You are on page 1of 14

Intern. J.

of Research in Marketing 24 (2007) 254 – 267


www.elsevier.com/locate/ijresmar

Calibration of consumer knowledge of the web


Kishore Gopalakrishna Pillai a,⁎, Charles Hofacker b,⁎
a
Department of Marketing, Leeds University Business School, University of Leeds, Leeds, LS2 9JT, United Kingdom
b
Department of Marketing, College of Business, Florida State University, Tallahassee, FL 32306, USA

Abstract

Calibration of consumer knowledge of the web refers to the correspondence between accuracy and confidence in knowledge of the web. Being
well-calibrated means that a person is realistic in his or her assessment of the level of knowledge that he or she possesses. This study finds that
involvement leads to better calibration and that calibration is higher for procedural knowledge and common knowledge, as compared to
declarative knowledge and specialized knowledge. Neither usage, nor experience, has any effect on calibration of knowledge of the web. No
difference in calibration is observed between genders. But, in agreement with previous findings, this study also finds that males are more confident
in their knowledge of the web. The results point out that calibration could be more a function of knowledge-specific factors and less that of
individual-specific factors. The study also identifies flow and frustration with the web as consequences of calibration of knowledge of the web and
draws the attention of future researchers to examine these aspects.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Calibration; Knowledge; Web; Involvement; Knowledge type

1. Introduction and subjective knowledge, which refers to consumers' percep-


tion of their own knowledge (Raju, Lonial, & Mangold, 1995).
Consumer knowledge has been a topic of considerable An aspect of knowledge that has been omitted from the
research interest in marketing for the past three decades (Alba & discourse in consumer research is knowledge calibration.
Hutchinson, 1987; Bettman & Park, 1980; Brucks, 1985; Knowledge calibration refers to the correspondence between
Mitchell & Dacin, 1996; Roy & Cornwell, 2004). The accuracy and confidence in knowledge. Notions of correct and
importance of consumer knowledge can be located in the fact incorrect knowledge tap into only the first dimension, i.e.,
that knowledge is central to understanding consumer behaviors accuracy. A person can be well calibrated even when possessing
such as information search and information processing. inaccurate knowledge. Being well-calibrated means that a
Researchers have examined the antecedents and consequences person is realistic in his or her assessment of the level of
of prior knowledge of consumers (Alba, 1983; Chase & Simon, knowledge that he or she possesses. This, in turn, is likely to
1973) and the different types of consumer knowledge such as facilitate appropriate actions such as information search or
product class knowledge, price knowledge, knowledge of the optimal decision making. Hence, examination of this construct
World Wide Web, etc. (Brucks, 1985; Estelami, Lehmann, & is likely to unravel processes that have not been brought to light
Holden, 2001; Page & Uncles, 2004). Researchers have also by the research on consumer knowledge. The latter construct
attempted to distinguish between objective knowledge, which has attracted considerable research, while research on calibra-
refers to the absolute knowledge possessed by the consumer, tion of consumer knowledge is relatively new. Alba and
Hutchinson (2000) highlighted the importance of this construct
and its implications in consumer research and called for
research in this area. This paper responds to the call by
⁎ Corresponding authors. K.G. Pillai is to be contacted at tel.: +44 113
examining the calibration of consumer knowledge of the web, a
3432636. C. Hofacker, tel.: +1 850 6447864. topic that is of considerable relevance, given the increasing
E-mail addresses: kgp@lubs.leeds.ac.uk (K.G. Pillai), importance of the web for consumers. The paper proposes a
Chofacker@cob.fsu.edu (C. Hofacker). model of calibration of knowledge of the web and examines the
0167-8116/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijresmar.2007.02.001
K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267 255

antecedent roles of involvement with the web, experience, 1974; Murphy & Winkler, 1977), whereas doctors' diagnoses
usage, knowledge type, and gender on calibration of consumer have not (Christen-Szalanski & Busyhead, 1981; Oskamp,
knowledge of the web. 1962). Explanations for this disparate finding include (a) the
The paper's contribution is threefold. One, it seeks to repetitiousness of forecasting with immediate, unambiguous
empirically examine knowledge calibration in the consumer feedback which enables validation and correction, facilitating
domain. Second, in doing so, it adds a new dimension to the calibration (Lichtenstein, Fischoff, & Philips, 1982), (b) dif-
research on knowledge and usage of the web. Third, it contrib- ferent payoff matrices or loss functions for forecasters and
utes to the larger stream of research on knowledge calibration doctors, whereby doctors perceive the penalty for false negative
through the examination of knowledge type as an antecedent of as more devastating and hence state their opinions with greater
calibration. conviction (Keren, 1991; O'Connor, 1989). The several studies
We start by introducing the concept of knowledge calibra- across various domains point to the fact that high calibration is
tion, followed by a brief review of the literature on knowledge rarely achieved and moderate levels that include some degree of
calibration. The hypotheses for the study are presented after- systematic bias, usually overconfidence, are the norm (Pillai &
wards. Subsequently, the method and the results of the study Goldsmith, 2006).
are presented. The results indicate the antecedent effects of Noteworthy among other domains that have examined the
involvement and knowledge type in enhancing calibration. issue of the confidence–accuracy relationship is forensic
Broadly, the results imply that calibration of knowledge of psychology. In criminal investigations, witnesses often attempt
the web in particular and calibration of knowledge in general to identify a culprit and give a statement about the certainty
could be determined by knowledge variables such as types of of the identification, expressed as a confidence judgment. It
knowledge. The study discounts the role of individual specific has been demonstrated that a confident witness is regarded as
factors, other than involvement, as antecedents of knowledge a credible witness, with a great impact on the outcome of a
calibration. trial (Olsson, 2000). Yet many experimental studies of witness
identification have shown that the confidence–accuracy
2. Literature review and conceptual framework relationship is not reliably different from zero (Loftus, 1974;
Narby, Cutler, & Penrod, 1996). Consequently, the issue of
2.1. Knowledge calibration calibration or “realism in confidence assessments” has attracted
considerable scholarly attention in this field.
As defined earlier, calibration of knowledge refers to the As mentioned earlier, this study is an attempt to apply the
correspondence between accuracy and confidence in knowl- construct of knowledge calibration to consumer research. In this
edge. High accuracy and high confidence together lead to high study, we examine the calibration of consumers' knowledge of
calibration. Similarly, low accuracy and low confidence lead the web. The latter construct is dealt with in the following
to high calibration. Lack of correspondence between the two section.
lead to low calibration; thus, low accuracy along with high
confidence (overconfidence), and high accuracy along with low 2.2. Consumers' knowledge of the web
confidence (underconfidence) are instances of poor calibration,
termed as miscalibration. The following diagram, adapted from As noted earlier, examination of consumer knowledge has
Goldsmith and Pillai (2006), captures the above distinctions in a a rich tradition in the marketing literature. Specifically, the
2 × 2 matrix. influence of product knowledge on consumers' information
Several scientists, including statisticians, meteorologists, and processing has been of great interest to researchers (e.g., Alba &
psychologists, have sought to address the measurement and Hutchinson, 1987; Bettman & Park, 1980; Brucks, 1985).
explanation of judgments of confidence and their relation to Product knowledge has been found to be influential in con-
accuracy (e.g., Gigerenzer, Hoffrage, & Kleinbolting, 1991; sumers' product evaluations (Blair & Innis, 1996), precision of
Harvey, 1997; Yates, 1990). A recurring finding from many of price estimates (Biswas & Sherrell, 1993), and processing of
these studies is that people are systematically overconfident advertising messages (Maheswaran & Sternthal, 1990). Re-
about the accuracy of their knowledge and judgment (Pillai search on consumers' knowledge of the web is relatively recent,
& Goldsmith, 2006). Overconfidence can be defined as the as the web itself is relatively recent.
overestimation of the probability of an event. Thus, in a knowl- During the past decade, pre-existing firms have sought to
edge assessment task, if the subject estimates that 70% of his/ realize benefits for themselves and for their customers online by
her answers are correct, but only 50% are actually so, the providing alternative purchasing channels, inexpensive self-
subject is overconfident. service options, complementary services, longer product lines,
Reviewing research on calibration, Spence (1996) states that recommendation agents and interactive advertisements. Other
much research indicates that individuals are overconfident, firms have sprung up to provide new services like online auc-
though certain professions have displayed high calibration in tions, interactive driving directions, video upload and exchange
judgments. For example, weather forecasters and odds-makers and e-mail. While we tend to think of virtual channels as being
(those who calculate and set betting odds based on the less expensive than physical channels, these new interfaces
prediction of the result of a contest, such as a horserace or an have their own costs in terms of the consumer knowledge that
election) have been shown to be well calibrated (Hoerl & Fallin, is required. Not all consumers are ready for this onslaught of
256 K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267

technology (Parasuraman, 2000). The examination of consumer


knowledge of the web and the calibration of such knowledge
becomes relevant in this context.
Knowledge has been defined as the body of facts and prin-
ciples (i.e., information and understanding) accumulated by
mankind (i.e., stored in memory) about a domain (Delbridge
& Bernard, 1998). It follows that consumers' knowledge of
the web refers to the information and understanding of the
web stored in the consumers' memory. Examples of consumer
knowledge of the web include information stored in a consum-
er's memory about how to use a search engine, what cookies
are, or the benefit of ‘what's new' links.
Researchers in the past have used proxy measures to infer
consumer knowledge of the web, such as number of hours
spent online or months and years that the medium has been
used (Diaz, Hammond, & McWilliam, 1997). Self-perception Fig. 1. Accuracy, confidence, and calibration matrix.
judgments of Web skill have also been used to measure
consumer skill on the web (Novak, Hoffman, & Yung, 2000).
Recently, Page and Uncles (2004) proposed a scale to measure knowledge on the one hand and common and specialized
consumers' web knowledge content. Because the vast array knowledge on the other. The selection of the antecedents was
of knowledge that consumers could possess about the web determined, in part, by the research focus on these constructs.
imposes limitations on valid measurement, these authors For example, involvement is a variable that has attracted con-
restricted their focus to knowledge arising from machine siderable attention as a determinant of consumer behavior. As
interactivity of the web. While person-to-person interaction discussed below, experience, usage, and gender have also been
occurs frequently online, we also restrict our study to machine shown to be of relevance to the stream of research on knowl-
interactivity (MI), defined as “the extent to which users can edge. The antecedent role of knowledge type is one of the key
participate in modifying the form and content of a mediated contributions of this study, and we chose those distinctions
environment in real time” (Steuer, 1992, p. 84). The focus on that are of considerable import to the study of web knowledge.
machine interactivity is justified by the fact that e-services, Fig. 2 shows the model of the antecedents of calibration of
online retailing and other self-service technologies, as well as consumer knowledge of the web. We start with the discussion of
web advertising, all involve interaction between the consumer individual-specific factors.
and technology, rather than human-to-human interaction.
As detailed in the Measures section, we use the scale 2.3.1. Involvement
proposed by Page and Uncles (2004) to measure consumers' Involvement has been defined in terms of personal rele-
knowledge of the web. In the following paragraphs, we discuss vance of the product or the advertisement to the subject.
each of the antecedent variables of the study and offer propo- Zaichkowski (1985) defines involvement as a person's per-
sitions. A model of antecedents of calibration of knowledge is ceived relevance of the object based on inherent needs, values,
proposed. While the model is not comprehensive, many of the and interests. Involvement is generally considered to be a
more important variables that affect calibration are represented. function of (a) individual characteristics such as interests,
This is followed by a test of the model. values, and goals, (b) situational factors such as the purchase
It should be noted that the hypotheses discussed in the occasion or perceived risk, and (c) characteristics of the object
following sections are derived from the interpretation of knowl- or stimulus (Laurent & Kapferer, 1985). Outcomes associated
edge calibration as the realism of perceptions on knowledge. with high involvement include more time and effort spent in
The four quadrants presented earlier (see Fig. 1) are not dealt search-related activities (Bloch, Sherrel, & Ridgway, 1986),
with separately, for two reasons. The first is that accuracy (at an more extensive decision making, greater perceived differences
aggregate, between subjects, level) and confidence are contin- in product attributes, and a greater likelihood of establishing
uous variables, and the dichotomous division was intended brand preferences (Zaichkowski, 1985).
for conceptual clarification. More importantly, respondents in The effects of involvement on the processes mediating the
the study exhibited high levels of accuracy, rendering the effects of a persuasive communication about an issue have been
distinction between the two types of calibration redundant. extensively investigated (Maheswaran & Sternthal, 1990;
Miniard, Bhatla, Lord, Dickson, & Unnava, 1991; Okechuku,
2.3. Antecedents of calibration of knowledge of the web 1992; Verlegh, Steenkamp, & Meulenberg, 2005). A robust
finding of this literature is that uninvolved message recipients,
We discuss two different types of antecedent variables: compared to those who are highly involved, are characterized
(a) individual-specific factors such as involvement, experience, by (a) less attention to and less cognitive elaboration of attribute
usage, and gender, and (b) knowledge-specific factors which information, and (b) more reliance on peripheral cues available
includes the distinctions between declarative and procedural in the message (Celsi & Olson, 1988; Batra & Ray, 1985; Petty,
K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267 257

Fig. 2. Model of antecedents of calibration of knowledge of the web. Dashed arrows indicate that no formal hypotheses are developed, but the relationships are tested as
important empirical issues.

Cacioppo, & Schumann, 1983). Starting with Ray et al. (1973), a testing task where accuracy and confidence are measured, as
large numbers of studies have reported that consumers involved in the case of this study. More involved consumers are likely to
in a situation or product are more active processors of cognitive collect more information about the product, and will be more
information about it. Celsi and Olson (1988) tested several motivated to form realistic and correct notions regarding their
hypotheses concerning the effects of felt involvement on the knowledge of the domain. Hence,
amount of attention and comprehension effort, the focus of
H1. Higher the involvement with the world wide web, higher
attention and comprehension processes, and the extent of
the calibration of knowledge of the web.
cognitive elaboration during comprehension. These authors
found that involved consumers attend to and comprehend more
information about shopping situations and products and 2.3.2. Cumulative online experience
produce more elaborate meanings and inferences about them Consumers' ability to understand and represent web-based
(Celsi & Olson, 1988). information is structured and constrained according to their
Research in memory-based product judgments has shown existing domain experience (Moreau, Lehmann, & Markman,
that individuals are cognitive misers who attempt to minimize 2001), which is driven by their tenure online, or duration of time
search effort (Wyer & Srull, 1986), and involvement during a they have used the web. Consumer experience may include both
memory-based judgment influences this processing objective direct experiences such as information searching, evaluation,
and affects search intensity (Park & Hastak, 1994). Compared to purchase, and consumption of interactive products, and indirect
uninvolved consumers, involved consumers are more motivated experiences such as advertising exposure and the observation of
to form a relatively accurate judgment and hence they search others' consumption (Alba & Hutchinson, 1987).
more intensely for judgment-relevant information. Considerable evidence suggests that cumulative consumer
The above discussion highlights the role of involvement in experience increases knowledge (e.g., Park, Mothersbaugh, &
enhancing effort by consumers. Involvement can also be Feick, 1994), and thus some researchers measured experience as
expected to increase familiarity with the domain, and practice a proxy of consumer knowledge (e.g., Bettman & Park, 1980).
in dealing with specific issues concerning the domain. Effort Alba and Hutchinson (1987) proposed that consumer knowl-
and practice are cues associated with performance, which refers edge consists of two dimensions — familiarity and expertise.
to knowledge accuracy and correct assessment of confidence in The existing literature indicates that domain experience is
258 K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267

closely related to both familiarity and expertise. Consumer clear, and the individual's abilities are up to the challenge, then
experience is considered as a direct antecedent to familiarity he or she becomes involved in the activity and is intrinsically
(Alba & Hutchinson, 1987), and research has shown that motivated.
experience is related to expertise as well (Mitchell & Dacin, Thus, it can be seen that TAM posits a goal-directed,
1996). We argue that the effect of experience on familiarity extrinsically motivated route to usage of the web, while the flow
will be stronger than the effect on expertise, leading to the theory argues the case of an intrinsically motivated path. We
feeling of knowing. Of course expertise should increase as a argue that the extrinsically motivated route is likely to operate in
function of cumulative time logged online, but not nearly as a manner similar to experience. That is, higher usage will lead
fast as their sense of familiarity. Research on the concept of to people developing a feeling of knowing, more than what
confidence that has examined practice – a variable closely is warranted by an actual increase in knowledge. While the
related to experience – supports this argument. This research accuracy increases, calibration does not. But for the intrinsically
suggests that practice produces dissociation between confidence motivated people, higher usage will lead to, as well as follow
and performance because of elevations in confidence that are from, higher involvement, and this is likely to lead to more
unaccompanied by corresponding increases in accuracy (Paese realistic assessments of their knowledge of the web; i.e., better
& Sniezek, 1991). Similar themes were echoed by Fischoff and calibration. Examining the disparate effects of different types of
Slovic (1980), who argued that even minimal familiarity with motivation is beyond the scope of this research. Previous
a task or tentative rules of performance in the task can produce research has noted that some people are motivated by extrinsic
surprising amounts of conviction, and by Heath and Tversky factors while others by intrinsic factors (Darley, 1999), and
(1991), who noted that familiarity and knowledge are accom- often both coexist in individuals (Roberts, Hann, & Slaughter,
panied by feelings of competence. In the information systems 2006). Given the null effect from extrinsically motivated indi-
field, research has recorded the role of experience in increasing viduals and the positive effect from intrinsically motivated
users' confidence in their ability to master and use computers individuals, we expect a weak positive effect of usage on
supporting their task (DeLone, 1988). Based on these, we calibration of consumer knowledge.
reason that experience will be positively related to confidence
H3. Higher the usage, higher the calibration of knowledge of
in knowledge, but will not be related to calibration, since it
the web.
produces a dissociation between confidence and accuracy.
H2. Higher the experience, higher the level of confidence in
2.3.4. Gender
knowledge.
Research has recorded the role of gender on confidence.
Males have been found to have higher confidence than females
2.3.3. Weekly usage in various domains of knowledge (Estes & Hosseini, 1988;
Here, usage refers to the extent of current weekly use of the Goldsmith & Goldsmith, 1997; Kalaian & Freeman, 1994).
World Wide Web. The dominant paradigm that informs us of Estes and Hosseini (1988) examined confidence in investment
adoption of the web is the technology acceptance model (TAM). decisions in the stock market and found that men are more
Recent research has employed TAM to explain continued usage confident than women. Sutton's (1987) research on the illusion
of the web (Kim & Malhotra, 2005). In the context of web of control in insurance also corroborates these findings. Sutton
usage, TAM suggests that there exists a direct and positive (1987) found that women are better prospects than men for auto
effect between attitude towards web usage, usage intention, and insurance because of a more realistic perception of their risk of
actual usage. Perceived usefulness and ease of use determine the auto accidents. We generalize from these findings and expect
attitudes toward using the web. In turn, usage intentions are males to have higher confidence in their knowledge of the web.
determined by these attitudes and perceived usefulness. Finally, Also, previous research has noted a gender gap in computer
usage intentions lead to usage. These relationships have found literacy in which males are more literate than females (Chen,
strong empirical support (Davis, 1993; Venkatesh & Davis, 1986; Sachs & Bellisimo, 1993). It has also been found that, in
2000). However, a significant body of theoretical evidence general, men tend to have more favorable attitudes toward IT
underscores the importance of intrinsic motivators in web (Schumacher & Morahan-Martin, 2001; Simon, 2001). Some
usage. Researchers have become increasingly aware of the research has also noted that women tend to be less skilled in IT
relevance of the non-extrinsic motives of use such as intrin- (Harrison & Rainer, 1992). In the light of this substantial body
sically enjoyable experiences (i.e., flow) in understanding of finding, we expect males to have higher accuracy than
attitudes and behaviors. Flow is defined as the holistic sensation females in their knowledge of web. Higher accuracy along with
that people feel when they act with total involvement higher confidence results in no difference in calibration between
(Csikszentmihalyi, 1977). Flow is a positive, highly enjoyable genders. Hence,
state of consciousness that occurs when our perceived skills
H4. Males will have higher confidence in their knowledge of
match the perceived challenges we are undertaking. When this
the web, compared to females.
occurs, an individual derives intrinsic enjoyment from the ac-
tivity and tends to continue with it. If the task is too easy, the Following the examination of individual-specific factors, we
person becomes bored. If the work demands skills beyond the examine the antecedent effects of knowledge-specific factors
capabilities of the individual, anxiety is created. When goals are (refer Fig. 2).
K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267 259

2.3.5. Knowledge type — procedural versus declarative known information of the domain of interest required to perform
Based on the literature from cognitive sciences and mar- general and common domain related tasks successfully. Con-
keting, a distinction can be drawn between procedural and versely, specialized knowledge is defined as skilled and/or
declarative knowledge. Declarative knowledge is defined as the extraordinary information about facts, terms, attributes, and
‘factual information that is somewhat static in nature which is dynamic information underlying skillful actions about a domain
usually describable' (Best, 1989, p. 7); for instance, information required to perform skilled domain related tasks successfully
about the attributes, terminology, evaluative criteria, facts, and (Page & Uncles, 2004). The definition of common knowledge is
usage situations of the domain of interest (Brucks, 1986). a non-technical one and should be distinguished from the
Procedural knowledge, by contrast, refers to the dynamic infor- technical definition of the term in economics and game theory,
mation underlying skillful actions (Best, 1989; p 7); that is, the where it denotes a piece of information that is available to
knowledge of rules or procedures for taking action believed to agents A and B, and both A and B know that the other has the
be stored and organized into production systems (Brucks, 1986). information, and both A and B know that the other knows that
Anderson (1983) discussed the impact of these different the other has the information, and so on ad infinitum (Lewis,
types of knowledge content on behavior, suggesting that 1969).
although both types can guide behavior, procedural knowledge Given the widespread usage of the web, consumers are likely
content is considered to be of greater influence on actual to possess a high degree of common knowledge. They are
behavior. For example, to send an e-mail, knowing how to send also likely to have higher familiarity with common knowledge,
an e-mail (procedural) will have more influence on the final enabling realistic assessments of confidence. Specialized knowl-
behavior than knowing what an e-mail is (declarative). edge questions are more technical in nature and consumers
Research on procedural and declarative information in soft- are likely to be less familiar with these. Also, given the wide-
ware manuals have found that users prefer procedural to declar- spread use of computers and the web, consumers could have
ative information (Ummelen, 1997). Such preference for been exposed to some of these technical aspects during formal
procedural information is likely to result in greater familiarity training or in their own usage of the web. We posit that the
with procedural information and better realism in perceptions modicum of familiarity arising from such exposure leads to a
regarding procedural knowledge, compared to declarative perception that they know these topics, even when they do not.
knowledge. Research on the feeling of knowing has recorded how such an
Procedural knowledge has been recognized as a link between intuition often produces dissociation between accuracy and
experience and performance (e.g., Bonner & Walker, 1994). confidence (Koriat, 1998; Nelson et al., 1986). Such dissocia-
Anderson's (1983) ACT⁎ theory (ACT is not an acronym, tions will lead to a miscalibration of knowledge. It is argued that
though it has been translated as Adaptive Control of Thought), this phenomenon is more likely in the case of specialized
a general theory of cognition that focuses on memory processes, knowledge than in the case of common knowledge, leading to
offers a view that practice of a cognitive task results in pro- higher calibration in the latter, compared to the former. Hence,
cedural knowledge, which enhances performance. In the con-
H6. Calibration will be higher for common knowledge,
text of knowledge of the web, practice will also help verify the
compared to specialized knowledge.
correctness or incorrectness of the knowledge, since incorrect
knowledge will hinder effective navigation of the web. That is,
web users operate in an environment where immediate and real- 3. Method
time feedback on their procedural knowledge is available. Such
a feedback is not available in the case of declarative knowledge 3.1. Sample
of the web. Research has recorded the role of feedback in
improving calibration (e.g., Lichtenstein et al., 1982). Higher We used two samples to test the hypotheses. Sample 1
calibration has been observed in certain professions, such as comprised 151 students enrolled in undergraduate business
weather forecasting, where immediate feedback is available. courses in a large university in the southeastern United States,
Given the availability of feedback for procedural knowledge of who participated in the study for course credit. Sample 2
the web, we posit that web users will be better calibrated in their comprised 153 adults in a British city. The adults included
procedural knowledge compared to declarative knowledge. (a) community members (40%) who were contacted in libraries,
Therefore, parks, residential apartments, and retail stores, (b) nonacademic
staff members (25%) and faculty members (5%) in a leading
H5. Calibration will be higher for procedural knowledge of the
British university and (c) postgraduate and undergraduate stu-
web, compared to declarative knowledge.
dents in the same university (30%). All respondents were
contacted individually and were paid 2 pounds for participation.
2.3.6. Knowledge type — common versus specialized The respondents filled out a structured questionnaire that
Knowledge literature also distinguishes between experts measured consumer knowledge, involvement with the web,
and novices, and the familiar and the unfamiliar (Alba & experience, and usage. In order to control for order effects,
Hutchinson, 1987; Page & Uncles, 2004). This distinction leads roughly half of the sample answered questionnaires which
to two types of knowledge — common and specialized. had procedural knowledge questions first, followed by declar-
Common knowledge can be defined as general and/or publicly ative knowledge questions, while the other half answered
260 K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267

questionnaires which had declarative knowledge questions first, and confidence rating for each item provided the measure
followed by procedural knowledge questions. Common and of calibration at the respondent level. In keeping with the
specialized knowledge questions were distributed within these. definition of calibration as ranging from 0 (nil calibration) to 1
We conducted t-tests to assess differences between the (perfect calibration), we restricted the lower bound of the
samples. No difference was observed in mean calibration correlation measure to zero. That is, negative correlations
( p = .58) while a difference was observed in confidence observed in a few cases (which were not statistically significant)
( p = .00). The latter could be attributed to the difference in age, were restricted to zero.
in that the first sample comprised only undergraduate students,
whereas the second comprised a distribution of adults of varying 3.2.2. Involvement
age groups. A simple regression model in the second sample Involvement was measured using the ten-item scale pro-
showed a significant effect (beta = −.21; p = .01) of age on posed by Zaichkowsky (1994). Respondents were asked to
confidence. Moreover, regression models between confidence indicate their perceptions of the web on a 7-point bipolar scale
and the independent variables showed comparable beta param- anchored by adjectives such as important–unimportant, fasci-
eter estimates and significances across the two samples. Con- nating–mundane, relevant–irrelevant, etc. Involvement was
sequently, we decided to combine the two samples, with sample measured as the mean of the ten items, after recoding the
included as a covariate. Results are reported for the combined negatively framed items.
sample as well as the separate samples. Given the similarity
between American and British cultures, and the technical nature 3.2.3. Experience
of the questions, cross cultural issues are discounted. Experience was measured by a single item that asked
Information on gender was missing from 28 questionnaires respondents to specify the number of years they have been using
in the first sample and hence the sample size for multivariate World Wide Web.
tests was correspondingly lower. In addition, in some cases web
involvement could not be computed due to missing data, which 3.2.4. Usage
also resulted in a few dropouts. When calibration was computed Usage was measured by a single item that asked respondents
for subscales, there were a few instances where correlation to specify the number of hours per week they spend browsing or
could not be computed because one variable was constant, using the web. It should be noted here that online researchers
which resulted in slight reductions in sample sizes for the have employed single-item measures to measure experience/
specific analyses using the subscales. usage (e.g., Rodgers, Negash, & Suk, 2005). Also, the use of
self-reported usage measures have been validated by research
3.2. Measures that suggests that self-reported usage measures correlate well
with actual usage measures (Venkatesh & Davis, 2000).
3.2.1. Consumer knowledge calibration
As mentioned earlier, consumer knowledge of the web was 4. Results
measured using a 38-item scale developed by Page and Uncles
(2004). Based on qualitative and quantitative analyses, these 4.1. Scale reliability
authors developed four scales that measure common declarative
knowledge, specialized declarative knowledge, common pro- The ten-item involvement scale had acceptable reliability
cedural knowledge, and specialized procedural knowledge. The (alpha = .875). The reliability of the knowledge of the web scale,
number of items for each of the aforementioned scales was 10, computed using KR 20, fell below the threshold of .7 (overall
11, 6, and 11. Most of the items employed a true/false format to .52; specialized = .83; common = .13; procedural = .35; declara-
elicit responses, while a few had multiple choice responses. tive = .39). This is not deemed to be a cause for concern due to
A combined scale, comprising all 38 items (21 declarative the following reasons: (a) the construct of knowledge of the web
knowledge items and 17 procedural knowledge items) was used is necessarily multidimensional and a scale that attempts to tap
to measure consumer knowledge of the web in this study. The into the various dimensions will have to compromise on
few multiple choice items (3 items) were converted to dichoto- reliability indices and (b) the focal construct is calibration of
mous (true/false) format to ensure correspondence. Respon- knowledge of the web and not knowledge of the web per se. As
dents indicated their choice of answer for each item, and then long as the scale has face and content validity, it serves to be an
indicated their confidence that the answer is correct, on a scale acceptable instrument to measure calibration. It should also be
ranging from 50% to 100%. We chose a restricted scale ranging noted that previous research on consumer knowledge has used a
from 50% to 100%, as against a scale ranging from 0% to 100%, set of items chosen judgmentally by researchers, for which
because of the dichotomous response format (true or false) reliabilities have not been reported (e.g., Capraro, Broniarczyk,
which provides a 50% chance of success even in a random & Srivastava, 2003).
choice.
Knowledge calibration was measured through point biserial 4.2. Descriptive statistics
correlation between accuracy and confidence (e.g., Lin, Moore,
& Zabrucky, 2001). Calibration was measured at the within Table 1 shows the descriptive statistics. Calibration ranged
subjects level. Correlation between accuracy (either 0 or 1) from 0 to .72, with a mean of .34. Mean level of involvement
K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267 261

Table 1
Descriptive statistics — means and standard deviations
Overall N American N British N Overall mean (S.D.) American mean (S.D.) British mean (S.D.)
Calibration 304 151 153 .34 (.16) .34 (.16) .33 (.15)
Web involvement 295 149 146 6.15 (.84) 6.32 (.71) 5.97 (.93)
Experience 302 150 152 8.31 (2.52) 8.66 (2.15) 7.97 (2.80)
Usage 301 150 151 11.26 (7.85) 12.45 (8.04) 10.07 (7.51)
Accuracy 304 151 153 80% (8%) 81% (7%) 79% (9%)
Confidence 304 151 153 81% (9%) 83% (8%) 79% (10%)

with the web was 6.15, indicating a high level of involvement lambda = .94; F = 3.98; p = . 02; b = .78; p = .02; British —
with the web. Experience with the web ranged from 1 to Wilks' lambda = .96; F = 3.25; p = .04; b = .62; p = .02) and
15 years, with a mean of 8.3 years. Usage ranged from 1 to 40 h found no relationship between experience and calibration
per week, with a mean of 11.26 h. Accuracy ranged from 55% to (American — Wilks' lambda = .94; F = 3.98; p = .02; b = .01;
97%, with an average of 80%. Hence, the sample exhibited high p = .16; British — Wilks' lambda = .96; F = 3.25; p = .04;
levels of accuracy. This finding leads to the discounting of b = .00; p = .65).
instances of calibration arising from low accuracy and low The third hypothesis stated that usage will have a positive
confidence. effect on calibration. The overall sample did not support this
hypothesis (Wilks' lambda = .92; F = 10.92; p = .00; b = .00;
4.3. Tests of hypotheses p = .17). The American sample also did not support the
hypothesis (Wilks' lambda = .98; F = 1.36; p = .26; b = .00;
We employed a general linear model procedure to test the p = .95). The British sample showed an effect, albeit in the
hypotheses, with calibration and confidence as the dependent negative direction (Wilks' lambda = .84; F = 12.73; p = .00; b =
variables and web-involvement, experience, usage, and gender − .004; p = .04).
as the independents. Sample was a covariate, when the overall Hypothesis 4 stated that males will have higher confidence in
sample was tested. Comparison of means was employed to test their knowledge of the web, compared to females. We also
hypotheses 5 and 6. examined the relationship between gender and calibration as
The first hypothesis stated that higher involvement will lead an empirical issue. The overall sample supported H4 (Wilks'
to higher calibration of knowledge. This hypothesis was sup- lambda = .84; F = 24.71; p = .00; b = − 6.65; p = .00) and found
ported by the overall sample (Wilks' lambda = .96; F = 5.11; no relationship between gender and calibration (Wilks'
p = 01; b = .03; p = .03) and the American sample (Wilks' lambda = .84; F = 24.71; p = .00; b = .02; p = .29). Similar effects
lambda = .93; F = 4.13; p = .02; b = .05; p = .03), but not by the were obtained in American (H4 — Wilks' lambda = .90;
British sample (Wilks' lambda = .97; F = 1.94; p = .15; b = .01; F = 6.65; p = .00; b = − 4.73; p = .00; gender–calibration —
p = .33). Table 2a shows the multivariate test results (Wilks' Wilks' lambda = .90; F = 6.65; p = .00; b = .00 p = .96) and British
lambda, F, and p values), while Table 2b shows the parameter samples (H4 — Wilks' lambda = .79; F = 18.42; p = .00; b =
estimates and their p values. − 8.12; p = .00; gender–calibration — Wilks' lambda = .79;
The second hypothesis stated a positive relationship between F = 18.42; p = .00; b = .03; p = .24). It is to be noted that gender
experience and confidence. Though no hypothesis was offered, was entered as a dummy variable in the general linear model
we examined the relationship between experience and calibra- with values of 1 for males and 2 for females.
tion. The overall sample supported H2 (Wilks' lambda = .96; The fifth hypothesis stated that calibration will be higher for
F = 6.16; p = .00; b = .69; p = .00), while no relationship was procedural knowledge compared to declarative knowledge. We
found between experience and calibration (Wilks' lambda = .96; computed calibration scores for declarative and procedural
F = 6.16; p = .00; b = .00; p = .69). Similarly, American and subscales for each respondent. This was done by measuring the
British samples also supported H2 (American — Wilks' correlations between accuracy and confidence in the subscales

Table 2a
Multivariate tests for H1, H2, H3, and H4
Effect Overall sample American sample British sample
Value F Error df Sig. Value F Error df Sig. Value F Error df Sig
Intercept Wilks' Lambda .43 172.68 259 .00 .44 73.37 116 .00 .40 101.98 138 .00
Experience Wilks' Lambda .96 6.16 259 .00 .94 3.98 116 .02 .96 3.25 138 .04
Usage Wilks' Lambda .92 10.92 259 .00 .98 1.36 116 .26 .84 12.73 138 .00
Gender Wilks' Lambda .84 24.71 259 .00 .90 6.65 116 .00 .79 18.42 138 .00
Web involvement Wilks' Lambda .96 5.11 259 .01 .93 4.13 116 .02 .97 1.94 138 .15
Sample Wilks' Lambda .96 5.44 259 .01
262 K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267

Table 2b
Parameter estimates
Overall sample American sample British sample
Dependent variable Parameter B Standard error t Sig. B Standard error t Sig. B Standard error t Sig.
Calibration Intercept .16 .08 1.94 .05 − .04 .14 −.26 .79 .26 .09 2.79 .01
Experience .00 .00 .40 .69 .01 .01 1.41 .16 .00 .01 − .46 .65
Usage − .00 .00 − 1.39 .17 .00 .00 −.07 .95 .00 .00 − 2.11 .04
Gender .02 .02 1.05 .29 .00 .03 .05 .96 .03 .03 1.17 .24
Web involvement .03 .01 2.15 .03 .05 .02 2.20 .03 .01 .01 .98 .33
Sample − .00 .02 − .114 .91
Confidence Intercept 75.02 4.04 18.57 .00 72.45 5.96 12.16 .00 73.21 5.12 14.29 .00
Experience .69 .19 3.51 .00 .78 .33 2.39 .02 .62 .25 2.44 .02
Gender − 6.65 .96 − 6.93 .00 −4.73 1.29 −3.66 .00 − 8.12 1.40 − 5.79 .00
Sample − 3.22 .98 − 3.30 .00

for each respondent. The hypothesis was tested by comparing Results were similar to those for the entire scale for hypotheses
the mean calibration of procedural knowledge (.38) among all 2–4, as well as the empirical relationships between (a) expe-
respondents with the mean calibration of declarative knowledge rience and calibration and (b) gender and calibration. Regarding
(.31). A paired sample t-test indicates that the difference is hypothesis 1, it was found that web involvement led to high
significant (p = .00). Table 3 shows the test results. These show calibration in the cases of declarative (b = .03; p = .04) and
that respondents are better calibrated in their procedural specialized knowledge (b = .04; p = .01), but not in the cases of
knowledge compared to their declarative knowledge. The procedural (b = .02; p = .33) and common knowledge (b = .00;
American sample also supported the hypothesis (p = .00), p = .89).
while the British sample fell marginally short of significance Since it could be argued that the constructs of experience
( p = .055). (number of years the person has been using the web) and usage
The sixth hypothesis stated that calibration will be higher (number of hours spent browsing the web) are inherently
for common knowledge compared to specialized knowledge. related, we ran the model with (a) calibration and (b) confidence
Similar to the procedure mentioned above, paired sample t-tests as dependents, and (a) web involvement, (b) gender and (c) the
were used to test the hypothesis. The overall sample supported product of experience and usage as independents, with sample
the hypothesis (mean calibration common = .39; special- as the covariate. The results remained the same. Only web
ized = .28; p = .00). The American as well as the British samples involvement was related to calibration (p = .04).
provided similar effects (American p = .00; British p = .01)
(Table 4). 5. Discussion
To throw further light on the gender differences, we also
examined the differences in accuracy. It was found that the We organize the discussion section into two parts. In the first
accuracy levels are higher for males compared to females part, we continue the discussion of the model, examining some
(p = .00). Hence, although no difference was observed in potential antecedents and, more importantly, the consequences
calibration, the results lend support to previous findings regard- of calibration of consumer knowledge of the web. In the second
ing gender differences in accuracy and confidence. part, we examine the implications of the study and outline future
We also examined whether the hypothesized relationships research directions.
between individual-specific factors and calibration and confi-
dence differ between the various types of knowledge. Four 5.1. Extension of the model-antecedent factors
models, where calibration and confidence were computed for
the subscales (procedural knowledge, declarative knowledge, Several other antecedents of consumer knowledge of the web
common knowledge, and specialized knowledge), were exam- can be identified. Two potential antecedents are distorted
ined for the entire sample (including sample as a covariate). memory and counterfactual thinking. Notions of “availability,”

Table 3
Comparison between calibration of procedural and declarative knowledge results of t-test
Mean (S.D.) N Paired differences t df Sig.
Mean S.D. s.e. (mean)
Calibration — procedural O — .38 (.24) O — 286
A — .41(.27) A — 149 O .07 .29 .08 4.44 285 .00
B — .35(.21) B — 137 A .11 .32 .03 4.14 148 .00
Calibration—declarative O — .31 (.19) O —286 B .04 .26 .02 1.94 136 .055
A — .30 (.19) A — 149
B — .31 (.18) B — 137
O — Overall; A — American; B — British.
K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267 263

Table 4
Comparison between calibration of common and specialized knowledge results of t-test
Paired differences t df Sig.
Mean (S.D.) N Mean S.D. s.e. (mean)
Calibration — common O — .39 (.27) O — 279
A — .41 (.28) A — 134 O .10 .32 .02 5.38 278 .00
B — .36 (.25) B — 145 A .13 .32 .03 4.83 133 .00
O —.28 (.18) O — 279 B .07 .31 .03 2.78 144 .01
Calibration—specialized A — .28 (.19) A — 134
B — .29 (.18) B — 145
O — Overall; A — American; B — British.

“dominance of the given,” “salience effect” and so forth (Hoffman & Novak, 1996), or to restate slightly, when a firm's
formalize the idea that memory and judgment have a simple web interface is not sufficiently easy to use. In addition to the
direct relationship. Empirical studies support the expectation of key role that ease of use has with the TAM described previously,
a direct relationship between memory and judgment. Beyth- numerous other studies (e. g., Collier & Bienstock, 2006; Yoo &
Marom and Fischoff (1977) provided strong evidence on the Donthu, 2001; Yang & Jun, 2002) have uncovered a critical role
strength of memory–judgment relationship, and Gabrielcik and for ease of use in perceptions of online service quality. Over-
Fazio (1984) demonstrated that ease of retrieval exerts a causal confidence could make the consumer more likely to conclude
(not just correlational) influence on frequency estimates. Based that the site is hard to use and to attribute e-service failure to
on the evidence, we reason that distorted and incomplete the site, not to him or herself, with potentially negative con-
memory diminishes calibration of knowledge of the web. Con- sequences for the retention of that customer.
versely, undistorted and complete memory promotes calibra- Several firms plan to move customer service to the web in
tion. Another potential antecedent factor is counterfactual order to lower costs and provide faster service. Frustration
thinking. It has been suggested that individuals who are arising from poor calibration could lead to the disconfirmation
required to consider contradictory evidence have lower over- of expected service quality. This in turn will lead to lower levels
confidence than individuals who are not required to do so of exclusive loyalty and repeat-purchase rates. The frustration
(Hoch, 1985). Experimental studies have demonstrated that could be attributed to self (“I am not as good as I thought”) or to
individuals automatically generate supporting reasons as a the firm (“the website is not good”). Neither of these augurs
byproduct of the question–answering process, but must be well for the firm. Thus, calibration of web knowledge could be
explicitly instructed to consider the counterfactual (Koriat, a mediator between firm's e-service offering and customer
Lichenstein, & Fischoff, 1980). Reviewing the relevant satisfaction/loyalty.
psychology literature, Faust (1986) identifies a series of rules While calibration of knowledge leads to optimal informa-
or necessary conditions that help improve calibration. Among tion search, overconfident people (cell 1) might act presump-
the rules suggested are to (1) present evidence that contradicts, tuously and commit errors in their web use. Underconfident
disconfirms, and refutes one's position and (2) generate com- people (cell 4) might spend considerable time searching for
peting alternative hypotheses. It follows that those who are the relevant information, even while they are in the know.
given to considering opposite and alternative perspectives Both of these are likely to result in frustration with web use.
regarding their knowledge of the web are likely to be more Calibrated people who possess the right amount of knowledge
calibrated in their knowledge than those who do not consider (cell 2) spend little time searching for such information, while
such counterfactual perspectives (Pillai & Goldsmith, 2006). calibrated people who don't possess the right kind of knowl-
Hence, counterfactual thinking could lower overconfidence edge spend appropriate and optimal amount of time searching
and enhance calibration of knowledge of the web. Several for the relevant information. Some research has examined
other antecedent variables could be identified. The interested specific antecedents of user frustration, including lack of
reader is referred to the review article by Alba and Hutchinson navigation standards (Neerincx, Lindenberg, & Pemberton,
(2000). 2001), disorientation in hyperspace and associated cognitive
overload (Nielsen, 2000; Patel, Drury, & Shalin, 1998), and
5.2. Consequence of calibration of knowledge of the web speed, accuracy, and preference of navigational systems
(Kiger, 1984). Apart from the direct effect of miscalibration
We discuss two potential consequences of calibration of web on frustration as argued above, miscalibrated individuals
knowledge — frustration with web and flow experience. are likely to feel disorientated, which also contributes to
frustration.
5.2.1. Frustration with web
A lack of calibration could contribute to the frustrations 5.2.2. Flow
ordinary people experience when dealing with the paradoxes of Flow is the process of optimal experience (Csikszentmihalyi,
web technology (Mick & Fournier, 1998). Frustration is the 1977). Flow experience in a computer mediated environment
natural outcome when the technology challenge is too great has been defined as the state occurring during network
264 K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267

navigation which is (1) characterized by a seamless sequence of of knowledge of the web. It was reasoned that intrinsically
response facilitated by machine interactivity, (2) intrinsically motivated people might be more calibrated in their knowledge
enjoyable, (3) accompanied by a loss of self-consciousness, of the web than extrinsically motivated people. This is an area
and (4) self-reinforcing (Hoffman & Novak, 1996). In the that merits further investigation.
flow experience, consumers are so acutely involved in the act It is also shown that calibration is higher for procedural
of network navigation in the hypermedia environment that knowledge, compared to declarative knowledge. This finding
“nothing else seems to matter” (Csikszentmihalyi, 1990, p.4). can add to the research on designing computer and software
Flow is assumed to have a number of positive outcomes for the manuals. In particular it adds a new dimension to the debate on
firm (Hoffman & Novak, 1996). the optimum contents of software manuals. Some researchers
Calibration of knowledge of the web will facilitate the flow (Carroll, 1990; Van der Maij & Carroll, 1995) have asserted that
experience. As noted earlier, flow occurs when perceived skills manuals should be reduced to contain only procedural, task-
match perceived challenges. A realistic assessment of the skill is supporting information. The finding regarding the differences
required to achieve flow. Since knowledge of the web is a skill in calibration of procedural and declarative knowledge of the
in the context of navigation, better calibration leads to better web can inform this debate. If people are more likely to be
flow. Overconfident individuals will face bottlenecks in their miscalibrated in their declarative knowledge, there is a case to
navigation and wouldn't know how to remedy the situation include such knowledge in manuals. Similar reasoning would
since they are unaware of their ignorance. They will also lead to the argument that, apart from common knowledge,
undertake tasks that are too challenging, which results in a manuals should focus on specialized knowledge as well. These
reduction in flow experience. Underconfident individuals will findings will also contribute to the design of contents for
dither, which inhibits their navigation. They are likely to engage training modules on the World Wide Web.
in navigation that is too simple to be enjoyable. Calibrated The study found no difference in calibration between the
individuals are likely to experience flow in the strict sense of the genders. However, the previously reported finding that males
term. are more confident than females was supported. Though we did
not offer formal hypotheses, data on age and education was
5.3. Implications of the findings collected in the British sample. A simple regression model
indicated that neither age (p = .17) nor education (p = .65) had
The results of the study point to the role of involvement any effect on calibration. Education, however, was found to be
with the web in promoting calibration of knowledge of the positively correlated with confidence (.22).
web. Higher calibration implies greater realism of judgment Generalizing from the above findings, it can be stated that
in the level of one's knowledge. Greater realism enables one calibration of knowledge is more a function of type of knowl-
to judge correctly what one is capable of doing. It also facili- edge and less a function of demographic variables. This is an
tates optimal information search, which in the context of important finding to the program of research on consumer
knowledge of the web could include attending training pro- knowledge calibration.
grams, reading manuals, or seeking information from collea- This study contributes to the development of a program of
gues. And, as argued above, calibration lowers frustration with research on knowledge calibration in consumer sciences. A
web usage and enhances flow experience. Hence, calibration large body of work on knowledge and confidence calibration,
of knowledge of the web will have performance implications its antecedents and consequences informs the dialogue in
for the web user. several areas of enquiry such as psychology, education, etc.
The study suggests that getting involved with the web is one Surprisingly, little research on knowledge calibration exists in
way to achieve better performance while using the web. Apart consumer research. It is hoped that this study acts as a catalyst
from the possible direct effect of involvement on performance, for enquiry in this important area.
calibration will have a mediating role on the involvement– Future research needs to extend the proposed model by
performance relationship. In a dynamic environment such as the examining other antecedents and consequents. As mentioned
web, technological changes and innovations by companies earlier, the mediating role of calibration between e-service
result in the creation of new knowledge, which in turn requires offering and satisfaction/loyalty is of considerable importance
that users continuously update their knowledge. In such an and needs to be examined. Research also needs to investigate
environment, the mediating role of calibration can be partic- the roles of other antecedent factors in influencing calibration of
ularly relevant. Knowing what they know and what they don't knowledge of the web. Further research can also distinguish
know and being realistic in their perceptions enables web users between the different types of high and low calibration (cells 1
to (a) devote the optimal amount of effort in adding to their to 4) and examine the differential effects of antecedent factors
knowledge of the web and (b) anticipate potential usability and consequences arising from these different instances. In
problems arising from their level of knowledge and take appro- particular, the two types of calibration will have different effects
priate steps to counter them, thus minimizing frustration and on consequences such as information search. As discussed
dissatisfaction. before, high accuracy and high confidence were observed in this
The study found that neither experience nor usage has an study, rendering such distinctions redundant.
effect on calibration of knowledge of the web. This underscores At a general level, the model can be applied to the domain of
the need to focus on involvement to enhance calibration specific knowledge of consumers and their use of the web.
K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267 265

Considerable research has examined online information search. Celsi, R. L., & Olson, J. C. (1988). The role of involvement in attention and
Researchers have suggested several antecedents to online comprehension processes. Journal of Consumer Research, 15, 210−224.
Chase, W. G., & Simon, H. A. (1973, January). Perception in chess. Cognitive
information search, such as income, experience with the web, Psychology, 4, 55−81.
and education (Klein & Ford, 2003). Models of online search Chen, M. (1986). Gender and computers: The beneficial effects of experience on
that combine the benefit–cost perspective and the ability– attitudes. Journal of Educational Computing Research, 2, 265−282.
motivation paradigm have been proposed (Kulviwat, Guo, & Christen-Szalanski, J. J. J., & Busyhead, J. B. (1981). Physician's use of
probabilistic information in a real clinic setting. Journal of Experimental
Engchanil, 2004). Calibration of knowledge of the domain
Psychology. Human Perception and Performance, 7, 928−935.
(product category) could be an antecedent to online search Collier, J. E., & Bienstock, C. C. (2006). Measuring service quality in e-
(those in cells 1 and 2 perform lower search compared to those retailing. Journal of Service Research, 8(3), 260−275.
in cells 3 and 4). More importantly, calibration could be a Csikszentmihalyi, M. (1977). Beyond boredom and anxiety. San Francisco:
determinant of the effectiveness of search, whereby calibrated Jossey-Bass.
individuals will perform more effective search compared to Csikszentmihalyi, M. (1990). Flow: The psychology of optimal experience.
New York: Harper and Row.
miscalibrated individuals. Darley, W. K. (1999). The relationship of antecedents of search and self-esteem
to adolescent search effort and perceived product knowledge. Psychology
Acknowledgments and Marketing, 16(5), 409−427.
Davis, F. D. (1993). User acceptance of information technology: system
The authors thank Professor Hubert Gatignon and three characteristics, user perceptions and behavioral impacts. International
Journal of Man–Machine Studies, 38, 475−487.
anonymous reviewers for their valuable comments throughout Delbridge, A., & Bernard, J. R. L. (Eds.). (1998). The concise Macquarie
the review process. English dictionary, 3rd ed. Sydney: Macquarie University.
DeLone, W. H. (1988). Determinants of success for computer usage in small
business. MIS Quarterly, 12(1), 51−61.
References
Diaz, A. N., Hammond, K., & Mcwilliam, G. (1997). A study of web use and
attitudes amongst novices, moderate users, and heavy users. In 25th EMAC
Alba, J. W. (1983). The effects of product knowledge on the comprehension, conference proceedings, Vol 4. (pp. 1624−1633).
retention, and evaluation of product information. In R. P. Bagozzi, & A. M. Estelami, H., Lehmann, D. R., & Holden, A. C. (2001). Macro-economic
Tybout (Eds.), Advances in consumer research, vol. 10. (pp. 577−580). Ann determinants of consumer price knowledge: A meta-analysis of four decades
Arbor, MI: Association for Consumer Research. of research. International Journal of Research in Marketing, 18(4), 341−355.
Alba, J. W., & Hutchinson, J. W. (1987). Dimensions of consumer expertise. Estes, R., & Hosseini, J. (1988). The gender gap on Wall Street: An empirical
Journal of Consumer Research, 13, 411−454. analysis of confidence in investment decision making. Journal of
Alba, J. W., & Hutchinson, J. W. (2000, September). Knowledge calibration: Psychology: Interdisciplinary and Applied, 122(6), 577−590.
What consumers know and what they think they know. Journal of Faust, D. (1986). Learning and maintaining rules for decreasing judgment
Consumer Research, 27, 123−156. accuracy. Journal of Personality Assessment, 50(4), 585−600.
Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: Harvard Fischoff, B., & Slovic, P. (1980). A little learning…: Confidence in multicue
University Press. judgment tasks. In R. S. Nickerson (Ed.), Attention and performance VIII
Batra, R., & Ray, M. L. (1985). How advertising works at contact. In L. Alwitt, (pp. 779−800). Hillsdale, NJ: Erlbaum.
& A. A. Mitchell (Eds.), Psychological processes and advertising effects: Gabrielcik, A., & Fazio, R. H. (1984). Priming and frequency estimation: A
Theory, research, and applications (pp. 13−43). Hillsdale, NJ: Erlbaum. strict test of the availability heuristic. Personality and Social Psychology
Best, J. B. (1989). Cognitive psychology, 2nd ed. New York: West Publishing. Bulletin, 10, 85−89.
Bettman, J. R., & Park, C. W. (1980, December). Effects of prior knowledge and Gigerenzer, G., Hoffrage, U., & Kleinbolting, H. (1991). Probabilistic mental
experience and phase of the choice process on consumer decision processes: models: A brunswikian theory of confidence. Psychological Review, 103,
A protocol analysis. Journal of Consumer Research, 7, 234−248. 650−669.
Beyth-Marom, R., & Fischoff, B. (1977). Direct measures of availability and Goldsmith, E. B., & Goldsmith, R. E. (1997, February). Gender differences in
judgments of category frequency. Bulletin of the Psychonomic Society, 9, perceived and real knowledge of financial investments. Psychological
236−238. Reports, 80, 236−238.
Biswas, A., & Sherrell, D. L. (1993). The influence of product knowledge and Goldsmith, R. E., & Pillai, K. G. (2006). Knowledge calibration. In D. Schwartz
brand name on internal price standards and confidence. Psychology and (Ed.), Encyclopedia of knowledge management (pp. 311−316). Hershey,
Marketing, 10, 31−46. PA: Idea Group Publishing.
Blair, M. E., & Innis, D. E. (1996). The effects of product knowledge on the Harrison, A. W., & Rainer, R. K. (1992). The influence of individual differences
evaluation of warranted brands. Psychology and Marketing, 13(5), on skill in end-user computing. Journal of Management Information
445−456. Systems, 9(1), 93−111.
Bloch, P. H., Sherrel, D. L., & Ridgway, N. M. (1986). Consumer search: An Harvey, N. (1997). Confidence in judgment. Trends in Cognitive Sciences, 1,
extended framework. Journal of Consumer Research, 13, 119−126. 78−82.
Bonner, S. E., & Walker, P. L. (1994). The effects of instruction and experience Heath, C., & Tversky, A. (1991). Preference and belief: Ambiguity and
on the acquisition of auditing knowledge. The Accounting Review, 69, competence in choice under uncertainty. Journal of Risk and Uncertainty, 4
157−178. (1), 5−28.
Brucks, M. (1985, June). The effects of product class knowledge on information Hoch, S. J. (1985, October). Counterfactual reasoning and accuracy in
search behavior. Journal of Consumer Research, 12, 1−16. predicting personal events. Journal of Experimental Psychology: Learning,
Brucks, M. (1986). A typology of consumer knowledge content. Advances in Memory, and Cognition, 11, 719−731.
Consumer Research, 13, 58−63. Hoerl, A. E., & Fallin, H. K. (1974). Reliability of subjective evaluations in a
Capraro, A. J., Broniarczyk, S., & Srivastava, R. K. (2003). Factors influencing high incentive situation. Journal of the Royal Statistical Society, 137,
the likelihood of customer defection: The role of consumer knowledge. 227−230.
Journal of the Academy of Marketing Science, 31(2), 164−175. Hoffman, D. L., & Novak, T. P. (1996). Marketing in hypermedia computer-
Carroll, J. M. (1990). The Nurnberg funnel: Designing minimalist instruction mediated environments: Conceptual foundations. Journal of Marketing, 60
for practical computer skill. Cambridge, MA: MIT Press. (3), 50−68.
266 K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267

Kalaian, H. A., & Freeman, D. J. (1994). Gender differences in self confidence Novak, T. P., Hoffman, D. L., & Yung, Y. F. (2000). Measuring the customer
and educational beliefs among secondary teacher candidates. Teaching and experience in online environments: A structural modeling approach. Mar-
Teacher Education, 10(6), 647−658. keting Science, 19, 22−44.
Keren, G. (1991). Calibration of probability judgments: Conceptual and O'Connor, M. J. (1989). Models of human behavior and confidence in
methodological issues. Acta Psychologica, 77, 217−273. judgment: A review. International Journal of Forecasting, 5, 159−169.
Kiger, J. I. (1984). The depth/breadth tradeoff in the design of many-driven Okechuku, C. (1992). The relationships of prior knowledge and involvement to
interfaces. International Journal of Man–Machine Studies, 20(2), advertising recall and evaluation. International Journal of Research in
201−213. Marketing, 9(2), 115−130.
Kim, S. S., & Malhotra, N. K. (2005). A longitudinal model of continued IS use: Olsson, N. (2000). A comparison of correlation, calibration, and diagnosticity as
An integrative view of four mechanisms underlying postadoption phenom- measures of the confidence–accuracy relationship in witness identification.
ena. Management Science, 51(5), 741−755. Journal of Applied Psychology, 85(4), 504−511.
Klein, L. R., & Ford, G. T. (2003). Consumer search for information in the Oskamp, S. (1962). The relationship of clinical experience and training methods
digital age: An empirical study of prepurchase search for automobiles. to several criteria of clinical prediction. Psychological Monographs, 76 (28,
Journal of Interactive Marketing, 17(3), 29−49. Whole no. 547).
Koriat, A. (1998). Metamemory: The feeling of knowing and its vagaries. In M. Paese, P. W., & Sniezek, J. A. (1991, February). Influences on the appro-
Sabourin, F. Craik, & M. Robert (Eds.), Advances in psychological science priateness of confidence in judgment: Practice, effort, information, and
Biological and cognitive aspects, vol. 2. (pp. 461−479) Hove, England: decision-making. Organizational Behavior and Human Decision Processes,
Psychology Press/ Erlbaum (UK) Taylor & Francis. 48, 100−130.
Koriat, A., Lichtenstein, S., & Fischoff, B. (1980, March). Reasons for Page, K., & Uncles, M. (2004). Consumer knowledge of the World Wide Web:
confidence. Journal of Experimental Psychology. Human Learning and Conceptualization and measurement. Psychology and Marketing, 21(8),
Memory, 6, 107−118. 573−591.
Kulviwat, S., Guo, C., & Engchanil, N. (2004). Determinants of online Park, C. W., Mothersbaugh, D. L., & Feick, L. (1994). Consumer knowledge
information search: A critical review and assessment. Internet Research, 14 assessment. Journal of Consumer Research, 21(1), 71−82.
(3), 245−253. Park, J., & Hastak, M. (1994). Memory-based product judgments: Effects of
Laurent, G., & Kapferer, J. N. (1985). Measuring consumer involvement involvement at encoding and retrieval. Journal of Consumer Research, 21,
profiles. Journal of Marketing Research, 22, 41−53. 534−547.
Lewis, D. (1969). Convention; a philosophical study. Cambridge, MA: Harvard Parasuraman, A. (2000). Technology Readiness Index (Tri): A multiple-item
University Press. scale to measure readiness to embrace new technologies. Journal of Service
Lichtenstein, S., Fischoff, B., & Philips, L. D. (1982). Calibration of subjective Research, 2(4), 307−320.
probabilities. The state of the art up to 1980. In D. Kahnemann, P. Slovic, & Patel, S. C., Drury, C. G., & Shalin, V. (1998). Effectiveness of expert semantic
A. Tversky (Eds.), Judgments under uncertainty: Heuristics and biases (pp. knowledge as a navigational aid within hypertext. Behavioir & Information
306−334). New York: Cambridge University Press. Technology, 17(6), 313−324.
Lin, L., Moore, D., & Zabrucky, K. M. (2001). An assessment of students' Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983, September). Central and
calibration of comprehension and calibration of performance using multiple peripheral routes to advertising effectiveness: The moderating role of
measures. Reading Psychology, 22, 111−128. involvement. Journal of Consumer Research, 10, 135−146.
Loftus, E. F. (1974). Reconstructing memory: The incredible eyewitness. Psy- Pillai, K. G., & Goldsmith, R. E. (2006). Calibrating managerial knowledge of
chology Today, 8(7), 116−119. customer feedback measures: A conceptual model. Marketing Theory, 6(2),
Maheswaran, D., & Sternthal, B. (1990). The effects of knowledge, motivation, 223−243.
and type of message on ad processing. Journal of Consumer Research, 5, Raju, P. S., Lonial, S. C., & Mangold, W. G. (1995). Differential effects of
115−133. subjective knowledge, objective knowledge, and usage experience on
Mick, D. G., & Fournier, S. (1998). Paradoxes of technology: Consumer decision making: An exploratory investigation. Journal of Consumer
cognizance, emotions and coping strategies. Journal of Consumer Research, Psychology, 4, 153−180.
25(2), 123−143. Ray, M. L., Sawyer, A. G., Rothschild, M. L., Heeler, R. M., Strong, E. C., &
Miniard, P. W., Bhatla, S., Lord, K. R., Dickson, P. R., & Unnava, H. R. (1991, Reed, J. B. (1973). Marketing communications and the hierarchy of effects.
June). Picture-based persuasion processes and the moderating role of In P. Clarks (Ed.), New models of mass communication research (pp.
involvement. Journal of Consumer Research, 18, 92−107. 147−176). Beverly Hills, CA: Sage.
Mitchell, A. A., & Dacin, P. F. (1996, December). The assessment of alternative Roberts, J., Hann, I., & Slaughter, S. A. (2006). Understanding the motivations,
measures of consumer expertise. Journal of Consumer Research, 23, 219−240. participation, and performance of open source software developers: A longi-
Moreau, C. P., Lehmann, D. R., & Markman, A. B. (2001). Entrenched tudinal study of the Apache projects. Management Science, 52(7), 984−999.
knowledge structures and consumer response to new products. Journal of Rodgers, W., Negash, S., & Suk, K. (2005). The moderating effect of online
Marketing Research, 38, 14−29. experience on the antecedents and consequences of online satisfaction.
Murphy, A. H., & Winkler, P. (1977). Can weather forecasters formulate reliable Psychology and Marketing, 22(4), 313−331.
probability forecasts of precipitation and temperature? National Weather Roy, D. P., & Cornwell, T. B. (2004). The effects of consumer knowledge on
Digest, 2, 2−9. responses to event sponsorships. Psychology and Marketing, 21(3), 185−207.
Narby, D. J., Cutler, B. L., & Penrod, S. D. (1996). The effects of witness, target, Sachs, C., & Bellisimo, Y. (1993). Attitudes toward computers and computer
and situational factors on eyewitness identifications. In S. L. Sporer, R. S. use: The issue of gender. Journal of Research on Computing in Education,
Malpass, & G. Koehnken (Eds.), Psychological issues in eyewitness 26, 256−270.
identification (pp. 117−153). Mahwah, NJ: Erlbaum. Schumacher, P., & Morahan-Martin, J. (2001). Gender, internet, and computer
Neerincx, M. A., Lindenberg, J., & Pemberton, S. (2001). Support concepts for attitudes and behavior. Computers in Human Behavior, 17, 95−110.
web navigation: A cognitive engineering approach. Proceedings of the tenth Simon, S. J. (2001). The impact of culture and gender on web sites: An empirical
international conference on the World Wide Web (pp. 119−128). Hong study. Database for Advances in Information Systems, 32(1), 18−37.
Kong: ACM. Spence, M. T. (1996). Problem–problem solver characteristics affecting the
Nelson, T. O., Leonesio, R. J., Landwehr, R. S., & Narens, L. (1986). A calibration of judgments. Organizational Behavior and Human Decision
comparison of three predictors of an individual's memory performance: The Processes, 67(3), 271−279.
individual's feeling of knowing versus the normative feeling of knowing Steuer, J. (1992). Defining virtual reality: Dimensions of determining
versus base-rate item difficulty. Journal of Experimental Psychology. telepresence. Journal of Communication, 42, 73−93.
Learning, Memory, and Cognition, 12(2), 279−287. Sutton, N. A. (1987). Life insurance and the consumer: An illusion of control.
Nielsen, J. (2000). Designing web usability. Indianapolis: New Riders. Best's Review, 88(3), 54−57.
K.G. Pillai, C. Hofacker / Intern. J. of Research in Marketing 24 (2007) 254–267 267

Ummelen, N. (1997). Procedural and declarative information in software Yang, Z., & Jun, M. (2002). Consumer perception of e-service quality: From
manuals: Effects on information use, task performance, and knowledge. internet purchaser and non-purchaser perspectives. Journal of Business
Amsterdam: Rodopi. Strategies, 19(1), 19−41.
Van der Maij, H., & Carroll, J. M. (1995). Principles and heuristics for designing Yates, J. F. (1990). Judgment and decision making. Englewood Cliffs, NJ:
minimalist instruction. Technical Communication, 42, 243−246. Prentice Hall.
Venkatesh, V., & Davis, F. D. (2000). Theoretical extension of the technology Yoo, B., & Donthu, N. (2001). Developing a scale to measure the perceived
acceptance model: Four longitudinal field studies. Management Science, 46 quality of an internet shopping site (sitequal). Quarterly Journal of
(2), 186−204. Electronic Commerce, 2(1), 31−46.
Verlegh, P. W. J., Steenkamp, J. -B. E. M, & Meulenberg, M. T. G. (2005). Zaichkowski, J. L. (1985). Measuring the involvement construct. Journal of
Country-of-origin effects in consumer processing of advertising claims. Consumer Research, 12, 341−352.
International Journal of Research in Marketing, 22(2), 127−139. Zaichkowsky, J. L. (1994). The personal involvement inventory: Reduction,
Wyer, R. S., Jr, & Srull, T. K. (1986). Human cognition in its social context. revision, and application to advertising. Journal of Advertising, 23(4),
Psychological Review, 93(3), 322−332. 59−70.

You might also like