You are on page 1of 37

Assessing "Neighborhood Effects": Social Processes and New Directions in Research

Author(s): Robert J. Sampson, Jeffrey D. Morenoff, Thomas Gannon-Rowley


Source: Annual Review of Sociology, Vol. 28 (2002), pp. 443-478
Published by: Annual Reviews
Stable URL: http://www.jstor.org/stable/3069249
Accessed: 29/12/2009 05:03

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=annrevs.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Annual Reviews is collaborating with JSTOR to digitize, preserve and extend access to Annual Review of
Sociology.

http://www.jstor.org
Annu. Rev. Sociol. 2002. 28:443-78
doi: 10.1146/annurev.soc.28.110601.141114
Copyright? 2002 by AnnualReviews. All rightsreserved

ASSESSING
"NEIGHBORHOOD
EFFECTS":
Social
Processesand New Directionsin Research

RobertJ.Sampson,1JeffreyD. Morenoff,2and
Thomas Gannon-Rowley'
'Departmentof Sociology, Universityof Chicago, Chicago, Illinois 60637;
e-mail: rjsam@src.uchicago.edu,tpgannon@uchicago.edu
2Departmentof Sociology, Universityof Michigan,AnnArbor,Michigan48106;
e-mail: morenoff@isr.umich.edu

Key Words neighborhood effects, ecology, problem behavior, social processes,


mechanisms
* Abstract This paper assesses and synthesizes the cumulative results of a new
"neighborhood-effects" literature that examines social processes related to problem
behaviors and health-relatedoutcomes. Our review identified over 40 relevant studies
published in peer-reviewed journals from the mid-1990s to 2001, the take-off point for
an increasing level of interest in neighborhood effects. Moving beyond traditionalchar-
acteristics such as concentratedpoverty, we evaluate the salience of social-interactional
and institutional mechanisms hypothesized to account for neighborhood-level varia-
tions in a variety of phenomena (e.g., delinquency, violence, depression, high-risk
behavior), especially among adolescents. We highlight neighborhood ties, social con-
trol, mutual trust, institutional resources, disorder, and routine activity patterns. We
also discuss a set of thorny methodological problems that plague the study of neigh-
borhood effects, with special attention to selection bias. We conclude with promising
strategies and directions for future research, including experimental designs, taking
spatial and temporal dynamics seriously, systematic observational approaches, and
benchmark data on neighborhood social processes.

INTRODUCTION

At the outset of the 1990s, Jencks & Mayer (1990, Mayer & Jencks 1989) ar-
gued that if growing up in a poor neighborhood mattered, intervening processes
such as collective socialization, peer-group influence, and institutional capacity
were presumably part of the reason. Their influential assessment of the so-called
neighborhood-effects literature was ultimately pessimistic, however, for few stud-
ies could be found that measured and identified social processes or mechanisms.
A major reason is that the data sources traditionally relied upon by neighborhood
researchers-the U.S. Census and other government statistics-typically provide

0360-0572/02/0811-0443$14.00 443
444 * GANNON-ROWLEY
SAMPSON MORENOFF

informationon the sociodemographiccomposition of statisticalareas (e.g., the


povertyrate or racialmakeupof census tracts)ratherthanthe dynamicprocesses
hypothesizedto shape child and adolescentwell-being. Then and now, Jencks &
Mayer'scritiquewas formidable.
The good news is thatthe decade since theirreview markeda periodof major
advances in neighborhood-levelresearch, as researchersbegan to explore new
methods and ideas for understandingwhat makes places more or less healthy,
particularlyfor young people. A large numberof studies were also launchedin a
shortperiod,so manythatthe studyof neighborhoodeffects, forbetterorworse,has
become somethingof a cottageindustryin the social sciences. Figure 1 documents
this strikingtrend.After spurtsin the 1960s and 1970s followed by a decline, the
mid 1990s to the year 2000 saw more thana doublingof neighborhoodstudiesto
the level of about 100 papers per year. The bad news is that this recent spurtin
quantityhas not been equally matchedin quality;much hardwork remainsto be
done.
The purposeof this paperis to synthesizethe resultsof the recentgenerationof
neighborhoodstudiesthatfocus on social andinstitutionalprocesses,especially as
relatedto problembehavioramongyoung people. We begin with a brief overview
of two longstandingconcerns-how researcherstypically define local communi-
ties (what is a neighborhood?)and the persistentpatternsthat link problem-and
health-relatedbehaviors with concentratedpoverty and other indicatorsof resi-
dential differentiation.The heartof our assessmentthen turnsto advancesin the

120

100

< 80

-60-

20 f

0 I i ,I 1I T 9 I ,1,9
i6 1 2 1 I , r T 19 1 1,, 1
1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 I982 1984 1986 1988 19901992 1994 1996 1998

and"SocialCapital"in title:SocialCitationIndex.
Figure 1 Articleswith"Neighborhood"
NEIGHBORHOOD
EFFECTS 445

measurementof neighborhoodsocial and institutionalprocesses. Ourreview cov-


ers researchin the secondhalf of the 1990s, the take-offpointfor anincreasinglevel
of activity (Figure 1). We also evaluatea set of thornymethodologicalproblems
that plague the study of neighborhoodeffects, the most notable being selection
bias. We conclude with strategiesto addressthese challenges and promisingdi-
rectionsfor futureresearch,includingexperimentaldesigns, spatialand temporal
models, systematic observationalapproaches,and collecting benchmarksurvey
data on neighborhoodsocial processes.

DEFINING NEIGHBORHOOD

RobertParkand ErnestBurgess laid the foundationfor urbansociology by defin-


ing local communities as "naturalareas"that developed as a result of competi-
tion between businesses for land use and between populationgroups for afford-
able housing. A neighborhood,accordingto this view, is a subsectionof a larger
community-a collection of bothpeople andinstitutionsoccupyinga spatiallyde-
fined areainfluencedby ecological, cultural,and sometimespolitical forces (Park
1916, pp. 147-154). Suttles(1972) laterrefinedthis view by recognizingthatlocal
communitiesdo not form theiridentitiesonly as the resultof free-marketcompe-
tition. Instead,some communitieshave their identity and boundariesimposed on
them by outsiders. Suttles also arguedthat the local communityis best thought
of not as a single entity, but ratheras a hierarchyof progressivelymore inclusive
residentialgroupings.In this sense, we can think of neighborhoodsas ecological
units nested within successively largercommunities.
In practice, most social scientists and virtually all studies of neighborhoods
we assess rely on geographicboundariesdefined by the Census Bureauor other
administrativeagencies (e.g., school districts,police districts).Although admini-
stratively defined units such as census tracts and block groups are reasonably
consistent with the notion of overlappingand nested ecological structures,they
offerimperfectoperationaldefinitionsof neighborhoodsforresearchandpolicy.As
we discuss later,researchershave thusbecome increasinglyinterestedin strategies
to define neighborhoodsthat respect the logic of street patternsand the social
networksof neighborinteractions(e.g., Grannis1998).

NEIGHBORHOOD DIFFERENTIATION

Building on a long history of sociological research on urbancommunities, the


study of neighborhoodeffects has generateda multidisciplinaryresearchagenda
with a strong focus on child and adolescent development.Spurredin large part
by Wilson's (1987) seminal book, The TrulyDisadvantaged, modem neighbor-
hood researchhas attendedprimarilyto structuraldimensions of neighborhood
disadvantage,especially the geographicisolation of poor, African-American,and
446 SAMPSON* MORENOFF GANNON-ROWLEY

single-parentfamilies with children(Small & Newman 2001). The rangeof child


and adolescentoutcomes associatedwith concentrateddisadvantageis quite wide
andincludesinfantmortality,low birthweight,teenagechildbearing,droppingout
of high school, child maltreatment,and adolescentdelinquency(for an overview,
see Brooks-Gunnet al. 1997a,b).Thereis also independentevidence thata number
of health-relatedindicatorsclusterspatially,includinghomicide, infantmortality,
low birthweight,accidental injury, and suicide (Almgren et al. 1998, Sampson
2001). The weight of evidence thus suggests thatthereare geographic"hotspots"
for crime and problem-relatedbehaviorsand that such hot spots arecharacterized
by the concentrationof multipleforms of disadvantage.
To a lesser extent, the social-ecological literaturehas considered aspects of
neighborhood differentiationother than concentrateddisadvantage, including
life-cycle status,residentialstability,home ownership,density,and ethnic hetero-
geneity.The evidence on these factorsis decidedlymoremixed, especially for pop-
ulation density and ethnic heterogeneity(Brooks-Gunnet al. 1997a,b, Morenoff
et al. 2001). Perhapsthe most extensive areaof inquiryafterdisadvantage,dating
back to the early Chicago School, concernsresidentialstabilityand home owner-
ship. Althoughthe evidence here is also mixed (e.g., Ross et al. 2000), residential
instabilityandlow ratesof home ownershiparedurablecorrelatesof manyproblem
behaviors(Brooks-Gunnet al. 1997a,b).A more recentbutunderstudiedobject of
inquiryis concentratedaffluence(Massey 1996). Brooks-Gunnet al. (1993) argue
that it is the positive influence of concentratedsocioeconomic resources,rather
than the presence of low-income neighbors,that mattersmost for adolescentbe-
haviors. The common tactic of focusing on concentrateddisadvantagemay thus
obscurethe potentialprotectiveeffects of affluentneighborhoods.
In short, empirical research on social-ecological differentiationhas estab-
lished a reasonablyconsistent set of neighborhoodfacts relevantto childrenand
adolescents.
* First, there is considerablesocial inequalityamong neighborhoodsin terms
of socioeconomic and racial segregation.There is strong evidence on the
connection of concentrateddisadvantagewith the geographic isolation of
AfricanAmericans.
* Second, a numberof social problemstend to come bundledtogetherat the
neighborhoodlevel, including, but not limited to, crime, adolescent delin-
quency,social andphysicaldisorder,low birthweight,infantmortality,school
dropout,and child maltreatment.
* Third,these two sets of clusters are themselvesrelated-neighborhood pre-
dictorscommon to many child and adolescentoutcomes include the concen-
trationof poverty,racialisolation, single-parentfamilies, and rates of home
ownershipand length of tenure.
* Fourth,empiricalresults have not variedmuch with the operationalunit of
analysis. The place stratificationof local communitiesin Americansociety
by factorssuch as social class, race, andfamily statusis a robustphenomenon
NEIGHBORHOOD
EFFECTS 447

thatemergesat multiplelevels of geography,whetherlocal communityareas,


census tracts,or otherneighborhoodunits.
* Fifth, the ecological concentrationof povertyappearsto have increasedsig-
nificantlyduringrecent decades, as has the concentrationof affluenceat the
upperend of the income scale.
Neighborhoods and residential differentiation thus remain persistent in
American society. As real estate agents and homeowners(especially those with
children)often declare,location seems to matter.The next logical questions are:
Why does neighborhoodmatter,for what,andto whatdegree?The cumulativefacts
on neighborhooddifferentiationyield a potentiallyimportantclue in thinkingabout
these questions.If numerousandseeminglydisparateoutcomesarelinkedtogether
empiricallyacross neighborhoodsand are predictedby similar structuralcharac-
teristics,theremay be commonunderlyingcauses. We thus assess this possibility,
along with alternativeinterpretationsthatquestionthe existence of neighborhood
effects altogether.

BEYOND POVERTY:SOCIAL PROCESSES


AND MECHANISMS

Duringthe 1990s, a numberof scholarsmoved beyond the traditionalfixationon


concentratedpoverty and began to explicitly theorize and directly measure how
neighborhoodsocial processes bearon the well-being of childrenandadolescents.
Unlike the more staticfeaturesof sociodemographiccomposition(e.g., race, class
position), social processes or mechanismsprovideaccountsof how neighborhoods
bring abouta change in a given phenomenonof interest(Sorensen 1998, p. 240).
Although concern with neighborhoodmechanismsgoes back at least to the early
ChicagoSchool of sociology, only recentlyhave we witnesseda concertedattempt
to theorize and empiricallymeasurethe social-interactionaland institutionaldi-
mensions that might explain how neighborhoodeffects are transmitted.
This review focuses on the resultingturnto social processes in neighborhood-
effects research.We performeda systematic search for studies that investigated
variationsin some aspect of social processes or mechanismsacross ecologically
definedunitsof analysis(e.g., census tracts,block groups).1We limitedourreview
to quantitativestudies published in peer-reviewedsocial or behavioral science
journals beginning in the latter half of the 1990s (1996) and running through
summer2001. This periodmaps onto the upswing in action seen in Figure 1 and

'Giventhisframing,we didnotattemptto evaluatethe school-effects


literature.Although
the connectionof schoolsandneighborhoodsis clearlyimportant
(Jencks& Mayer1990)
andconsiderable progresshasbeenmadein recentresearch(e.g.,Welshet al. 1999),space
limitations
precluded ourtakingonthenexusof schoolandneighborhood socialprocesses.
448 SAMPSON
m MORENOFF GANNON-ROWLEY

follows in sequence the research epoch covered in Gephardt's(1997) review.2


Initially,we decidedto limit ourfocus to problem-relatedor health-compromising
behaviorsamong childrenand adolescents,such as delinquency,droppingout of
high school, andteenchildbirth.3In conductingourreview,however,we foundvery
few neighborhood-effectsstudiesthatrestrictedtheirattentionsolely to childrenor
adolescents.Moreover,this criterionexcludedmany studiesthatshed new light on
neighborhoodsocial mechanisms.Wethushighlightstudiesof child andadolescent
development,whereverpossible, but cast a wider net in orderto capturestudies
of problem-andhealth-relatedoutcomesthatcover a varietyof ages (e.g., ratesof
crime,adultdepression),as long as theyexaminesome dimensionof neighborhood
or interveningsocial processes.4
We organize our assessment by implementing a classification based on re-
searchdesign and level of analysis. We included studies in our review that fit any
of the three following categories:(a) neighborhood-levelstudies with neighbor-
hood process measures,in which both the dependentand independentvariables
are expressed as aggregate scales, counts, or rates across ecologically defined
areas that are akin to neighborhoods;(b) multilevel studies with neighborhood
process measures,in which sample members are nested within ecologically de-
fined neighborhoods,the dependentvariableis measuredat the individuallevel,
andthe independentvariablesincludeboth individual-levelfactorsandaggregate-
level measuresof neighborhoodcharacteristics(both structureand process); and
(c) multilevelstudieswith pseudo or proxy neighborhood-processmeasures,iden-
tical to the previous category except that social processes are actually measured
at the individual level. Although studies in the third category usually make in-
ferences aboutneighborhood-levelvariations,they only marginallyfulfill our se-
lection criteriabecause analyticallythey treatsocial processes as individual-level

reviewswithdifferentfoci. Forexample,
2Weencouragereadersto consultindependent
Gephardt(1997) emphasizesthe role of structural such as concentrated
characteristics
poverty.Burton& Jarrett (2000)examinefamilyprocessesin neighborhood-based studies
of childdevelopment,witha specificfocusonminoritypopulations. Duncan&Raudenbush
(1999)andSobel(2001)outlinemainlymethodological issuesandresearchstrategiesin
neighborhood-level studiesof childandyouthdevelopment. Robert(1999)reviewsthe
betweencommunitysocioeconomic
relationship contextandvarioushealthoutcomes(in-
cludingphysical),whileLeventhal& Brooks-Gunn (2000) andEarls& Carlson(2001)
focusprimarily onyoungchildren,families,andneighborhoods. Finally,we donotattempt
to covera burgeoning ethnographic or qualitative in
literature anydetail,orcorresponding
culturalaccountsof neighborhood effects.Fora recenteffortalongtheselines,see Small
& Newman(2001).
3Althoughwe appreciatecriticismsof the somewhatarbitrary designationsof whatare
deemed"non-normative" or "problem" behaviors,we follow the spiritof the reviewed
authorsby includingbehaviorssuchas teenagepremarital childbirth
andage of onsetof
firstsex.
4Although we alsoincludedstudiesthatfocusonfamily
outsideourinitialselectioncriteria,
orpeer-group interveningprocesses(e.g.,South&Baumer2000)becauseof theirtheoretical
relevanceto understanding neighborhood effects(seeJencks& Mayer1990).
NEIGHBORHOOD
EFFECTS 449

characteristicsratherthanas emergentpropertiesof neighborhoods.Forexample,


this categorywould include a neighborhoodsocial process measurederivedfrom
a single individual'sreportof his or her neighborhood(e.g., Aneshensel & Sucoff
1996, Geis & Ross 1998), even though such a strategymay yield an unreliable
neighborhood-levelmeasurewith considerablemeasurementerror.5On theoret-
ical groundswe nonetheless include such studies in our review, but only if they
analyze data where individualsare nested within ecologically defined neighbor-
hoods andstructuralcharacteristics(e.g., poverty)areexpressedas aggregate-level
measures.

TakingStock:Summaryof Results
In Tables 1-3 we summarizethe major findings from 40 studies that met our
selection criteria, ordered according to our three-fold classification scheme-
neighborhood-level(Table 1, N = 15 studies), multilevel (Table 2, N = 8), and
individual-levelmeasuresof socialprocesses(Table3, N = 17).We referthereader
to these tablesfor the relevantdetailsof each study(e.g., unitof analysis,measures,
key results);our focus from here on out is thematicsynthesis across the range of
studies.6Ourassessmentleads us to the following synthesis.
* Advances in ResearchDesign and Measurement.One of the most important
first-orderfindings from recent research is that community-basedsurveys
can yield reliableandvalid measuresof neighborhoodsocial andinstitutional
processes.However,unlikeindividual-levelmeasurements,which arebacked
up by decades of psychometricresearchinto their statisticalproperties,the
methodologyneeded to evaluateneighborhoodmeasuresis not widespread.
Raudenbush& Sampson (1999) thus proposedmoving towarda science of
ecological assessment,which they call "ecometrics,"by developingsystem-
atic proceduresfor directly measuringneighborhoodmechanisms, and by
integratingand adaptingtools from psychometricsto improvethe qualityof
neighborhood-levelmeasures.Leavingaside statisticaldetails,the important
point is thatneighborhoodprocesses can and shouldbe treatedas ecological

5An alternative foundin the neighborhood-level


approach, andmultilevelprocessstudies
(categoriesa andb) is to surveymultiplerespondents livingin the sameecologicalareas
anduse theircollectiveassessmentto buildneighborhood indicators(e.g., Elliottet al.
1996,Cooket al. 1997,Sampsonet al. 1997).Otherstrategies,describedbelow,include
standardized observationalassessments.See footnote7 for furtherjustificationregarding
thisclassification.
6In eachtablewe reportfindingsprimarily as theyrelateto neighborhood processmea-
suresandoutcomes.Wetryto maintainfidelityto eachstudy'sinterpretations orinterests,
butoftentheydivergefromthe presentpaper'sfocus.All findingsreportedweredeemed
significantby studyauthorsunlessotherwisenoted.Also, in orderto standardize report-
ing of findings,"full(y)"and "partial(ly)"
mediationreferto a significantdirecteffect
beingreducedto nonsignificance (NS) or to significantbutsubstantiallyreducedlevels,
respectively.
TABLE 1 Neighborhood-levelstudies of social processes and mechanisms, 1996-2001
Data sources Neighborhood Process (P) and
Author(s) and sample units of analysis Outcome (0) measures Main fin
Wamer& SeattleVictim Tracts(N= 100) P: local social ties * ties -*
Rountree Survey(N = 5,302), 0: burglary,assault * stabili
(1997) Seattlepolice, but onl
1990 Census c ties --
white t
Rountree& Same as above Same as above P: male, female * female
Warner social ties with lo
(1999) 0: violent crime " stabili
by mal
* male ti
Bellair Police Services Police beats P: measuresof social * infrequ
(1997) Study (N = 12,019): (N= 100) interaction(10) -> (-)
Rochester,St. Louis, 0: burglary,vehicle theft, * infrequ
Tampa/St.Petersburg robbery,propertycrime overal
index, total crime index on tota
* structu
fully o
effect p
Bellair SeattleVictim Survey Tracts(N = 100) P: informalsurveillance, * survei
(2000) (N = 5,302) unsupervisedteens, * teens ->
Seattlepolice, neighboring Reciproc
1990 Census 0: robbery/stranger * model
assault(R/SA), burglary survei
effect o
* model
R/SA->
-> (-)
* fear do
surve
* model
strong
-> (+)
Sampson Chicago Community Neighborhood P: informal(child) * social
(1997) Survey(N = 3,864), clusters social control contro
1990 Census (N =80) 0: delinquency(gang * stabil
fights, graffiti,causing media
troublein groups)
Morenoff Chicago Community Neighborhood P: collective efficacy * inequ
et al. (2001) Survey(N = 8,782), clusters (CE), organizations, for pr
Chicagopolice, vital (N=343) voluntaryassociations, effect
statistics, 1990 Census kin/friendshipties homic
0: homicide * spatia
* otherp
Sampson& Chicago Community Tracts P: collective efficacy * CE ->
Raudenbush Survey (N = 3,864), (N = 196) (CE), observeddisorder contro
(1999) systematicobservations, 0: burglary,homicide, * CE -+
Chicago police, 1990 robbery,victimization outco
Census * observ
but m
* observ
* disord
struct
Recipro
* CE is
obser
recipr
TABLE 1 (Continued)
Data sources Neighborhood Process (P) and
Author(s) and sample units of analysis Outcome (0) measures Main fin

Markowitzet al. BritishCrimeSurvey Political P: disorder,cohesion, * cohesio


(2001) (3 waves): constituencies fear some s
N = 11,030, (N = 151) 0: burglary, * disord
N= 11,741, victimization struct
N= 10,059 Reciproc
* model
* model
disorde
* model
robber
Petersonet al. Columbuspolice, FBI Tracts P: neighborhood * extrem
(2000) UniformCrimeReports, (N = 177) institutions(4) crime,
variouspublic records/ 0: violent crime center
local sources, 1990 (robbery,assault, * bars -+
Census homicide, rape) * recreat
extrem
Veysey & BritishCrime Political P: networks,unsupervised * peer gr
Messner Survey constituencies peer groups, organization * org. pa
(1999) N =10,905 (N=238) participation (-) vic
0: victimization * SES h
throug
Hirschfield& Merseyside(Eng.) Enumeration P: social cohesion (police calls, * social
Bowers police/govt., districts Homewatchorganizations)
(1997) BritishConsumer (N = 2,800) 0: assault,robbery,sexual
Survey, 1991 Census offences, burglary
(N= 150-200
households/unit)
Smith et al. Police, local sources Face blocks P: severalroutineactivity (RA) * "l
(2000) (mid-size southern (N = 7,931) measures:"landuse" measures * ow
city), 1990 Census (e.g., motels, bars);guardianship 2S
(i.e., owner occupied) o
O: streetrobbery * "l
di
* gu
m
* R
af
ro
* ev
st
LaGrange EdmontonPolice/ Enumerationareas P: proximityto schools, * pr
(1999) Transit/Park (N =654) malls sc
Department O: mischief, vandalism * m
(park,transit) sc
C
va
Scribneret al. New Orleanspublic Tracts(N = 155) P: routineactivitymeasures * al
(1998) healthrecords/local (alcohol outlets) w
sources,1994 Census O: sexual risk behavior "o
(i.e., gonorrhea)
Cohen et al. New Orleanspublic Block groups P: neighborhooddisorder * di
(2000) healthrecords/local =
(N 55) O: same as above * po
sources, systematic m
observations,1990
Census
TABLE 2 Multi-levelstudies of social processes and mechanisms, 1996-2001
Data sources Neighborhood Process (P) and
Author(s) and sample units of analysis Outcome (0) measures
Elliott et al. Chicago (N = 887) and Chicago, tracts(N = 58) P: informalcontrol,
(1996) Denver(N = 820) survey Denver,block groups social integration,
of youth-caretakerpairs, (N= 33) informalnetworks
1990 Census 0: prosocial
competence (PC),
conventionalfriends (CF),
problembehavior(PB)

Sampsonet al. Chicago Community Neighborhoodclusters P: collective efficacy (CE)


(1997) Survey (N = 8,872), (N = 343) 0: perceivedviolence,
1990 Census homicide events,
victimization
Gorman- Chicago Youth Tractclusters P: neighborliness, Clusteringused
Smith et al. DevelopmentStudy (N = 275) safety concerns based on neigh
(2000) (N = 288, 11-14-year- 0: delinquency: * "structuralri
old boys), archival chronicminor, poverty,hete
sources, 1990 Census escalating, * "social proce
serious chronic, high safety c
non-offenders * structurerisk
escalatingde
* neighborhoo
Perkins& Surveyof Baltimore Neighborhoods P: perceivedphysical/ Individualleve
Taylor residents(N =412), (N=50) social disorder, * Model 1, 2, 3
(1996) systematicobservations, observeddisorder disorder -> (
archivaldata, 1980 0: fear of crime Neighborhood
Census disorder,2-3 m
* observeddis
* perceiveddis
? perceiveddis
Steptoe & Surveyof Londonadults Postal sectors P: neighborhood * cohesion, inf
Feldman (N = 658), Census (year (N = 37) problems,social * neighborhoo
(2001) not reported) cohesion, informal areasonly, co
social control control (NS)
0: psychological Neighborhood
distress * neighborhoo
TABLE 2 (Continued)
Data sources Neighborhood Process (P) and
Author(s) and sample units of analysis Outcome (0) measures Main finding

Rountree& Seattle Victim Neighborhoods P: incivilities, Individuallev


Land(1996) Survey(N ==5,090) (N = 300) integration, * exposure,ta
Seattlepolice, severalroutine Neighborhoo
1990 Census activity measures * incivilities,
(e.g. "exposure, * exposure,ta
targetattractiveness, * previousvi
guardianship") with low in
0: fear of burglary
Coultonet al. Interviewsof parents Block groups P: social control Individuallev
(1999) (N=400), (N = 20) (disorder,lack of * support --
Child Services, controlof children), Neighborhoo
1990 Census communityresources, * control,res
social support
0: child maltreatment
Cutronaet al. Same as above Same as above P: communitydeviance, Clusteringus
(2000) (caregiversonly, cohesion, disorder Model 1:
N = 709 women) 0: distress (general * disorder -
distress, anxiety) * social supp
* cohesion, d
Model 2:
* disorder-e
neighborh
* cohesion ->
neighborh
NEIGHBORHOOD
EFFECTS 457

or collective phenomenaratherthanas individual-levelperceptionsor traits.


Again, we believe this distinctionis crucialfor the advancementof research.7
* DisparatebutConvergingMeasuresof NeighborhoodMechanisms.Wefound
very little consistency across studies in Tables 1-3 in the way neighborhood
social and institutionalprocesses were operationalizedor theoreticallysitu-
ated.Moreover,manyindicatorsof neighborhoodmechanismsareintercorre-
lated,raisingthe questionof how manyindependentandvalidconstructsthere
really are (see also Cook et al. 1997, Furstenberget al. 1999, Sampsonet al.
1999). Forexample,is thereonly one higher-ordersocial process,or arethere
multiplesubdimensions?Sifting throughthe myriadoperationaldefinitions,
empirical findings, and theoreticalorientationsrepresentedin Tables 1-3,
we believe thatfour classes of neighborhoodmechanisms,althoughrelated,
appearto have independentvalidity.
1. Social Ties/Interaction:One of the drivingforces behind much of the
researchon neighborhoodmechanismshas been the concept of social
capital,which is generallyconceptualizedas a resourcethatis realized
throughsocial relationships(Coleman1988, Leventhal& Brooks-Gunn
2000). The studies we reviewed include measuresthat tap several di-
mensions of social relations, such as the level or density of social ties
betweenneighbors(Rountree& Warner1999, Elliottet al. 1996,Veysey
& Messner 1999, Morenoffet al. 2001), the frequencyof social inter-
action among neighbors (Bellair 1997), and patternsof neighboring
(Warner& Rountree1997, Bellair 2000).
2. Norms and Collective Efficacy:Although social ties are important,the
willingness of residentsto interveneon behalf of childrenmay depend,
in larger part, on conditions of mutual trust and sharedexpectations
amongresidents.One is unlikelyto intervenein a neighborhoodcontext
where the rules are unclearand people mistrustor fear one another.It
is the linkage of mutual trust and the sharedwillingness to intervene
for the public good that captures the neighborhoodcontext of what
Sampsonet al. (1997) termcollective efficacy.Sampsonandcolleagues
constructeda measure of collective efficacy by combining scales of
the capacity for informal social control (see also Elliott et al. 1996,
Steptoe & Feldman 2001) and social cohesion (see also Rountree&

7Raudenbush & Sampson(1999) demonstrate thatindividual-level


reliabilitiesarefun-
damentallydifferentandoftenvery discrepantfrom,the aggregate-level reliabilityof a
surveymeasureto detectbetween-neighborhood differences.Themainfactorsincreasing
theaggregatereliabilityof a measurearethenumberof respondents withineachneighbor-
hood(with25 a rule-of-thumb goal),thenumberof neighborhoods, andtheproportion of
totalvariancebetweenneighborhoods relativeto theamountwithinneighborhoods. Simi-
larly,observational
reliabilitiesdependon thenumberof ecologicalunitsassessedandthe
between-unitvariance.
458 SAMPSON? MORENOFFPGANNON-ROWLEY

Land 1996, Markowitzet al. 2001). Othermeasuresrelatedto the idea


of sharedexpectationsfor social controlincludeinformalsurveillanceor
guardianship(Bellair2000) and the monitoringof teenage peer groups
(Veysey & Messner 1999, Bellair 2000).
3. InstitutionalResources,at least in theory,referto the quality,quantity,
and diversityof institutionsin the communitythataddressthe needs of
youth, such as libraries,schools and otherlearningcenters,child care,
organizedsocial and recreationalactivities, medical facilities, family
supportcenters, and employmentopportunities.In practice,however,
empiricalmeasureshave been limitedto the merepresenceof neighbor-
hood institutionsbased on survey reports(Coulton et al. 1999, Elliott
et al. 1996) and archivalrecords (Petersonet al. 2000). A few stud-
ies have used surveys to tap levels of participationin neighborhood
organizations(Veysey & Messner 1999, Morenoffet al. 200 1).8
4. Routine Activities: A concern for institutionssuggests a fourth,often
overlookedfactorin discussionsof neighborhoodeffects-how landuse
patternsand the ecological distributionsof daily routineactivitiesbear
on children'swell-being.The locationof schools, the mix of residential
with commercialland use (e.g., stripmalls, bars),public transportation
nodes, and large flows of nighttimevisitors, for example, are relevant
to organizinghow and when children come into contact with peers,
adults, and nonresidentactivity.Like studies of institutions,however,
direct measures of social activity patternsare mostly absent. Studies
of routine activities typically measuretypes of land use in the neigh-
borhood, such as the presence of schools, stores and shoppingmalls,
motels and hotels, vacantlots, bars,restaurants,gas stations,industrial
units, andmultifamilyresidentialunits (e.g., LaGrange1999, Sampson
& Raudenbush1999, Smith et al. 2000, Petersonet al. 2000, Scribner
et al. 1998).
* StrongestEvidenceLinksNeighborhoodProcesses to Crime.To date, most
research on neighborhoodinteractionaland institutionalprocesses has fo-
cused on crime outcomes, especially police records of homicide, robbery,
and strangerassault, and survey reportsof violent and propertyvictimiza-
tion. This focus is not surprisinggiven the influenceof social disorganization
theoryin criminology,motivatingresearchon crime ratesandneighborhood
mechanisms (Morenoff et al. 2001). The studies summarizedin Tables 1
and 2 suggest that crime rates are relatedto neighborhoodties and patterns
of interaction(Warner& Rountree 1997, Rountree& Warner1999, Veysey
& Messner 1999, Bellair 1997), social cohesion and informalsocial control
(Elliottet al. 1996, Sampsonet al. 1997, Hirschfield& Bowers 1997, Morenoff

8Moststudiesunderreviewthusdo notdistinguish dimensionsof


well betweenstructural
(e.g.,theirdensity)andmediatinginstitutional
institutions processes.
EFFECTS
NEIGHBORHOOD 459

et al. 2001, Bellair 2000), institutionalresources(Veysey & Messner 1999,


Petersonet al. 2000), androutineactivitypatterns,especially mixed land use
and proximityto schools and malls (LaGrange1999, Smith et al. 2000).
* Activationof Social Ties. Thereis evidence, however,suggestingthat strong
social ties may not be as critical for child well-being and general safety as
the sharedexpectationthatneighborswill interveneon behalf of the neigh-
borhood.One can imagine situationswhere strongties may impedeeffortsto
establish social control, as when dense local ties foster the growth of gang-
relatednetworks(Pattillo-McCoy1999). Moreover,weak ties-less intimate
connections between people based on more infrequentsocial interaction-
maybe essentialforestablishingsocial resourcessuchasjob referralsbecause
they integratethe communityby bringingtogether otherwise disconnected
subgroups(Granovetter1973, Bellair 1997). Two generalresearchfindings
supportthis line of thinking. First, some studies have shown that the as-
sociation of ties with crime is largely mediatedby informal social control
and social cohesion (Elliottet al. 1996, Morenoffet al. 2001). Second, other
studies have qualified the relationshipbetween ties and crime by suggest-
ing that crime is related only to certain patternsof neighborhoodties and
social interaction,such as social ties among women (Rountree& Warner
1999) or moderatefrequencyof social interactionamong neighbors(Bellair
1997). These findings suggest that the activationof social ties to achieve
shared expectations for action, or what Sampson et al. (1997, 1999) pro-
pose is a generalconstructof collective efficacy,may be a criticalingredient
for understandingneighborhoodcrime and general aspects of community
well-being.
* Social Mechanismsand Health. In a related matter,a growing numberof
studieshave expandedthe scope of neighborhoodinquiryto considermental
health outcomes such as depressionand psychological distress (Ross 2000,
Cutronaet al. 2000), and high-risk adolescentbehaviorssuch as early sex-
ual initiation,teen childbearing,and conductdisorder(e.g., Upchurchet al.
1999, South & Baumer 2000, Lanctot & Smith 2001). As Table 3 under-
scores, these studies overwhelminglymeasure social processes at the indi-
vidual ratherthanneighborhoodlevel, makingit difficultto offer a summary
assessmentof which, if any,neighborhood-levelmechanismsare important.
Overall, however, it appears that concentratedpoverty, disorder,and low
neighborhoodcohesion arelinkedto greatermentaldistress(e.g., Ross 2000,
Elliott 2000, Cutronaet al. 2000, Geis & Ross 1998, Aneshensel & Sucoff
1996), risk taking and deviant peer affiliation among adolescents (Brody
et al. 2001, Kowaleski-Jones2000, Lanctot & Smith 2001), and indicators
of high-risk sex (Cohen et al. 2000, Baumer & South 2001). Some stud-
ies show that peer-groupfactors (e.g., deviant attitudes)mediate the effect
of neighborhooddisadvantageon teenage behaviors(e.g., South & Baumer
2000).
TABLE 3 Individual-levelstudiesof social processes and mechanisms, 1996-2001
Data sources Neighborhood Process (P) and
Author(s) and sample units Outcome (0) measures Main find
Lanctot& RochesterYouth Tracts (N = 80) P: neighborhood * disorga
Smith Development disorganization, * risky tim
(2001) Study:surveyof "deviantinfluences": * gang m
girl-caretakerpairs risky time with friends, drug/a
(N = 196 girls age gang member
14-17); various 0: early sexual activity,
local data sources pregnancy,status
offenses, drug/alcohol
use
South & National Survey Zip codes P: peers' prochildbearing Models: s
Baumer of Children (N = 202) attitudesand behaviors, * peers' a
(2000) (N = 562 women), parentsknow child's fully m
1980 Census friends * own su
0: firstteenage premarital mediat
birth * parentk
Final Mod
* own att
* affluen
Baumer& same studyas above Zip codes P: same as above * peers' a
South (2001) (N= 1,111 young (N = 278) 0: sex activities:first partial
adults,age 18-22) premaritalsex, disadva
frequency,# partners, * parents
unprotectedsex * disadva
remain
control
Ramirez- Michigansurvey Tracts P: prosocial activities * prosocial
Valles et al. of sexually active (N not reported) (i.e., school, church, * family S
(1998) students(N = 370 community risk beha
adolescents), organizations) * neighbo
1990 Census 0: sex risk behavior controll
(firstsex, # partners, activitie
condom use)
Upchurch Los Angeles County Tracts(N = 49) P: perceivedambient Neighborh
et al. (1999) survey(N = 870, hazards(AH) * AH -> (
age 12-17), (e.g., threats,disorder, * middle-c
1990 Census disintegration) -> (+) b
0: first sex * working
first sex,
* "structu
in under
only afte
Aneshensel same as above same as above P: social cohesion, AH Clustering
& Sucoff (N = 877) (see above) * cohesion
(1996) 0: emotionalhealth: * AH -* (
depression,anxiety, depressio
oppositionaldefiant
disorder,conduct
disorder
TABLE 3 (Continued)
Data sources Neighborhood Process (P) and
Author(s) and sample units Outcome (0) measures Main finding

Stiffman YouthServices Tracts P: perceived * perceived p


et al. (1999) Project:St. Louis (N not reported) neighborhood by exposur
surveyof adolescents problems,peer mediated)b
(N= 792, age14-18), influence,violence * "objective
1990 Census exposure
0: mentalhealth
(6 indicators)
Shumow Milwaukeesurvey Tracts(N = 93) P: perceivedneigh. * "structura
et al. (1998) (N = 168) (5thgraders): danger(PR, CR) mediatedb
studentreports(SR), 0: child distress, * risk -- (+)
parentreports(PR), child misconduct parentperc
teacherreports(TR), (PR, CR, TR) * parentperc
1990 Census misconduc
* child perce
* parent,chil
misconduc
Geis & Ross Community,Crime, Tracts P: perceiveddisorder * ties - (-)
(1998) and HealthSurveyof (N= 1,169) (physical, social), * disorder ->
Illinois adults social ties with
(N = 2,482), neighbors
1990 Census 0: perceived
powerlessness
Ross (2000) Same as above Same as above P: perceiveddisorder * disorder ->
0: depression * disadvanta
by disorde
Ross et al. Same as above Same as above P: perceiveddisorder, * stability->
(2000) informalties with stability--
neighbors,fear, * perceived
powerlessness informalt
0: distress (depression, * ties -> (-)
anxiety) separatel
* disorder,p
interactio
(excludin
Elliott Nevada survey of Zip codes, P: social integration * integratio
(2000) adults(N = 361), (N not reported) 0: depression neighborh
1990 Census
Silver PittsburghViolence Tracts(N = 145) P: perceivedsupport * support -t
(2000) Risk Assessment 0: patientviolence * disadvanta
Study of psychiatric social sup
patients (N = 270);
1990 Census
Wikstrom PittsburghYouth Neighborhood P: peer delinquency * peer delin
& Loeber Study,surveyof units (i.e., tract 0: serious offending * disadvant
(2000) fourthand seventh aggregates) high or ba
gradeboys (N = 90) holds for l
(N= 1,014), * disadvant
1990 Census * risk profil
areas
TABLE 3 (Continued)
Data sources Neighborhood Process (P) and
Author(s) and sample units Outcome (0) measures Main
Kowaleski- NationalLong. Survey Zip codes P: "perceivedcontext": * nei
Jones of YouthMerged (N not reported) neighborhoodproblems beh
(2000) Mother-ChildFiles 0: risk taking, aggressive * nei
(N=860, age 14-18), behaviorindex beh
1990 Census * sta
Brodyet al. Family and Community Community P: collective socialization, Clust
(2001) HealthStudy,survey clusters communitydeviance meas
of youth-caretaker (N =41) (caregiver,child reports) * com
pairsin Iowa and 0: affiliationwith deviant * coll
Georgia(N = 867), peers (child report) str
1990 Census * abo
chi
Seidman Surveyof New York Tracts P: "experiential Exper
et al. City middle/high (N = 203) neighborhood":daily * neg
(1998) school students hassles (e.g., fear), ant
(N = 754), cohesion, social activities, * mo
1990 Census peer values beh
0: antisocialbehaviorindex stu
(delinquency,alcohol use, nei
negativepeer involvement)
NEIGHBORHOOD
EFFECTS 465

* ConcentratedPovertyand StructuralCharacteristicsStill Matter.Although


some studies show that social and institutionalprocesses mediate the asso-
ciation of neighborhoodstructuralfactors with crime and other aspects of
well-being, in many cases they do not explain all or even most of the tra-
ditional correlations.Factors such as concentrateddisadvantage,affluence,
and stability remain direct predictorsof many outcomes (Morenoff et al.
2001, South& Baumer2000, Petersonet al. 2000). Moreover,neighborhood
mechanismsare not producedin a vacuum;some social processes, particu-
larly those relatedto the idea of collective efficacy,appearto emerge mainly
in environmentswith a sufficient endowment of socioeconomic resources
and residentialstability(Sampsonet al. 1999).
* Disorder-Explanatory Mechanismor Outcome?The key process indica-
tors proposedby a numberof studies relate to social and physical disorder
or neighborhoodincivilities (e.g., Perkins& Taylor 1996, Rountree& Land
1996, Cohen et al. 2000, Markowitzet al. 2001). Much of the interestin dis-
orderwas stimulatedby the theoryof "brokenwindows"(Wilson & Kelling
1982), which suggests thatphysical signs of disorder-such as brokenwin-
dows, public drinking,and graffiti-signal the unwillingnessof residentsto
confront strangers,intervenein a crime, or call the police. However,there
is evidence that the direct link between disorderand crime is not as strong
as the brokenwindows theory would suggest, and that disorderis predicted
by the same characteristicsas crime itself, inducing a spuriousrelationship
(Sampson& Raudenbush1999, Markowitzet al. 2001). This does not nec-
essarily mean that disorderis irrelevant.Because signs of disorderare stark
visual remindersof neighborhooddeterioration,they may trigger institu-
tional disinvestment,out-migration,and a general malaise among residents
(Sampson& Raudenbush1999, Ross 2000, Perkins& Taylor1996). Further
researchis needed to determinewhetherdisorderis etiologically analogous
to crime,a cause of crime(see brokenwindowstheory),a mechanismthathas
independentconsequencesfor mentalhealth,or some combinationthereof.

METHODOLOGICAL
CHALLENGES

Despite recentprogress,a dauntingnumberof complex challenges remainin as-


sessing neighborhoodeffects. Indeed,methodologicalissues such as the differen-
tial selection of individualsinto communities,indirectpathwaysof neighborhood
effects, measurementerror,and simultaneitybias (what is causing what?) rep-
resent serious obstacles to drawing definitive conclusions on the causal role of
neighborhoodsocial context (Duncan& Raudenbush1999, Sobel 2001, Winship
& Morgan 1999). The ubiquitous use of the phrase "neighborhoodeffects" is
thus quiteproblematicfrom a methodologicalstandpoint.Neighborhoodsarealso
much moreheterogeneousinternallyand less monolithicthancommonlybelieved
466 * GANNON-ROWLEY
SAMPSON. MORENOFF

(Cook et al. 1997),9 and as noted earlier(see also Raudenbush& Sampson 1999),
far too many studies simply treatneighborhoodprocesses as one more variableto
tag onto individuals.

Selection Bias and ExperimentalDesigns


Althougha full discussion is beyond the scope of this paper,the issue of selection
bias is probablythe biggest challengefacingneighborhood-levelresearch.How do
we know that the area differencesin any outcome of interest, such as adolescent
delinquency,are the result of neighborhoodfactors ratherthan the differential
selection of adolescents or their families into certainneighborhoods?10A recent
body of researchhas directlytakenup this issue by examiningan ongoing housing
programin five majorcities across the United States. The Moving to Opportunity
(MTO) demonstrationis a U.S. Departmentof Housing and UrbanDevelopment
projectin Boston, Baltimore,New YorkCity,Chicago,andLos Angeles (see Katz
et al. 2001 and http://www.wws.princeton.edu/-kling/mto/).Based on findings
fromthe Gautreauxprogramin Chicago showing improvedoutcomesfor children
and adults (Rosenbaum 1995), the MTO programwas designed to test whether
familieswho movedfrominner-city,high-povertyareasto low-povertyareascould
attainthe apparentimprovementsseen in Chicago.
The MTO programutilized an experimentaldesign by randomly assigning
eligible applicantsto one of three groups, two of which received some form of
treatmentin the formof Section 8 vouchersor certificates,anda controlgroupthat
receivedno experimentaltreatment.11 Thisprocessof randomassignmentprovides
an almost unique opportunity separatethe role of neighborhoodcontext from
to
the selection bias thatmay arise from residentialmobility decisions. A key issue,
however,is that not all subjects take up the experimentaltreatment.To address

9Itis oftennoted,in this regard,thatmoreof the variancein almostany outcomelies


withinratherthanbetweenneighborhoods. Yetlargeneighborhood differences(andpoten-
tial interventioneffects) arenot incompatiblewith the low between-neighborhoodvariance
componentsthat are commonly observed(see Duncan& Raudenbush1999).
"0Economistshave been most forthrightin addressingselection bias and individualchoice.
For an extended discussion of selection and identificationproblems in neighborhood-ef-
fects research,see Manski (1993). For a non-experimentalapproachto the endogeneityof
neighborhoodprocesses and social interactions,see Durlauf(2001). Excellent sociological
approachesto statisticalinferenceare Wmship& Morgan(1999) and Sobel (2001).
"Families were deemed eligible if they were public housing or Section-8 assistedhousing
residentswith childrenandlived in a census tractwith a povertyrategreaterthanor equalto
40%.Eligible applicantswere randomlyselectedfrom a waitinglist andrandomlyassigned
to one of three groups:a control group whose membersreceived no change in assistance;
a Section-8 comparisongroup (S-8) that received rental certificatesor vouchers without
geographicalrestrictionsand no special assistance;and an experimentalgroup(MTO)that
receivedvouchersor housingcertificateswith a requirementthatthey move to a low poverty
area(less than 10%).MTOparticipantsalso receivedcounseling and housing assistance.
NEIGHBORHOOD
EFFECTS 467

this issue, MTO Treatment-on-Treated (TOT)analysis comparesthe outcomes of


families who actuallyreceived the treatment(those who actually move, whether
in the location-restrictedor -unrestrictedgroups)to the outcomesof control-group
families who would have acceptedthe treatmenthad it been offered. The Intent-
to-Treat(ITT) analysis comparesthe averageoutcomes for eithertreatmentgroup
with those of the controlgroup,estimatingthe effect of being offeredthe treatment,
regardlessof whetherthe familydecidedto acceptthe certificate/voucherandmove
(Katz et al. 2001).
In Table4 we summarizethe findingsfromthe experimentalliterature.Prelimi-
naryevidenceis generallypositiveforthe outcomesof moversto low-incomeareas,
in accordancewith the early GautreauxProject.Generally,families thatmoved to
low-povertyareasexperiencedimprovedoutcomes vis-a-vis overallhealth(phys-
ical andmental),safety,boys' problembehavior,andwell-being (Katzet al. 2001,
Ludwig et al. 2001, Rosenbaum& Harris2001). A reductionin behaviorprob-
lems among boys was found in Boston (Katz et al. 2001) as well as in Baltimore
(Ludwig et al. 2001). The large reductionin juvenile arrestsfor violent offenses
in Baltimorewas accompaniedby an increasein juvenile arrestsfor propertyof-
fenses, althoughthe latterfindingpertainsonly to the intent-to-treatspecification
and did not hold up when preprogramcharacteristicswere adjusted.
Despite the importanceof experiments,we should not lose sight of their limi-
tations. Selection bias must still be consideredin the form of differentialtake-up
rates and dropoutfrom the program.Most importantfrom our perspective, the
randomassignmentof housing vouchersdoes not addresscausalprocesses of why
neighborhoodsmatter.When MTO families move from one neighborhoodto an-
other,entirebundlesof variableschange at once, makingit difficultto disentangle
the change in neighborhoodpoverty from simultaneouschanges in social pro-
cesses (Katz et al. 2001, p. 621). The clear tendency has been to interpretMTO
resultsin termsof the effects of changingconcentratedpoverty,but for the reasons
expressedin this paper,such an assertionis arbitrary-any numberof changes in
social processes associatedwith povertymay accountfor the result.Note also that
MTOdoes not randomlyallocateneighborhoodconditionsto participants;voucher
recipients can choose to live in any numberof middle-class neighborhoodcon-
ditions. Thus, while MTO may provide policy makerswith evidence on whether
offeringhousingvoucherscan improvethe lives of poorchildren,it is less satisfac-
tory to social scientists interestedin explainingthe mechanismsof neighborhood
effects.

Overcontroland IndirectPathways
Much research on neighborhoodsis inconsistent with the logical expectations
set forth by contextual theories that stress enduringeffects and developmental
pathways(Sampson2001). The most commonstrategyin multilevelneighborhood
researchis to estimatea directeffects model wherebya host of individual,familial,
peer, and school variablesare enteredas controlsalongside currentneighborhood
TABLE 4 Summaryof Moving To Opportunity(MTO) findings.
Site and Data sources and
Study sample Outcomes (0) Main findings

Ludwig Baltimore MTO Baseline Survey, TOT:MTO & S8: (-) violent crime; S8
et al. N = 336 Dept. JuvenileJustice ITT:MTO:(-) violent crime (adjustedo
(2001)a (age 11-16) 0: juvenile arrests (unadjustedonly); S8: (-) violent crim
(adjustedonly, incidence only)
Katzet al. Boston MTO Baseline Survey, TOT/ITT:MTO & S8: (NS) received AF
(2001)a N = 612 Boston Follow-up Survey, adulthealth, adultscalm and peaceful; (
(age 6-15) administrativerecords, close friend
N = 540 1990 Census TOT/ITT': MTO:(-) injuries,asthmaatt
(adults) 0: multiple-child S8: (+) safety (on 3 of 7 measures)
(behavioral,health) and
adult(income, health)
Rosenbaum Chicago Baseline Survey, Chicago * MTO & S8 Movers: (-) social/physica
& Harris N = 120 Follow-up Survey,Urban feelings of safety
(2001)b (adults) InstituteUnderclass * MTO Movers v. S8 Movers:(-) socia
Database teenagers;(-) ratingsof transportatio
a(-) denotesa significantdecreaseon noted outcomes and (+) a significantincrease,both in comparisonto controlgroups."Unadjus
individualcharacteristics;"adjusted"includes such controls. See text for distinctionsbetween "TOT"and "ITT".
bFindingsare basedon MTO-and S8-Movers'comparisonsof destinationneighborhoodsto neighborhoodsof origin.
NEIGHBORHOOD
EFECTS 469

characteristicsof residence. But this strategyconfoundsthe potentialimportance


of both long-termcommunityinfluences and mediatingdevelopmentalpathways
regardingchildren'spersonaltraitsand dispositions,learningpatternsfrompeers,
family socialization,school climate, and more. Put differently,static models that
estimatethe directeffect of currentneighborhoodcontext on a particularoutcome
(e.g., delinquency,level of academicachievement)may be partitioningoutrelevant
variance in a host of mediating and developmentalpathways of influence. The
generalmisuse of controlvariablesin sociology (Lieberson1985) thus appearsto
be exacerbatedin the case of neighborhoodeffects.12

Event-BasedModels
Anotherdisconnectbetween theory and design is tied to the common practicein
neighborhood-effectsresearchof looking solely at the characteristicsof the in-
dividual's place of residence. Although seemingly natural,a problem with this
approachis thatmany behaviorsof interest(e.g., stealing, smoking,takingdrugs)
unfold in places (e.g., schools, parks,center-cityareas) outside of the residential
neighborhoodsin which the individuals involved in these behaviors live. Con-
siderthe natureof routineactivitypatternsin modem U.S. cities, where residents
traversethe boundariesof multipleneighborhoodsduringthe courseof a day.Ado-
lescents occupy manydifferentneighborhoodcontextsoutsideof home, especially
in the companyof peers. Even childrenexperiencemore residentialenvironments
than we commonly expect (Burton et al. 1997, p. 135). This is a problematic
scenario for neighborhoodresearch seeking to explain contextualeffects on in-
dividualdifferencesin behavior.For example, it is possible for the prevalenceof
participationin some crimes to be spreadfairlyevenly acrossindividualsliving in
many neighborhoods,even as crime events are highly concentratedin relatively
few neighborhoods.This sort of neighborhoodeffect on events (typical of drug
markets,for example, where buyers often come from afar) is obscured in cur-
rentpractice.It thus pays to take seriously contextualtheoriesthatfocus more on
behavioralevents than individualdifferences-for example, how neighborhoods
fare as units of guardianshipor socialization over their own public spaces. The
crime-rateliteratureoften takes this strategyby locating the incidence of crime
events ratherthanthe residence of offenders(e.g., see Table 1).

12n a recentstudythatchartsa welcomechangein pace (butwith an outcomefalling


outsideourselectioncriteria),Axinn& Yabiku(2001)examinedtherelationship between
macro-level socialchangeandindividual (orfertility)behaviorinNepal.They
child-bearing
foundevidenceof bothenduringeffectsof childhoodcommunity contextandindependent
effectsof adultcommunitycontextonwomen'schildbearing behavior. Fromthisfinding,the
authorsrightlyadvocatefora morecomprehensive viewof enduringcontextual influences
(Axinn& Yabiku2001:1252),accounting formultipledevelopmental influencesacrossthe
life course.
470 SAMPSON * MORENOFF GANNON-ROWLEY

NEW DIRECTIONS IN STUDYING


NEIGHBORHOOD PROCESSES

The basic argumentthat unites our assessment is that researchneeds to take se-
riously the measurementand analysis of neighborhoodsas importantunits of
analysis in their own right, especially with regardto social-interactionaland in-
stitutionalprocesses. We focus on five directionsfor designing researchon the
neighborhoodcontext of child and adolescent well-being that build on the idea
of takingneighborhoodsocial processes, and hence ecometrics,seriously:(a) re-
definingneighborhoodboundariesin ways thataremore consonantwith social in-
teractionsandchildren'sexperiences,(b) collecting dataon the physicalandsocial
propertiesof neighborhoodenvironmentsthroughsystematicsocial observations,
(c) taking accountof spatialinterdependenceamong neighborhoods,(d) analyz-
ing the dynamics of change in neighborhoodsocial processes, and (e) collecting
benchmarkdata on neighborhoodsocial processes.

Neighborhood Boundaries
Although predominantin the literature,the strategy of defining neighborhoods
based on Census geography and using tracts or higher geographical aggrega-
tions as proxies for neighborhoodsis problematicfrom the standpointof studying
social processes. The micro-dimensionsof neighborhoodinteractionmay be par-
ticularlyimportantfor child well-being because of the spatialconstraintson chil-
dren's patternsof daily activities. A new approachto defining neighborhoods,
as seen in Grannis's(1998, 2001) recent studies of Los Angeles, San Francisco,
Pasadena,CA, and Ithaca,NY, delineatesecological contexts based on the geog-
raphyof streetpatterns.Using a GeographicInformationSystem (GIS), Grannis
(1998, 2001) defines residentialunits that he calls "tertiarycommunities"by de-
lineating aggregationsof streetblocks that are reachableby pedestrianaccess-
meaning that pedestrians can walk through the area without having to cross
over a major thoroughfare.Granniscompares communitiesdefined by residen-
tial streetpatternsto dataon the social networksof neighbors,includingresidents'
cognitive maps of their neighborhoodsand areas of social interaction(see also
Coulton et al. 2001). He finds that residents interact more with people living
withintheirtertiarycommunitiesthanwithpeople who live nearbybutacrossmajor
thoroughfares.
The micro-ecologyof pedestrianstreetsbearsdirectlyon patternsof interaction
that involve childrenand families. Parentsare generally concernedwith demar-
cating territoryoutside of which theirchildrenshouldnot wanderunaccompanied
by an adult, to ensure that their children stay in areas that are safe for play and
conduciveto adultmonitoring.To the extentthatthese limited spaces of children's
daily activitiesusuallydo not cross majorthoroughfares,definingtertiarycommu-
nities may providea foundationfor constructingneighborhoodindicatorsof child
well-being and social processes more generally.
NEIGHBORHOODEFFECTS 471

SystematicSocial Observation
Anothermovementin neighborhoodresearchis to collect data that more directly
reflectthe sights, sounds, and feel of the streets.The motivationbehindcollecting
observationaldata is that there are physical and social featuresof neighborhood
environmentsthat cannot be reliably capturedin surveys but that provide very
tangiblecontextsfor child development.Considerthe exampleof using systematic
social observationof streetblocks (Sampson& Raudenbush1999). Between June
and October 1995, observers drove a sport utility vehicle (SUV) at about 3-5
miles per hour down every street within a sample of Chicago neighborhoods.To
observe each block face (one side of a street within a block), a pair of video
recordersand a pair of trained observers (one of each located on each side of
the SUV) simultaneouslycapturedsocial activities and physical featuresof both
block faces. The observersrecordedtheirobservationsonto a writtenlog for each
block face, also makingcommentariesinto the videotapeaudiowheneverrelevant.
Applyingthese procedures,a totalof 23,816 block faces were observedandvideo-
recorded.
By observingblock faces, datacan be aggregatedto anylevel of analysisdesired
(e.g., block, block group,housing project,or neighborhood)to characterizesocial
and physical characteristics.Such datacan be exploited to build new measuresof
micro-neighborhoodcontexts.For example, flexible neighborhoodindicatorscan
be constructedthatbearon child well-being, includingsuch validatedmeasuresas
physical disorder(e.g., the presenceor absenceof cigarettesin the street,garbage,
empty beer bottles, graffiti, abandonedcars); social disorder (e.g., presence or
absence of adultsloitering,drinkingalcohol in public, public intoxication,adults
fighting,prostitution);physical conditionof housing (e.g., vacanthouses, burned
out houses or businesses, dilapidatedparks), and alcohol and tobacco influence
(e.g., presenceof alcohol signs and tobacco signs on a block, presenceof barsand
liquor stores on a block). Direct measuresof street-levelsocial interactions(e.g.,
adultsplaying with children)can also be constructed.
A limitationof systematicsocial observationis thatit is relativelyexpensiveand
tedious to videotapeblock faces and then code the resultingtapes. However,one
mightimplementthismethodologyon a widerscale by havinginterviewersobserve
and rate city blocks on foot while they are out in the field conductinginterviews.
If this methodology,which is substantiallycheaperthan using videotapes,yields
comparablyreliablemeasures,it could serve as a model for integratingsystematic
social observationwith traditionalsurveys.

SpatialDynamics of Child Well-Being


A thirdtrendin neighborhoodresearchis the expansionof communitycontext to
include nearby areas outside of the formal boundariesof a given neighborhood,
however defined. The general idea is that social behavioris influencednot only
by what happens in one's immediate neighborhood,but also by what happens
in surroundingareas (Morenoffet al. 2001, Smith et al. 2000). For example, the
472 SAMPSON* MORENOFF GANNON-ROWLEY

benefits of collective efficacy may accruenot just to the residentsof a particular


neighborhood,butpotentiallyto residentsin adjacentareasas well (Sampsonet al.
1999). Parentswho send theirchild to play with friendsin a nearbyneighborhood,
whereresidentstend to engage in collective supervisionand monitoring,derive a
spatialadvantagemuch in the same way thatthey would benefitfrom living next
to a parkor a good school. By contrast,neighborhoodswith minimalexpectations
for social control and sparseinterfamilyexchange producespatialdisadvantages
for parentsand childrenwho live in adjoiningareas.
This frameworkhas implications for understandingresidential stratification.
For example, if African-Americanneighborhoodsare embeddedin more disad-
vantagedenvironmentsthanare similarlyendowedwhite neighborhoods,then the
consequences of racial segregationmay be greaterand more systemic than pre-
viously thought. Patillo-McCoy's (1999) ethnographicstudy of "Groveland,"a
communityin Chicago, suggests thatblack middle-classfamilies face such a spa-
tial (and structural)disadvantage.Despite networksof social control, she found
that black middle-classfamilies must constantlystruggleto escape the problems
of drugs, violence, and disorderthat spill over from neighboringcommunities.
The clear implicationof such spatialdynamicsfor the studyof child well-being is
thatcommunityindicatorsthatfocus on processes or outcomes internalto a given
neighborhoodare getting only partof the story.Newly developed techniquesfor
analyzing and displaying the connection of social and spatialprocesses are thus
importantfor progressin neighborhood-effectsresearch.13

Dynamics of Change
In addition to spatial dynamics there is a clear need for rigorous longitudinal
studies of neighborhoodtemporaldynamics.Just as individualschange, develop,
and are sometimes transformed,so too neighborhoodsare dynamicentities. One
of the appealingfeaturesof the recent focus on social processes, at least from a
theoreticalperspective,is the recognition that processes such as social control,
reciprocalexchange,andepidemicsarerootedin dynamicaspectsof social life, as
comparedto the morecommonfocus on static,compositionalcharacteristics(such
as race)thatarenot fundamentallycausalvariables(Winship& Morgan1999). It is
painfullyironic,however,thatmost studiesof social processare,methodologically
at least, cross-sectionalin nature.Wehave scantinformationon how neighborhood
processesevolve overtime,orhow theyinteractwith allegedoutcomes.The limited
researchthat does exist (e.g., Bellair 2000) points to complex interactionsand
feedbackprocesses amongstructuralconstraints,social ties, andbehaviorssuch as
crime.Researchersthus need to redoubletheireffortsto investigateneighborhood
social processes in trulydynamic,interactivefashion.

13See http://phdcn.harvard.edu/res-pubs/maproom/index.htmlfor examples of maps dis-


playing spatialadvantagesand disadvantagesthat transcendneighborhoodboundaries.
NEIGHBORHOOD
EFFECTS 473

Towarda Benchmarkof EcometricData


A final step in fosteringprogresswould be to supportthe systematiccollection of
benchmarkdata on social environmentsthat can be comparedacross communi-
ties. The goal would be to develop a standardizedapproachto the collection and
disseminationof data thatindividualcommunitiescan use to evaluatewhere they
stand in regardto nationaland/orregional norms. Similarto school reportcards
thattrackthe progressof educationalreform,a standardizedapproachto assessing
collective propertieswould eventuallyallow local communitiesto gauge how well
or poorly they are doing. For example, the SustainableSeattle projecthas com-
bined archivalrecords,census data, and surveys to compile sustainabilitytrends
acrosscommunitiesin diverseareasof concern(e.g., economic resources,literacy,
low birthweight,neighborliness).The Leaders Roundtablein Portland,Oregon
has undertakena similar initiative to collect data on communityhealth using a
combinationof focus groups, surveys,key stakeholderinterviews,and document
reviews. More ambitiously,RobertPutnamrecently launcheda benchmarksur-
vey both nationwide and in about 40 American communities, with the goal of
assessing baseline levels of social capital and eventuallychanges over time (see
http://www.cfsv.org/communitysurvey/).

CONCLUSION
Thereis little doubtthatnumerousproblemshinderthe estimationof neighborhood
effects. Manyof these complex challengeshave been discussedin this paper.Still,
we would conclude on a positive note by arguingthatwe now know quite a bit. As
reviewed,the evidence is solid on the ecological differentiationof Americancities
along socio-economic and racial lines, which in turn correspondsto the spatial
differentiationof neighborhoodsby multiple child, adolescent, and adult behav-
iors. These conditionsareinterrelatedandappearto varyin systematicandtheore-
tically meaningfulways with hypothesized social mechanisms such as informal
social control,trust,institutionalresourcesand routines,peer-groupdelinquency,
and perceived disorder(Tables 1-3). An importanttake-awayof our assessment
is that these and other neighborhood-levelmechanisms can be measured reli-
ably with survey,observational,and archivalapproaches.Anotherfindingis that
extra-localneighborhoodmechanismsappearwith considerablestrength,suggest-
ing that spatialexternalitiesoperateabove and beyond the internalneighborhood
characteristicsof traditionalconcern.
Despite progress,fundamentalquestions remain.Even when directly focused
on social processes, the many differences in research design and measurement
across studies in Tables 1-4 make it difficult to provide an overall estimate of
the magnitudeof associations.We also know little aboutthe causes of key social
processes or whether they are responsive to neighborhoodpolicy interventions.
For example, what produces or can change collective efficacy and institutional
capacity? Although much effort has been put into understandingthe structural
474 SAMPSON MORENOFF
E GANNON-ROWLEY

backdropto neighborhoodsocial organization,we need a deeperfocus on cultural,


normative,andcollective-actionperspectivesthatattachmeaningto how residents
frametheircommitmentto places. Anotherlimitationof neighborhood-effectsre-
searchhas been its lack of attentionto measuringpeernetworksandthe connection
of neighborhoodsand school processes.
Perhapsthe mainthreatto neighborhood-effectsresearchis individualselection
bias, althougheven herewe wouldview the news as somewhatencouraging.As the
nascentexperimentalliterature(Table4) has demonstrated,when randomization
is invokedwe still find evidence for the apparentinfluence of place. We applaud
the MTO experimentalturn,but caution that the task remainsto specify the ex-
act mechanisms of transmission.An ideal, albeit difficult, strategywould be to
combineexperimentalassignmentof neighborhoodconditionswith a longitudinal
assessment of changes in social processes and individualbehaviors.We would
also cautionagainstthe common tendencyto view selection bias as an individual
traitand a nuisanceto be controlled.When individualsselect neighborhoods,they
appearto do so based on social characteristicssuch as neighborhoodracial segre-
gation, economic status,and friendshipties. Researchneeds to betterunderstand
the mutualinterplayof neighborhoodselection decisions, structuralcontext, and
social interactions(e.g., Durlauf2001).
Armed with methodological advances in ecometrics that are improving our
prospects for measuring neighborhoodsocial processes, we are optimistic re-
gardingthe next generationof researchthat takes up these and other challenges.
When combinedwith advancesin definingmicro-neighborhoodcontextsbasedon
streetpatterns,systematicallyobservingpublic spaces, longitudinal-experimental
designs, and detecting spatial dynamics, contextual research on the dynamic
sources of child, adolescent, and even adult development has a bright future
indeed.

ACKNOWLEDGMENTS
We thankSandyJencks and JohnGoeringfor theircriticalfeedback.

The Annual Reviewof Sociologyis online at http://soc.annualreviews.org

LITERATURE
CITED
AlmgrenG, GuestA, Imerwahr G, SpittelM. Am.J. Sociol.106(5):1219-
tilitylimitation.
1998.Joblessness,familydisruption,andvi- 61
olent deathin Chicago,1970-1990. Soc. BaumerEP, SouthSJ.2001.Community effects
Forces76(4):1465-93 on youthsexualactivity.J. MarriageFam.
AneshenselCS, SucoffCA. 1996.Theneigh- 63:540-54
borhoodcontextof adolescent andcom-
mentalhealth. BellairPE. 1997.Socialinteraction
J. HealthSoc.Behav.37:293-310 munitycrime:examiningthe importance of
AxinnWG,YabikuST. 2001. Socialchange, neighbornetworks.Criminology 35(4):677-
of families,andfer-
the socialorganization 703
NEIGHBORHOODEFFECTS 475

Bellair PE. 2000. Informal surveillance and CutronaCE, Russell DW,Hessling RM, Brown
street crime: a complex relationship.Crim- PA, MurryV. 2000. Direct and moderating
inology 38(1):137-67 effects of community context on the psy-
Brody GH, Ge X, Conger R, Gibbons FX, chological well-being of African American
Murry VM, et al. 2001. The influence of women. J. Pers. Soc. Psychol. 79:1088-101
neighborhood disadvantage, collective so- DuncanGJ, RaudenbushSW. 1999. Assessing
cialization,and parentingon African Amer- the effects of context in studies of child and
ican children'saffiliationwith deviantpeers. youth development.Educ. Psychol. 34:29-
Child Dev. 72(4):1231-46 41
Brooks-Gunn J, Duncan GJ, Klebanov PK, DurlaufS. 2001. A frameworkfor the study of
Sealand N. 1993. Do neighborhoodsinflu- individualbehavior and social interactions.
ence child and adolescentdevelopment?Am. Sociol. Methodol.31(1):47-87
J. Sociol. 99(2):353-95 EarlsF, CarlsonM. 2001. The social ecology of
Brooks-Gunn J, Duncan GJ, Aber JL, eds. child healthand well-being. Annu.Rev.Pub.
1997a. NeighborhoodPoverty:Vol. I: Con- Health 22:143-66
text and Consequencesfor Children. New ElliottM. 2000. The stressprocessin neighbor-
York:Russell Sage Found. hood context.Health Place 6:287-99
Brooks-Gunn J, Duncan GJ, Aber JL, eds. Elliott DS, Wilson WJ, Huizinga D, Sampson
1997b. NeighborhoodPoverty:Vol. II:Pol- RJ, Elliott A, Rankin B. 1996. The effects
icy Implicationsin StudyingNeighborhoods. of neighborhooddisadvantageon adolescent
New York:Russell Sage Found. development.J. Res. CrimeDelinq. 33:389-
Burton LM, JarrettRL. 2000. In the mix, yet 426
on the margins:the place of families in ur- FurstenbergFF Jr, Cook TD, Eccles J, Elder
ban neighborhoodandchild developmentre- GH Jr, SameroffA. 1999.Managingto Make
search.J. MarriageFam. 62:1114-35 It: Urban Families and AdolescentSuccess.
Burton LM, Price-SpratlenT, Spencer MB. Chicago:Univ. Chicago Press
1997. On ways of thinking about measur- Geis KJ, Ross CE. 1998. A new look at urban
ing neighborhoods:implications for study- alienation: the effect of neighborhooddis-
ing context and developmentaloutcomesfor orderon perceivedpowerlessness.Soc. Psy-
children.See Brooks-Gunnet al. 1997b, pp. chol. Q. 61:232-46
132-44 GephartMA. 1997. Neighborhoodsand com-
CohenD, SpearS, ScribnerR, KissingerP,Ma- munities as contexts for development. See
son K, Widgen J. 2000. "Brokenwindows" Brooks Gunnet al. 1997a, pp. 1-43
andthe risk of gonorrhea.Am.J. Pub.Health Gorman-SmithD, TolanPH, HenryDB. 2000.
90:230-36 A developmental-ecologicalmodel of the
ColemanJS. 1988. Social capitalin the creation relation of family functioning to patterns
of humancapital.Am.J. Sociol. 94: S95-120 of delinquency.J. Quant. Crim. 16(2):169-
Cook TD, Shagle SC, Degirmencioglu SM. 98
1997. Capturing social process for testing Grannis R. 1998. The importance of trivial
mediationalmodels of neighborhoodeffects. streets:residentialstreetsandresidentialseg-
See Brooks-Gunn et al. 1997b, pp. 94-119 regation.Am.J. Sociol. 103(6):1530-64
Coulton CJ, Korbin J, Chan T, Su M. 2001. Grannis R. 2001. Street islands: social net-
Mappingresidents'perceptionsof neighbor- works, street networks, and segregation.
hoodboundaries:a methodologicalnote.Am. Work.Pap., Dep. Sociol., Comell Univ.
J. CommunityPsychol. 29:371-83 GranovetterMS. 1973. The strengthof weak
CoultonCJ, KorbinJE, Su M. 1999. Neighbor- ties. Am.J. Sociol. 78(6):1360-80
hoods and child maltreatment:a multi-level Hirschfield A, Bowers KJ. 1997. The effect
study.ChildAbuseNegl. 23:1019-40 of social cohesion on levels of recorded
476 SAMPSON MORENOFF * GANNON-ROWLEY

crime in disadvantagedareas. Urban Stud. Morenoff JD, Sampson, RJ, RaudenbushSW.


34(8):1275-95 2001. Neighborhood inequality, collective
Jencks C, Mayer S. 1990. The social conse- efficacy, and the spatial dynamics of homi-
quences of growing up in a poor neighbor- cide. Criminology39(3):517-60
hood. In Inner-CityPoverty in the United ParkR. 1916. Suggestionsforthe investigations
States, ed. L Lynn, Jr.,MGH McGeary,pp. of humanbehaviorin the urbanenvironment.
111-85. Washington,DC: Natl. Acad. Press Am. J. Sociol. 20(5):577-612
Katz LF, Kling J, LiebmanJB. 2001. Moving Pattillo-McCoyM. 1999. Black PicketFences:
to Opportunityin Boston: early results of a Privilege and Peril Amongthe Black Middle
randomizedmobilityexperiment.Q. J. Econ. Class. Chicago:Univ. Chicago Press
116(2):607-54 Perkins DD, TaylorRB. 1996. Ecological as-
Kowaleski-JonesL. 2000. Staying out of trou- sessmentsof communitydisorder:theirrela-
ble: communityresources and problem be- tionshipto fear of crime and theoreticalim-
havioramong high-riskadolescents.J. Mar- plications. Am. J. CommunityPsychol. 24:
riage Fam. 62:449-64 63-107
LaGrangeTC. 1999. The impact of neighbor- PetersonRD, KrivoLJ, HarrisMA. 2000. Dis-
hoods, schools, and malls on the spatialdis- advantageand neighborhoodviolent crime:
tributionof propertydamage.J. Res. Crime Do local institutionsmatter?J. Res. Crime
Delinq. 36:393-422 Delinq. 37:31-63
Lanctot N, Smith CA. 2001. Sexual activity, Ramirez-VallesJ, ZimmermanMA, Newcomb
pregnancy,and deviance in a representative MD. 1998. Sexual risk behavior among
urbansample of African American girls. J. youth: modeling the influence of proso-
YouthAdol. 30(3):349-72 cial activities and socioeconomic factors.J.
LeventhalT, Brooks-GunnJ. 2000. The neigh- Health Soc. Behav. 39:237-53
borhoods they live in: the effects of neigh- RaudenbushSW, SampsonRJ. 1999. 'Ecomet-
borhood residence on child and adolescent rics': Towarda science of assessing ecolog-
outcomes.Psychol. Bull. 126(2):309-37 ical settings, with applicationto the system-
LiebersonSL. 1985. MakingIt Count:TheIm- atic social observationof neighborhoods.So-
provementof Social Research and Theory. ciol. Methodol.29:1-41
Berkeley:Univ. Calif. Press RobertSA. 1999. Socioeconomic position and
Ludwig J, HirschfieldP, DuncanGJ. 2001. Ur- health:the independentcontributionof com-
ban poverty and juvenile crime: evidence munity socioeconomic context. Annu. Rev.
from a randomizedhousing-mobilityexper- Sociol. 25:489-516
iment. Q. J. Econ. 116(2):665-79 Rosenbaum JE. 1995. Changing the geogra-
ManskiCF. 1993. Identificationof endogenous phy of opportunityby expanding residen-
social effects-the reflection problem. Rev. tial choice-lessons from the Gautreaux
Econ. Stud.60:531-42 program. Housing Policy Debate 6:231-
MarkowitzFE, Bellair PE, Liska AE, Liu J. 69
2001. Extending social disorganizationthe- RosenbaumJE, HarrisLE. 2001. Low-income
ory: modeling the relationships between families in theirnew neighborhoods.J. Fam.
cohesion, disorder, and fear. Criminology Issues 22(2):183-210
39(2):293-319 Ross CE. 2000. Neighborhood disadvantage
MasseyDS. 1996. The age of extremes:concen- and adult depression.J. Health Soc. Behav.
tratedaffluence and poverty in the twenty- 41:177-87
firstcentury.Demography33:395-412 Ross CE, Reynolds JR, Geis KJ. 2000. The
Mayer S, Jencks C. 1989. Growingup in poor contingentmeaning of neighborhoodstabil-
neighborhoods:How much does it matter? ity for residents' psychological well-being.
Science 243:1441-45 Am. Sociol. Rev.65:581-97
NEIGHBORHOODEFFECTS 477

RountreePW, Land KC. 1996. Perceived risk mental illnesses. Criminology 38:1043-
versus fear of crime: empirical evidence 74
of conceptually distinct reactions in surveySmallML, NewmanK. 2001. Urbanpovertyaf-
data.Soc. Forces 74:1353-76 terTheTrulyDisadvantaged:therediscovery
RountreePW,WamerBD. 1999. Social ties and of the family,the neighborhood,andculture.
crime: Is the relationshipgendered?Crimi- Annu.Rev.Sociol. 27:23-45
nology 37(4):789-813 SmithWR, FrazeeSG, Davison EL. 2000. Fur-
Sampson RJ. 1997. Collective regulation of theringthe integrationof routineactivityand
adolescent misbehavior: validation results social disorganizationtheories:smallunitsof
from Chicago neighborhoods.J. Adolescent analysis and the study of streetrobberyas a
Res. 12(2):227-44 diffusionprocess. Criminology38:489-523
Sampson RJ. 2001. How do communities un- Sobel M. 2001. Whatare neighborhoodeffects?
dergirdor underminehuman development? Paper presented at "Poverty Traps"Conf.,
Relevantcontexts and social mechanisms.In SantaFe Inst., SantaFe, NM
Does It Takea Village? CommunityEffects Sorensen AB. 1998. Theoretical mechanisms
on Children,Adolescents, and Families, ed. and the empiricalstudy of social processes.
A. Booth,N. Crouter,pp. 3-30. Mahwah,NJ: In Social Mechanisms: An Analytical Ap-
L Erlbaum proach to Social Theory, ed. P Hedstrom,
SampsonRJ, MorenoffJD, Earls F. 1999. Be- R Swedberg, pp. 238-66. Cambridge,UK:
yond social capital:spatialdynamicsof col- CambridgeUniv. Press
lective efficacy for children.Am. Sociol. Re-
South SJ, BaumerEP.2000. Decipheringcom-
view 64:633-60 munity and race effects on adolescent pre-
Sampson RJ, RaudenbushSW. 1999. System- maritalchildbearing.Soc. Forces 78:1379-
atic social observation of public spaces: a 1407
new look at disorderin urbanneighborhoods. Steptoe A, Feldman PJ. 2001. Neighborhood
Am. J. Sociol. 105(3):603-51 problems as sources of chronic stress: dev-
Sampson RJ, RaudenbushSW, Earls F. 1997. elopment of a measure of neighborhood
Neighborhoods and violent crime: a mul- problems, and associations with socioeco-
tilevel study of collective efficacy. Science nomic status and health. Ann. Behav. Med.
277:918-24 23(3):177-85
ScribnerRA, Cohen DA, Farley TA. 1998. A StiffmanAR, Hadley-IvesE, Elze D, Johnson
geographic relation between alcohol avail- S, Dore P. 1999. Impact of environmenton
abilityandgonorrhearates.Sex. Transm.Dis. adolescentmentalhealthandbehavior:struc-
25:544-48 turalequationmodeling.Am.J. Orthopsychi-
Seidman E, YoshikawaH, RobertsA, Chesir- atr.69:73-86
Teran D, Allen L, et al. 1998. Structural Suttles GD. 1972. The Social Constructionof
and experientialneighborhoodcontexts, de- Communities.Chicago:Univ. Chicago Press
velopmental stage, and antisocial behavior Upchurch DM, Aneshensel CS, Sucoff CA,
among urban adolescents in poverty. Dev. Levy-Storms L. 1999. Neighborhood and
Psychopathol. 10:259-81 family contextsof adolescentsexual activity.
Shumow L, VandellDL, Posner J. 1998. Per- J. MarriageFam. 61:920-33
ceptions of danger: a psychological medi- Veysey BM, MessnerSF. 1999. Furthertesting
ator of neighborhooddemographiccharac- of social disorganizationtheory:an elabora-
teristics.Am. J. Orthopsychiatry68(3):468- tion of Sampson and Groves's "community
78 structureand crime."J. Res. Crime Delinq.
Silver E. 2000. Extending social disorga- 36:156-74
nization theory: a multilevel approach to Warner BD, Rountree PW. 1997. Local so-
the study of violence among persons with cial ties in a community and crime model:
478 SAMPSON MORENOFF * GANNON-ROWLEY

questioningthe systemic natureof informal tors, andneighborhoodcontext.Criminology


social control.Soc. Prob. 44(4):520-36 38:1109-42
Welsh WN, Greene JR, Jenkins PH. 1999. Wilson JQ, Kelling G. 1982. The police and
School disorder:the influencesof individual, neighborhoodsafety: broken windows. At-
institutional,and communityfactors.Crimi- lantic Month. 127:29-38
nology 37(1):73-115 Wilson WJ. 1987. The TrulyDisadvantaged:
Wikstrom POH, Loeber R. 2000. Do disad- The Inner City, the Underclass, and Public
vantagedneighborhoodscause well-adjusted Policy. Chicago:Univ. Chicago Press
childrento become adolescent delinquents? WinshipC, MorganSL. 1999. The estimationof
A study of male juvenile serious offend- causaleffects fromobservationaldata.Annu.
ing, individual risk and protective fac- Rev. Sociol. 25:659-706

You might also like