sensitive to an ever-present tension between searcher examining the family in the past or a
agency-the active moving fluid side of people distant country needs to be aware of the full so
changing social reality-and structure-the cial context (e.g., the nature of work, forms of
fixed regularities and patterns that shape social communication, transportation technology,
[ or historical life. For both types of research social reality si etc.). He or she looks at maps and gets a feel for
multaneously is what people create and some the laws in effect, the condition of medical care,
motives as causal thing that imposes restrictions on human and common social practices. For example, the
choice.' meaning of "a visit by a family member" is af
Sixth, generalization and theory are limited fected by conditions such as roads of dirt and
d due to a in field and H-C research. Historical and cross mud, the inability to call ahead of time, and the
cultural knowledge is incomplete and provi lives ofpeople who work on a farm with animals
ers of social reality sional, based on selective facts and limited that need constant watching.
xts for more questions. Neither deduces propositions or tests A reconstruction of the past or another cul
hypotheses in order to uncover fixed laws. Like ture is easily distorted. Compared to the people
wise, replication is unrealistic because each re- being studied, a researcher is usually more aware
310 PART THREE / CONDUCTING QUALITATIVE RESEARCH
of events occurring prior to the time studied, tion, proletarianization, tightened labor mar
events occurring in places other than the loca kets, and the like.
tion studied, and events that occurred after the Historical-comparative researchers shift be
period studied. This awareness gives the re tween details of specific context and making a
searcher a greater sense of coherence than was general comparison. A researcher examines spe
experienced by those living in the past or in an cific contexts, notes similarities and differences,
isolated social setting that he or she guards then generalizes. Comparative researchers com
against in a reconstruction. pare across cultural-geographic units (e.g., ur
Historical-comparative researchers recog ban areas, nations, societies, etc.)." Historical
nize the capacity of people to learn, make deci researchers investigate past contexts, usually in
sions, and act on what they learn to modify the one culture (e.g., periods, epochs, ages, eras,
course of events. For example, if a group of peo etc.), for sequence and comparison. Of course, a
ple are aware of or gain consciousness of their researcher can combine both to investigate mul
own past history and avoid the mistakes of the tiple cultural contexts in one or more historical
past, they may act consciously to alter the course contexts. Yet, each period or society has its
of events. Of course, people will not necessarily unique causal processes, meaning systems, and
learn or act on what they have learned, and if social relations, which may lack equivalent ele
they do act they will not necessarily be suc ments across the units. This produces a creative
cessful. Nevertheless, people's capacity to learn tension between the concrete specifics in a con
introduces indeterminacy into historical text and the abstract ideas a researcher uses to
comparative explanations. make links across contexts.
An H-C researcher wants to find out The use of transcultural concepts in com
whether people viewed various courses of action parative analysis is analogous to the use of tran
as plausible. Thus, the worldview and knowledge shistorical ones in historical research' In
of the people under study shaped what they saw comparative research, a researcher translates the
as possible or impossible ways to achieve goals. specifics of a context into a common, theoretical
The researcher asks whether people were con language. In historical research, theoretical con
scious of certain things. For example, if an army cepts are applied across time.
knew an enemy attack was coming and so de
cided to cross a river in the middle of the night,
the action "crossing the river" would have a dif
STEPS IN A HISTORICAL
ferent meaning than in the situation where the
COMPARATIVE RESEARCH
army did not know the enemy was approaching.
PROJECT
A historical-comparative researcher inte
grates the micro (small-scale, face-to-face inter In this section, we turn to the process of doing
action) and macro (large-scale social structures) H-C research. Conducting historical-compara
levels. The H-C researcher describes both levels tive research does not involve a rigid set of steps
or layers of reality and links them to each other. and, with only a few exceptions, it does not use
For example, an H-C researcher examines the complex or specialized techniques.
details of individual biographies by reading di
aries or letters to get a feel for the individuals:
Conceptualizing the Object of Inquiry
the food they ate, their recreational pursuits,
their clothing, their sicknesses, their relations An H -C researcher begins by becoming familiar
with friends, and so on. He or she links this with the setting and conceptualizing what is be
micro-level view to macro-level processes: in ing studied. He or she may start with a loose
creased immigration, mechanization of produc model or a set of preliminary concepts and apply
CHAPTER 12 I HISTORICAL-COMPARATIVE RESEARCH 311
!led labor mar- them to a specific setting. The provisional con A researcher adjusts initial concepts, questions,
cepts contain implicit assumptions or organiz or focus on the basis of what he or she discovers
archers shift be ing categories to "package" observations and in the evidence and considers a range of research
t and making a guide a search through evidence. reports at different levels of analysis (e.g., general
:r examines spe If a researcher is not already familiar with context and detailed narratives on specific
and differences, the historical era or comparative settings, he or topics).
esearchers com she conducts an orientation reading (reading
; units (e.g., ur several general works). This will help the re
Evaluating Quality of Evidence
tc.)." Historical searcher grasp the specific setting, assemble or
itexts, usually in ganizing concepts, subdivide the main issue, and The H-C researcher gathers evidence with two
ichs, ages, eras, develop lists of questions to ask. It is impossible questions in mind: How relevant is the evidence
son. Of course, a to begin serious research without a framework to emerging research questions and evolving
investigate mul of assumptions, concepts, and theory. Concepts concepts? How accurate and strong is the
r more historical and evidence interact to stimulate research. For evidence?
. society has its example, Skocpol (1979) began her study of rev As the focus of research shifts, evidence that
ing systems, and olution with puzzles in macro-sociological the was not relevant can become relevant. Likewise,
k equivalent ele ory and the histories of specific revolutions. The some evidence may stimulate new avenues of in
oduces a creative lack of fit between histories of revolutions and quiry and a search for additional confirming ev
pecifics in a con existing theories stimulated her research. idence. An H -C researcher reads evidence for
esearcher uses to three things: the implicit conceptual frame
works, particular details, and empirical general
locating Evidence
:oncepts in com izations. He or she evaluates alternative
o the use of tran Next, a researcher locates and gathers evidence interpretations of evidence and looks for "si
al research.P In through extensive bibliographic work. A re lences," or cases where the evidence fails to ad
.her translates the searcher uses many indexes, catalogs, and refer dress an event, topic, or issue. For example,
nmon, theoretical ence works that list what libraries contain. For when examining a group of leading male mer
1, theoretical con-
comparative research, this means focusing on chants in the 1890s, a researcher finds that the
specific nations or units and on particular kinds evidence and documents about them ignore
of evidence within each. The researcher fre their wives and many servants.
quently spends many weeks searching for
sources in libraries, travels to several different
Organizing Evidence
l specialized research libraries, and reads dozens
CH (if not hundreds) of books and articles. Com As a researcher gathers evidence and locates new
parative research often involves learning one or sources, he or she begins to organize the data.
: process of doing more foreign languages. Obviously, it is unwise to take notes madly and
storical-compara As the researcher masters the literature and let them pile up haphazardly. A researcher be
a rigid set of steps takes numerous detailed notes, he or she com gins a preliminary analysis by noting low-level
ns, it does not use pletes many specific tasks: creating a bibliogra generalizations or themes. Next, a researcher or
ques. phy list (on cards or computer) with complete ganizes evidence, using theoretical insights to
citations, taking notes that are neither too stimulate new ways to organize data and for new
skimpy nor too extensive (i.e., more than one questions to ask of evidence.
ct of Inquiry sentence but less than dozens of pages of The interaction of data and theory means
becoming familiar quotes), leaving margins on note cards for that a researcher goes beyond a surface examina
alizing what is be adding themes later on, taking all notes in the tion of the evidence to develop new concepts by
start with a loose same format (e.g., on cards, paper, etc.), and de critically evaluating the evidence based on the
.oncepts and apply veloping a file on themes or working hypotheses. ory. For example, a researcher reads a mass of
312 PART THREE I CONDUCTING QUALITATIVE RESEARCH
evidence about a protest movement. The pre mountains of evidence into exposition and pre
liminary analysis organizes the evidence into a pares extensive footnotes. She or he must also
theme: People who are active in protest interact weave together the evidence and arguments to
with each other and develop shared cultural communicate a coherent, convincing picture or
meanings. He or she examines theories of cul "tell a story" to readers.
ture and movements, then formulates a new
concept: "oppositional movement subculture."
The researcher then uses this concept to reexam
ine the evidence. DATA AND EVIDENCE IN
HISTORICAL CONTEXT
ound in archives (a
: stored), in private
source. Researchers find them in their original ple, it was illegal for slaves in the United States to
:s, and in museums
form or preserved in microfiche or on film. They read or write, and thus written sources on the
uments and objects
are often the only surviving record of the words, experience of slavery have been indirect or diffi
rams, commercials,
thoughts, and feelings of people in the past. cult to find.
be primary sources
Written documents are helpful for studying so The written word on paper was the main
mple of a classic pri
cieties and historical periods with writing and medium of communication prior to the wide
ellowed letters writ
literate people. A frequent criticism of written spread use of telecommunications, computers,
war to his wife and
sources is that they were largely written by elites and video technology to record events and
cher.
or those in official organizations; thus, the views ideas. In fact, the spread of forms of communi
shed written docu
of the illiterate, the poor, or those outside official cation that do not leave a permanent physical
ant type of primary
social institutions may be overlooked. For exam- record (e.g., telephone conversations, computer
314 PART THREE / CONDUCTING QUALITATIVE RESEARCH
records, and television or radio broadcasts), and Limitations of secondary historical evidence
which have largely replaced letters, written include problems of inaccurate historical ac
ledgers, and newspapers, may make the work of counts and a lack of studies in areas of interest.
future historians more difficult. Such sources cannot be used to test hypotheses.
Post facto (after-the-fact) explanations cannot
Secondary Sources. Primary sources have re meet positivist criteria of falsifiability, because
alism and authenticity, but the practicallimita few statistical controls can be used and replica
tion of time can restrict research on many tion is impossible. Yet, historical research by
primary sources to a narrow time frame or loca others plays an important role in developing
tion. To get a broader picture, many H-C re general explanations, among its other uses. For
searchers use secondary sources, the writings of example, such research substantiates the emer
specialist historians who have spent years study gence and evolution of tendencies over time.
ing primary sources.
Potential Problems. The many volumes of
Running Records. Running records consist of secondary sources present a maze of details and
files or existing statistical documents maintained interpretations for an H-C researcher. He or she
by organizations. An example of a running must transform the mass of descriptive studies
record is a file in a country church that contains into an intelligible picture that is consistent with
a record of every marriage and every death from the richness of the evidence. It also must bridge
1910 to the present. the many specific time periods or locales. The re
searcher faces potential problems with sec
Recollections. The words or writings of indi ondary sources.
viduals about their past lives or experiences One problem is that historians rarely pre
based on memory are recollections. These can be sent theory-free, objective "facts." They implic
in the form of memoirs, autobiographies, or in itly frame raw data, categorize information, and
terviews. Because memory is imperfect, recollec shape evidence using concepts. The historian's
tions are often distorted in ways that primary concepts are a mixture drawn from journalism,
sources are not. For example, Blee (1991) inter the language ofhistorical actors, ideologies, phi
viewed a woman in her late eighties about being losophy, everyday language in the present, and
in the Ku Klux Klan (see Box 12.1). social science. Most are vague, applied inconsis
In gathering oral history, a type of recollec tently, and not mutually exclusive nor exhaus
tion, a researcher conducts unstructured inter tive. For example, a historian describes a group
views with people about their lives or events in of people in a nineteenth-century town as upper
the past. This approach is especially valuable for class, but never defines the term and fails to link
nonelite groups or the illiterate. The oral history it to any theory of social classes. The historian's
technique began in the 1930s and now has a pro implicit theories constrain the evidence and the
fessional association and scholarly journal de social researcher may be looking for evidence for
voted to it. explanations that are contrary to ones implicitly
being used by historians in secondary sources.
Historians also select some information
Research with Secondary Sources
from all possible evidence. Yet, the H-C re
Usesand Limitations. Social researchers often searcher does not know how this was done.
use secondary sources, the books and articles Without knowing the selection process, a histor
written by historians, as evidence of past condi ical-comparative researcher must rely on the
tions." Secondary sources have limitations and historian's judgments, which can contain bi
need to be used with caution. ases.f For example, a historian reads 10,000
CHAPTER 12 I HISTORICAL-COMPARATIVE RESEARCH 315
y historical evidence pages of newspapers, letters, and diaries, then Also, narratives rarely explicitly indicate
urate historical ac boils down this information into summaries and how combination or interaction effects operate,
:in areas of interest. selected quotes in a 100-page book. An H-C re or the relative size of different factors. For exam
d to test hypotheses. searcher does not know whether information ple, the historian discusses three conditions as
rxplanations cannot that the historian left out is relevant for his or causing an event. Yet, rarely do readers know
alsifiability, because her purposes. which is most important or whether all three
)e used and replica The typical historian's research practice also conditions must operate together to have a
storical research by introduces an individualist bias. A heavy reliance causal impact, but no two conditions alone, or
role in developing on primary sources and surviving artifacts com no single condition alone, creates the same
g its other uses. For bines with an atheoretical orientation to pro impact.l''
stantiates the emer duce a narrow focus on the actions of specific The narrative organization creates difficul
encies over time. people. This particularistic, micro-level view di ties for the researcher using secondary sources
rects attention away from integrating themes or and creates conflicting findings. The H -C re
, many volumes of patterns. This emphasis on the documented ac searcher must read though weak concepts, un
: maze of details and tivities of specific individuals is a type oftheoret known selection criteria, and unclear casual
esearcher. He or she ical orientation.? logic. Theory may reside beneath the narrative
f descriptive studies Another problem is in the organization of but it remain implicit and hidden.
iat is consistent with the evidence. Tradional historians organize evi Two last problems are that a historian is
'. It also must bridge dence as a narrative history. This compounds influenced by when he or she is writing and
ds or locales. The re problems of undefined concepts and the selec historiographic schools. Various schools of his
iroblems with sec tion of evidence. In the historical narrative, ma toriography (e.g., diplomatic, demographic,
terial is chronologically organized around a ecological, psychological, Marxist, intellectual,
istorians rarely pre single coherent "story." Each part of the story is etc.) have their own rules for seeking evidence
'facts." They implic connected to each other part by its place in the and asking questions, and they give priority to
ize information, and time order of events. Together, all the parts form certain types of explanatory factors. Likewise, a
epts. The historian's a unity or whole. Conjuncture and contingency historian writing today will examine primary
vn from journalism, are key elements of the narrative form-that is, materials differently from how those writing in
tors, ideologies, phi if X (or X plus Z) occurred, then Y would occur, the past, such as 1920s, did.
~ in the present, and and if X (or X plus Z) had not occurred, some
ue, applied inconsis thing else would have followed. The contingency
rclusive nor exhaus Research with Primary Sources
creates a logical interdependency between ear
ill describes a group lier and later events. The historian is the major issue when using sec
ritury town as upper A difficulty of the narrative is that the pri ondary sources. When using primary sources,
term and fails to link mary organizing tool-time order or position in the biggest concern is that only a fraction of
isses. The historian's a sequence of events-does not denote theoreti everything written or used in the past has sur
the evidence and the calor historical causality. In other words, the vived into the present. Moreover, what survived
king for evidence for narrative meets only one of the three criteria for is a nonrandom sample of what once existed.
rry to ones implicitly establishing causality-that of temporal order. Historical-comparative researchers attempt
secondary sources. Moreover, narrative writing frequently obscures to read primary sources with the eyes and as
some information causal processes. This occurs when a historian in sumptions of a contemporary who lived in the
e. Yet, the H -C re cludes events in the narrative to enrich the back past. This means "bracketing," or holding bad
lOW this was done. ground or context, to add color, but that have no knowledge of subsequent events and modem
ion process, a histor causal significance. Likewise, he or she presents values. For example, when reading a source pr0
IT must rely on the events with a delayed causal impact, or events duced by a slave holder, moralizing aga.inst~
iich can contain bi that are temporarily "on hold" with a causal im ery or faulting the author for not s.eei.ng In en ti
torian reads 10,000 pact occuring at some unspecified later time. not worthwhile. The H-C researcber ~~ ha.:X
316 PART THREE / CONDUCTING QUALITATIVE RESEARCH
moral judgments and becomes a moral relativist events, sources, or people mentioned in the doc
while reading primary sources. ument and asks whether they can be verified. He
Another problem is that locating primary or she examines implicit assumptions or value
documents is a time-consuming task. A re positions, and the relevant conditions under
searcher must search through specialized in which the document was produced is noted
dexes and travel to archives or specialized (e.g., during wartime or under a totalitarian
libraries. Primary sources are often located in a regime). The researcher also considers language
dusty, out-of-the-way room full of stacked card usage at the time and the context of statements
board boxes containing masses of fading docu within the document to distill a meaning.
ments. These may be incomplete, unorganized, In an H-C study of Chinese migrant net
and in various stages of decay. Once the docu works in Peru, Chicago, and Hawaii early in the
ments or other primary sources are located, the twentieth century, McKeown (2001) used both
researcher evaluates them by subjecting them to primary and secondary historical sources and
external and internal criticism (see Figure 12.1). running records. He considered events over
External criticism means evaluating the au nearly a century of history and in three nations,
thenticity of a document itself to be certain that and everything from major international events
it is not a fake or a forgery. Criticism involves and national laws to individual family biogra
asking: Was the document created when it is phies. He relied on secondary sources for major
claimed to have been, in the place where it was national or international events. Although his
supposed to be, and by the person who claims to study was primarily historical and qualitative, he
be its author? Why was the document produced also examined quantitative data from running
to begin with, and how did it survive? records and provided graphs, charts, and tables
Once the document passes as being authen of statistics. His evidence also included geo
tic, a researcher uses internal criticism, an exam graphic maps and photographs, quotes from
ination of the document's contents to establish 100-year-old telegrams, official government
credibility. A researcher evaluates whether what documents, original newspaper reports, and se
is recorded was based on what the author di lections from personal letters in three languages.
rectly witnessed or is secondhand information. By comparing Chinese migrants over a long his
This requires examining both the literal mean torical period and in divergent social-cultural
ing of what is recorded and the subtle connota settings, he could trace the formation and oper
tions or intentions. The researcher notes other ation of transnational communities and social
------------------------.
FIGURE 12.1 Internal and External Criticism
External Criticism Internal Criticism
Eyewitness or
When
Secondhand Account?
Written?
Where Was Why Was
It Written? It Written?
Literal
Why Did It Meaning?
Survive?
Internal
Consistency?
in Context
Who Was the Real Author? Connotations?
CHAPTER 12 / HISTORICAL-COMPARATIVE RESEARCH 317
tioned in the doc identities. He learned that networks with links across several social units or settings are less
an be verified. He back to villagesin China and crossing several na likely to apply only to a specific culture or set
mptions or value tional borders helped to sustain a vibrant, inter ting. It is difficult for a researcher to detect hid
:onditions under acting community. The network was held den biases, assumptions, and values until he or
'oduced is noted together by social relations from the village of she applies a concept in different cultures or set
Ier a totalitarian origin, clan, family, business transactions, and tings. Different social settings provide a wider
onsiders language shared language and customs. One of McKe range of events or behavior, and the range in one
text of statements own's major arguments is that a perspective culture is usually narrower than for human be
a meaning. based solely on nations can limit a researcher's havior in general. Thus, research in a single cul
lese migrant net ability to see a social community that is transna ture or setting focuses on a restricted range of
Iawaii early in the tional and the hybrid of multiple cultures. Many possible social activity. For example, two re
(2001) used both aspects of the transnational community devel searchers, Hsi-Ping and Abdul, examine the re
oped in reaction to specific interactions that oc lationship between the age at which a child is
rical sources and
ered events over curred locally. weaned and the onset of emotional problems.
i in three nations, Hsi-Ping looks only at U.S. data, which show a
ternational events range from 5 to 15months at weaning, and indi
cate that emotional problems increase steadily as
ial family biogra
COMPARATIVE RESEARCH age of weaning increases. She concludes that late
sources for major
weaning causes emotional problems. Abdul
nts. Although his Types of Comparative Research
looks at data from 10 cultures and discovers a
and qualitative, he
ata from running A Comparative Method. Comparative re range from 5 to 36 months at weaning. He finds
charts, and tables search is more of a perspective or orientation that the rate of emotional problems rises with
lso included geo than a separate research technique. In this sec age of weaning until 18 months; it then peaks
phs, quotes from tion, we consider its strengths. and falls to a lower level. Abdul arrives at more
icial government Problems in other types of research are accurate conclusions: Emotional problems are
er reports, and se magnified in a comparative study. 11 Holt and likely for weaning between the ages of 6 and 24
in three languages. Turner (1970:6) said, "In principle, there is no months, but weaning either earlier or later re
ats over a long his difference between comparative cross-cultural duces the chances of emotional problems. Hsi
ent social-cultural
research and research conducted in a single soci Ping reached false conclusions about the
nnation and oper ety. The differences lie, rather, in the magnitude relationship because of the narrow range of
iunities and social of certain types of problems." A comparative weaning age in the United States.
perspective exposes weaknesses in research de The way comparative research raises new
----.
sign and helps a researcher improve the quality
of research. The focus of comparative research is
on similarities and differences between units.
Comparative research helps a researcher
questions and stimulates theory building is a
major strength. For example, Lamont (2000)
compared samples of blue-collar and lower
white-collar workers in France and the United
identify aspects of social life that are general States for their justifications and forms of argu
less or across units (e.g., cultures), as opposed to being ment used to explain racial differences. She drew
d Account? limited to one unit alone. All researchers want to random samples from telephone directories of
Why Was generalize to some degree. Positivist researchers Whites and Blacks in the suburbs of Paris and
It Written? are interested in discovering general laws or pat New York City and interviewed respondents for
E Literal
Meaning?
Internal
Consistency?
itions?
terns of socialbehavior that hold across societies.
But most positivist research is not comparative.
The comparative orientation improves
measurement and conceptualization. Concepts
developed by researchers who conduct research
two hours. Lamont found that the arguments of
racists and antiracists alike differed widely be
tween France and the United States. People use
arguments and rationales closely tied to the
dominant cultural themes of their society. For
318 PART THREE / CONDUCTING QUALITATIVE RESEARCH
example, in the United States, there is a long his H-C research, rigorous theory testing or experi
tory of using biological inferiority to explain mental research is rarely possible. For example, a
racial differences. This declined greatly but it still researcher interested in the effects of economic
exists, yet such a rationale is absent in France. In recessions cannot cause one group of countries
the United States, the market has near-sacred to have a recession while others do not. Instead,
status and both racist and antiracists frequently the researcher waits until a recession occurs and
used the market and personal economic success then looks at other characteristics of the country
in their arguments, but the market factor was or unit.
absent in France because it is not viewed as a fair
and efficient mechanism for allocating re
The Units Being Compared
sources. The French use cultural arguments,
egalitarianism, and the universality of all hu Culture versus Nation. For convenience, com
mans much more than Americans. In fact, the parative researchers often use the nation-state as
idea of a fundamental equality among all human their unit of analysis. The nation-state is the ma
beings was nearly absent among the justifica jor unit used in thinking about the divisions of
tions given in the United States. Such a discrep people across the globe today. Although it is a
ancy stimulates researchers to seek explanations dominant unit in current times, it is neither an
for the relationship and to develop new research inevitable nor a permanent one; in fact, it has
questions. been around for only about 300 years.
Comparative research also has limitations. The nation-state is a socially and politically
It is more difficult, more costly, and more time defined unit. In it, one government has sover
consuming than research that is not compara eignty (i.e., military control and political author
tive. The types of data that can be collected and ity) over populated territory. Economic relations
problems with equivalence (to be discussed) are (e.g., currency, trade, etc.), transportation
also frequent problems. routes, and communication systems are inte
Another limitation is the number of cases. grated within territorial boundaries. The people
Comparative researchers can rarely use random of the territory usually share a common lan
sampling. Sufficient information is not available guage and customs, and there is usually a com
for all of the approximately 150 nations in the mon educational system, legal system, and set of
world. It is unavailable for a nonrandom subset political symbols (e.g., flag, national anthem,
(poor countries, nondemocratic countries, etc.). etc.). The government claims to represent the in
In addition, can a researcher treat all nations as terests of all people in the territory under its
equal units when some have over a billion peo control.
ple and others only 100,000? The small number The nation-state is not the only unit for
of cases creates a tendency for researchers to par comparative research. It is frequently a surrogate
ticularize and see each case as unique, limiting for culture, which is more difficult to define as a
generalization. For example, a researcher exam concrete, observable unit. Culture refers to a
ines five cases (e.g., countries), butthe units dif common identity among people based on
fer from each other in 20 ways. It is difficult to shared social relations, beliefs, and technology.
test theory or determine relationships when Cultural differences in language, customs, tradi
there are more different characteristics than tions, and norms often follow national lines. In
units. fact, sharing a common culture is a major factor
A third limitation is that comparative re causing the formation of distinct nation-states.
searchers can apply, not test, theory, and can The boundaries of a nation-state may not
make only limited generalizations. Despite the match those of a culture. In some situations, a
ability to use and consider cases as wholes in single culture is divided into several nations; in
CHAPTER 12 / HISTORICAL-COMPARATIVE RESEARCH 319
y testing or experi other cases, a nation-state contains more than Canada, France, and the United States; a re
ible. For example, a one culture. Over the past centuries, boundaries searcher discovers a strong association between
effects of economic between cultures and distinct vibrant cultures speaking English and having the dollar as cur
group of countries have been destroyed, rearranged, or diffused as rency, or speaking French and using the franc as
ers do not. Instead, territory around the world was carved into currency. Obviously, the association exists be
ecession occurs and colonies or nation-states by wars and conquest. cause the units of analysis (i.e., states or
istics ofthe country For instance, European empires imposed arbi provinces) are subparts oflarger units (i.e., na
trary boundaries over several cultural groups in tions). The features of the units are due to their
nations that were once colonies. Likewise, new being parts of larger units and not to any rela
~d
immigrants or ethnic minorities are not always tionship among the features. Social geographers
assimilated into the dominant culture in a na also encounter this because many social and cul
convenience, com
tion. For example, one region of a nation may tural features diffuse across geographic space.
~ the nation-state as
have people with a distinct ethnic backgrounds, Galton's problem is an important issue in
ion-state is the ma
languages, customs, religions, social institutions, comparative research because cultures rarely have
rut the divisions of
and identities (e.g., the province of Quebec in clear, fixed boundaries. It is hard to say where one
IY. Although it is a
Canada). Such intranational cultures can create culture ends and another begins, whether one
nes, it is neither an
regional conflict, since ethnic and cultural iden culture is distinct from another, or whether the
one; in fact, it has
tities are the basis for nationalism. features of one culture have diffused to another
100years.
The nation-state is not always the best unit over time. Galton's problem occurs when the re
ially and politically
for comparative research. A researcher should lationship between two variables in two different
rrnment has sover
ask:What is the relevant comparative unit for my units is actually due to a common origin, and they
nd political author
research question-the nation, the culture, a are not truly distinct units (see Figure 12.2).
Economic relations
small region, or a subculture? For example, a re Galton's problem originated with regard to
.), transportation
search question is: Are income level and divorce comparisons across cultures, but it applies to
1 systems are inte
related (i.e., are higher-income people less likely historical comparisons also. It arises when a re
rdaries. The people
to divorce?)? A group of people with a distinct searcher asks whether units are really the same
re a common lan
culture, language, and religion live in one region or different in different historical periods. For
.e is usually a com
of a nation. Among them, income and divorce example, is the Cuba of 1875 the same country
usystem, and set of
are not related; elsewhere in the nation, however, as the Cuba of 2005? Do 130 years since the end
, national anthem,
where a different culture prevails, income and di ofSpanish colonialism, the rise of U.S. influence,
to represent the in
vorce are related. If a researcher uses the nation independence, dictatorship, and a communist
territory under its
state as his or her unit, the findings could be revolution fundamentally change the unit?
ambiguous and the explanation weak. Instead of
the only unit for assuming that each nation-state has a common
quently a surrogate Data in Cross-Cultural Research
culture, a researcher may find that a unit smaller
ficult to define as a than the nation-state is more appropriate. Comparative Field Research. Comparative re
:ulture refers to a searchers use field research and participant ob
people based on Galton's Problem. The issue of the units of servation in cultures other than their own.
fs, and technology. comparison is related to a problem named after Anthropologists are specially trained and pre
ige, customs, tradi Sir Francis Galton (1822-1911). When re pared for this type of research. The exchange of
N national lines. In searchers compare units or their characteristics, methods between anthropological and field re
ire is a major factor they want the units to be distinct and separate search suggests that there are small differences
inct nation-states. from each other. If the units are not different but between field research in one's own society and
tion-state may not are actually the subparts of a larger unit, then re in another culture. Field research in a different
some situations, a searchers will find spurious relationships. For culture is usually more difficult and places more
several nations; in example, the units are the states and provinces in requirements on the researcher.
320 PART THREE I CONDUCTING QUALITATIVE RESEARCH
----------.
FIG U R E 1 2.2 Galton's Problem
them have been studied by social researchers. A
valuable source of ethnographic data on differ
Galton's problem occurs when a researcher ent cultures is the Human Relations Area Files
observes the same social relationship (represented (HRAF) and the related Ethnographic Atlas, 12
by X) in different settings or societies The HRAF is a collection of field research re
(represented as A, B, and C) and falsely concludes ports that bring together information from
that the social relationship arose independently in ethnographic studies on various cultures, most
these different places. The researcher may believe of which are primitive or small tribal groupings.
he or she has discovered a relationship in three Extensive information on nearly 300 cultures
separate cases. But the actual reason for the has been organized by social characteristics or
occurrence of the social relation may be a shared practices (e.g., infant feeding, suicide, childbirth,
or common origin that has diffused from one etc.). A study on a particular culture is divided
setting to others. This is a problem because the up, and its information on a characteristic is
researcher who finds a relationship (e.g., a marriage grouped with that from other studies. This
pattern) in distinct settings or units of analysis makes it easy to compare many cultures on the
(e.g., societies) may believe it arose independently same characteristic. For example, a researcher
in different units. This belief suggests that the
interested in inheritance can learn that of 159
relationship is a human universal. The researcher
different cultures in which it has been studied,
may be unaware that in fact it exists because
119 have a patrilineal form (father to son), 27
people have shared the relationship across units.
matrilineal (mother to daughter), and 13 mixed
inheritance.
A B c Researchers can use the HRAF to study rela
tionships among several characteristics of differ
ent cultures. For example, to find out whether
sexual assault against women, or rape, is associ
ated with patriarchy (i.e., the holding of power
and authority by males), a researcher can exam
ine the presence of sexual assault and the
strength of patriarchy in many cultures.
Using the HRAF does have limitations,
however. First, the quality of the original re
search reports depends on the initial researcher's
length of time in the field, familiarity with the
language, and prior experience, as well as on the
explicitness of the research report. Also, the
Existing Sources of Qualitative Data. Com range of behavior observed by the initial re
parative researchers can use secondary sources. searcher and the depth of inquiry can vary. In
For example, a researcher who conducts a com addition, the categorization of characteristics in
parative study of the Brazilian, Canadian, and the HRAF can be crude. Another limitation in
Japanese educational systems can read studies by volves the cultures that have been studied. West
researchers from many countries, including ern researchers have made contact with and
Brazil, Canada, and Japan, which describe the conducted field research on a limited number of
education systems in the three nations. cultures prior to these cultures' contact with the
There may have been 5,000 different cul outside world. The cultures studied are not a
tures throughout human history; about 1,000 of representative sample of all the human cultures
CHAPTER 12 I HISTORICAL-COMPARATIVE RESEARCH 321
social researchers. A that existed. In addition, Galton's problem (dis of refusal. Researchers must tailor the sampling
phic data on differ cussed earlier) can be an issue. unit to the culture and consider how basic units,
Relations Area Files such as the family, are defined in that culture.
Iinographic Atlas. 12 Cross-National Survey Research. Survey re Special samples or methods for locating people
if field research re- search was discussed in a previous chapter. This for a sample may be required.
information from section examines issues that arise when a re Questionnaire writing problems in the re
ious cultures, most searcher uses the survey technique in other cul searcher's own culture are greatly magnified
.alltribal groupings. tures. The limitations of a cross-cultural survey when studying a different culture. A researcher
nearly 300 cultures are not different in principle from those of a sur needs to be especially sensitive to question word
al characteristics or vey within one culture. Nevertheless, they are ing, questionnaire length, introductions, and
, suicide, childbirth, usually much greater in magnitude and severity. topics included. He or she must be aware oflocal
.r culture is divided Survey research in a different culture re norms and of the topics that can and cannot be
l a characteristic is quires that the researcher possess an in-depth addressed by survey research. For example, open
rther studies. This knowledge of its norms, practices, and customs. questions about political issues, alcohol use, reli
any cultures on the Without such an in-depth knowledge, it is easy gion, or sexuality may be taboo. In addition to
imple, a researcher to make serious errors in procedure and inter these cultural issues, translation and language
n learn that of 159 pretation. Knowing another language is not equivalency often pose serious problems (see
.t has been studied, enough. A researcher needs to be multicultural Equivalence in Historical-Comparative Re
(father to son), 27 and thoroughly know the culture in addition to search). Techniques such as back translation (to
ater), and 13 mixed being familiar with the survey method. Substan be discussed) and the use of bilingual people are
tial advance knowledge about the other culture helpful, but often it is impossible to ask the exact
-IRAF to study rela is needed prior to entering it or planning the same question in a different language.
acteristics of differ survey. Close cooperation with the native people Interviewing requires special attention in
D find out whether of the other culture is also essential. cross-cultural situations. Selection and training
1, or rape, is associ A researcher's choice of the cultures or na of interviewers depends on the education,
e holding of power tions to include in a cross-cultural survey should norms, and etiquette of the other culture. The
searcher can exam be made on both substantive (e.g., theoretical, interview situation raises issues such as norms of
1 assault and the research question) and practical grounds. Each privacy, ways to gain trust, beliefs about confi
lY cultures. step of survey research (question wording, data dentiality, and differences in dialect. For exam
have limitations, collection, sampling, interviewing, etc.) must be ple, in some cultures, an interviewer must spend
of the original re tailored to the culture in which it is conducted. a day in informal discussion before achieving the
~ initial researcher's One critical issue is how the people from the rapport needed for a short formal interview.
amiliarity with the other culture experience the survey. In some cul
ce, as well as on the tures, the survey and interviewing itself may be a Existing Sources ofQuantitative Data. Quan
report. Also, the strange, frightening experience, analogous to a titative data for many variables are available for
by the initial re police interrogation. different nations. In addition, large collections
quiry can vary. In Sampling for a survey is also affected by the of quantitative data have been assembled. They
.f characteristics in cultural context. Comparative survey re gather information on many variables from
ther limitation in searchers must consider whether accurate sam other sources (e.g., newspaper articles, official
een studied. West pling frames are available, the quality of mail or government statistics, United Nations reports).
contact with and telephone service, and transportation to remote There are significant limitations on existing
limited number of rural areas. They need to be aware ofsuch factors cross-national data, many of which are shared
s' contact with the as how often people move, the types of dwellings by other existing statistics. The theoretical defin
studied are not a in which people live, the number of people in a ition of variables and the reliability of data col
ie human cultures dwelling, the telephone coverage, or typical rates lection can vary dramatically across nations.
322 PART THREE I CONDUCTING QUALITATIVE RESEARCH
Missing information is a frequent limitation. In corporation owners, both unemployment and
tentional misinformation in the official data imprisonment rates rise compared to times
from some governments can be a problem. An when low-income people and workers have
other limitation involves the nations on which greater political power and influence.
data are collected. For example, during a 35-year
period, new nations come into existence and
others change their names or change their hor
ders. EQUIVALENCE IN HISTORICAL
The existing data are available in major na COMPARATIVE RESEARCH
tional data archives in a form that computers
The Importance of Equivalence
can read, and researchers can conduct secondary
analysis on international existing statistics data. Equivalence is a critical issue in all research. It is
For example, Sutton (2004) conducted a quanti the issue of making comparisons across diver
tative, statistical study on 15 nations between gent contexts, or whether a researcher, living in
1960 and 1990. Researchers have long observed a specific time period and culture, correctly
that imprisonment rates do not closely follow reads, understands, or conceptualizes data about
changes in crime rates. Sutton tested the Rusche people from a different historical era or culture.
and Kirchheimer thesis. It says that unemploy Without equivalence, a researcher cannot use
ment rates cause a rise in imprisonment rates be the same concepts or measures in different cul
cause imprisonment is a government attempt to tures or historical periods, and this makes com
control a surplus of unemployed working-class parison difficult, if not impossible. It is similar to
males in the population who could become un the problems that arise with measurement valid
rulyand dangerous to the social order. Basically, ity in quantitative research.
it predicts that prisons will be filled when many
workers are out of work and will empty out
Types of Equivalence
when the economy is booming. Sutton gathered
data from government statistical yearbooks of The equivalence issue has implications for H-C
the 15 countries, from publications by interna research. A researcher might misunderstand or
tional organizations such as the World Health misinterpret events in a different era or culture.
Organization and the International Labor Orga Assuming that the interpretation is correct, a re
nization, and from prior social science studies searcher may find it difficult to conceptualize
that identified features of several nations, such as and organize the events to make comparisons
their unionization pattern, political party struc across times or places. If he or she fully grasps
ture, and so forth. Sutton found only limited another culture, a researcher may still find it dif
support for the original thesis, but he docu ficult to communicate with others from his or
mented a strong effect from several other fac her own time and culture. The equivalence issue
tors. He argued that the effect of unemployment can be divided into four subtypes: lexicon equiv
on imprisonment was probably spurious (see alence, contextual equivalence, conceptual
the discussion of a spurious relationship in equivalence, and measurement equivalence.
Chapters 2, 4, and 10 ofthis book). Sutton found
that specific features ofthe nation's political or Lexicon Equivalence. Lexicon equivalence is
ganization and labor market structure appeared the correct translation of words and phrases, or
to cause both specific unemployment patterns finding a word that means the same thing as an
and different imprisonment policies. In short, other word. This is clearest between two lan
when low-income people and workers were po guages. For example, in many languages and
litically weak compared to wealthy people and cultures there are different forms of address and
CHAPTER 12 / HISTORICAL-COMPARATIVE RESEARCH 323
employment and pronouns for intimates (e.g., close friends and Lexicon equivalence can be significant in
npared to times family members) and subordinates (e.g., historical research because the meaning of words
id workers have younger persons and lower-status people) from changes over time, even in the same language.
luence, those used in unknown or public settings or for The greater the distance in time, the greater the
persons of higher social status. There are no di chance that an expression will have a different
rectly equal linguistic forms ofspeech in English, meaning or connotation. For example, today the
although the idea of close personal versus public word weed refers to unwanted plants or to mar
)RICAL relations exists in English-speaking cultures. In ijuana, but in Shakespeare's era, the word meant
CH such languages, switching pronouns when say clothing (see Box 12.3).
ing, "How are you today?" might indicate a
ence change in status or in the social relationship. Contextual Equivalence. Contextual equiva
n all research. It is One would have to indicate it in another, per lence is the correct application of terms or con
.ons across diver haps nonverbal, way if speaking in English. In cepts in different social or historical contexts. It
searcher, living in cultures where age is an important status (e.g., is an attempt to achieve equivalence within spe
culture, correctly Japan), many status-based words exist that are cific contexts. For example, in cultures with dif
ualizes data about absent in English. One cannot say, for example, ferent dominant religions, a religious leader
ical era or culture. "my brother" without indicating whether one is (e.g., priest, minister, or rabbi) can have differ
ircher cannot use speaking of an older or younger brother, and ent roles, training, and authority. In some con
:s in different cul separate words are used for "my younger texts, priests are full-time male professionals
:1 this makes com brother" or "my older brother." who are wealthy, highly esteemed, well-educated
ible. It is similar to Comparative researchers often use a tech community leaders and also wield political
ieasurement valid- nique called back translation to achieve lexicon power. In other contexts, a priest is anyone who
equivalence. In back translation, a phrase or rises above others in a congregation on a tempo
question is translated from one language to an rary basis but is without power or standing in
other and then back again. For example, a
phrase in English is translated into Korean and
plications for H-C then independently translated from Korean
misunderstand or back into English. A researcher then compares
'ent era or culture. the first and second English versions. For exam
ion is correct, a re ple, in a study to compare knowledge of inter
t to conceptualize national issues by U.S. and Japanese college
nake comparisons students, the researchers developed a question The meaning of a statement or answer to a question
or she fully grasps naire in English. They next had a team of Iapan often depends on the customs of a culture, the social
nay still find it dif ese college faculty translate the questionnaire situation, and the manner in which the answer is spo
others from his or into Japanese. Some changes were made in the ken. The manner of answering can reverse the differ
eequivalence issue questionnaire. When they used back transla ent meanings of the same answer based on the
pes: lexicon equiv tion, they discovered "30 translating errors, in manner in which the answer was spoken.
ence, conceptual cluding some major ones" (Cogan et al., 1988:
It equivalence. 285).
Back translation does not help when words
con equivalence is for a concept do not exist in a different language
ds and phrases, or (e.g., there is no word for trust in Hindi, for Polite No Yes
~ same thing as an loyalty in Turkish, for privacy in Chinese, or for Emphatic Yes No
between two lan good quarrel in Thai). Thus, translation may re
ny languages and quire complex explanations, or a researcher may Source: Adapted from Hymes (1970:329).
rms of address and not be able to use certain concepts.
324 PART THREE / CONDUCTING QUALITATIVE RESEARCH
the community. Priests in such a context may be tions of being in a particular class, and class cat
less well educated, have low incomes, and be egories or boundaries differ across societies,
viewed as foolish but harmless people. A re making the study of social class across societies
searcher who asks about "priests" without notic difficult.
ing the context could make serious errors in At times, the same or a very similar concept
interpretations. exists across cultures but in different forms or
Context also applies across historical eras. degrees of strength. For example, in many Asian
For example, attending college has a different societies, there is a marked difference between
meaning today than in a historical context in the outward, public presentation and definition
which only the richest 1 percent of the popula of selfand the private, personal presentation and
tion attended college, most colleges had fewer the definition of self. What one reveals and
than 500 students, all were private all-male insti shows externally is often culturally detached
tutions that did not require a high school from true, internal feelings. Some languages
diploma for entry, and a collegecurriculum con mark this linguistically, as well. The idea of a dis
sisted of classical languages and moral training. tinct self for public, nonfamily, or nonprivate
Attending college 100 years ago was not the situations exists in Western cultures, as well, but
same as it is today; the historical context has al it is much weaker and less socially significant. In
tered the meaning of attending college. addition, many Western cultures assume that
the inner self is "real" and should be revealed, an
Conceptual Equivalence. The ability to use the assumption that is not always shared cross
same concept across divergent cultures or his culturally.
torical eras is conceptual equivalence. Researchers At other times, there is no direct cultural
live within specific cultures and historical eras. equivalent. For example, there is no direct West
Their concepts are based on their experiences ern conceptual equivalent for the Japanese ie. It
and knowledge from their own culture and era. is translated as family system, but this idea was
Researchers may try to stretch their concepts by created by outsiders to explain Japanese behav
learning about other cultures or eras, but their ior. The ie includes a continuing line of familial
views of other cultures or eras are colored by descent going back generations and continuing
their current life situations. This creates a persis into the future. Its meaning is closer to a Euro
tent tension and raises the question: Can a re pean lineage "house" among the feudal nobility
searcher create concepts that are simultaneously than the modern household or even an extended
true reflections of life experiences in different family. It includes ancestors, going back many
cultures or eras and that also make sense to him generations, and future descendants, with
or her? branches created by noninheriting male off
The issue of a researcher's concept is a spe spring (or adopted sons). It can also include a
cial case of a larger issue, because concepts can religious identity and property-holding dimen
be incompatible across different time periods or sions (as land or a business passed down for gen
cultures. Is it possible to create concepts that are erations). It can include feelings of obligation to
true, accurate, and valid representations of so one's ancestors and feelings to uphold any com
ciallife in two or more cultural or historical set mitments they may have made. The ie is also
tings that are very different? For example, the embedded in a web of hierarchical relationships
word class exists in many societies, but the sys with other ie and suggests social position or sta
tem of classes (i.e., the role of income, wealth, tus in a community.
job, education, status, relation to means of pro Conceptual equivalence also applies to the
duction), the number of classes, the connota study of different historical eras. For example,
CHAPTER 12 / HISTORICAL-COMPARATIVE RESEARCH 325
class, and class cat measuring income is very different in a historical offensive within a culture. Sensitivity means
er across societies, era with a largely noncash society in which most showing respect for the traditions, customs, and
lass across societies people grow their own food, make their own meaning of privacy in a host country. For exam
furniture and clothing, or barter goods. Where ple, it may be taboo for a man to interview a mar
'ery similar concept money is rarely used, it makes no sense to mea ried woman without her husband present.
different forms or sure income by number of dollars earned. In general, a researcher who visits another
rple, in many Asian Counting hogs, acres of land, pairs of shoes, ser culture wants to establish good relations with the
difference between vants, horse carriages, and the like may be more host country's government. He or she will not
rtion and definition appropriate. take data out of the country without giving
tal presentation and something (e.g., results) in return. The military
It one reveals and Measurement Equivalence. Measurement equiv or political interests ofthe researcher's home na
ulturally detached alence means measuring the same concept in dif tion or the researcher's personal values may con
s, Some languages ferent settings. If a researcher develops a concept flict with official policy in the host nation. A
11. The idea of a dis appropriate to different contexts, the question researcher may be suspected of being a spy or
nily, or nonprivate remains: Are different measures necessary in dif may be under pressure from his or her home
ultures, as well, but ferent contexts for the same concept? The mea country to gather covert information.
cially significant. In surement equivalence issue suggests that an H-C Sometimes, the researcher's presence or
ltures assume that researcher must examine many sources of partial findings may cause diplomatic problems. For
iuld be revealed, an evidence in order to measure or identify a theo example, a researcher who examines health care
rays shared cross- retical construct. When evidence exists in frag practices in a country, then declares that official
mentary forms, he or she must examine government policy is to ignore treating a serious
no direct cultural extensive quantities of indirect evidence in or illness can expect serious controversy. Likewise,
e is no direct West der to identify constructs. a researcher who is sympathetic to the cause of
r the Japanese ie. It groups who oppose the government may be
I, but this idea was threatened with imprisonment or asked to leave
in Japanese behav the country. Social researchers who conduct re
ETHICS
ling line of familial search in any country should be aware of such is
ms and continuing Historical-comparative research shares the ethi sues and the potential consequences of their
is closer to a Euro cal concerns found in other nonreactive research actions.
the feudal nobility techniques. The use of primary historical
Ireven an extended sources occasionally raises special ethical issues.
. going back many First, it is difficult to replicate research based on
CONCLUSION
lescendants, with primary material. The researcher's selection cri
aeriting male off teria for use of evidence and external criticism In this chapter, you have learned methodologi
can also include a of documents places a burden on the integrity of cal principles for an inquiry into historical and
ty-holding dimen the individual researcher. comparative materials. The H -C approach is ap
ssed down for gen Second, the right to protect one's privacy propriate when asking big questions about
IgS of obligation to may interfere with the right to gather evidence. A macro-level change, or for understanding social
) uphold any com person's descendants may want to destroy or hide processes that operate across time or are univer
ide. The ie is also private papers or evidence of scandalous behav sal across several societies. Historical-compara
:hical relationships ior. Even major political figures (e.g., presidents) tive research can be carried out in several ways,
ial position or sta want to hide embarrassing official documents. but a distinct qualitative H-C approach is similar
Comparative researchers must be sensitiveto cul to that of field research in important respects.
also applies to the tural and political issues of cross-cultural interac Historical-comparative research involves a
eras, For example, tion. They need to learn what is considered different orientation toward research more than
326 PART THREE I CONDUCTING QUALITATIVE RESEARCH
), and Stinchcombe
sion.
ee Sewell (1987).
ission of difficulties in
ee Whiting (1968).
e discussed by others,
Introduction
Comparing Methods of Data Analysis
uished between the
Similarities
Differences
1:76-87), Runciman
Successive Approximation
-96).
The Illustrative Method
nd limitations of com
Other Techniques
rson (1973), Holt and
Network Analysis
327
328 PART THREE! CONDUCTING QUALITATIVE RESEARCH
intitative analysis terns or relationships, but they begin analysis build new theory to create a realistic picture of
ocess. Researchers early in a research project, while they are stillcol social life and stimulate understanding more
er data and in so lecting data. The results of early data analysis than to test a causal hypothesis. Explanations
rs what they did. guide subsequent data collection. Thus, analysis tend to be rich in detail, sensitive to context, and
-ct large amounts is less a distinct final stage of research than a di capable of showing the complex processes or se
d document how mension of research that stretches across all quences of social life. The explanations may be
it. The degree to stages. causal, but this is not always the case. The re
dized and visible Another difference is the relationship be searcher's goal is to organize specific details into
reveal their study tween data and social theory. Quantitative re a coherent picture, model, or set of interlocked
searchers manipulate numbers that represent concepts.
I on comparison. empirical facts to test theoretical hypotheses. By A qualitative researcher divides explana
atures of the evi contrast, qualitative researchers create new con tions into two categories: highly unlikely and
rnally or with re cepts and theory by blending together empirical plausible. The researcher is satisfied by building
identify multiple evidence and abstract concepts. Instead of test a case or supplying supportive evidence. He or
, or mechanisms ing a hypothesis, a qualitative analyst may illus she may eliminate some theoretical explanations
len look for pat trate or color in evidence showing that a theory, from consideration while increasing the plausi
mces, aspects that generalization, or interpretation is plausible. bility of others because only a few explanations
litative and quan The fourth difference is the degree of ab will be consistent with a pattern in the data.
rvoid errors, false straction or distance from the details of social Qualitative analysis can eliminate an explana
~ inferences. Re life. In all data analysis, a researcher places raw tion by showing that a wide array of evidence
sible fallacies or il data into categories that he or she manipulates contradicts it. The data might support more
ious explanations, in order to identify patterns. Quantitative than one explanation, but all explanations will
and evaluate mer researchers assume that social life can be repre not be consistent with it. In addition to elimi
~ authentic, valid, sented by using numbers. When they manipu nating less plausible explanations, qualitative
late the numbers according to the laws of data analysis helps to verify a sequence of events
statistics, the numbers reveal features of social or the steps of a process. This temporal ordering
life. Qualitative analysis does not draw on a is the basis of finding associations among vari
large, well-establishedbody of formal knowledge ables, and it is useful in supporting causal
rrs from quantita from mathematics and statistics. The data are in arguments.
:irst, quantitative the form of words, which are relatively impre
cialized, standard cise, diffuse, and context-based, and can have
iques. Hypothesis more than one meaning.
CODING AND CONCEPT
s vary little across
FORMATION
ects. Quantitative
Explanations and Qualitative Data
md builds on ap Qualitative researchers often use general ideas,
it, qualitative data Qualitativeexplanationstake many forms. A qual themes, or concepts as analytic tools for making
he wide variety in itativeresearcherdoes not have to choose between generalizations. Qualitative analysis often uses
d by the many ap a rigid idiographic/nomothetic dichotomy nonvariable concepts or simple nominal-level
that is, between describing specifics and verify variables.
it quantitative re ing universal laws.Instead, a researcher develops
malysis until they explanations or generalizations that are close
Conceptualization
a and condensed to concrete data and contexts but are more
-n manipulate the than simple descriptions. He or she usually uses Quantitative researchers conceptualize and re
terns or relation a lower-level, less abstract theory, which is fine variables in a process that comes before data
can look for pat- grounded in concrete details. He or she may collection or analysis. Bycontrast, qualitative re
330 PART THREE / CONDUCTING QUALITATIVE RESEARCH
searchers form new concepts or refine concepts sures of variables, which are in the form of num
that are grounded in the data. Concept forma bers, into a machine-readable form for statistical
tion is integral to data analysis and begins during analysis.
data collection. Conceptualization is how a qual Coding data has a different meaning in
itative researcher organizes and makes sense of qualitative research. A researcher codes by orga
the data. nizing the raw data into conceptual categories
A qualitative researcher organizes data into and creates themes or concepts. Instead of a sim
categories on the basis of themes, concepts, or ple clerical task, coding is an integral part of data
similar features. He or she develops new con analysis guided by the research question. Coding
cepts, formulates conceptual definitions, and ex encourages higher-level thinking about the data
amines the relationships among concepts. and moves a researcher toward theorical gener
Eventually, he or she links concepts to each alizations.
other in terms of a sequence, as oppositional sets Coding is two simultaneous activities: me
(X is the opposite of Y) or as sets of similar cate chanical data reduction and analytic data cate
gories that he or she interweaves into theoretical gorization. Coding data is the hard work of
statements. Qualitative researchers conceptual reducing mountains of raw data into manage
ize or form concepts as they read through and able piles. In addition to making a large mass of
ask critical questions of data (e.g., field notes, data manageable, it is how a researcher imposes
historical documents, secondary sources, etc.). order on the data. Coding also allows a re
The questions can come from the abstract vo searcher to quickly retrieve relevant parts of the
cabulary of a discipline such as sociology-for data. Between the moments of thrill and inspira
example: Is this a case of class conflict? Was role tion' coding qualitative data, or filework, can be
conflict present in that situation? Is this a social wearisome and tedious.
movement? Questions can also be logical-for
example: What was the sequence of events? How Open Coding. Open coding is performed dur
does the way it happened here compare to over ing a first pass through recently collected data.
there? Are these the same or different, general or The researcher locates themes and assigns initial
specificcases? Researchers often conceptualize as codes or labels in a first attempt to condense the
they code qualitative data. mass of data into categories. He or she slowly
In qualitative data analysis, ideas and evi reads field notes, historical sources, or other
dence are mutually interdependent. This applies data, looking for critical terms, key events, or
particularly to case study analysis. Cases are not themes, which are then noted. Next, he or she
given preestablished empirical units or theoreti writes a preliminary concept or label at the edge
cal categories apart from data; they are defined by of a note card or computer record and highlights
data and theory. By analyzing a situation, the re it with brightly colored ink or in some similar
searcher organizes data and applies ideas simul way. The researcher is open to creating new
taneously to create or specify a case. Making or themes and to changing these initial codes in
creating a case, called casing, brings the data and subsequent analysis. A theoretical framework
theory together. Determining what to treat as a helps if it is used in a flexible manner.
case resolves a tension or strain between what the Open coding brings themes to the surface
researcher observes and his or her ideas about it. from deep inside the data. The themes are at a
low level of abstraction and come from the re
searcher's initial research question, concepts in
Coding Qualitative Data
the literature, terms used by members in the so
A quantitative researcher codes after all the data cial setting, or new thoughts stimulated by im
have been collected. He or she arranges mea mersion in the data.
CHAPTER 13 / ANALYSIS OF QUALITATIVE DATA 331
In the form of num An example of this is found in LeMasters's the Knights movement. Temperance is a new
e form for statistical (1975) field research study of a working-class and unexpected category. The researcher codes
tavern when he found that marriage came up in the notes with the label "temperance" and in
ferent meaning in many conversations. If he open coded field cludes it as a possible theme.
'cher codes byorga notes, he might have coded a block offield notes Qualitative researchers vary in the units they
nceptual categories with the theme marriage. Following is an exam code. Some code every line or every few words;
Its. Instead of a sim ple of hypothetical field notes that can be open others code paragraphs and argue that much of
integral part of data coded with the theme marriage: the data are not coded and are dross or left over.
.h question. Coding The degree of detail in coding depends on the
king about the data I wore a tie to the bar on Thursday because I research question, the "richness" ofthe data, and
ird theorical gener had been at a late meeting. Sam noticed it the researcher's purposes.
immediately and said. "Damn it, Doc. I wore Open-ended coding extends to analytic
eous activities: me one of them things once-when I got mar notes or memos that a researcher writes to him
analytic data cate ried-and look what happened to me! By self or herself while collecting data. Researchers
the hard work of God, the undertaker will have to put the next should write memos on their codes (see the later
data into manage one on." I ordered a beer, then asked him, discussion in Analytic Memo Writing).
cing a large mass of "Why did you get married?" He replied,
researcher imposes "What the hell you gain' to do? You just can't Axial Coding. This is a "second pass" through
: also allows a re go on shacking up with girls all your life-I the data. During open coding, a researcher fo
elevant parts of the did plenty of that when I was single"with a cuses on the actual data and assigns code labels
)fthrill and inspira smile and wink. He paused to order another for themes. There is no concern about making
or filework, can be beer and light a cigarette, then continued, "A connections among themes or elaborating the
man, sooner or later, likesto have a home of concepts that the themes represent. By contrast,
his own, and some kids, and to have that, you in axialcoding, the researcher begins with an or
~ is performed dur have to get married. There's no way out of ganized set of initial codes or preliminary con
ntly collected data. it-they got you hooked." I said, "Helen [his cepts. In this second pass, he or she focuses on
s and assigns initial wife] seemslike a nice person." He returned, the initial coded themes more than on the data.
rpt to condense the "Oh, hell, she's not a bad kid, but she's a god Additional codes or new ideas may emerge dur
;. He or she slowly damn woman and they get under my skin. ing this pass, and the researcher notes them; but
sources, or other They piss me off. If you go to a party, just his or her primary task is to review and examine
IDS, key events, or when you start having fun, the wifesays'let's initial codes. He or she moves toward organizing
ed. Next, he or she go home.'" (Adapted from LeMasters, ideas or themes and identifies the axis of key
or label at the edge 1975:36-37) concepts in analysis.
cord and highlights During axial coding, a researcher asks about
or in some similar Historical-comparative researchers also use causes and consequences, conditions and inter
n to creating new open coding. For example, a researcher studying actions, strategies and processes, and looks for
se initial codes in the Knights of Labor, an American nineteenth categories or concepts that cluster together. He
retical framework century movement for economic and political or she asks questions such as: Can I divide exist
manner. reform, reads a secondary source about the ac ing concepts into subdimensions or subcate
mes to the surface tivities of a local branch of the movement in a gories? Can I combine several closely related
he themes are at a specific town. When reading and taking notes, concepts into one more general one? Can I orga
come from the re the researcher notices that the Prohibition party nize categories into a sequence (i.e., A, then B,
estion, concepts in was important in local elections and that tem then C), or by their physical location (i.e., where
nembers in the so perance was debated by members of the local they occur), or their relationship to a major
stimulated by im- branch. The researcher's primary interest is in topic of interest? For example, a field researcher
the internal structure, ideology, and growth of studying working-class life divides the general is
332 PART THREE I CONDUCTING QUALITATIVE RESEARCH
sue of marriage into subparts (e.g., engagement, core generalizations or ideas. For example, a re
weddings). He or she marks all notes involving searcher studying working-class life in a tavern
parts of marriage and then relates marriage to decides to make gender relations a major theme.
themes of sexuality, division of labor in house In selective coding, the researcher goes through
hold tasks, views on children, and so on. When his or her field notes, looking for differences in
the theme reappears in different places, the re how men and women talk about dating, engage
searcher makes comparisons so he or she can see ments, weddings, divorce, extramarital affairs,
new themes (e.g., men and women have differ or husband/wife relations. He or she then com
ent attitudes toward marriage). pares male and female attitudes on each part of
In the example of historical research on the the theme of marriage.
Knights of Labor, a researcher looks for themes Likewise, the researcher studying the
related to temperance. He or she looks for dis Knights of Labor decides to make the move
cussions of saloons, drinking or drunkenness, ment's failure to form alliances with other polit
and relations between the movement and polit ical groups a major theme. The researcher goes
ical parties that support or oppose temperance. through his or her notes looking for compro
Themes that cluster around temperance could mise and conflict between the Knights and other
also include drinking as a form of recreation, political parties, including temperance groups
drinking as part of ethnic culture, and differ and the Prohibition party. The array of concepts
ences between men and women regarding and themes that are related to temperance in ax
drinking. ial coding helps him or her discover how the
Axial coding not only stimulates thinking temperance issue facilitated or inhibited al
about linkages between concepts or themes but liances.
it also raises new questions. It can suggest drop During selective coding, major themes or
ping some themes or examining others in more concepts ultimately guide the researcher's
depth. In addition, it reinforces the connections search. He or she reorganizes specific themes
between evidence and concepts. As a researcher identified in earlier coding and elaborates more
consolidates codes and locates evidences, he or than one major theme. For example, in the
she finds evidence in many places for core working-class tavern study, the researcher exam
themes and builds a dense web of support in the ines opinions on marriage to understand both
qualitative data for them. This is analogous to the theme of gender relations and the theme of
the idea of multiple indicators described with re different stages of the life cycle. He or she does
gard to reliability and measuring variables. The this because marriage can be looked at both
connection between a theme and data is ways. Likewise, in the Knights of Labor study,
strengthened by multiple instances of empirical the researcher can use temperance to under
evidence. stand the major theme of failed alliances and
also to understand another theme, sources of di
Selective Coding. By the time a researcher is vision within the movement that were based on
ready for this last pass through the data, he or ethnic or religious differences among members
she has identified the major themes of the re (see Figure 13.1).
search project. Selective coding involves scanning
data and previous codes. Researchers look selec
Analytic Memo Writing
tively for cases that illustrate themes and make
comparisons and contrasts after most or all data Qualitative researchers are compulsive note-tak
collection is complete. They begin after they ers. Their data are recorded in notes, they write
have well-developed concepts and have started comments on their research strategy in notes,
to organize their overall analysis around several and so on. They keep their notes organized in
CHAPTER 13 I ANALYSIS OF QUALITATIVE DATA 333
er studying the
make the move
s with other polit
ae researcher goes
Step 1: Open Coding
rnperance groups
earray of concepts
temperance in ax
discover how the Organization of Codes
! or inhibited al
. major themes or
the researcher's
es specific themes
id elaborates more
r example, in the
e researcher exam Step 2: Axial Coding
) understand both Organize all the codes created during open coding into a structure by separating
s and the theme of them into major or minor levels and showing relations among the codes.
.le, He or she does
oe looked at both
its of Labor study,
»erance to under
~iled alliances and
erne, sources of di
that were based on
s among members
mpulsive note-tak
n notes, they write Step 3: Selective Coding
I strategy in notes, Take the organized codes from the axial coding process and review the codes in the original
notes organized in data notes to select the best illustrations for entering them into a final report.
files, and often have many files with different good-quality analytic memos can become sec
kinds of notes: a file on methodological issues tions of the final report.
(e.g., locations of sources or ethical issues), a file The technology involved in writing analytic
of maps or diagrams, a file on possible overall memos is simple: pen and paper, a few note
outlines of a final report or chapter, a file on spe books, computer files, and photocopies of notes.
cific people or events, and so on. There are many ways to write analytic memos;
The analytic memo is a special type of note. each researcher develops his or her own style or
It is a memo or discussion of thoughts and ideas method. Some researchers make multiple copies
about the coding process that a researcher writes of notes, then cut them and place selections into
to himself or herself. Each coded theme or con an analytic memo file. This works well if the data
cept forms the basis of a separate memo, and the fIles are large and the analytic memos are kept
memo contains a discussion of the concept or distinct within the file (e.g., on different-colored
theme. The rough theoretical notes form the be paper or placed at the beginning). Other re
ginning of analytic memos. searchers link the analytic memo file locations to
The analytic memo forges a link between the data notes where a theme appears. Then it is
the concrete data or raw evidence and more ab easy to move between the analytic memo and the
stract, theoretical thinking (see Figure 13.2). It data. Because data notes contain links or marked
contains a researcher's reflections on and think themes, it is easy to locate specific sections in the
ing about the data and coding. The researcher data. An intermediate strategy is to keep a run
adds to the memo and uses it as he or she passes ning list of locations where a major theme ap
through the data with each type of coding. The pears in the raw data.
memos form the basis for analyzing data in the As a researcher reviews and modifies ana
research report. In fact, rewritten sections from lytic memos, he or she discusses ideas with col
----------------------.
I
I
I
f
I
I I
I~
I
Data Analytic
Other I I
Notes 1.-' Memos
Files ,---.J
.-'
~ Filed by Concept
orTheme Filed by Purpose
Organized by Date
or Data Source
J--:/
Final
Report
CHAPTER 13 I ANALYSIS OF QUALITATIVE DATA 335
llOS can become sec leagues, and returns to the literature with a focus Some argue that the narrative approach is a
on new issues. Analytic memos may help to gen presentation of data without analysis. There can
ed in writing analytic erate potential hypotheses, which can be added be analysis in a narrative, but it is "light" and
:l paper, a few note and dropped as needed, and to develop new subtle. In the narrative method, a researcher as
photocopies of notes. themes or coding systems. sembles the data into a descriptive picture or ac
'lite analytic memos; count of what occurred, but he or she largely
is or her own style or leaves the data to "speak for themselves." He or
make multiple copies she interjects little in the form of new systematic
ANALYTIC STRATEGIES FOR
I place selections into concepts, external theories, or abstract models.
QUALITATIVE DATA
works well if the data The explanation resides not in abstract concepts
ytic memos are kept Techniques ofcoding and memo writing are ap and theories, but in a combination of specific,
.on different-colored proaches to the analysis of qualitative data. Most concrete details. The researcher presents or re
.ginning), Other re qualitative researchers use these techniques to veals the social reality as members in a field set
nemo file locations to some degree, often combined with a more spe ting experience it, or the worldview of specific
ne appears. Then it is cific strategy for the analysis of qualitative data. historical actors at a particular point in time. By
ialytic memo and the In this section you will learn about four strate using little commentary, a researcher tries to
ntain links or marked gies researchers use to analyze qualitative data: convey an authentic feel for life's complexity as
pecificsections in the the narrative, ideal types, successive approxima experienced by particular people in specific cir
egy is to keep a run tion, and the illustrative method. cumstances, and does not derive abstract princi
e a major theme ap Compared to the analysis of quantitative ples or identify generalizable analytic patterns.
data, strategies for qualitative data are more di In the narrative, data are "analyzed" or "ex
's and modifies ana verse, less standardized, and less explicitly out plained" in the terminology and concepts of the
usses ideas with col- lined by researchers. Only in the past decade people being studied. The analysis appears in
have researchers started to explain and outline how a researcher organizes the data for presen
exactly how they analyze qualitative data. tation and tells the story. It appears in a greater
In general, data analysis means a search for attention to particular people, events, or facts,
patterns in data-recurrent behaviors, objects, and it relies on literary devices-the creative se
or a body of knowledge. Once a pattern is iden lection of particular words to tell a story, de
tified, it is interpreted in terms of a social theory scribe a setting, show character development,
or the setting in which it occurred. The qualita and present dramatic emphasis, intrigue, or
tive researcher moves from the description of a suspense.
historical event or social setting to a more gen Researchers debate the usefulness of the
eral interpretation of its meaning. narrative strategy. On the one hand, it provides
I rich concrete detail and clearly demonstrates the
Other
Files :--, The Narrative
temporal ordering of processes or specific
events. It captures a high degree of complexity
You encountered the narrative in the last chap and conveys a nuanced understanding of how
by Purpose ter on historical-comparative research. In field particular events or factors mutually affect each
,/ research, it is also called a natural history or other. The narrative allows the researcher to as
realist tale approach. The narrative is a largely semble very specific concrete details (i.e., the
atheoretical description. The researcher-author names, actions, and words of specific people and
"disappears" from the analysis and presents the the detailed descriptions of particular events at
concrete details in chronological order as if they specific times) that may be idiosyncratic but that
were the product of a unique and "naturally un contribute to a complete explanation. On the
folding" sequence of events. He or she simply other hand, many researchers criticize the narra
"tells a story" of what occurred. tive approach for being too complex, particular,
336 PART THREE I CONDUCTING QUALITATIVE RESEARCH
and idiosyncratic. It does not provide general or "reality" can be compared. An ideal type is a
izations. The narrative may present an over device used for comparison, because no reality
whelming array of particular details, but not ever fits an ideal type. For example, a researcher
provide a general explanation that researchers develops a mental model of the ideal democracy
can apply to other people, situations, or time pe or an ideal collegebeer party. These abstractions,
riods (see Box 13.1). with lists of characteristics, do not describe any
specific democracy or beer party; nevertheless,
they are useful when applied to many specific
Ideal Types
cases to see how well each case measures up to
Max Weber's ideal type is used by many qualita the ideal. This stage can be used with the illus
tive researchers. Ideal types are models or men trative method described earlier.
tal abstractions of social relations or processes. Weber's method of ideal types also comple
They are pure standards against which the data ments John Stuart Mill's method of agreement.
Many qualitative researchers, especially feminist re The narrative is found in literature, artistic ex
searchers, use the narrative because they believe it pressions, types of therapy, judicial inquiries, social
best enables them to retain a richness and authen or political histories, biography and autobiography,
ticity from their original data sources (i.e., individual medical case histories, and journalistic accounts. As a
personal stories or events in ethnographies, or spe way to organize, analyze, and present qualitative so
cific historical events). In simple terms, the narrative cial science data, the narrative shares many features
is story telling. In it, an author presents two or more with other academic and cultural communication
events in temporal and causal sequences. Some nar forms, but it differs from the positivist model for or
ratives are complex, with elements such as (1) a sum ganizing and reporting on data. The positivist model
mary statement of the entire story; (2) an emphasizes using impersonal, abstract, "neutral" lan
orientation that identifies specific times, places, per guage and a standardized analytic approach.
sons, and situations; (3) complicating actions or Many qualitative researchers argue that re
twists in the plot of "what happened"; (4) an evalua searchers who adopt the positivist model are sim
tion or emotional assessment of the narrative's ply using an alternative form of narrative, one with
meaning or signifigance; (5) a resolution or what oc specialized conventions. These conventions en
cured after a dramatic high point that resolves a sus courage formal analytic models and abstract theo
penseful climatic event; and (6) a coda or signal that ries, but such models or theories are not necessarily
the narrative is ending. superior to a story-telling narrative. Positivist data
People frequently tell one another stories in daily analysis and reporting conventions have two nega
life. They usually structure or organize their narra tive effects. First, they make it easier for researchers
tives into one of several recognized patterns, often to lose sight of the concrete actual events and per
recounting it with visual clues, gestures, or voice in sonal experiences that comprise social science data.
tonations for dramatic emphasis. The structure may Second, they make it more difficult for researchers
include plot lines, core metaphors, and rhetorical de to express ideas and build social theories in a for
vices that draw on familiar cultural and personal mod mat that most people find to be familiar and com
els to effectively communicate meanings to others. fortable.
CHAPTER 13 / ANALYSIS OF QUALITATIVE DATA 337
:i. An ideal type is a With the method of agreement, a researcher's at about patterns in data by referring to something
, because no reality tention is focused on what is common across that is already known or an experience familiar
ample, a researcher cases, and he or she looks for common causes in to the reader. They can describe relationships
:he ideal democracy cases with a common outcome. By itself, the buried deep within many details and are a short
These abstractions, method of agreement implies a comparison hand method for seeing patterns in a maze of
[o not describe any against actual cases. This comparison of cases specific events. They also make it easier to com
party; nevertheless, could also be made against an idealized model. A pare social processes across different cases or set
:d to many specific researcher could develop an ideal type of a social tings. For example, a researcher says that a room
ase measures up to process or relationship, then compare specific went silent after person X spoke: "A chill like a
used with the illus cases to it. cold gust of air" spread through the room. This
Iier. Qualitative researchers have used ideal types does not mean that the room temperature
Itypes also comple in two ways: to contrast the impact of contexts dropped or that a breeze was felt, but it suc
thod of agreement. and as analogy. cinctly expresses a rapid change in emotional
tone. Likewise, a researcher reports that gender
Contrast Contexts. Researchers who adopt a relations in society Y were such that women
strongly interpretive approach may use ideal were "viewed like property and treated like
types to interpret data in a way that is sensitive to slaves." This does not mean that the legal and so
the context and cultural meanings of members. cial relations between genders were identical to
They do not test hypotheses or create a general those of slave owner and slave. It implies that an
terature, artistic ex izable theory, but use the ideal type to bring out ideal type of a slave-and-master relationship
:Jicial inquiries, social the specifics of each case and to emphasize the would show major similarities to the evidence
, and autobiography, impact of the unique context. on relations between men and women if applied
ialistic accounts. Asa Researchers making contrasts between con to society Y.
resent qualitative so texts often choose cases with dramatic contrasts The use of analogies to analyze qualitative
shares many features or distinctive features. For example, in Work and data serves as a heuristic device (i.e., a device that
ural communication Authority in Industry, Reinhard Bendix (1956) helps one learn or see). It can represent some
isitivist model for or compared management relations in very differ thing that is unknown and is especially valuable
The positivistmodel ent contexts: Czarist Russia and industrializing when researchers attempt to make sense of or
istract, "neutral"lan England. explain data by referring to a deep structure or
tic approach. When comparing contexts, researchers do an underlying mechanism. Ideal types do not
.ers argue that re not use the ideal type to illustrate a theory in dif provide a definitive test of an explanation.
.ivist model are sirn
ferent cases or to discover regularities. Instead, Rather, they guide the conceptual reconstruc
f narrative, one with
they accentuate the specific and the unique. tion of the mass of details into a systematic
5e conventions en
Other methods of analysis focus on the general format.
; and abstract theo
s are not necessarily and ignore peculiarities. By contrast, a re
stive, Positivist data searcher who uses ideal types can show how
Successive Approximation
ions have two nega unique features shape the operation of general
asier for researchers processes. Successive approximation involves repeated it
tual events and per erations or cycling through steps, moving to
, social science data. Analogies. Ideal types are used as analogies to ward a final analysis. Over time, or after several
cult for researchers organize qualitative data. An analogy is a state iterations, a researcher moves from vague ideas
al theories in a for ment that two objects, processes, or events are and concrete details in the data toward a com
,e familiar and com- similar to each other. Researchers use analogies prehensive analysis with generalizations. This is
to communicate ideas and to facilitate logical similar to the three kinds of coding discussed
comparisons. Analogies transmit information earlier.
338 PART THREE / CONDUCTING QUALITATIVE RESEARCH
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _- m l. .
Get
Bowl
Preheat Grease
Oven Pan
Gather
Ingredients
L... ~
PourPan
into Batter ~ ...J
IseJmerl
of ways people understand the world. For exam of qualitative analysis. In addition to tax
ple (Canter et al., 1985:90), a gambler sorts a list onomies, maps, and lists, they suggested the use
of eight gambling establishments five times. of flowcharts, organizational charts, causal dia
Each sort has three to four categories. One of the grams, and various lists and grids to illustrate
sorts organized them based on "class of casino" analysis (see Figure 13.4).
(high to low). Other sorts were based on "frills,"
"size of stake," "make me money," and "per
sonal preference." By examining the sorts, the
SOFTWARE FOR QUALITATIVE
researcher sees how others organize their worlds.
DATA
Quantitative researchers have used computers
Diagrams
for nearly 40 years to generate tables, graphs, and
Qualitative researchers have moved toward pre charts to analyze and present numerical data. By
senting their data analysis in the form of dia contrast, qualitative researchers moved to com
grams and charts. Diagrams and charts help puters and diagrams only in the past decade. A
them organize ideas and systematically investi researcher who enters notes in a word-process
gate relations in the data, as well as communi ing program may quickly search for words and
cate results to readers. Researchers use spatial or phrases that can be adapted to coding data and
temporal maps, typologies, or sociograms. linking codes to analytic memos. Word process
Quantitative researchers have developed ing can also help a researcher revise and move
many graphs, tables, charts, and pictorial devices codes and parts of field notes.
to present information. Miles and Huberman New computer programs are continuously
(1994) argued that data display is a critical part being developed or modified, and most come
--~.
;j
FIG U R E 1 3.4
CHAPTER 13 / ANALYSIS OF QUALITATIVE DATA
~
;j
~ John Yes Yes N/A No
Mary Yes DK No Yes
Martin No Yes N/A Yes
Yoshi Yes No Yes Yes
Halfway (Jamie)
House Nurse
addition to tax
Police
Clinic Hospital I
y suggested the use
Dept.
charts, causal dia
l grids to illustrate
ITATIVE
with highly detailed and program-specific user rowboat, schooner, vessel, yacht, steamer, ocean
manuals. The review here does not go into detail liner, tug, canoe, skiff, cutter, aircraft carrier,
re used computers about specific software. It covers only the major dinghy, scow, galley, ark, cruiser, destroyer, flag
:tables, graphs, and approaches to qualitative data analysis at this ship, and submarine. In addition, some pro
numerical data. By time. grams permit the combination of words or
ers moved to com Some programs perform searches of text phases using logical terms (and, or, not) in what
the past decade. A documents. What they do is similar to the are called Boolean searches. For example, a re
in a word-process searching function available in most word-pro searcher may search long documents for when
ITch for words and cessing software. The specialized text retrieval the keywords college student and drinking and
to coding data and programs are faster and have the capability of smoking occur within four sentences of one an
lOS. Word process fmding close matches, slight misspellings, simi other, but only when the word fraternity is not
-r revise and move lar-sounding words, and synonyms. For exam present in the block of text. This Boolean search
ple, when a researcher looks for the keyword uses and to seek the intersection of college stu
s are continuously boat, the program might also tell whether any of dent with either of two behaviors that are con
d, and most come the following appeared: ship, battleship, frigate, nected by the logical term or, whereas the logical
342 PART THREE / CONDUCTING QUALITATIVE RESEARCH
search word not excludes situations in which the branches, and twigs) among the codes, and
term fraternity appears. among the data to which the codes refer. The
Most programs show the keyword or phrase qualitative data are rearranged in the program
and the surrounding text. The programs may based on the researcher's codes and the relations
also permit a researcher to write separate memos among codes that a researcher specifies.
or add short notes to the text. Some programs
count the keywords found and give their loca
tion. Most programs create a very specific index
CONCLUSION
for the text, based only on the terms of interest
to the researcher. In this chapter, you have learned how re
Textbase managers are similar to text re searchers analyze qualitative data. In many re
trieval programs. The key difference is their abil spects, qualitative data are more difficult to deal
ity to organize or sort information about search with than data in the form of numbers. Num
results. They allow researchers to sort notes by a bers have mathematical properties that let a re
key idea or to add factual information. For ex searcher use statistical procedures. Qualitative
ample, when the data are detailed notes on inter analysis requires more effort by an individual re
views, a researcher can add information about searcher to read and reread data notes, reflect on
the date and length of the interview, gender of in what is read, and make comparisons based on
terviewee, location of interview, and so on. The logic and judgment.
researcher can then sort and organize each inter Most forms of qualitative data analysis in
view or part of the interview notes using a com volve coding and writing analytic memos. Both
bination of key words and added information. are labor-intensive efforts by the researcher to
In addition, some programs have Hypertext read over data carefully and think about them
capability. Hypertext is a way oflinking terms to seriously. In addition, you learned about meth
other information. It works such that clicking ods that researchers have used for the analysis of
the mouse on one term causes a new screen (one qualitative data. They are a sample of the many
that has related information) to appear. The re methods of qualitative data analysis. You also
searcher can identify keywords or topics and link learned about the importance of thinking about
them together in the text. For example, a field negative evidence and events that are not pre
researcher wants to examine the person Susan sent in the data.
and the topic of hair (including haircuts, hair This chapter ends the section of the book on
styles, hair coloring, and hats or hair covering). research design, data collection, and data analy
The researcher can use Hypertext to connect all sis. Social research also involves preparing re
places Susan's name appears to discussions of ports on a research project, which is addressed in
hair. By the mouse clicking on Susan's name, the next chapter.
one block oftext quickly jumps to another in the
notes to see all places where Susan and the hair
topic appear together. Key Terms
Code-and-retrieve programs allow a re
searcher to attach codes to lines, sentences, para axial coding
graphs, or blocks of text. The programs may empty boxes
permit multiple codes for the same data. In ad illustrative method
dition to attaching codes, most programs also al narrative history
low the researcher to organize the codes. For open coding
example, a program can help a researcher make selectivecoding
outlines or "trees" of connections (e.g., trunks, successive approximation
ng the codes, and
re codes refer. The
~ed in the program
les and the relations
er specifies. Writing the Research
Report
~ learned how re
~ data. In many re
lore difficult to deal
of numbers. Num
perties that let a re Introduction
edures. Qualitative
The Research Report
by an individual re
latanotes, reflect on Why Write a Report?
Steps in Writing
:e of thinking about
ts that are not pre
343
344 PART FOUR I WRITING A RESEARCH REPORT
mity. Other rea ture. They want a condensed, detailed descrip A research report should be objective, accu
o fulfill a class or tion of research design. They pay close attention rate, and clear. Check and recheck details (e.g.,
Hgation to an or to how variables are measured and the methods page references in citations) and fully disclose
carch, to persuade of data collection. Scholars like a compact, how you conducted the research project. If read
ecific aspects of a tightly written, but extensive section on data ers detect carelessness in writing, they may ques
public about find analysis, with a meticulous discussion of results. tion the research itself. The details of a research
~ general public is Practitioners prefer a short summary ofhow project can be complex, and such complexity
r communication the study was conducted and results presented in means that confusion is always a danger. It
'i'a second stage of a few simple charts and graphs. They like to see makes clear writing essential. Clear writing can
an outline of alternative paths of action implied be achieved by thinking and rethinking the re
by results with the practical outcomes of pursu search problem and design, explicitly defining
ing each path. Practitioners must be cautioned terms, writing with short declarative sentences,
not to overgeneralize from the results of one and limiting conclusions to what is supported by
al writers say: Al study. It is best to place the details of research the evidence.
writing. This is be design and results in an appendix.
'e effective when it When writing for the general public, use Organizing Thoughts. Writing does not hap
ence. You should simple language, provide concrete examples, pen magically or simply flow out of a person
mtly depending on and focus on the practical implications of find when he or she puts pen to paper (or fingers to
:e is an instructor, ings for social problems. Do not include details keyboard) although many people have such an
scientists, practi of research design or of results, and be careful illusion. Rather, it is hard work, involving a se
ic, It goes without not to make unsupported claims when writing quence ofsteps and separate activities that result
1be clear, accurate, for the public. Informing the public is an impor in a final product. Writing a research report is
tant service, which can help nonspecialists make not radically different from other types of writ
rt for different rea better judgments about public issues. ing. Although some steps differ and the level of
ments on how it is complexity may be greater, most of what a good
rs want to see writ Style and Tone. Research reports are written writer does when writing a long letter, a poem, a
reflect clear, logical in a narrow range of styles and have a distinct set of instructions, or a short story applies to
auld demonstrate a tone. Their purpose is to communicate clearly writing a research report.
nd methodological the research method and findings. First, a writer needs something about which
this is to use techni Style refers to the types of words chosen by to write. The "something" in the research report
appropriate; they the writer and the length and form of sentences includes the topic, research question, design and
Iy or incorrectly. or paragraphs used. Tone is the writer's attitude measures, data collection techniques, results,
nts, it is best to de or relation toward the subject matter. For exam and implications. With so many parts to write
l each part ofthe re ple, an informal, conversational style (e.g., collo about, organization is essential. The most basic
proceed in a logical, quial words, idioms, cliches, and incomplete tool for organizing writing is the outline. Out
rany specific exam sentences) with a personal tone (c.g., these are lines help a writer ensure that all ideas are in
anguage to explain my feelings) is appropriate for writing a letter to cluded and that the relationship between them is
l the various steps of a close friend, but not for research reports. Re clear. Outlines are made up of topics (words or
strategy is to begin search reports have a formal and succinct (say phrases) or sentences. Most of us are familiar
hen structure the re ing a lot in few words) style. The tone expresses with the basic form of an outline (see Figure
distance from the subject matter; it is profes 14.1).
finitions of technical sional and serious. Field researchers sometimes Outlines can help the writer, but they can
hy standard proce use an informal style and a personal tone, but also become a barrier if they are used improp
ig) were used. They this is the exception. Avoid moralizing and flow erly. An outline is simply a tool to help the writer
-arch is linked to ab ery language. The goal is to inform, not to advo organize ideas. It helps (1) put ideas in a se
ndings in the litera- cate a position or to entertain. quence (e.g., what will be said first, second, and
346 PART FOUR I WRITING A RESEARCH REPORT
----------------------1.
FIGURE 14.1 Form of Outline
I. First major topic One of the most important
A. Subtopic of topic I Second level of importance
1. Subtopic of A Third level of importance
a. Subtopic of 1 Fourth level of importance
b. Subtopic of 1
(1) Subtopic of b Fifth level of importance
(2) Subtopic of b
(a) Subtopic of (2) Sixth level of importance
(b) Subtopic of (2)
i. Subtopic of (b) Seventh level of importance
ii. Subtopic of (b)
2. Subtopic of A Third level of importance
B. Subtopic of topic I Second level of importance
II. Second major topic One of the most important
third); (2) group related ideas together (e.g., Back to the Library. Few researchers finish
these are similar to each other but they differ their literature review before completing a re
from those); and (3) separate the more general, search project. The researcher should be familiar
or higher-level, ideas from more specific ideas, with the literature before beginning a project,
and the specific ideas from very specific details. but will need to return to the literature after
Some students feel that they need a com completing data collection and analysis, for sev
plete outline before writing, and that once an eral reasons. First, time has passed between the
outline is prepared, deviations from it are im beginning and the end of a research project, and
possible. Fewwriters begin with a complete out new studies may have been published. Second,
line. The initial outline is sketchy because until after completing a research project, a researcher
you write everything down, it is impossible to will know better what is or is not central to the
put all ideas in a sequence, group them together, study and may have new questions in mind
or separate the general from the specific. For when rereading studies in the literature. Finally,
most writers, new ideas develop or become when writing the report, researchers may find
clearer in the process of writing itself. that notes are not complete enough or a detail is
A beginning outline may differ from the fi missing in the citation of a reference source (see
nal outline by more than degree of complete Box 14.1). The visit to the library after data col
ness. The process of writing may not only reveal lection is less extensive and more selective or fo
or clarify ideas for the writer but it will also stim cused than that conducted at the beginning of
ulate new ideas, new connections between ideas, research.
a different sequence, or new relations between When writing a research report, researchers
the general and the specific. In addition, the frequently discard some of the notes and sources
process of writing may stimulate reanalysis or a that were gathered prior to completing the re
reexamination of the literature or findings. This search project. This does not mean that the ini
does not mean beginning allover again. Rather, tial library work and literature review were a
it means keeping an open mind to new insights waste of time and effort. Researchers expect that
and being candid about the research project. some of the notes (e.g., 25 percent) taken before
--- .. CHAPTER 14 I WRITING THE RESEARCH REPORT 347
sented at the annual meeting of the American So nalof Sociology April 1988, Vol. 3. Retrieved Jan
ciological Association, August, New York City. uary 16, 1999. http://www.sociology.org/con
tent/volOO 3 .00 3 / sosteric.htmI
Internet Sources
[Note: The date retrieved is the date that the reader Newspaper Article
Lee, Don. 1999. "State's Job Growth Hits Unex
located and read the work on the Internet.]
pected Cold Spell." Los Angeles Times (January
Announcement or Personal Home Page 16). Retrieved January 16, 1999. http://www.
American Sociological Association 1999. Journals latimes.com/HOME/BUSINESS/topstory.html
and Newsletters. Retrieved January 16, 1999.
http://www.asanet.org/Pubs/publicat.html
Journal Abstract or Book Review
Stanbridge, Karen. 2005. Review of The
On-Line Journal Article New Transnational Activism by Sidney Tarrow.
Sosteric, Mike, Mike Gismondi and Gina Ratkovic. Canadian Journal of Sociology Online. Retrieved
1998. "The University, Accountability, and Mar January 1 2, 2006. http://www.cjsonline.ca/
ket Discipline in the Late 19905." Electronic Jour- reviews/transnatl.html.
completing the project will become irrelevant as paraphrased. It means more than replacing an
the project gains focus. They do not include other's words with synonyms; paraphrasing is
notes or references in a report that are no longer borrowing an idea, boiling it down to its essence,
relevant, for they distract from the flow of ideas and giving credit to the source.
and reduce clarity.
Returning to the library to verify and ex
Steps in Writing
pand references focuses ideas. It also helps avoid
plagiarism. Plagiarism is a serious form of cheat Writing is a process. The way to learn to write
ing, and many universities expel students caught is by writing. It takes time and effort, and it
engaging in it. If a professional ever plagiarizes in improves with practice. There is no single cor
a scholarly journal, it is treated as a very serious rect way to write, but some methods are asso
offense. Take careful notes and identify the exact ciated with good writing. The process has
source of phrases or ideas to avoid unintentional three steps:
plagiarism. Cite the sources of both directly
quoted words and paraphrased ideas. For direct 1. Prewriting. Prepare to write by arranging
quotes, include the location of the quote with notes on the literature, making lists of ideas,
page numbers in the citation. outlining, completing bibliographic cita
Using another's written words and failing to tions, and organizing comments on data
give credit is wrong, but paraphrasing is less analysis.
clear. Paraphrasing is not using another's exact 2. Composing. Get your ideas onto paper as a
words; it is restating another's ideas in your own first draft by freewriting, drawing up the
words, condensing at the same time. Researchers bibliography and footnotes, preparing data
regularly paraphrase, and good paraphrasing re for presentation, and forming an introduc
quires a solid understanding of what is being tion and conclusion.
CHAPTER 14 / WRITING THE RESEARCH REPORT 349
3. Rewriting. Evaluate and polish the report by clarity. For the rest of us mortals, writing means
improving coherence, proofreading for me that rewriting-and rewriting again-is neces
chanical errors, checking citations, and re sary. For example, Ernest Hemingway is re
viewing voice and usage. ported to have rewritten the end of Farewell to
101. 3. Retrieved Jan Arms 39 times. It is not unusual for a profes
{.sociology.org/con Many people find that getting started is dif sional researcher to rewrite a report a dozen
tml ficult. Beginning writers often jump to the sec times. Do not become discouraged. If anything,
ond step and end there, which results in rewriting reduces the pressure; it means you can
Growth Hits Unex poor-quality writing. Prewriting means that a start writing soon and get out a rough draft that
geles Times (January writer begins with a file folder full of notes, out you can polish later. Plan to rewrite a draft at
1999. http://www. lines, and lists. You must think about the form of least three or four times. A draft is a complete re
:SS/topstory.html the report and audience. Thinking time is im port, from beginning to end, not a few rough
portant. It often occurs in spurts over a period of notes or an outline.
time before the bulk of composing begins. Rewriting helps a writer express himself or
Review of The
Some people become afflictedwith a strange herself with a greater clarity, smoothness, preci
by Sidney Tarrow.
ailment called writer's block when they sit down sion, and economy of words. When rewriting,
~ Online. Retrieved
to compose writing. It is a temporary inability to the focus is on clear communication, not
//www.cjsonline.ca/
write when the mind goes blank, the fingers pompous or complicated language. Rewriting
freeze, and panic sets in. Writers from beginners means slowlyreading what you have written and,
through experts occasionally experience it. If if necessary, reading out loud to see if it sounds
you experience it, calm down and work on over right. It is a good idea to share your writing with
coming it. others. Professional writers often have others
than replacing an Numerous writers begin to compose by read and criticize their writing. New writers soon
1S; paraphrasing is freewriting-that is, they sit down and write learn that friendly, constructive criticism is very
fawn to its essence, down everything they can as quickly as it enters valuable. Sharing your writing with others may
:e. the mind. Freewriting establishes a link between be difficult at first because it means exposing
a rapid flow of ideas in the mind and writing. your written thoughts and encouraging criticism.
When you freewrite, you do not stop to reread Yet, the purpose of the criticism is to clarify writ
what you wrote, you do not ponder the best ing, and the critic is doing you a favor.
ly to learn to write word, you do not worry about correct grammar, Rewriting involves two processes: revising
and effort, and it spelling, or punctuation. You just put ideas on and editing. Revising is inserting new ideas,
re is no single cor paper as quickly as possible to get and keep the adding supporting evidence, deleting or chang
methods are asso- creative juices or ideas flowing. You can later ing ideas, moving sentences around to clarify
The process has clean up what you wrote. meaning, or strengthening transitions and links
Writing and thinking are so intertwined between ideas. Editing means cleaning up and
that it is impossible to know where one ends and tightening the more mechanical aspects of writ
vrite by arranging the other begins. This means that if you plan to ing, such as spelling, grammar, usage, verb tense,
laking lists of ideas, sit and stare at the wall, the computer output, sentence length, and paragraph organization.
iibliographic cita the sky, or whatever until all thoughts become When you rewrite, go over a draft and revise it
omments on data totally clear before beginning, you may not get brutally to improve it. This is easier if some time
anything written. Writing itself can ignite the passes between a draft and rewriting. Phrases
as onto paper as a thinking process. that seemed satisfactory in a draft may look
g, drawing up the fuzzy or poorly connected after a week or two
tes, preparing data Rewriting. Perhaps one in a million writers is a (see Box 14.2).
ming an introduc- creative genius who can produce a first draft that Even if you have not acquired typing skills,
communicates with astounding accuracy and it is a good idea to type and print out at least one
350 PART FOUR I WRITING A RESEARCH REPORT
1. Mechanics. Check grammar, spelling, punctua Is there a transition between paragraphs within
tion, verb agreement, verb tense, and verb/sub the report?
ject separation with each rewrite. Remember 5. Repetition. Remove repeated ideas, wordiness.
that each time new text is added, new errors can and unnecessary phrases. Ideas are best stated
creep in. Mistakes are not only distracting but once, forcefully, instead of repeatedly in an un
they also weaken the confidence readers place clear way. When revising, eliminate deadwood
in the ideas you express. (words that add nothing) and circumlocution
2. Usage. Reexamine terms, especially key terms, (the use of several words when one more precise
when rewriting to see whether you are using the word will do). Directness is preferable to wordi
exact word that expresses your intended mean ness. The wordy phrase, To summarize the above.
ing. Do not use technical terms or long words it is ourconclusion in light of the data that X has Q
unnecessarily. Use the plain word that best ex positive effectof considerable magnitude on the oc
presses meaning. Get a thesaurus and use it. A currence of Y, notwithstanding the fact that Yoc
thesaurus is an essential reference tool, likea dic curs only on rare occasions, is better stated, In
tionary, that contains words of similar meaning sum, we conclude that X has a large positive effect
and can help you locate the exact word for a on Y, but Yoccurs infrequently.
meaning you want to express. Precise thinking 6. Structure. Research reports should have a trans
and expression requires precise language. Do parent organization. Move sections around as
not say average if you use the mean. Do not say necessary to fit the organization better, and use
mankind or policeman when you intend people or headings and subheadings. A reader should be
police officer. Do not use principal for principle. able to follow the logical structure of a report.
3. Voice. Writers of research reports often make 7. Abstraction. A good research report mixes ab
the mistake of using the passive instead of the stract ideas and concrete examples. A long
active voice. It may appear more authoritative, string of abstractions without the specifics is dif
but passive voice obscures the actor or subject ficult to read. Likewise, a mass of specific con
of action. For example, the passive, The relation crete details without periodic generalization also
ship between grade in school and more definite ca loses readers.
reer plans was confirmed by the data is better
8. Metaphors. Many writers use metaphors to ex
stated as the active, The data confirm the rela
press ideas. Phrases like the cutting edge, thebot
tionship between grade in school and more definite
tom line, and penetrating to the heart are used to
career plans. The passive, Respondent attitude to
express ideas by borrowing images from other
ward abortion was recorded byan interviewer reads
contexts. Metaphors can be an effective method
easier in the active voice: An interviewer recorded
of communication, but they need to be used
respondent attitude toward abortion. Also avoid
sparingly and with care. A few well-chosen, con
unnecessary qualifying language, such as seems
sistently used, fresh metaphors can communi
to or appears to.
cate ideas quickly and effectively; however, the
4. Coherence. Sequence, steps, and transitions excessive use of metaphors, especially overused
should be logically tight. Try reading the entire metaphors (e.g., the bottom line), is a sloppy.
report one paragraph at a time. Does the para unimaginative method of expression.
graph contain a unified idea? A topic sentence?
CHAPTER 14 / WRITING THE RESEARCH REPORT 351
draft before the final draft. This is because it is journal articles have abstracts that are printed on
easier to see errors and organization problems in the first page of the article. The abstract has in
a clean, typed draft. Feel free to cut and paste, formation on the topic, the research problem,
cross out words, or move phrases on the printed the basic findings, and any unusual research de
een paragraphs within copy. sign or data collection features.
Good keyboarding skills and the ability to Reports of applied research that are written
ted ideas, wordiness,
use a word processor are extremely valuable for practitioners have a longer summary called
Ideas are best stated
when writing reports and other documents. Se the executive summary. It contains more detail
f repeatedly in an un rious professionals find that the time they invest than an article abstract and includes the implica
eliminate deadwood
into building keyboard skills and learning to use tions of research and major recommendations
) and circumlocution
a word processor pays huge dividends later. made in the report. Although it is longer than an
zhen one more precise
Word processors not only make editing much abstract, an executive summary rarely exceeds
s preferable to wordi
easier but they also check spelling and offer syn four or five pages.
o summarize the above, onyms. In addition, there are programs that Abstracts and executive summaries serve
f the data that X has a check grammar. You cannot rely on the com several functions: For the less interested reader,
'e magnitude on the ac puter program to do all the work, but it makes they tell what is in a report; for readers looking
ing the fact that Y oc writing easier. The speed and ease that a word for specific information, they help the reader de
i, is better stated, In
processor offers is so dramatic that few people termine whether the full report contains impor
s a large positive effect
who become skilled at using one ever go back to tant information. Readers use the abstract or
'tly.
writing by hand or typing. summary to screen information and decide
5 should have a trans
One last suggestion: Rewrite the introduc whether the entire report should be read. It gives
e sections around as
tion and title after completing a draft so that they serious readers who intend to read the full re
zation better, and use
accurately reflect what is said. Titles should be port a quick mental picture of the report, which
s, A reader should be
short and descriptive. They should communi makes reading the report easier and faster.
tructure of a report.
cate the topic and the major variables to readers.
rch report mixes ab They can describe the type of research (e.g., "An Presenting the Problem. The first section of
le examples. A long
experiment on ...") but should not have unnec the report defines the research problem. It can
nrt the specifics is dif
essary words or phrases (e.g., "An investigation be placed in one or more sections with titles such
mass of specific con
into the ..."). as "Introduction," "Problem Definition," "Lit
:fie generalization also
erature Review," "Hypotheses," or "Background
Assumptions." Although the subheadings vary,
rse metaphors to ex The Quantitative Research Report
the contents include a statement of the research
~ cutting edge, the bot
The principles of good writing apply to all re problem and a rationale for what is being exam
the heart are used to
ports, but the parts of a report differ depending ined. Here, researchers explain the significance
g images from other
~ an effective method
on whether the research is quantitative or quali of and provide a background to the research
ey need to be used tative. Before writing any report, read reports on question. They explain the significance of the re
ew well-chosen, con the same kind of research for models. search by showing how different solutions to the
ohors can communi We begin with the quantitative research re problem lead to different applications or theo
ctively; however, the port. The sections of the report roughly follow retical conclusions. Introductory sections fre
, especially overused the sequence of steps of a research project. quently include a context literature review and
1m line), is a sloppy, link the problem to theory. Introductory sec
(pression. Abstract or Executive Summary. Quantitative tions also define key concepts and present con
research reports usually begin with a short sum ceptual hypotheses.
mary or abstract. The size ofan abstract varies; it
can be as few as 50 words (this paragraph has 90 Describing the Methods. The next section of
words) or as long as a full page. Most scholarly the report describes how the researcher designed
352 PART FOUR I WRITING A RESEARCH REPORT
the study and collected the data. It goes by sev own interpretations. Detailed summary statistics
eral names (e.g., "Methods," "Research Design," belong in appendixes.
or "Data") and may be subdivided into other
parts (e.g., "Measures," "Sampling," or "Manip Discussion. In the discussion section, re
ulations"). It is the most important section for searchers give the reader a concise, unam
evaluating the methodology of the project. The biguous interpretation of its meaning. The
section answers several questions for the reader: discussion is not a selectiveemphasis or partisan
interpretation; rather, it is a candid discussion of
1. What type of study (e.g., experiment, sur what is in the "Results" section. The "Discus
vey) was conducted? sion" section is separated from the results so that
2. Exactly how were data collected (e.g., study a reader can examine the data and arrive at dif
design, type of survey, time and location of ferent interpretations.
data collection, experimental design used)? Beginning researchers often find it difficult
3. How were variables measured? Are the mea to organize the "Discussion" section. One ap
sures reliable and valid? proach is to organize the discussion according to
4. What is the sample? How many subjects or hypotheses, discussing how the data relate to
respondents are involved in the study? How each hypothesis. In addition, researchers should
were they selected? discuss unanticipated findings, possible alterna
5. How were ethical issues and specific con tive explanations of results, and weaknesses or
cerns of the design dealt with? limitations.
:summary statistics communicate the research process and the data theoretical generalizations and data are not sep
collected through the process. arated into distinct sections. Generalizations are
Quantitative reports present hypotheses and intertwined with the evidence, which takes the
ssion section, re evidence in a logicallytight and condensed style. form of detailed description with frequent
a concise, un am Bycontrast, qualitative reports tend to be longer, quotes.
its meaning. The and book-length reports are common. The Researchers balance the presentation of data
nphasis or partisan greater length is for five reasons: and analysis to avoid an excessiveseparation of
andid discussion of data from analysis, called the error of segregation.
tion. The "Discus 1. The data in a qualitative report are more This occurs when researchers separate data from
n the results so that difficult to condense. Data are in the form of analysis so much that readers cannot see the
:a and arrive at dif words, pictures, or sentences and include connection. 1
many quotes and examples. The tone of fieldresearch reports is lessobjec
ften find it difficult 2. Qualitative researchers try to create a sub tiveand formal, and more personal. Fieldresearch
" section. One ap jective sense of empathy and understanding reports may be written in the first person (i.e., us
ussion according to among readers in addition to presenting ing the pronoun I) because the researcher was di
. the data relate to factual evidence and analytic interpreta rectly involved in the setting, interacted with the
researchers should tions. Detailed descriptions of specific set people studied, and was the measurement "instru
~s, possible alterna tings and situations help readers better ment." The decisions or indecisions, feelings, re
and weaknesses or understand or get a feel for settings. Re actions, and personal experiencesof the researcher
searchers attempt to transport the reader are parts of the field research process.
into the subjective world view and meaning Field research reports often face more skep
earchers restate the system of a social setting. ticism than quantitative reports do. This makes
arizefindings in the 3. Qualitative researchers use less standardized it essential to assess an audience's demands for
summarize the re techniques of gathering data, creating ana evidence and to establish credibility. The key is
-d "Summary." lytic categories, and organizing evidence. to provide readers with enough evidence so that
the conclusion are The techniques applied may be particular to they believe the recounted events and accept the
s. The "References" individual researchers or unique settings. interpretations as plausible. A degree of selective
5 that were referred Thus, researchers explain what they did and observation is accepted in field research, so the
·eport. Appendixes, why, because it has not been done before. critical issue is whether other observers could
itional information 4. Exploring new settings or constructing new reach the same conclusion if they examined the
on (e.g., question theory is a common goal in qualitative re same data.
;., descriptive statis search. The development of new concepts Field researchers face a data reduction
otes in quantitative and the examination of relationships among dilemma when presenting evidence. Most data
elaborate on infor them adds to the length of reports. Theory are in the form of an enormous volume of field
.ers use them spar flows out of evidence, and detailed descrip notes, but a researcher cannot directly share all
. information that tions demonstrate how the researcher cre the observations or recorded conversations with
stract from the flow ated interpretations. the readers. For example, in their study of med
5. Qualitative researchers may use more var ical students, Boys in White, Becker and col
ied and literary writing styles, which in leagues (1961) had about 5,000 pages of
Report
creaseslength. They have greater freedom to single-spaced field notes. Field researchers in
employ literary devices to tell a story or re clude only about 5 percent of their field notes in
esearch, it is more count a tale. a report as quotes. The remaining 95 percent is
[ualitative social re not wasted; there is just no room for it. Thus,
less structure. Nev Field Research. Field research reports rarely writers select quotes and indirectly COD\-~- the
Ie same: to clearly follow a fixed format with standard sections, and rest of the data to readers.
354 PART FOUR / WRITING A RESEARCH REPORT
There is no fixed organization for a field re setting, the means of gaining access, the role of
search report, although a literature review often the researcher, and the subject-researcher rela
appears near the beginning. There are many ac tionship into the discussion of evidence and
ceptable organizational forms. Lofland (1976) analysis. This is intensified if the writer adopts
suggests the following: what Van Maanen (1988:73) called a "confes
sional" style of writing.
1. Introduction A chronological, zoom lens, or theme-based
a. Most general aspects of situation organization allows placing the data collection
b. Main contours of the general situation method near the beginning or the end. In book
c. How materials were collected length reports, methodological issues are usually
d. Details about the setting discussed in a separate appendix.
e. How the report is organized Field research reports can contain transcrip
2. The situation tions of tape recordings, maps, photographs, or
a. Analytic categories charts illustrating analytic categories. They sup
b. Contrast between situation and other sit plement the discussion and are placed near the
uations discussion they complement. Qualitative field
c. Development of situation over time research can use creative formats that differ
3. Strategies from the usual written text with examples from
4. Summary and implications field notes. Harper's (1982) book contains many
photographs with text. The photographs give a
Devices for organizing evidence and analysis visual inventory of the settings described in the
also vary a great deal. For example, writers can text and present the meanings of settings in the
organize the report in terms of a natural history, terms of those being studied. For example, field
an unfolding of events as you discovered them, research articles have appeared in the form of all
or as a chronology, following the developmental photographs, a script for a play, or a documen
cycle or career ofan aspect of the setting or peo tary film.'
ple in it. Another possibility is to organize the re Direct, personal involvement in the inti
port as a zoom lens, beginning broadly and then mate details of a social setting heightens ethical
focusing increasingly narrowly on a specific concerns. Researchers write in a manner that
topic. Statements can move from universal protects the privacy of those being studied and
statements about all cultures, to general state helps prevent the publication of a report from
ments about a specific cultures, to statements harming those who were studied. They usually
about a specific cultural scene, to specific state change the names of members and exact loca
ments about an aspect of culture, to specific tions in field reports. Personal involvement in
statements about specific incidents. field research leads researchers to include a short
Field researchers also organize reports by autobiography. For example, in the appendix to
themes. A writer chooses between using abstract Street Corner Society, the author, William Foote
analytic themes and using themes from the cate Whyte (1955), gave a detailed account of the oc
gories used by the people who were studied. The cupations ofhis father and grandfather, his hob
latter gives readers a vivid description of the set bies and interests, the jobs he held, how he
ting and displays knowledge of the language, ended up going to graduate school, and how his
concepts, categories, and beliefs of those being research was affected by his getting married.
written about.?
Field researchers discuss the methods used Historical-Comparative Research. There is no
in the report, but its location and form vary. One single way to write a report on historical-com
technique is to interweave a description of the parative research. Most frequently, researchers
CHAPTER 14 / WRITING THE RESEARCH REPORT 355
charts, tables, and so forth supplement a discus convince reviewers that you, the researcher, are
.gs of settings in the
sion or give the reader a better feel for the places capable of successfully conducting the proposed
l. For example, field
and people being described. They are used in con research project. Reviewers have more confi
ed in the form of all
junction with frequent quotes as one among sev dence that a planned project will be successfully
llay, or a documen eral types of evidence. Historical-comparative completed if the proposal is well written and
reports rarely summarize data to test specific hy organized, and if you demonstrate careful
rement in the inti
potheses as quantitative research does. Instead, planning.
t1g heightens ethical
the writer builds a web of meaning or descriptive The proposal is similar to a research report,
e in a manner that
detail and organizes the evidence itself to convey but it is written before the research project be
e being studied and
interpretations and generalizations. gins. A proposal describes the research problem
m of a report from
There are two basic modes of organizing and its importance, and gives a detailed account
udied. They usually
H-C research reports: by topic and chronologi of the methods that will be used and why they
iers and exact loca
cally. Most writers mix the two types. For exam are appropriate.
mal involvement in
ple, information is organized chronologically The proposal for quantitative research has
-rsto include a short
within topics, or organized by topic within most of the parts of a research report: a title, an
:, in the appendix to
chronological periods. Occasionally other forms abstract, a problem statement, a literature
thor, William Foote
of organization are used-by place, by individ review, a methods or design section, and a bibli
d account of the oc
ual person, or by major events. If the report is ography. It lacks results, discussion, and conclu
randfather, his hob
truly comparative, the writer has additional op sion sections. The proposal has a plan for data
s he held, how he
tions, such as making comparisons within top collection and analysis (e.g., types of statistics). It
school, and how his
ics. Box 14.3 provides a sample of some frequently includes a schedule of the steps to be
getting married.
techniques used by historical-comparative re undertaken and an estimate of the time required
searchers to organize evidence and analysis. for each step.
search. There is no Some H-C researchers mimic the quantita Proposals for qualitative research are more
on historical-com tive research report and use quantitative re difficult to write because the research process it
luently, researchers search techniques. They extend quantitative self is less structured and preplanned. The re
356 PART FOUR! WRITING A RESEARCH REPORT
me relativelyaccessi 1 O. Synthesis. Historical-comparative researchers of shifting popular beliefs and problems with gov
ten synthesize many specific events and details ernment finances into a compact, coherent ex
-generalization is aI into a comprehensive whole. Synthesis results planation. Researchers using the narrative form
I historical-cornpara from weaving together many smaller generaliza summarize the argument in an introduction or
iers seek rigid, fixed tions and interpretations into coherent main conclusion. It is a motif or theme embedded
ve explanation. They themes. For example, a researcher studying the within the description. Thus, theoretical gener
strict determination. French Revolution synthesizes specific general alizations are intertwined with the evidence and
snket statement that izations about changes in social structure, inter appear to flow inductively out of the detailed
ive cultures in areas national pressures, agricultural dislocation, evidence.
~s was the inevitable
echnological culture,
specific factors that
searcher prepares a problem statement, litera results. The degree of competition for a grant
-struction in particu
ture review, and bibliography. He or she demon varies a great deal, depending on the source.
strates an ability to complete a proposed Some sources fund more than 3 out of 4 propos
f association is used qualitative project in two ways. First, the pro als they receive, others fund fewer than 1 in 20.
:h. As in other areas,
posal is well written, with an extensive discus The researcher needs to investigate funding
archers identify fac
sion of the literature, significance of the sources and ask questions: What types of pro
n time and place. For
problem, and sources. This shows reviewers fa jects are funded-applied versus basic research,
nining a city's nine
miliarity with qualitative research and the ap specific topics, or specific research techniques?
rsks whether years of
city are associated propriateness of the method for studying the What are the deadlines? What kind (e.g., length,
d whether those ar problem. Second, the proposal describes a qual degree of detail, etc.) of proposal is necessary?
immigrants. itative pilot study. This demonstrates motiva How large are most grants? What aspects (e.g.,
tion, familiarity with research techniques, and equipment, personnel, travel, etc.) of a project
nt to place events in
ability to complete a report about unstructured are or are not funded? There are many sources of
storical-comparative
between parts of a research. information on funding sources. Librarians or
vent and the larger officials who are responsible for research grants
1. For example, a re Proposals to Fund Research. The purpose of a at a college are good resource people. For exam
liar political ritual in research grant is to provide the resources needed ple, private foundations are listed in an annual
:ting describes how to help complete a worthy project. Researchers publication, The Foundation Directory. The
iteenth-century po whose primary goal is to use funding for per Guide to Federal Fundingfor Social Scientists lists
sonal benefit or prestige, to escape from other sources in the U.S. government. In the United
seful. The overuse of activities, or to build an "empire" are less suc States, there are many newsletters on funding
appropriate analogy cessful. The strategies of proposal writing and sources and two national computerized data
a researcher exam getting grants has become an industry called bases, which subscribers can search for funding
n country X and de grantsmanship. sources. Some agencies periodically issue
igs about death" in There are many sources of funding for re requests for proposals (RFPs) that ask for propos
Jiresa description of search proposals. Colleges, private foundations, als to conduct research on a specific issue. Re
untry Y. and government agencies have programs to searchers need to learn about funding sources
award grants to researchers. Funds may be used because it is essential to send the proposal to an
to purchase equipment, to pay your salary or appropriate source in order to be successful.
that of others, for research supplies, for travel to Researchers should show a track record of
collect data, or for help with the publication of past success in the proposal, especially ifthey are
358 PART FOUR / WRITING A RESEARCH REPORT
going to be in charge of the project. The re details on how funds were spent, the findings,
searcher in charge of a research project is the and an evaluation of whether the project met its
principal investigator (PI) or project director. objectives. Failure to spend funds properly.
Proposals usually include a curriculum vitae or complete the project described in the proposal.
academic resume, letters of support from other or file a final report may result in a researcher
researchers, and a record of past research. Re being barred from receiving future funding or
viewers feel safer investing funds in a project facing legal action. A serious misuse of funds
headed by someone who already has research ex may result in the banning of others at the same
perience than in a novice. One can build a track institution from receiving future funding.
record with small research projects or by assist The process of reviewing proposals after
ing an experienced researcher before seeking they are submitted to a funding source takes
funding as a principal investigator. anywhere from a few weeksto almost a year, de
The reviewers who evaluate a proposal pending on the funding source. In most cases,
judge whether the proposal project is appropri reviewers rank a large group of proposals, and
ate to the funding source's goals. Most funding only highly ranked proposals receive funding. A
sources have guidelines stating the kinds of pro proposal often undergoes a peer review in which
jects they fund. For example, programs that fund the reviewers know the proposer from the vitae
basic research have the advancement of knowl in the proposal, but the proposer does not know
edge as a goal. Programs that fund applied re the reviewers. Sometimes a proposal is reviewed
search often have improvements in the delivery by nonspecialists or nonresearchers. Instruc
of services as a goal. Instructions specify page tions on preparing a proposal indicate whether
length, number of copies, deadlines, and the like. to write for specialists in a field or for an edu
Follow all instructions exactly. cated general audience.
Proposals should be neat and professional If a proposal is funded, celebrate, but only
looking. The instructions usually ask for a de for a short time. If the proposal is rejected, which
tailed plan for the use of time, services, and per is more likely, do not despair. Most proposals
sonnel. These should be clearly stated and are rejected the first or second time they arc sub
realistic for the project. Excessively high or low mitted. Many funding sources provide written
estimates, unnecessary add-ens, or omitted es reviewer evaluations of the proposal. Always re
sentials will lower how reviewers evaluate a pro quest them if they are provided. Sometimes, a
posal. Creating a budget for a proposed project courteous talk on the telephone with a person at
is complicated and usually requires technical as the funding source will reveal the reasons for re
sistance. For example, pay rates, fringe benefit jection. Strengthen and resubmit a proposal on
rates, and so on that must be charged may not be the basis of the reviewer's comments. Most
easy to obtain. It is best to consult a grants offi funding sources accept repeated resubmissions
cer at a college or an experienced proposal of revised proposals, and proposals that have
writer. In addition, endorsements or clearances been revised may be stronger in subsequent
of regulations are often necessary (e.g., IRE ap competitions.
proval). Proposals should also include specific If a proposal has been submitted to an ap
plans for disseminating results (e.g., publica propriate funding source and all instructions arc
tions, presentations before professional groups, followed, reviewers are more likelyto rate it high
etc.) and a plan for evaluating whether the pro when:
ject met its objectives.
The proposal is a kind of contract between • It addresses an important research question.
researcher and the funding source. Funding It builds on prior knowledge and represents
agencies often require a final report, including a substantial advance of knowledge for basic
.nt, the findings, research. It documents a major social prob
le project met its lem and holds promise for solutions for ap
e. In most cases, the results and evaluating whether the pro grantsmanship
eceive funding. A • The project is well designed and shows seri plagiarism
er does not know • The researcher has the necessary experience request for proposals (RFPs)
.mitted to an ap
II instructions are
relyto rate it high
esearch question.
~e and represents
owledge for basic