Sei sulla pagina 1di 22

Journal of Retailing 90 (2, 2014) 233254

The Antecedents and Moderators of Offline Information Search:


A Meta-Analysis
Moutusy Maity a, , Mayukh Dass b,1 , Naresh K. Malhotra c,d
a

Marketing, Indian Institute of Management Lucknow, 1309 Faculty Block, Plot B1, Sector 62, Noida, UP 201307, India
b Marketing, Rawls College of Business, Texas Tech University, MS 2101, Lubbock, TX, USA
c Senior Research Fellow, Nelson Mandela Metropolitan University, South Africa
d Regents Professor Emeritus, Georgia Institute of Technology, Scheller College of Business, 800 West Peachtree St NW, Atlanta, GA 30308-1149, USA

Abstract
An examination of consumer information search literature reveals that the effects of several antecedents (cost, price dispersion, knowledge,
prior experience) on ofine information search vary greatly in terms of the directions and magnitudes of the determinants effect sizes, particularly
indicating the possibility of inverted-U shaped relationships. Also, despite a wide range of studies undertaken, there remains a need for identifying
an overarching framework for consumer information search. This article synthesizes previous empirical studies on ofine information search
literature, and attempts to provide and test a framework that advances our understanding of the current state of this literature. The review identifies
significant antecedents to offline information search, and also tests the moderating effects of demographic variables and study conditions on the
direct effects. In the context of the main effects, findings suggest that the nature of relationships of several antecedents with the central construct
is indeed inverted-U shaped. Further, moderator analyses identify the boundary conditions under which specific theories hold. The closing section
provides future research directions.
2014 New York University. Published by Elsevier Inc. All rights reserved.
Keywords: Offline information search; Offline consumer information search; Inverted-U shaped relationship; Moderators; Meta-analysis

Introduction
Consumer information search behavior has a long tradition
of being a useful mechanism for understanding consumer shopping behavior, including consumer choice, and choice processes
(e.g. Beatty and Smith 1987; Bettman 1974; Punj and Staelin
1983). Over the last few decades, multiple studies have repeatedly investigated the effects of different antecedent variables
on offline information search. However, the studies conclusions vary greatly in terms of the directions and magnitudes of
the determinants effect sizes. For example, past research findings on the relationship between cost and information search

Corresponding author. Fax: +91 120 668 1002.


E-mail addresses: mmaity@iiml.ac.in (M. Maity), mayukh.dass@ttu.edu
(M. Dass), naresh.malhotra@mgt.gatech.edu (N.K. Malhotra).
1 Fax: +1 806 742 2199.

yield both positive (e.g. Laroche, Cleveland, and Brown 2004)


and negative (e.g. Roedder-John, Scott, and Bettman 1986)
directions to the relationship. While these studies advance our
understanding of the effects of various factors on offline information search, there has been little attempt to collate the findings
of past studies to report the direct effects (between antecedents
and information search) and the moderating effects of potential
moderators.
To reconcile these issues, and to provide a holistic understanding of these existing research findings, we perform a
formal meta-analytic review of extant literature in offline
information search, examine the direct effects, and investigate
the role of study conditions (i.e. moderators). Therefore, the
specific objectives of this study are: (a) to ascertain the strength
and direction of the direct effects of specific antecedents to
offline consumer information search, (b) to assess the impact of
various moderators on the direct effects; and (c) to identify and
test an overarching framework for categories of antecedents to

http://dx.doi.org/10.1016/j.jretai.2014.03.001
0022-4359/ 2014 New York University. Published by Elsevier Inc. All rights reserved.

234

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

information search. A search on the term consumer information


search on EBSCO confirms that recent research on the topic is
increasingly focused on the phenomenon in the online (personal
computer, tablet, mobile) context. While consumers are undertaking information search online (e.g. Pauwels et al. 2011),
offline information search remains an important component of
a consumers purchase behavior. Businesses are increasingly
interested in learning about a consumers entire journey along
the purchase path, which includes both offline and online search
(e.g. MSIs Shopper Marketing initiative). Therefore, while
online information search has gained more visibility in recent
years, a review of offline information search literature is timely.
Consumer information search in the context of this metaanalysis is the effort that a consumer expends in order to acquire
information from the external environment (e.g. Srinivasan and
Ratchford 1991) for a specific purchase (Beatty and Smith 1987).
Consumer information search has been largely measured along
two dimensions: (1) amount of search (defined as: number of
alternatives/brands, number of attributes, number of times a
particular source is examined, or the extent of search) (GreganPaxton and Roedder-John 1995; Jacoby, Speller, and Berning
1974; Malhotra 1982), and (2) time spent on search (defined
as: total time required to search) (Brucks 1985). Hunter and
Schmidt (1990) and several other researchers (e.g. Glass 1976)
suggest that the same construct (though measured variously) can
be aggregated in a single meta-analysis. Therefore, this study
pools together all eligible studies with the dependent variable
information search.
This paper provides insights that advance both theoretical
and conceptual knowledge of offline information search. In particular, in the context of direct effects, findings suggest that
the relationships between information search and antecedents
cost, price dispersion, knowledge, and prior experience (each)
are inverted-U shaped. Study findings also suggest that benet, instead of cost of search, is more appropriate for modeling
consumer information search. In the context of moderators, the
study finds that the moderating effects of age and product type
on direct effects are significant.
The paper addresses the following gaps in the information
search literature. First, this research conducts a much-needed
meta-analytic review of a mature area, and addresses the need
to reconcile inconsistent findings among studies conducted
in the past. Second, this study separates chance effects from
moderator effects in the offline information search literature
and identifies conditions under which specific theories hold.
Third, it identifies a framework with specific categories of variables that affect consumer information search. The rest of the
paper is organized as follows. First, we present the procedure
adopted for identifying studies, and report the sampling frame.
Next, we discuss the theoretical foundations of categories of
antecedents to information search, and present the hypotheses
on the direct effects. Third, we present the research methods,
and discuss the findings on the direct effects. Next, we put
forth the hypotheses on the moderators, present the data analysis procedures and discuss the findings on the moderators. We
conclude with research implications and directions for future
research.

Studies included in the review


For identifying eligible studies, a keyword search in the
ABI/Inform database was followed by a search for conference
papers. Finally, letters requesting published and unpublished
studies on this topic were mailed to approximately 100 authors.
This search resulted in approximately 350 studies, including
those that were purely theoretical, concentrated on econometric modeling, carried out estimation and simulation, and did
not directly measure information search. A search on Google
Scholar also yielded similar results. Studies that suggest models (without quantitatively testing them), present derivations
only, or simply suggest the hypothesized direction between an
antecedent and information search (e.g. Gemunden 1985), were
not included in this review. Sixty-two pertinent studies were
identified, which investigated 79 antecedent variables (the complete list can be obtained from the authors), and yielded a sample
size (n) of 94,302, with 357 effect sizes (r). A requirement for
a meaningful meta-analysis is a minimum of ten reported studies per variable (Hunter and Schmidt 1990). Ten antecedents
report ten or more effect sizes per variable, and were included
for further analysis.
In order to identify moderators of offline consumer information search four different moderator categories (Assmus, Farley,
and Lehmann 1984; Sultan, Farley, and Lehmann 1990) are
considered: (1) measurement methods, (2) research context,
(3) estimation procedure, and (4) model specification. Units of
analyses in this study are bivariate correlations, and therefore,
are unaffected by the estimation procedure or model specification. Hence, this review focuses on two moderator categories,
namely measurement methods (study design and respondent
type) and research context (product type, age, gender, education and income), and assesses their impact on the homogeneity
of direct effects. Research in marketing and consumer behavior
suggests that the direct or mediating effects of these variables are
small and that these variables are best treated as moderators (see
Chen 2012; Cheon, Cho, and Sutherland 2007 for the specific
use of moderators in meta-analyses).
Antecedents: Theoretical foundations
Researchers examining consumer information search have
tried to identify an overarching framework explaining this phenomenon. Several researchers have suggested specific categories
of antecedent variables that affect consumer information search
(see Table 1). The categories identified in these investigations
are usually prompted by the authors assessment of the current
state of information search literature and the variables that have
been investigated prior to the categorization attempts. For example, in their categorization, Moore and Lehmann (1980) split the
Properties of Choice Situation category suggested by Bettman
(1979) into two categories Market Environment and Situational
Variables, and Newmans (1977) Potential Payoff category into a
number of other categories. Beatty and Smith (1987), in turn, add
a category Cost of Search to the scheme suggested by Moore
and Lehmann (1980). Many studies also test holistic models of
offline information search (e.g. Srinivasan and Ratchford 1991;

Table 1
Categories of antecedents to information search.
Author(s)

Variables/categories of variables suggested in extant literature

Newman (1977)

Cost

Buying Strategies (brand


and store preference,
satisficing vs. optimizing,
strategy of information
acquisition)

Potential Payoff (price,


style, perceived
difference, perceived risk,
knowledge, experience,
education, income)

Situational Variables
(urgency, financial
pressure, special buying
opportunities, location of
residence)

Personality Variables

Bettman (1979)

Properties of Choice
Situation (availability of
information, difficulty of
choice task, time
pressure)
Market Environment
(number of alternatives,
complexity of
alternatives, information
availability)
Cost

Costs vs. Benets of


Information

Conict and Conict


Response Strategies

Knowledge

Situational Variables
(time pressure, social
pressure, financial
pressure, ease of access of
information)
Benet

Potential Payoff/Product
Importance (perceived
risk, price, social
visibility)

Individual Differences
(in-store vs.
prior-processing, abilities,
concern with optimality
of choice)
Knowledge and
Experience (usage of
product)

Individual Differences
(involvement, ability,
training)

Conict and
Conict-Resolution
Strategies

Market Environment
Market Environment
(number of alternatives,
complexity of
alternatives, information
availability)
Size of evoked set

Situational Factors
Situational Variables
(time pressure, social
pressure, financial
pressure, ease of access of
information)
Interest in cars

Knowledge and
Experience (usage of
product)

Individual Differences
(involvement, ability,
training)

Conict and
Conict-Resolution
Strategies

Perceived risk

Perceived benefits

Cost

Relative brand uncertainty

Involvement

Risk aversion

Costs vs. Benets of


Information (number of
alternatives, time
pressure, psychological
costs)

Potential Payoff
(perceived risk, financial
pressure, price, social
visibility)

Individual brand
uncertainty
Knowledge and
Experience (usage of
product)

Amount of
experience, positive
experience, product
knowledge
Search cost

Punj and Staelin


(1983)

Bloch et al. (1986)


Beatty and Smith
(1987)

Srinivasan and
Ratchford (1991)

Moorthy et al.
(1997)
Authors (based on
EoI theory)

Factors Unique to a
Consumer (e.g. usable
prior knowledge, prior
memory structure)
Involvement
Potential Payoff/Product
Importance (perceived
risk, price, social
visibility)

Cost of Search
(psychological
costs)

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

Moore and Lehmann


(1980)

Other Variables
(household role,
party of major
influence, social
class, occupation,
age, stage of life
cycle)

Experience

Individual Differences
(involvement, ability,
training)

Categories of variables as identified by the authors (in italics). Specific variables as identified/used by the authors (in regular font).

235

236

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

Moorthy, Ratchford, and Talukdar 1997), but do not identify


specific categories of variables. We suggest a framework that
is parsimonious and is explicitly grounded in the Economics of
Information (EoI) theory.
Despite these seemingly ad hoc categorizations, most of the
authors agree that the phenomenon of consumer information
search is primarily grounded in the costbenefit concept of Economics of Information (EoI) by Stigler (1961). Therefore, we
take the EoI as the framework for identifying categories of
antecedent variables that affect consumer information search,
and for guiding our theoretical discussion and empirical investigations. We identify four categories of antecedent variables
that affect information search: Cost vs. Benet of Information,
Potential Payoff, Knowledge and Experience and Individual Differences.
We present propositions about the relationships of the categories of antecedent variables with information search, and then
present the hypotheses for the ten variables that are included in
the meta-analysis (Table 2).
Cost versus benet of information
It is widely acknowledged that there are costs associated with
acquiring and processing information (e.g. Punj and Staelin
1983; Srinivasan and Ratchford 1991), and one of the central arguments of the EoI (or, cost theory) is that consumers
search for information till marginal benefits continue to exceed
marginal costs (Smith, Venkataraman, and Dholakia 1999). As
a consequence of these costs, Stigler (1961) argues that consumers will continue to search for additional information until
the benefit obtained from additional information is smaller than
the cost of obtaining it. Since the EoI theory is conceptualized
as a costbenefit trade-off that a consumer undertakes while
searching for information, therefore, in keeping with Bettman
(1979), we suggest the category of Cost vs. Benet of Information as one of the categories of variables affecting consumer
information search.
The overall directions of the relationships of the specific
antecedent variables (which may belong to this category),
with the central construct, need to be ascertained. Researchers
suggest that search costs can be categorized as internal or
external costs, where internal search costs are those that are
cognitive costs and can be controlled by the consumer, while
external search costs are not within the consumers direct control (Smith, Venkataraman, and Dholakia 1999), and may be
situational.
Extant literature suggests that cognitive cost (psychological
cost) is negatively related to information search (e.g. Beatty and
Smith 1987; Punj and Staelin 1983). Humans have limited cognitive capabilities, and as cognitive costs increase, an individual
is likely to limit him/herself to satisficing (Simon 1987, p. 243)
as opposed to optimizing. Information overload provides a similar argument, and occurs when the amount of input to a system
exceeds its processing capacity (Malhotra 1982; Milford and
Perry 1977; Simon 1979). When information overload occurs,
it is likely that a reduction in decision quality will occur, thus
leading to lower information search.

While the above-stated relationship is the dominant view in


information search literature, Bodur and Matyas (2008) find that
there is an inverted-U shaped relationship between assortment
size and consumers evaluation of choice tasks for promotion
focus individuals. Findings in psychology and conflict management suggest that sometimes cognitive cost (in the form of
cognitive conflict) has an inverted-U shaped relationship with
decision making (e.g. Parayitam and Dooley 2011). A highly
complex task will have a higher level of cognitive cost and lower
level of performance compared to a moderately complex task.
This phenomenon occurs because complex tasks involve greater
demand and greater memory load and hence, is more vulnerable
to breakdowns in the process. A simple task does not provide the
adequate motivation required in order to reach optimal levels of
decision-making, as a moderately complex task does. We may
represent the relationship between cost and benefit as a curve,
such that the rising (falling) part of the curve represents the portion of search where benefits (costs) outweigh costs (benefits),
which correspond to the postulations of the EoI theory. (The horizontal axis represents the cost of search while the vertical axis
represents the benefits accrued from search as well as the amount
of information search undertaken.) Therefore, as cognitive costs
rise, information search will increase and then decrease. Hence,
we propose that:
Proposition 1a. The relationship between information search
and internal cost of search is inverted-U.
The benefit obtained from information search, represents the
rising part of the curve. Therefore, we expect that:
Proposition 1b. The relationship between information search
and benet is positive.
The latter category of search costs is exogenous in nature,
and is influenced by the situation that a consumer is in. For
example, Smith, Venkataraman, and Dholakia (1999) identify
urgency (or, time pressure) as one such cost given which the consumer may choose to continue (discontinue) her search. External
search costs by definition are not under the consumers control. Therefore, it is expected that these kinds of search costs
will not allow the consumer to be in a situation where marginal
benefits exceed marginal costs. The consumer will experience
a decrease in the amount of information search undertaken as
costs increase.
A large number of studies on decision making have advanced
the important role of cognitive capacity (e.g. Miller 1956; Simon
1957) and the limited resource models of Kahneman (1973)
and Norman and Bobrow (1975). Research in decision making acknowledges that while arousal may be a basic operator,
reductions in cognitive ability are often proposed as the central
explanatory element. For example, Eysenck and Calvo (1992)
asserted that individuals under stress demonstrate lower cognitive efficiency due to intrusive cognitions competing for the
limited pool of cognitive resources. Other studies demonstrated
that time pressure (e.g. Payne, Bettman, and Luce 1996), and
noise (e.g. Hockey 1970), as well as naturally occurring lifeevent stressors (e.g. Baradell and Klein 1993), pose a burden
on cognitive capacity, increasing errors made on cognitive tasks

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

237

Table 2
Directions of the hypothesized relationships and the effects reported in studies included in the meta-analysis.
Antecedent variables

Theories explaining the


relationships

Cost vs. Benet of Information


Number of alternatives
Cost Theory; Information
Overload; Utility Theory
Cost
Cost Theory; Information
Overload; Utility Theory;
EoI Theory
Price dispersion
Cost Theory; EoI Theory
Cost theory
Time pressure
Potential Payoff
Perceived risk
Financial constraint
Uncertainty

Risk Theory
Risk Theory
Risk Theory

Knowledge and Prior Experience


Knowledge
Memory; Theory of
Learning; Involvement
Theory; Risk Theory;
Effort Theory;
Satisfaction Theory
Memory; Theory of
Prior experience
Learning; Involvement
Theory; Risk Theory;
Effort Theory;
Satisfaction Theory
Individual Difference
Involvement

Involvement Theory

Hypotheses

Hypothesized
relationship

Number of
effects
reported

Range of
reported
effects (r)

Cumulative n

H1

Inverted-U

11a /17b

0.31 to 0.60

5,561

H2

Inverted-U

33/37

0.61 to 0.54

9,264

H3
H4

Inverted-U

16/21
16/21

0.40 to 0.54
0.88 to 0.36

5,666
4,829

H5
H6
H7

+
+
+

13/16
24/26
18/22

0.33 to 0.33
0.38 to 0.43
0.060.53

3,567
10,872
4,821

H8

Inverted-U

30/43

0.34 to 0.55

12,090

H9

Inverted-U

24/36

0.54 to 0.52

10,360

H10

20/25

0.17 to 0.80

8,186

All summary statistics (e.g. F, t, Z, chi-square) obtained across studies are converted to the common correlation coefficient metric, r.
a Number of statistically significant (at *p < 0.05) effects reported.
b Total number of effects reported.

and altering the way information is processed. In his review of


the literature, Staal (2004) concluded that stressors lead to two
specific changes in decision making: causing information processing to become more rigid with fewer alternatives scanned
and inducing the tendency to persist with a method or problemsolving strategy even after it has ceased to be helpful (p. 68).
Hence, we propose that:
Proposition 1c. The relationship between information search
and external cost of search is negative.
Number of alternatives, cost. Past research (Jacoby et al.
1974; Malhotra 1982) has demonstrated that as the amount
of information available to consumers increase, the amount of
search undertaken decrease. Number of alternatives is defined
as the total number of alternatives available from which the consumer has to make a choice (e.g. Ratchford and Srinivasan 1993;
Srinivasan and Ratchford 1991). The variable is operationalized by directly asking consumers or, by objectively counting
the number of options that are available to the consumer. An
individual has limited information processing capabilities (finite
cognitive capability), which limit the number of alternatives considered (Roedder John and Cole 1986). Therefore, based on cost
theory (Stigler 1961), large number of alternatives should reduce
the level of information search, that is, have a negative effect on
information search.

Cost of search is the consumers perception of the effort


expended in information search, and has been measure through
multi-item scales (e.g. DeSarbo and Choi 1999; Dowling and
Staelin 1994), opportunity cost and wages obtained (e.g. Urbany,
Dickson, and Kalapurakal 1996) in extant literature. In our
meta-analysis, we find that the ranges of effect sizes for the
relationship between information search and the antecedent variables, number of alternatives (0.31 to 0.60) and cost (0.61 to
0.54) contain both negative and positive valences. An increase in
these two variables increases internal search costs. Such divergent findings indicate that the relationships of these variables
with information search are probably inverted U-shaped. Also,
as discussed above, some researchers find an inverted-U shaped
relationship between cognitive cost and decision-making (e.g.
Bodur and Matyas 2008; Parayitam and Dooley 2011). Therefore, information search is high when cost of search is medium,
and information search is low when cost of search is high or low.
We hypothesize that:
H1. The relationship between information search and number
of alternatives is inverted-U.
H2. The relationship between information search and cost is
inverted-U.
Price dispersion. Perceived price dispersion is the judgment
that a consumer makes about the relative width (i.e. the variance)

238

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

of the available prices from which the consumer can make a


choice (e.g. Urbany, Dickson, and Kalapurakal 1996). In the
studies included in the meta-analysis, price dispersion has been
variously conceptualized as inter-brand difference, and price
range for homogeneous as well as for heterogeneous goods.
Recall that we conceptualize information search as acquiring
information from the external environment, which encompasses
a variety of possible reasons for price dispersion, including product heterogeneity (e.g. Pan, Ratchford, and Shankar 2003). High
levels of perceived price dispersion will lead to higher expected
savings from the search activity (Claxton, Fry, and Portis 1974).
Perceived price dispersion, too, is an internal search cost.
Since it is a necessary condition for consumers to undertake
a search activity (Stigler 1961), perceived price dispersion or
information variation is a basic premise of the EoI theory. Consumers who perceive wide dispersions in list prices would search
more extensively than buyers who perceive a narrow dispersion
(Urbany 1986). We observe that a range of effect sizes have
been reported for price dispersion (0.40 to 0.54) in extant literature. Price dispersion has been measured through subjective
scales (items measured through likert scales or bi-polar adjective
scales) (e.g. Putrevu and Ratchford 1997; Urbany, Dickson, and
Kalapurakal 1996), and has also been manipulated in experimental settings (e.g. Urbany 1986). Low information asymmetry (i.e.
similar levels of information among sellers and buyers) reduces
perceived price dispersion in the markets, which reduces cost
of search (Rothschild 1974), and hence is expected to reduce
information search since benefits are limited from undertaking
more information search. Similarly, higher information asymmetry (i.e. the information sellers may have that buyers lack) is
expected to increase information search. We further argue that at
very high price dispersion the cost associated with information
search will increase to an extent where search costs outweigh
the benefits obtained, and information search will be low, which
may happen in the context of product heterogeneity or when
consumers have a high personal search cost (in case of homogeneous products). Therefore, we hypothesize that information
search will be the highest at medium levels of price dispersion:
H3. The relationship between information search and price
dispersion is inverted-U.
Time pressure. Time pressure has been conceptualized as the
consumers perception of whether she had adequate time to complete a task and has been operationalized through multi-item
scales (e.g. Avery 1996), or, has been manipulated in experiments. Time pressure is an external search cost (e.g. Smith,
Venkataraman, and Dholakia 1999). Hence we hypothesize that:
H4. The relationship between information search and time
pressure is negative.

information search. Stigler (1961) calls for systematically (p.


224) accounting for all those factors that may affect consumers
information search, attributing to her ignorance. Ignorance
may refer to both risk and uncertainty (Vriend 1996). Perceived
risk is the uncertainty and the consequences associated with a
choice; it affects the extent of information search (Bauer 1960).
We, therefore, offer Potential Payoff as a category of antecedent
variables affecting information search.
The uncertainty and risk experienced by a consumer while
making a decision will affect the potential payoff expected from
that decision. In such situations information is a valuable asset
that may be used to reduce uncertainty. Consumers perceive
a wide variety of risks in the marketplace. Perceived risk is
defined as the consumers perception of the uncertainty and the
adverse consequences of buying a product (Dowling and Staelin
1994). Risk theory predicts that consumers who perceive a high
risk involved in choosing a product, may proactively search for
more information thereby trying to reduce the risk associated
with the outcome of making that particular choice (Mitra, Reiss,
and Capella 1999). Therefore, according to risk theory (e.g.
Dowling and Staelin 1994), as perceived risk increases, there
is an increase in the amount of information search undertaken.
Hence, we propose that:
Proposition 2. The relationship between information search
and potential payoff is positive.
Perceived risk. Perceived risk can be defined as the probability of any loss (financial or otherwise) that can occur as a
result of using a product, multiplied by the importance of that
loss (Srinivasan and Ratchford 1991). Information search is an
attempt to reduce the risk associated with the choice process.
Perceived risk has been measured via multiple items as risk aversion (e.g. DeSarbo and Choi 1999), or, as perceived risk (e.g.
psychological risk, social risk, financial risk Dholakia 2001).
In the meta-analysis we consider perceived risk as a single construct and acknowledge that this construct may have various
dimensions. Based on the above discussion, we hypothesize that:
H5. The relationship between information search and perceived risk is positive.
Financial constraint, uncertainty. Financial constraint is conceptualized as having a budget within which the consumer has
to make a choice (e.g. Hill 2001; Laroche et al. 2000). Also, consumers may have a range within which they would like to restrict
their expenditures (Urbany, Dickson, and Kalapurakal 1996).
Some studies find that nancial constraint is positively related to
external information search (e.g. Claxton, Fry, and Portis 1974).
As per risk theory, this restriction would result in searching for
more information. Therefore, we expect that greater financial
constraint leads to greater amount of information search.

Potential payoff

H6. The relationship between information search and nancial


constraint is positive.

Stigler (1961) points out that consumers may suffer from


ignorance (p. 224) while undertaking information search.
However, the EoI does not identify the factors that may contribute to a consumers ignorance in the context of undertaking

Uncertainty is the amount of information that a buyer brings


to the search process (e.g. Urbany 1986). The variable is often
manipulated by providing (not providing) subjects with additional information (e.g. Urbany 1986), or measured in the

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

context of an actual purchase (e.g., Moorthy, Ratchford, and


Talukdar 1997). Uncertainty may be knowledge uncertainty
(i.e. uncertainty about the information on alternatives, Urbany,
Dickson, and Wilkie 1989) or choice uncertainty (i.e. uncertainty about which alternative to choose, Laroche et al. 2004;
Urbany, Dickson, and Wilkie 1989). Greater uncertainty should
lead to greater information search undertaken. Therefore, we
hypothesize that:
H7. The relationship between information search and uncertainty is positive.
Knowledge and prior experience
We theorize that the ignorance (Stigler 1961, p. 224)
referred to in EoI also refers to the level of knowledge or experience an individual may have regarding the decision-making
situation or task. Therefore, we offer Knowledge and Experience as another category of antecedent variables that affect
information search.
Product class knowledge has been defined as subjective
knowledge or information that consumers have related to the
particular product class (Brucks 1985). Knowledge and prior
experience have often been interchangeably used to mean a
consumers familiarity with a buying situation (Bettman and
Park 1980; Brucks 1995), and researchers argue that the two
are conceptually similar. Past research examining the relationship between knowledge/prior experience and information
search has yielded equivocal results. Researchers (e.g. Jacoby,
Chestnut, and Fisher 1978) have found a positive relationship between purchasing experience and information search.
Weiss and Heide (1993) argue that a lack of prior experience necessitates extensive search on the part of the buyer
(Anderson, Chu, and Weitz 1987). Contrary to these findings,
many other researchers (Katona and Mueller 1955; Newman
and Staelin 1971) have found that increased positive experience led to less information search. Moore and Lehmann (1980)
found no relationship between prior experience and information search, and a negative relationship with product knowledge
and the central construct. These conflicting findings and arguments can probably be explained by a curvilinear relationship
between knowledge/prior experience and information search
(Malhotra 1983). Johnson and Russo (1984) also postulate an
inverted-U shaped relationship, which is validated by Moorthy,
Ratchford, and Talukdar (1997). Bettman and Park (1980) provide a possible explanation for this relationship, Those low
in knowledge/prior experience lack the ability to acquire and
process much current information, while those high in knowledge/prior experience lack the motivation to do so. The moderate
group, however, may possess enough ability and have enough
motivation to devote substantial processing effort to the task
(Malhotra 1983, p. 129).
Information is acquired to satisfy a need for change in
uncertainty (Burnkrant 1976). Following this argument, those
consumers who have a high prior knowledge/experience, assume
that they do not need more information to change the state of
uncertainty they may be in, and therefore, are not motivated to

239

search for more information. For those consumers who have low
prior knowledge/experience, the cost of acquiring information
outweighs the benefit that they associate with acquiring the current information, and so undertake limited information search.
In contrast, those consumers who have moderate information
do assume that their state of uncertainty may change by acquiring more information, and are motivated to undertake a greater
amount of information search. Based on the above discussion,
we propose that the relationship between information search and
knowledge/prior experience will be inverted-U shaped:
Proposition 3. The relationship between information search
and knowledge/prior experience is inverted U-shaped.
In our review, we treat the two variables separately depending
upon how the authors of the specific studies have operationalized
them in their investigations. Therefore, we hypothesize:
H8. The relationship between information search and knowledge is inverted U-shaped.
H9. The relationship between information search and prior
experience is inverted U-shaped.
Individual difference variables
EoI acknowledges the difference in tastes (Stigler 1961,
p. 216) among consumers which may affect information
search. We offer the category individual differences as another
antecedent category that affect consumer information search.
We expect that the direction of the relationships of different
individual difference variables with information search will be
different, and hence we do not offer any specific proposition for
the overall category of variables.
Involvement. Involvement theory was originally introduced
in social psychology, but has been adopted extensively in consumer behavior. Laaksonen (1994) defines involvement as the
amount of arousal, interest, or activation a person experiences
in a specific situation. Involvement (i.e. an arousal or concern
about the purchase decision) figures prominently in consumer
search effort theories. Researchers have argued that involvement
can be conceived in behavioral terms such as the time or energy
spent on information search, the number of alternatives considered, and the extent of the decision process (Laaksonen 1994;
Stone 1984).
Mitchell (1981) reports that, at higher involvement levels,
consumers will aggressively search for relevant information.
Petty and Cacioppo (1981) assert that high product involvement
motivates consumers to examine carefully product advantages
and disadvantages, which means greater information search.
Consumers are thought to engage in systematic or comprehensive search under high-involvement conditions, but minimize
search activity and rely on simple schemas or cognitive heuristics
under low-involvement conditions (Engel, Kollat, and Blackwell
1973). Hence, greater extent of involvement will lead to greater
amount of information search. We hypothesize that:
H10. The relationship between information search and involvement is positive.

240

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

Antecedents: Findings
Antecedents (direct effects)
For each of the ten antecedent variables, Table 2 summarizes the underlying theories, the hypothesized relationships
with information search, number of effects reported, range of
reported effects, and the cumulative n. We do not include the
antecedent benet in the meta-analysis since the number of effect
sizes obtained for this antecedent is less than ten. The findings
on the direct effects are discussed below. The conceptualization
and measurement of all the ten variables are fairly consistent
across all the studies included in the meta-analysis.
Methodology: Direct effects
Although correlations are the most common metrics found
in the 62 studies, many studies also report F, t, and chi-square
statistics. We convert these summary statistics to the common
correlation coefficient metric, r (e.g. Wolf 1986) (Table 2),
therefore ensuring the use of bivariate correlations in the metaanalysis. In order to test the hypotheses on the five inverted-U
shaped relationships (H1 , H2 , H3 , H8 , H9 ), ideally, we need
to identify at least three levels (i.e. low, medium, and high) of
each antecedent. However, different levels are not operationalized for any of the five variables in the studies included in our
meta-analysis. We, therefore, need to create post hoc categories
in order to test our hypotheses.
The relationships of the antecedent variables number of alternatives, cost and price dispersion, with information search
are explained by cost theory. In order to test the hypotheses,
therefore, we need to identify proxy variables that meaningfully represent cost of search (since the observed effect sizes
by themselves are inadequate for testing an inverted-U shaped
relationship). For these three antecedents, the parameters we
choose for creating post hoc categories that represent varying
levels of these antecedents are product category (i.e. infrequently/frequently purchased), and the similarity/dissimilarity
among products in each product category included in the studies. We argue that for infrequently purchased product categories,
searching for an additional product gives the consumer a larger
benefit and thus a greater incentive to search. Therefore, consumers are expected to search more for infrequently purchased
products and search less for frequently purchased products.
We define the similarity/dissimilarity among products as the
extent to which the products are homogeneous/heterogeneous
in that particular product category. We argue that the cost of
search is lower for homogeneous products than that associated
with the search for heterogeneous products. Four post hoc categories emerge representing four levels of increasing cost for
each antecedent variable: frequently purchased and similar, frequently purchased and dissimilar, infrequently purchased and
similar, and infrequently purchased and dissimilar. It is expected
that the cognitive cost, and hence, the cost associated with information search for frequently purchased and similar products
(e.g. groceries) will be the lowest, and that associated with infrequently purchased and dissimilar products (e.g. cars, house)

will be the highest. The other two categories represent medium


search costs.
In order to test the hypotheses for knowledge and prior
experience, we need to identify proxy variables that meaningfully represent knowledge about the product class. For the
variables knowledge and prior experience, the parameters chosen for creating the pertinent categories are the extent of
knowledge/experience (high/low), as reported in the original
studies, and product type (infrequently purchased/frequently
purchased). If participants are regular buyers (or have made
a purchase) of a product category, they are designated as
high in knowledge/experience, and vice versa. The rationale
for including product type is that the definition of knowledge/prior experience in extant literature is in the context of
product class. Therefore, we choose a meaningful proxy variable that represents consumers knowledge about a product
class.
A consumer buying a frequently purchased product regularly is likely to be quite knowledgeable about the product
class (frequently purchased and high knowledge/experience),
and the level of knowledge is likely to be higher than in
case of an infrequently purchased product with which the user
may have prior experience (infrequently purchased and high
knowledge/experience). The level of knowledge is even lower
for an infrequently purchased product with which the user
has little experience (infrequently purchased and low knowledge/experience). The consumer is expected to have the least
knowledge about a frequently purchased product with which
the user has low experience (frequently purchased and low
knowledge/experience). These four categories represent different levels of decreasing knowledge/prior experience.
The weighted mean effect size is calculated for the four subcategories identified for each of the five antecedents H1 , H2 ,
H3 , H8 , and H9 , and for the other five antecedents H4 , H5 , H6 ,
H7 , H10 . This study weights the effect size r using the relevant
sample size information (Table 3). For computational details,
see Hedges and Olkin (1985).
Findings: Direct effects
Number of alternatives and cost. For cost, as expected,
the weighted r associated with each of the four categories
demonstrate an inverted-U shaped curve (CostF S = 0.190;
CostIF S = 0.035,
CostIF DS = 0.023),
CostF DS = 0.090,
whereby, information search is high for medium levels of cost,
and low for the high and low levels of cost, thus lending support
to H2 . For number of alternatives, contrary to expectation, we
do not find an inverted-U shaped curve, and hence, no support
for H1 . However, note that the effect sizes for the various levels
are significant and different from each other. One possible
reason for this finding could be that extant studies have failed
to include adequate number of alternatives that capture the
complete spectrum of search costs.
Price dispersion. In accordance with our hypothesis, the
relationship between price dispersion and information search
displays an inverted-U shape. Findings for three subgroups
(PriceDF S = 0.200, PriceDF DS = 0.320, PriceDIF DS = 0.146, no

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

241

Table 3
Antecedent variables testing hypotheses for direct effects.
Antecedent variables
(support for hypotheses)

Levels of antecedent
variables (where applicable)

Unweighted r
(observed)

Weighted r
(corrected)

Nature of
relationship

Number of Alternatives*
H1 (No)

NAlternativesF S

10

0.009

0.046**

Linear

NAlternativesF DS
NAlternativesIF S
NAlternativesIF DS

3
3

0.160
0.192

0.138
0.302

CostF S

10

0.236

0.190**

CostF DS
CostIF S
CostIF DS

10
7
10

0.077
0.017
0.013

0.090
0.035
0.023

PriceDF S

0.195

0.200**

PriceDF DS
PriceDIF S
PriceDIF DS

0.282

0.121

0.320

0.146

Time Pressure
H4 (Yes)

21

0.20

0.14

Linear

Perceived Risk
H5 (Yes)

16

0.11

0.13

Linear

Financial Constraint
H6 (Yes)

26

0.13

0.04

Linear

Uncertainty
H7 (Yes)

22

0.27

0.24

Linear

KnowledgeF LK

18

0.078

0.093**

Inverted-U

KnowledgeIF LK
KnowledgeIF HK
KnowledgeF HK

9
11
5

0.213
0.151
0.033

0.184
0.120
0.076

ExperienceF LK

0.281

0.276**

ExperienceIF LK
ExperienceIF HK
ExperienceF HK

2
19
9

0.135
.023
0.089

0.135
0.016
.080

25

0.30

0.26

Cost
H2 (Yes)

Price Dispersion
H3 (Yes)

Knowledge
H8 (Yes)

Prior Experience
H9 (Yes)

Involvement
H10 (Yes)

Inverted-U

Inverted-U

Inverted-U

Linear

indicates that no study has been found in these categories. *Product Category not reported in one study. K = No. of Studies. All weighted and unweighted rs
are significant at <0.0001. The extensions appearing in subscript indicate: For antecedents Number of Alternatives, Cost, and Price Dispersion: (a) F S: frequently
purchased; similar; (b) F DS: Frequently Purchased; Dissimilar; (c) IF S: Infrequently Purchased; Similar; (d) IF DS: Infrequently Purchased; Dissimilar For
antecedents Knowledge and Prior Experience, (a) F LK: Frequently Purchased; Low Product Knowledge; (b) IF LK: Infrequently Purchased; Low Product Knowledge; (c) IF HK: Infrequently Purchased; High Product Knowledge; (d) F LK: Frequently Purchased; High Product Knowledge. **The correlation coefficients
associated with the different levels are different (sig. at p < 0.05).

study was found for the category PriceDIF S ) bear out that consumers undertake different amounts of information search at
different levels of price dispersion, as hypothesized, providing
support for H3 . Medium levels of perceived price dispersion
lead to the highest amount of information search. Since cost
of search is a function of price dispersion, it is expected that
at medium levels of dispersion, consumers obtain the highest
amount of information than at high or low levels of price dispersion. Therefore, though the overall relationship between price
dispersion and information search is positive, following cost
theory, we find that the amount of information search is low
when price dispersion is high (PriceDIF DS = 0.146). It is likely
that consumers do not search for more information when they

perceive high price dispersion as they might find the cost of


search too high. There could be supply-side issues as well. If
the available alternatives are perceived to be similar to each
other, the amount of search will be limited, unless it is a search
for price. The amount of information search is high when dispersion is medium (PriceDF DS = 0.320), and low for low price
dispersion (PriceDF S = 0.200).
Time pressure. As expected, the direction of the relationship between time pressure and information search is negative
(r TP = 0.14). This finding provides support to H4 .
Perceived risk, nancial constraint, uncertainty. The direction of the relationships between each of the variables
perceived risk, nancial constraint, and uncertainty are positive

242

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

(rPR = 0.13, r FC = 0.04, r U = 0.24), as expected. These


results provide support to H5 , H6 , and H7 .
Knowledge, prior experience. For knowledge, the effect
sizes for the four categories demonstrate an inverted-U shaped
curve (KnowledgeF LK = 0.093, KnowledgeIF LK = 0.184,
KnowledgeIF HK = 0.120, KnowledgeF HK = 0.076), providing support for H8 . For frequently purchased products, the
risk associated with decision-making is low, and therefore,
even when the level of knowledge is low, users undertake
low information search (KnowledgeF LK = 0.093), while the
information search is even lower when level of knowledge
is high (KnowledgeF HK = 0.076), probably demonstrating a
lack of motivation for gathering more information. We obtain
similar findings for prior experience (ExperienceF LK = 0.276, ExperienceIF LK = -0.135, ExperienceIF HK = 0.016,
ExperienceF HK = -0.080), providing support for H9 .
Involvement. The relationship between information search
and involvement is positive and strong (r I = 0.26). We find
support for H10 .
We find support for nine out of the ten hypotheses tested. The
findings on the relationships between information search and
cost and price dispersion provide new insights. The findings on
knowledge and prior experience agree with extant literature, and
provide additional support to our findings.
Moderators: Theory and ndings
Moderator analyses provide deeper insight into the
antecedents relationships with the central construct. Recall that
four of the ten antecedents have inverted-U shaped relationships
with information search. Ideally, a moderator analysis should be
carried out on each of the levels of the antecedent variables. However, note that for testing the hypotheses on the direct effects,
we used proxy parameters for creating categories that represent
the cost of search for three antecedents, and the extent of knowledge/prior experience of the participants for two antecedents,
which reduces the number of studies in each sub-group, and
moderator analyses may not be meaningfully carried out.
Therefore, for conducting moderator analysis we follow
clustering of effect sizes, as discussed by Hedges and Olkin
(1985), where the authors suggest the use of overlapping or
non-overlapping clusters. We follow the latter approach to
decompose the effect sizes for each antecedent variable into
smaller subsets, by rank-ordering u (where r is transformed into
Fishers z, and then multiplied by the square root of the sample size to obtain u). We examine the ranges of the weighted
mean effect sizes obtained for each antecedent variable that
demonstrates an inverted-U shaped relationship with information search, and find that the effect sizes for the variables cost,
price dispersion and prior experience may be categorized into
two subgroups, and those associated with the antecedent knowledge/experience may be categorized into three subgroups (Fig.
A in Web Appendix A; the x-axis represents that effect sizes, and
the y-axis represents the studies included in the meta-analysis).
The effect sizes for the categories that emerge through the
clustering approach, display valences that are similar to those
obtained for testing the hypotheses of the direct effects. For

example, for the antecedent cost, we obtain positive and negative valences for the two clusters obtained (Table 4). This study
weights the effect size r using the relevant sample size information, and also estimates the associated file drawer N (Table 4).
(Note that the results reported in Table 4 are those based on all
studies that measure information search in terms of number of
brands and time. We report the results of a separate meta-analysis
conducted on the findings based on studies that measure information search in terms of time only in Web Appendix B. The
results of the two findings are similar, demonstrating that our
findings are robust.) Homogeneity tests are carried out on a total
of fifteen clusters of direct effects, following the recommendations of Hedges and Olkin (1985), and results indicate that
moderator analysis should be carried out on thirteen relationships (Web Appendices C and D).
We make specific suppositions regarding the different clusters of effect sizes that emerge for the four antecedent variables.
For cost, the cluster that displays a weighted r of 0.09
(CostHigh/Low ) is assumed to represent either high or low level of
cost of search (or, both), and a weighted r of 0.29 (CostMedium ) is
assumed to represent the medium level of cost of search. These
assumptions are reasonable since theory as well as our findings on direct effects demonstrate that the amount of search
undertaken is the highest when cost of search is medium. Similar suppositions are made for the variables price dispersion,
knowledge and prior experience.
Hypotheses: Effect of moderators
Hypotheses are presented on thirteen relationships that are
moderated by seven moderators.
Age. Extant literature on information processing and decision making has extensively investigated the effects of age,
with conflicting findings. For example, Otnes (1990) finds that
age positively affects information search, while others (e.g.
Schaninger and Sciglimpaglia 1981) report the opposite effect.
Examining the social psychology literature on the impact of cognitive aging on the ability of decision-making, Mata, Schooler,
and Rieskamp (2007) find that search costs are higher for older
adults. Age-related cognitive decline is attributed to decreased
attentional capabilities (McDowd 1997). Kirasic et al. (1996)
show that age negatively affects information-processing speed
and working memory. This finding is in line with the theory of
cognitive competence as outlined by Schaie and Willis (1993),
which suggests that the ability to perform tasks decline with age.
The studies that report the age of respondents are categorized
as follows: <30 years, and >30 years, as per extant practice
(Irwin and McClelland 2003). In line with the theory of cognitive
competence, we predict the following:
H11a. For the moderator age, the effect size for information
search and each of the antecedents is high for consumers <30
years and low for consumers >30 years.
On the other hand, older consumers tend to be more risk
averse and are expected to undertake more information search
than the younger ones. Also, there are studies that suggest that
younger adults take less time to perform tasks predominantly

Table 4
Search for moderators.
Antecedent variables

Number of Alternatives

16

Unweighted r
(observed)
0.072

Weighted r
(corrected)

Observed
variance

Sampling
error variance

Residual variance (%
age of total variance)

File drawer Nf
(rc = .05)

Q (d.f.)

0.19

0.082

0.003

0.079 (96.72)

45

486.488a

Moderator
analysis

Cost

CostHigh/Low
CostMedium

27
10

0.16
0.27

0.09
0.29

0.02
0.018

0.004
0.005

0.003 (76.60)
0.013 (72.51)

23
48

115.38a (26)
37.3a (9)

Price Dispersion

Price DispersionHigh/Low
Price DispersionMedium

15
6

0.10
0.41

0.08
0.41

0.073
0.006

0.004
0.002

0.03 (88.58)
0.005 (74.10)

11
42

131.33a (14)
22.90a (5)

21
16
26
22

0.20
0.11
0.13
0.27

0.14
0.13
0.04
0.24

0.06
0.01
0.05
0.01

0.004
0.004
0.002
0.004

0.05 (92.58)
0.01 (53.84)
0.05 (95.58)
0.01 (67.70)

39
26
44
82

283.13a
34.64a
587.36a
67.84a

Time Pressure
Perceived Risk
Financial Constraint
Uncertainty

(20)
(15)
(25)
(21)

Knowledge

KnowledgeHigh/Low
KnowledgeHigh/Low
KnowledgeMedium

12
26
5

0.14
0.16
0.51

0.13
0.15
0.51

0.097
0.003
0.005

0.004
0.003
0.003

0.07 (57.12)
0.000 (6.54)
0.002 (55.90)

18
51
46

27.97a (11)
27.74 (25)
7.84 (4)

Prior Experience

Prior ExperienceHigh/Low
Prior ExperienceMedium

28
8

0.18
0.25

0.14
0.24

0.009
0.019

0.003
0.002

0.006 (63.47)
0.017 (87.45)

53
31

76.59a (27)
63.39a (7)

25

0.30

0.26

0.03

0.003

0.03 (92.31)

103

322.72a (24)

Involvement

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

Levels of antecedent
variables (where
applicable)

a Heterogeneous.
Nf = Fail-Safe N. rc = critical value of r.

243

244

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

due to their adaptability in information search (Gregan-Paxton


and Roedder-John 1995). Some researchers (Davids and Falkoff
1975) attribute this behavior to the younger consumers impatience, while Lucas (1972) argues that the younger generations
are greater risk takers than the older generations. Therefore, we
offer a competing hypothesis:
H11b. For the moderator age, the effect size for information
search and each of the antecedents is high for consumers >30
years and low for consumers <30 years.
Respondent type. Extant studies in the marketing literature
widely use student subjects. Some critics suggest that students
as participants are different from adults in the general population
(e.g. James and Sonner 2001). Some observe that students have
superior skills at many of the exercises required in these studies,
may be more compliant with authority, may try harder to complete the experimental tasks and may be more conscious of doing
a good job (e.g. Sears 1986). This practice of using comparatively homogenous student samples in relational research leads
to restricted ranges of effect size (Peterson 2001). This limitation
results in a bias for stronger effects than will be found among
non-student samples. The heterogeneity of the non-student sample, in contrast, increases error variance, which attenuates the
magnitude of effect and possibly underestimates the effect size
(Fern and Monroe 1996). Therefore, this moderator variable is
divided into the following two groups: (a) Students and (b)
Non-Students, and we predict that:
H12. The relationship between information search and each
of the antecedents is stronger for student sample, and is weaker
for non-student sample.
Study design. An experimental design usually generates
less error variance in the denominator of the effect size and
produces larger effect sizes (Hunter and Schmidt 1990). In addition, excluding potential confounds generally leads to stronger
relationships in an experimental setting compared with a nonexperimental one. Fern and Monroe (1996) expect that effect
sizes will be different for studies employing fixed-effects model
versus those using survey methods. Therefore, the two groups
created for this moderator variable are: (a) Experiment and
(b) Non-Experiment. Hence, the following hypothesis:
H13. The relationship between information search and each
of the antecedents is stronger for experiments, and is weaker for
non-experiments.
Product type. Consumer decision-making is influenced by
product type, which is classified as: credence versus experience versus search goods (Darby and Karni 1973; Mitra, Reiss,
and Capella 1999), convenience goods versus shopping goods
versus specialty goods (Hassanein and Head 2006), infrequently
purchased and frequently purchased goods (e.g. Putrevu and
Ratchford 1997), and hedonic and utilitarian products (Laurent
and Kapferer 1985), among others. This review adopts the search
versus experience versus credence product categorization for
product type. Nelson (1970) distinguishes between products on
search versus experience characteristics. Search characteristics
can be evaluated prior to purchase. Experience attributes can

be discerned only after purchase and consumption. Darby and


Karni (1973) extend Nelsons framework to include credence
goods, suggesting that credence attributes cannot be judged confidently by the consumer even after purchase and consumption.
We present two alternate hypotheses based on risk and cost
theory, as both points of view have been espoused in extant
literature.
According to risk theory, consumers are expected to undertake high information search for credence and experience
products, and low information search for search products (e.g.
Mitra et al. 1999). We provide a sample of products that we categorize as search, experience, and credence in our analysis, in
line with extant literature (e.g. Nelson 1970; Mitra, Reiss, and
Capella 1999): (a) Search: computer, checking account, grocery
items, household appliances; (b) Experience: television, automobile, apparel, movie; (c) Credence: life insurance, exercise
club, medicine supplements, car repair.
We argue that there is an element of customization that may be
associated with experience and credence goods (Guiltinan 1987;
Zeithaml 1981). The variability associated with these products
contributes to the uncertainty of product performance. Therefore, in accordance with risk theory, the risk associated with
these products is high, explaining the need for extensive information search. Similarly, experience products include those that
cannot be evaluated before consumption, and consumers are
likely to undertake extensive information search for these kinds
of products/services as well. Experience products include those
that have to be purchased in order to be consumed (e.g. movie), as
well as those that offer trial prior to purchase (e.g. automobiles,
apparel, shoes). For the latter type of product it is commonly
observed that consumers do undertake extensive search for information through examination of alternatives that are available in
the market. As search goods and services can be evaluated before
consumption (Anderson and Sullivan 1993), the consumer perceives a low risk for search products as compared to that for
experience and credence products, and the information search
for search products is expected to be the lowest. Hence, we
present the following hypothesis:
H14a. For the moderator product type, as per risk theory, the
effect size for information search and each of the antecedents is
high for credence and experience products, and low for search
products.
Conversely, as per cost theory consumers are less likely to
undertake extensive information search for credence and experience products due to high search costs (Nelson 1970, 1974), low
perceived benefit (since there is little or no performance information) and greater shopping effort. In contrast, consumers will
typically undertake more information search for search products
because of lower associated search costs. The amount of knowledge available to the consumer, prior to purchase, varies with
the lowest for credence products and the highest for search products. Therefore, consumers will be most sceptical of claims for
experience and credence products because these claims cannot
be verified even after purchase and immediate consumption of
goods (Nelson 1974). Hence, we present the following alternate
hypothesis on product type:

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

H14b. For the moderator product type, as per cost theory, the
effect size for information search and each of the antecedents is
high for search products, and low for credence and experience
products.
Due to space limitations predictions and findings on the
moderators gender, education and income, are reported in Web
Appendix E.
Methodology and ndings: Moderator effects
Effect sizes (Table 4) are partitioned into subgroups based
on the moderator variables (Geyskens, Steenkamp, and Kumar
1998; Viswesvaran and Sanchez 1998) (Table 5). In order to
test the hypotheses, post hoc t tests of difference in mean effect
sizes between subcategories of the same variable (e.g. experiment vs. non experiment) are computed with the Bonferonni
correction for simultaneous tests. The hypothesized relationships, the underlying theories, and the kind of support obtained
for the hypotheses, are presented in Table 5.
Age (H11a /H11b ). Of the 13 hypotheses for the moderator age,
we test eight, and find support for all eight; five hypotheses could
not be tested due to insufficient data. Overall, we find strong support for H11a and H11b . The increased cognitive capacity of the
younger consumer probably provides an explanation for the findings for costHigh/Low (rage<30 = 0.12; rage>30 = 0.16), and on
price dispersionHigh/Low (rage<30 = 0.10; rage>30 = 0.003), as the
cost associated with searching for information is high, and older
consumers search for less information. A possible reason for the
findings on uncertainty (rage<30 = 0.32; rage>30 = 0.15) could be
that in the context of uncertainty while all consumers undertake
increased information search, younger consumers are able to
process more information. Another possible explanation could
be that younger consumers find an uncertain situation more
challenging than older consumers. A similar argument explains
the observation on involvement (rage<30 = 0.42; rage>30 = 0.30).
Findings on these four antecedents provide support for H11a .
The findings for time pressure (rage<30 = 0.29;
rage>30 = 0.11) is in line with Freuds (1911/1959) developmental theories and the work of Green, Fry, and Myerson
(1994) on intertemporal choice that suggest that the level
of patience changes over the lifespan of an individual, and
that young people are more impatient than older people, and
therefore, search for less information. The findings on perceived
risk (rage<30 = 0.12; rage>30 = 0.22) agree with the reasoning
that younger consumers are more impatient, and higher risk
takers. The findings on knowledgeHigh/Low (rage<30 = 0.14;
rage>30 = 0.06) and prior experienceHigh/Low (rage<30 = 0.24;
rage>30 = 0.11) also agree with the reasoning that younger
consumers are greater risk takers and so in the context of high
or low prior experience/knowledge they are not motivated to
look for more information and mitigate the risk associated with
making a choice. These findings provide support for H11b . The
analysis for the moderator age allows us to identify boundary
conditions under which specific theories hold.
Respondent type (H12 ). Of the 13 hypotheses for the
moderator respondent type, we test eight hypotheses, and find

245

support for six; five hypotheses could not be tested due to


insufficient data. Overall, we find strong support for H12 . The
findings on costHigh/Low (rStudent = 0.15; rNon-Student = 0.07),
price dispersionHigh/Low (rStudent = 0.14; rNon-Student = 0.07),
time
pressure
(rStudent = 0.30;
rNon-Student = 0.07),
knowledgeHigh/Low (rStudent = 0.14; rNon-Student = 0.12), prior
experienceHigh/Low (rStudent = 0.18; rNon-Student = 0.12),
and involvement (rStudent = 0.56; rNon-Student = 0.24) are
as per expectations: effect sizes strengthen for student
samples.
The findings on prior experienceMedium (rStudent = 0.16;
rNon-Student = 0.25) are contrary to the hypothesized directions.
One possible reason maybe that students rely less on personal
experience compared to non-students. Research shows that students are less willing to share personal experiences compared
to non-students (e.g. Gough 2011). The findings on perceived
risk (rStudent = 0.15; rNon-Student = 0.14), though in the expected
direction, are not significant.
Study design (H13 ). Of the 13 hypotheses for the moderator
study design, we test eight hypotheses, and find support for
seven; five hypotheses could not be tested due to insufficient
data. Overall, we find strong support for H13 . The findings on
costHigh/Low (rExperiment = 0.20; rNon-Experiment = 0.06), number of alternatives (rExperiment = 0.33; rNon-Experiment = 0.19),
time pressure (rExperiment = 0.38; rNon-Experiment = 0.09),
perceived risk (rExperiment = 0.18; rNon-Experiment = 0.14), nancial constraint (rExperiment = 0.22; rNon-Experiment = 0.03),
prior
experienceHigh/Low
(rExperiment = 0.42;
rNon-Experiment = 0.12), and involvement (rExperiment = 0.31;
rNon-Experiment = 0.25), are as per expectations, where the
findings from experiments display a strengthened effect size as
compared to those from non-experiments.
The findings on price dispersionHigh/Low (rExperiment = 0.02;
rNon-Experiment = 0.09) are not as per the hypothesis. Note that the
number of observations in the first cell (i.e. experiment) is one,
and therefore, sample size could be a possible reason for this
finding.
Product type (H14a /H14b ). Of the 13 hypotheses for
the moderator product type (search/experience/credence),
we test nine hypotheses, and find support for six; four
hypotheses could not be tested due to insufficient data.
The findings for price dispersionHigh/Low (rSearch = 0.03;
rExperience = 0.19; rCredence = 0.21), nancial constraint
(rSearch = 0.01; rExperience = 0.25; rCredence = 0.09) and involvement (rSearch = 0.23; rExperience = 0.35; rCredence = 0.32), provide
support for risk theory, with consumers undertaking greater
information search for experience and credence products.
The findings for number of alternatives (rSearch = 0.26;
rCredence = 0.18),
knowledgeHigh/Low
rExperience = 0.15;
(rSearch = 0.12; rExperience = 0.14; rCredence = 0.16) and
prior experienceHigh/Low (rSearch = 0.11; rExperience = 0.20;
rCredence = 0.19), provide support for cost theory, with
consumers undertaking greater information search for search
products. These findings provide support for our supposition
that the same product categories have different impact on
consumer information search based on the situation consumers
find themselves in.

246

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254


Table 5
Moderator analysis: hypotheses and findings.

*p < 0.05 (t-test); n.c.: not calculated. Multiple regression did not enter this variable in the regression model; n.a.: not available, since
enough information is not present due to non reporting; K in parentheses.
a Hypothesized direction of effect size for each moderator group; = High; = Low.
b Support obtained (Y)/not obtained (N).
CC Cognitive Competence.
c Read: cognitive competence provides an explanation for this relationship.

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

The
findings
on
costHigh/Low
(rSearch = 0.09;
rExperience = 0.19;
rCredence = 0.01)
and
costMedium
(rSearch = 0.30; rExperience = 0.19; rCredence = 0.32) are not
as per expectations. Note that the effect sizes for the two levels
of the antecedent cost display similar patterns: effect sizes
for experience products are the lowest. Experience products
are usually hedonic (e.g. meal at a restaurant), while search
and experience products are often utilitarian (e.g. opening a
bank account, buying insurance) in nature. Information on
product attributes is processed by consumers when undertaking
information search. We argue that search and credence products can be described in terms of attributes that are usually
utilitarian (and not hedonic) in nature. However, attributes
describing experience products are usually more hedonic (and
less utilitarian). These differences in the nature of attributes
increase the cost of processing this information associated
with experience products. The cognitive processes involved
in processing information on attributes for the different types
of products, therefore, probably provides an explanation for
the findings for the two levels of cost of information search.
The findings on perceived risk (rSearch = 0.16; rExperience = 0.13;
rCredence = 0.13) are not significant. Overall, we find good
support for H14a/b . The analysis for the moderator product type
also allows us to identify boundary conditions under which
specific theories hold.
Additional analysis. We undertake additional analysis on a
different type of product type, subcategorizing the effect sizes
by infrequently purchased/frequently purchased products (HP ,
Table 5). The reason for undertaking this additional analysis is
to test an alternative product type categorization that is widely
used and extensively adopted in extant literature, and is a means
for validating the findings obtained for the other moderators.
Infrequently purchased products are those that are not bought
often, and therefore, consumers are expected to undertake more
extensive information search on these products than on the ones
that are frequently purchased. Habit has been shown to exert a
strong influence in the case of frequently purchased products and
services. Also, infrequently purchased products usually call for
a higher monetary investment, which further explains the need
for undertaking more information search. The marginal benefit
associated with information search for an infrequently purchased
product is more than that associated with that of a frequently
purchased product. We expect that most of the hypothesized
relationships are dictated by perceived risk. However, we note
an exception. We expect that when consumers have low product
knowledge/prior experience they will undertake less information search for infrequently purchased products due to the cost
associated with searching for information.
Based on the above arguments, we make the following predictions:
HP . For the moderator product type (infrequently purchased/frequently purchased), the effect size for information
search and each of the antecedents is high for infrequently
purchased products, and low for frequently purchased products.
Of the 13 hypotheses for the moderator product type (infrequently purchased/frequently purchased product), we test 13

247

hypotheses, and find support for seven. Overall, we find


some support for HP . The finding for price dispersionHigh/Low
(rInfrequently purchased = 0.05; rFreq Purch = 0.11) is unexpected.
This finding may be because in the context of high dispersion, consumers find it costly to undertake extensive search,
while in the context of low perceived price dispersion they
do not expect any substantial gains by undertaking additional information search for infrequently purchased products.
Another possible explanation could be attributed to the presence
of individuals who have high search costs for homogeneous
products (e.g. Varian 1980). The finding for time pressure
(rInfreq Purch = 0.38; rFreq Purch = 0.10) probably indicates that
especially for infrequently purchased products (i.e. more complex decision-making situations), consumers find it difficult to
process information in situations where time is a constraint, and
undertakes lower information search. The finding for nancial
constraint (rInfreq Purch = 0.03; rFreq Purch = 0.16) is contrary to
expectations. One possible reason for such behavior may be
that in situations of financial constraint, consumers have access
to less number of choices when choosing infrequently purchased products, and access to a greater number of choices
when choosing frequently purchased products. Consumers
appear to display a greater involvement (rInfreq Purch = 0.22;
rFreq Purch = 0.26) with frequently purchased products. A possible reason may be that frequently purchased products provide
consumers with an opportunity for seeking variety than does the
infrequently purchased product category. The findings for perceived risk (rInfreq Purch = 0.14; rFreq Purch = 0.14) and costMedium
(rInfreq Purch = 0.30; rFreq Purch = 0.28) are not significant.
Multivariate analysis
In this section we present and test a holistic model on
consumer information search that includes several inverted-U
shaped variables as uncovered in the analysis of direct effects.
Such a model of offline information search should ideally
include all the ten significant variables identified in this analysis
as antecedents to information search, and the intercorrelations
among these ten antecedents (e.g. involvement-cost; dispersionnumber of alternatives) from past studies need to be identified for
testing such a model. A search for these intercorrelations among
the antecedents, however, yields intercorrelations for only four
of those variables (prior experience, knowledge, perceived risk,
and cost), and benet. We include the variable benet in this
validation model. (The rationale for including this variable is to
ensure the representation of benet in the costbenefit paradigm.
Note that we formulate PAc specifically on the variable benet,
but are not able to include it in the meta-analysis as the number
of effect sizes uncovered are less than ten (see Web Appendix
F).) A correlation matrix containing all these variables needs to
be constructed in order to test the holistic model.
In our review, we find that the relationships between information search and the antecedent variables knowledge, prior
experience, and cost are inverted-U, where the first variable has
three distinct effect sizes, while the later two variables have
two distinct effect sizes each. Therefore, the correlation matrix
should contain the intercorrelations among all the following

248

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

variables: information search, benet, knowledge1, knowledge2,


knowledge3, prior experience1, prior experience2, perceived
risk, cost1, and cost2. In extant literature, however, the relationships between the different levels of each of the three antecedents
that have inverted-U relationships with information search (e.g.
knowledge3, prior experience1), and the other antecedents are
not reported. Therefore, we construct a correlation matrix using
correlations reported in extant literature, and impute reasonable values specifically for relationships that have not been
investigated in extant literature (see Web Appendix G). Such
an approach is often followed in meta-analytic reviews, where
the findings obtained on direct effects are further validated
through multivariate approaches like multiple regression procedure, or structural equation modeling (Szymanski and Henard
2001). Note that the variables knowledge1, knowledge2, and
knowledge3 are the different levels representing the variable
knowledge. Similarly, prior experience (prior experience1 and
prior experience2), and cost (cost1 and cost2) are conceptualized
as having two levels each.
A baseline model is formulated where the relationships
between the antecedent variables (benet, knowledge1, knowledge2, knowledge3, prior experience1, prior experience2,
perceived risk, cost1, and cost2, as discussed above) and information search are considered (Model 1a) (Fig. 1 a) (see Grace
and Keeley 2006) for a discussion on nonlinear modeling). The
results of a structural equations model (using AMOS 18.0)
yield a significant chi-square (2 = 264.17, df = 31; p = 0.000)
for Model 1a, suggesting that the model be rejected. To assess
the overall fit of the model, we reviewed a number of goodnessof-fit indices, including RMSEA (.073), CFI (.843), TLI (.773),
indicating that the model does not have an acceptable fit. Modification indices indicate several antecedents relationships with
benet. Model 1b (Fig. 1a) is tested, and the fit indices suggest that the model has an acceptable fit (2 = 93.00, df = 25,
p = 0.000; RMSEA = 0.044; CFI = 0.954; TLI = 0.918) (Web
Appendix H). Two alternate models, Model 2a and Model 2b
(Fig. 1b), are tested with knowledge, prior experience and cost
as composite variables (Grace and Keeley 2006) (Web Appendix
I). Neither of these models provides an acceptable fit. Model 1b
displays the best fit.
In Model 1b, all estimated coefficients of the antecedents to
information search are significant, except knowledge2 (Fig. 1a).
The coefficients associated with knowledge1 ( = 0.225) and
knowledge3 ( = 0.482) indicate that they are both significant,
and that the directions of the coefficients are opposite to each
other, supporting the hypothesized nonlinear relationship with
information search. Similar findings are reported for prior
experience1 ( = 0.113), prior experience2 ( = 0.241), cost1
( = 0.189) and cost2 ( = 0.195). These findings validate our
hypotheses that the variables knowledge, prior experience and
cost have nonlinear relationship with information search. Note
that similar results regarding nonlinearity of the three variables
knowledge, prior experience and cost hold across all the four
tested models. In order to further test the importance of knowledge1, knowledge3, prior experience1, prior experience2, cost1
and cost2, each of these paths are deleted and Model 1b is tested
again. In each case, the chi-square value increased, indicating

the importance of considering the different levels of knowledge,


prior experience and cost separately.
General discussion
In conducting the meta-analysis, we make several contributions to the marketing literature. First, we present an overarching
framework for consumer information search, and test it. The
framework that we propose is based on the EOI theory and
presents a parsimonious categorization of antecedents to consumer information search. Second, this study demonstrates that
the relationships of the variables cost, price dispersion, knowledge and prior experience with information search are invertedU shaped. Note that the inverted-U shape of the relationships for
knowledge and prior experience are well-established in extant
literature. Our findings on moderator analysis demonstrate that
effect sizes are stronger for students (vs. non-students) and for
experiments (vs. surveys) results that are also well-established
in the marketing literature. As many of our findings agree with
those established in extant literature, they provide us with confidence that our findings on the direct effects and moderators
are robust. Therefore, this study reconciles inconsistent findings,
and urges researchers to re-examine their understanding of effect
sizes (and sometimes the directions) of specific antecedents to
offline information search, and offers an improved understanding of the general strength and variability of the relationships
between several antecedent variables and information search.
The inverted-U shaped relationships, especially for cost and
price dispersion, present a new, different and updated view on
how these factors affect consumer information search.
Third, this study conducts moderator analyses to examine
whether study conditions further explain the direct effects. Our
analyses identify boundary conditions under which specific theories hold. Note that of the 65 hypotheses for the moderators, we
are able to test 46, and find support for 36 (we do not find support
for ten hypotheses, and we cannot test nineteen hypotheses due
to insufficient data). A support for the majority of the hypotheses
that we were able to test further enhances our confidence in our
findings and makes a significant contribution to the literature.
Implications for retailing and e-tailing
Our findings have strong implications for retailing as well
as for e-tailing. We find that information search undertaken by
younger consumers is high under conditions of high uncertainty.
In order to reduce uncertainty, marketers may need to understand
the nature of uncertainty associated with the purchase process
and provide information accordingly. As Cox and Rich (1964)
suggest, the uncertainty experienced by a consumer may be associated with the product, the brand, and the place of purchase,
among others, and that the consumer may abandon the purchase
decision-making process if the level of uncertainty continues
to remain high. This finding is pertinent in the context of ecommerce with younger consumers increasingly making most
of their purchase decisions online.
The findings on the moderating effect of age on involvement, suggests that the younger consumers probably search for

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

249

(a)
Cost1
(0.137***)

-0.189***
(-0.192***)

Cost2
(0.128***)
Benefit
(R2=0.108)

0.195***
(0.198***)

0.354***
(0.361***)

(0.099***)

Knowledge1

-0.225***
(-0.229***)

Knowledge2

0.024
(0.025)

Knowledge3

0.482***
(0.491***)

(0.084***)
(0.197***)

(0.091***)

Information Search
R2=0.567
(R2=0.552)

0.086***
(0.087***)
Perceived Risk

-0.113***
(-0.114***)
Prior Exp1
0.241***
(0.245***)
Prior Exp2

(b)
Cost1

Cost2

0.757a
(0.790a)

Cost

(0.000***)

Benefit
(R2=0.044)

Knowledge1
(0.196***)

-0.738***
(-0.735***)

Knowledge2
Knowledge3

0.087***
(0.087***)

Perceived Risk

Prior Exp1

(0.074)
-0.419
(-0.438)
0.061a
(0.047a)

0.256***
(0.261***)
0.352***
(0.361***)

Knowledge

0.087***
(0.087***)

0.942
(0.938)

Information Search
R2=0.569
(R2=0.552)

0.236***
(0.269***)
Prior Exp

-0.429***
(-0.426***)
0.912a
(0.913a)

Prior Exp2
1. Solid lines represent relationships between antecedents and information search, as proposed by Models 1a and 2a. Results of Models 1a
and 2a are in parentheses. Dotted lines represent additional relationships suggested by modification indices, as suggested by Models 1b and
2b. Results of Models 1b and 2b are not in parentheses.
2. *** indicates significance at p < .01 level; ** indicates significance at p < .05 level; * indicates significance at p < .1 level

Fig. 1. (a) Model 1a and Model 1b. (b) Model 2a and Model 2b.

250

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

more information because of a higher level of involvement compared to older consumers. This finding suggests that in providing
information, retailers may use affect-based appeals for younger
consumers and cognition-based appeals for older consumers
online as well as offline. The finding on the moderating effect of
age on time pressure, suggests that promotion and product cues
that lay an increased emphasis on aspects that shorten the information search stage and help consumers make choices quickly,
may be appropriate for younger consumers.
Even though consumers generally undertake a greater amount
of information search for infrequently purchased products, under
the conditions of price dispersion (high or low), and time pressure they are likely to undertake a limited search for infrequently
purchased products compared to frequently purchased products.
This observation suggests that the increased cost of information
search is a likely reason. An implication for retailers is that they
may provide price-related information for some of the top competing brands in the product category, thereby encouraging the
consumer to buy from that particular retailer, as well as increasing consumer trust in the retailer. It is especially convenient to
provide this information on retailers websites.
Consumer interactions with stores and brands to obtain
information prior to purchasing a product/service are increasingly taking place on the Internet, often through a variety of
devices (accessed through desktop computers, laptops or mobile
devices). However, this assertion may not apply equally to all
product categories. For example, consumers information search
and purchase of frequently purchased products (e.g. grocery
items a product category that is strongly represented in the
studies included in the meta-analysis) from online stores are limited. Also, consumers reach out across various channels to shop.
Retailers, therefore, embrace a multichannel context, leveraging
traditional as well as new and evolving channels to reach out to
consumers. In sum, the traditional (i.e. offline) channel continues to be an important component in a retailers channel mix,
and consumer search behavior in this channel is an activity that
retailers may ignore at their own peril.
We expect that the categories of variables affecting consumer
information search investigated in this meta-analysis, are likely
to influence search undertaken across new and evolving channels as well. For example, Korgaonkar, Silverblatt, and Girard
(2006) suggest that in the online context, perceived risk is greater
for experience goods than for search goods. Zauberman (2003)
demonstrate that in the context of online consumer information search, cost of search is a key determinant of the amount
of search undertaken. Therefore, the categories of antecedent
variables that we investigate in our meta-analysis (e.g. cost, perceived risk, knowledge) as well as the moderating variables (e.g.
product type search, experience, credence), in the context of
offline information search, are likely to affect consumer information search in the online context as well (e.g. Konus, Verhoef,
and Neslin 2008).
Implications for theory and research
Main effects. First, the review makes it apparent that empirical investigations in the area of consumer information search

largely focus on studying the effects of a handful of antecedent


variables on the central construct. For example, the relationship
between cost and information search has been extensively modeled. However, the relationship between benet and information
search (weighted r = 0.40, Web Appendix F), despite displaying a strong effect size, remains under-investigated. This finding
is further corroborated through the validation of the overall
model (Model 1b). Future studies should explicitly include the
variable benet in the context of studying consumer information
search.
Second, our findings provide evidence that the relationships between several antecedents and information search are
inverted-U shaped. Specifically, for the antecedent cost, we
uncover both positive and negative relationships with consumer
information search. Our findings call attention for the need to
explicitly model the different levels of cost of search in future
studies.
Third, we also find an inverted-U shaped relationship for
price dispersion, suggesting that even though the overall effect
size between this antecedent and information search is positive,
there may be variations in the degree of the effect depending on
the underlying reason for the perceived dispersion. For example,
when price dispersion is high, consumers search for less information. One reason could be that consumers may learn where to
look for low prices as they learn that market prices are dispersed
(Urbany 1986). Another reason could be that there is less benefit
that is expected from this search. Studies included in the metaanalysis do not measure price dispersion after controlling for
the differences in attributes of products. Since price dispersion
has traditionally been used as an indicator of search benefit in
information economics (Tellis and Wernerfelt 1987), there is a
need for accounting for the different degrees of benefit that may
be obtained through search.
Fourth, our findings establish that cost and price dispersion have an inverted-U shaped relationship with information
search. Economics literature suggests that search cost and price
dispersion are related (e.g. Carlson and McAfee 1983; Stahl
1989). Consumers undertake limited information search in a
marketplace characterized by high price dispersion (when search
costs are high). Conversely, when the market is competitive
(i.e. limited price dispersion), consumers also undertake limited
information search. Similarly, if all consumers have the same
positive search costs, there will also be no price dispersion and
no information search (the Diamond paradox).
Fifth, for knowledge and prior experience, we are able to
demonstrate an inverted-U shaped relationship with information
search, where at medium level of the antecedent variables, the
amount of information search undertaken is the highest, which
agrees with the findings reported by Moorthy, Ratchford, and
Talukdar (1997). Sixth, the review also provides evidence that
the effect sizes between information search and the antecedents
including perceived risk, nancial constraint, uncertainty and
involvement are positive, and that between time pressure and the
central construct is negative, as per expectations.
Contingency effects. Moderating effects of several moderators including, study conditions (respondent type, study design),
product categories (search, experience, credence/infrequently

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

purchased, frequently purchased), and demographic characteristics (age) are tested on the direct effects. Overall, we find
strong support for all the moderators. Through the moderator
analysis we are able to identify specific conditions under which
certain theories hold (Table 5). For the moderator age, we offer
competing hypotheses based on the theory of cognitive competence and that of risk theory. Our findings suggest that theory of
cognitive competence explains the moderating effect of age for
the variables costHigh/Low , price dispersionHigh/Low , uncertainty
and involvement. Risk theory provides an explanation for the
moderating effect on the variables time pressure, perceived risk,
knowledgeHigh/Low and prior experienceHigh/Low.
Our findings on the moderator product type (search, experience, credence), identifies conditions in which cost theory or
risk theory provide explanation for the observed effect sizes. The
moderating effect of search/experience/credence product type
on consumer information search under the conditions of price
dispersionHigh/Low , nancial constraint and involvement may be
explained by risk theory. Similarly, the tenets of cost theory probably provide an explanation for the moderating effect on number
of alternatives, knowledgeHigh/Low and prior experienceHigh/Low.
Similarly, for another product type categorization (infrequently purchased/frequently purchased), we find that it is
risk theory that provides explanation for the moderating effect
on most of the antecedent variables. Our moderator analysis
also indicates that there are several relationships (e.g. price
dispersion search/experience/credence product type; prior
experience search/experience/credence product type) that have
not received adequate attention in the context of offline consumer
information search. The information search literature will benefit from undertaking studies on these relationships that have so
far received limited attention.

251

be embraced, which explicitly incorporates and manipulates the


different levels of the antecedent variables an approach that
is currently missing in literature. More experimental studies are
also called for as our meta-analysis finds that experiments display a strengthened effect size as compared to non-experiments.
Also, a number of other variables that are not relevant to
consumer information search in the offline context may impact
information search in the new and evolving channels. Many
of the new and emerging channels (for undertaking shopping)
have become a reality because of developments in technology.
Concomitantly, users of these channels are expected to interact
with technology for fulfilling information and shopping needs.
A set of factors that probably impact search in these channels
include variables that capture the users interaction with technology, such as self efcacy, afnity to technology, ease of use,
usefulness, and use of decision aids, to name a few. Also, the
phenomenon of word-of-mouth (WOM) is expected to have
an increased significance in the context of online information
search as compared to that of offline context. The enthusiastic
adoption of social media (including social networking websites,
review posting sites, product sites that include reviews of products), and the ensuing electronic (e-WOM) have implications for
the amount of information search undertaken by a consumer.
A separate meta-analysis needs to be conducted to investigate
the effects of antecedents on information search in online settings. All studies included in this meta-analysis are set in offline
traditional retailing settings, and the choice to focus the investigation to this specific format is conscious and logical. This article
hopes to provoke future research, especially collating consumer
information search behavior from newer retail formats, which
can then be compared to the findings presented here. Such comparisons will expand the understanding of information search
behavior of consumers across various retailing formats.

Limitations and directions for future research


The theoretical, substantive and managerial contributions
outlined earlier, notwithstanding, the study does have limitations. This study uses the results of bivariate findings in the
meta-analysis. Only those studies that are accessible in the
English-language databases are included in the study. Second,
intercorrelations among all the ten antecedent variables cannot
be located. Therefore, the holistic model including all the ten
variables cannot be tested.
Our findings open up several hitherto unrecognized avenues
for future research. It is evident from the reported findings that
several antecedents to information search are inverted-U shaped,
suggesting that it is imperative that these variables be so considered in future studies. Importantly, future investigations need to
consider the different levels of the same variable. Such attempts
will help in providing the much-needed data that we were unable
to find for the partially filled cells that put a constraint on
hypothesis testing. These future studies can generate the required
correlations to test a more comprehensive model that include
many or all of the antecedents identified in this meta-analysis.
We also note that studies in the information search literature
have mostly concentrated on the survey methodology, indicating that a greater use of the experimental methodology needs to

Acknowledgements
The authors wish to thank Professors Ruth Bolton, Shelby
Hunt, Brian Ratchford, and Dr. Leslie Vincent for their comments on an earlier version of this paper. The authors would
also like to thank the editors and the three anonymous reviewers
for providing a truly collegial review process.
Appendix A. Supplementary data
Supplementary material related to this article can be
found, in the online version, at http://dx.doi.org/10.1016/
j.jretai.2014.03.001.
References
Anderson, Erin, Wujin Chu and Barton A. Weitz (1987), Industrial Purchasing: An Empirical Exploration of the Buyclass Framework, Journal of
Marketing, 51 (July), 7186.
Anderson, Eugene W. and Mary W. Sullivan (1993), The Antecedents and
Consequences of Customer Satisfaction for Firms, Marketing Science, 12
(2), 12543.

252

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

Assmus, Gert, John U. Farley and Donald R. Lehmann (1984), How Advertising
Affects Sales: Meta-Analysis of Econometric Results, Journal of Marketing
Research, 21 (1), 6574.
Avery, Rosemary J. (1996), Determinants of Search for Nondurable Goods: An
Empirical Assessment of the Economics of Information Theory, Journal of
Consumer Affairs, 30 (2), 390420.
Baradell, Janet G. and Kitty Klein (1993), Relationship of Life stress and Body
Consciousness to Hypervigilant Decision Making, Journal of Personality
and Social Psychology, 64 (2), 26773.
Bauer, Raymond A. (1960), Consumer Behavior as Risk-Taking, in Dynamic
Marketing for a Changing World, Hancock R. S. ed. Chicago: American
Marketing Association, 38998.
Beatty, Sharon E. and Scott M. Smith (1987), External Search Effort: An Investigation across Several Product Categories, Journal of Consumer Research,
14 (June), 8395.
Bettman, James R. (1974), Toward a Statistics for Consumer Decision Net
Models, Journal of Consumer Research, 1 (1), 7180.
(1979), An Information Processing Theory of Consumer
Choice, Reading, MA: Addison-Wesley.
Bettman, James R. and Whan C. Park (1980), Effects of Prior Knowledge
and Experience and Phase of the Choice Process on Consumer Decision
Processes: A Protocol Analysis, Journal of Consumer Research, 7 (3),
23448.
Bloch, Peter H., Daniel L. Sherrel and Nancy M. Ridgeway (1986), Consumer Search: An Extended Framework, Journal of Consumer Research,
13 (June), 11926.
Bodur, H. Onur and Lissa Matyas (2008), When Do Consumers Prefer More
Choice? Moderating Effects of Regulatory Focus, Advances in Consumer
Research, 35, 7878.
Brucks, Merrie (1985), The Effects of Product Class Knowledge on Information
Search Behavior, Journal of Consumer Research, 12 (June), 116.
Burnkrant, Robert E. (1976), A Motivational Model of Information Processing
Intensity, Journal of Consumer Research, 3, 2130.
Carlson, John A. and R. Preston McAfee (1983), Discrete Equilibrium Price
Dispersion, Journal of Political Economy, 91 (3), 48093.
Chen, Chien-Chung (2012), Detecting Moderator Effects On Construct Relationships In Empirical Sales Research: A Meta-Analysis, Journal of
Management & Marketing Research, 11 (September), 118.
Cheon, Hongsik John, Chang-Hoan Cho and John Sutherland (2007), A MetaAnalysis of Studies on the Determinants of Standardization and Localization
of International Marketing and Advertising Strategies, Journal of International Consumer Marketing, 19 (4), 10947.
Claxton, John, Joseph N. Fry and Bernard Portis (1974), A Taxonomy of Prepurchase Information Gathering Patterns, Journal of Consumer Research,
1 (December), 3542.
Cox, Donald F. and Stuart U. Rich (1964), Perceived Risk and Consumer
Decision-Making: The Case of Telephone Shopping, Journal of Marketing
Research, 1 (4), 329.
Davids, Anthony and Bradley B. Falkoff (1975), Juvenile Delinquents Then and
Now: Comparisons of Findings from 1959 and 1974, Journal of Abnormal
Psychology, 84 (April), 1614.
Darby, Michael R. and Edi Karni (1973), Free Competition and the Optimal Amount of Fraud, Journal of Law and Economics, 16 (April),
6788.
DeSarbo, Wayne S. and Jungwhan Choi (1999), A Latent Structure Double
Hurdle Regression Model for Exploring Heterogeneity in Consumer Search
Patterns, Journal of Econometrics, 89, 42355.
Dholakia, Utpal M. (2001), A Motivational Process Model of Product Involvement and Consumer Risk Perception, European Journal of Marketing, 35
(11/12), 134060.
Dowling, Grahame R. and Richard Staelin (1994), A Model of Perceived Risk
and Intended Risk-Handling Activity, Journal of Consumer Research, 21
(June), 11934.
Engel, James F., David T. Kollat and Roger L.D. Blackwell (1973), Consumer
Behavior, 21st ed. New York: Free.
Eysenck, Michael W. and Manuel G. Calvo (1992), Anxiety and Performance: The Processing Efficiency Theory, Cognition and Emotion, 6,
40934.

Fern, Edward F. and Kent B. Monroe (1996), Effect Size Estimates: Issues
and Problems in Interpretation, Journal of Consumer Research, 23 (2),
89105.
Freud, Sigmund (1911/1959), Formulations Regarding the Two Principles of
Mental Functioning, New York: Basic Books.
Gemunden, Hans Georg (1985), Perceived Risk and Information Search: A
Systematic Meta-analysis of the Empirical Evidence, International Journal
of Research in Marketing, 2, 79100.
Geyskens, Inge, Jan-Benedict E.M. Steenkamp and Nirmalya Kumar (1998),
Generalizations About Trust in Marketing Channel Relationships Using
Meta-Analysis, International Journal of Research in Marketing, 15,
22348.
Glass, Gene V. (1976), Primary, Secondary, and Meta-Analysis of Research,
Educational Research, 5 (10), 38.
Grace, James B. and Jon E. Keeley (2006), A Structural Equation Model
Analysis of Postfire Plant Diversity in California Shrublands, Ecological
Applications, 16, 50314.
Green, Leonard, Astrid F. Fry and Joel Myerson (1994), Discounting of Delayed
Rewards: A Life-Span Comparison, Psychological Science, 5, 336.
Gregan-Paxton, Jennifer and Deborah Roedder-John (1995), Are Young Children Adaptive Decision Makers? A Study of Age, Journal of Consumer
Research, 21 (4), 56780.
Gough, Matthew (2011), Looking After Your Pearls: The Dilemmas of Mental
Health Self-Disclosure in Higher Education Teaching, Journal of Mental
Health Training, Education & Practice, 6 (4), 20310.
Guiltinan, Joseph P. (1987), The Price Bundling of Services: A Normative
Framework, Journal of Marketing, 51, 7485.
Hassanein, Khaled and Milena Head (2006), The Impact of Infusing Social
Presence in the Web Interface: An Investigation Across Product Types,
International Journal of Electronic Commerce, 10 (2), 3155.
Hedges, Larry V. and Ingram Olkin (1985), Statistical Methods for MetaAnalysis, Orlando, FL: Academic Press.
Hill, C. Jeanne (2001), The Nature of Problem Recognition and Search in the
Extended Health Care Decision, Journal of Services Marketing, 15 (6/7),
45479.
Hockey, G. Robert J. (1970), Effect of Loud Noise on Attentional Selectivity,
Quarterly Journal of Experimental Psychology, 22, 2836.
Hunter, John E. and Frank L. Schmidt (1990), Methods of Meta-Analysis:
Correcting Error and Bias in Research Findings, Newbury Park: Sage Publications.
Irwin, Julie R. and Gary H. McClelland (2003), Negative Consequences
of Dichotomizing Continuous Predictor Variables, Journal of Marketing
Research, 60 (August), 36671.
Jacoby, Jacob, Donald E. Speller and Carol K. Berning (1974), Brand Choice
Behavior as a Function of Information Load: Replication and Extension,
Journal of Consumer Research, 1, 3342.
Jacoby, Jacob, Robert A. Chestnut and William A. Fisher (1978), A Behavioral
Process Approach to Information Acquisition in Nondurable Purchasing,
Journal of Marketing Research, 15 (November), 53244.
James, William L. and Brenda S. Sonner (2001), Just Say No to Traditional Student Samples, Journal of Advertising Research, 41 (SeptemberOctober),
6371.
Johnson, Eric J. and J. Edward Russo (1984), Product Familiarity and Learning
New Information, Journal of Consumer Research, 11 (June), 54250.
Kahneman, Daniel (1973), Attention and Effort, Englewood Cliff, NJ: PrenticeHall.
Katona, George and Eva Mueller (1955), A Study of Purchase Decisions, in
Consumer Behavior: The Dynamics of Consumer Reaction, Clark L. H. ed.
New York: New York University Press, 3087.
Kirasic, Kathleen C., Gary L. Allen, Shannon H. Dobson and Katherine S.
Binder (1996), Aging, Cognitive Resources, and Declarative Learning,
Psychology and Aging, 11 (4), 65870.
Konus, Umut, Peter C. Verhoef and Scott A. Neslin (2008), Multichannel
Shopper Segments and Their Covariates, Journal of Retailing, 84 (4),
398413.
Korgaonkar, Pradeep, Ronnie Silverblatt and Tulay Girard (2006), Online
Retailing, Product Classifications, and Consumer Preferences, Internet
Research, 16 (3), 26788.

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254


Laaksonen, Pirjo (1994), Consumer Involvement: Concepts and Research, London, UK: Routledge.
Laroche, Michael, Gad Saad, Mark Cleveland and Elizabeth Browne (2000),
Gender Differences in Information Search Strategies for a Christmas Gift,
Journal of Consumer Marketing, 17 (6/7), 50024.
Laroche, Michel, Mark Cleveland and Elizabeth Browne (2004), Exploring
Age-Related Differences in Information Acquisition for a Gift Purchase,
Journal of Economic Psychology, 25 (1), 6195.
Laurent, Gilles and Jean-Noel Kapferer (1985), Measuring Consumer Involvement Profiles, Journal of Marketing Research, 49, 4153.
Lucas, Robert E. (1972), Expectations and the Neutrality of Money, Journal
of Economic Theory, 4, 10324.
Malhotra, Naresh K. (1982), Information Load and Consumer Decision Making, Journal of Consumer Research, 8 (4), 41931.
(1983), On Individual Differences in Search Behavior
for a Nondurable, Journal of Consumer Research, 10 (June), 12531.
Mata, Rui, Lael J. Schooler and Jorg Rieskamp (2007), The Aging Decision
Maker: Cognitive Aging and the Adaptive Selection of Decision Strategies,
Psychology and Aging, 22 (4), 796810.
McDowd, Joan M. (1997), Inhibition in Attention and Aging, Journals of
Gerontology Series B: Psychological Sciences and Social Sciences, 52,
26573.
Milford, James T. and Raymond P. Perry (1977), A Methodological Study of
Overload, Journal of General Psychology, 97, 1317.
Miller, George A. (1956), The Magical Number Seven, Plus or Minus Two:
Some Limits on Our Capacity for Processing Information, Psychological
Review, 63, 8197.
Mitchell, Andrew A. (1981), The Dimensions of Advertising Involvement, In
Advances in Consumer Research, Vol. 8, Monroe Kent B. ed. Provo, UT:
Association for Consumer Research, 2530.
Mitra, Kaushik, Michelle C. Reiss and Louis M. Capella (1999), An Examination of Perceived Risk, Information Search and Behavioral Intentions in
Search, Experience and Credence Services, Journal of Services Marketing,
13 (3), 20828.
Moore, William L. and Donald R. Lehmann (1980), Individual Differences in
Search Behavior for a Nondurable, Journal of Consumer Research, 7 (3),
296307.
Moorthy, Sridhar, Brian T. Ratchford and Debabrata Talukdar (1997), Consumer Information Search Revisited: Theory and Empirical Analysis,
Journal of Consumer Research, 23 (4), 26377.
Nelson, Philip (1970), Information and Consumer Behavior, Journal of Political Economy, 78 (MarchApril), 31129.
(1974), Advertising as Information, Journal of Political
Economy, 82 (JulyAugust), 72954.
Newman, Joseph W. and Richard Staelin (1971), Multivariate Analysis of Differences in Buyer Decision Time, Journal of Marketing Research, 8 (May),
1928.
Newman, Joseph W. (1977), Consumer External Search: Amount and Determinants, in Consumer and Industrial Buying Behavior, Woodside Arch, Sheth
Jagdish and Bennet Peter, eds. New York: North-Holland
Norman, Donald A. and Daniel J. Bobrow (1975), On Data-Limited and
Resource-Limited Processes, Cognitive Psychology, 7 (1), 4464.
Otnes, Cele C. (1990), A Study of Consumer External Search Strategies Pertaining to Christmas Shopping, (unpublished doctoral dissertation). TN:
University of Tennessee.
Pan, Xing, Brian T. Ratchford and Venkatesh Shankar (2003), The Evolution of Price Dispersion in Internet Retail Markets, Advances in Applied
Microeconomics, 12, 85105.
Parayitam, Satyanarayana and Robert S. Dooley (2011), Is too much Cognitive Conflict in Strategic Decision-Making Teams too Bad?, International
Journal of Conict Management, 22 (4), 34257.
Pauwels, Koen, Peter S.H. Leeflang, TeerlingF Marije L. and K.R. Eelko Huizingh (2011), Does Online Information Drive Offline Revenues? Only for
Specific Products and Consumer Segments!, Journal of Retailing, 87 (1),
117.
Payne, John W., James M. Bettman and Mary F. Luce (1996), When Time is
Money: Decision Behavior Under Opportunity-Cost Pressure, Organizational Behavior and Human Decision Processes, 66 (2), 13152.

253

Peterson, Robert A. (2001), On the Use of College Students in Social Science Research: Insights from a Second-Order Meta-Analysis, Journal of
Consumer Research, 28 (3), 45061.
Petty, Richard E. and John T. Cacioppo (1981), Issue Involvement as a Moderator of the Effects on Attitude of Advertising Content and Context, In
Advances in Consumer Research, Vol. 8, Monroe K. B. ed. Provo, UT:
Association for Consumer Research, 204.
Punj, Girish N. and Richard Staelin (1983), A Model of Consumer Information
Search Behavior for New Automobiles, Journal of Consumer Research, 9
(March), 36680.
Putrevu, Sanjay and Brian T. Ratchford (1997), A Model of Search Behavior
with an Application to Grocery Shopping, Journal of Retailing, 73 (4),
46386.
Ratchford, Brian T. and Narasimhan Srinivasan (1993), An Empirical Investigation of Returns to Search, Marketing Science, 12 (1), 7387.
Roedder-John, Deborah, Carol A. Scott and James R. Bettman (1986), Sampling Data for Covariation Assessment: The Effect of Prior Beliefs on Search
Patterns, Journal of Consumer Research, 13 (1), 3847.
Roedder John, Deborah and Catherine A. Cole (1986), Age Differences in
Information Processing: Understanding Deficits in Young and Elderly Consumers, Journal of Consumer Research, 13 (3), 297315.
Rothschild, Michael (1974), Searching for the Lowest Price when the Distribution of Prices is Unknown, Journal of Political Economy, 82 (4),
689711.
Schaie, K. Warner and Sherry L. Willis (1993), Age Difference Patterns of
Psychometric Intelligence in Adulthood: Generalizability Within and Across
Domains, Psychology and Aging, 8 (1), 4455.
Schaninger, Charles M. and Donald Sciglimpaglia (1981), The Influence of Cognitive Personality Traits and Demographics on Consumer
Information Acquisition, Journal of Consumer Research, 8 (2),
20816.
Sears, David O. (1986), College Sophomores in the Laboratory: Influences of a
Narrow Data Base on Social Psychologys View of Human Nature, Journal
of Personality and Social Psychology, 51 (3), 51530.
Simon, Herbert A. (1957), Models of Man: Social and Rational, New York: John
Wiley & Sons.
(1979), Rational Decision Making in Business Organizations, American Economic Review, 69 (4), 493513.
(1987), Satisficing, In The New Palgrave: A Dictionary
of Economics, Vol. 4, Eatwell , John , Milgate M. and Newman P. eds. New
York: MacMillan, 2435.
Smith, Gerald E., Meera P. Venkataraman and Ruby R. Dholakia (1999),
Diagnosing the Search Cost Effect: Waiting Time and the Moderating
Impact of Prior Category Knowledge, Journal of Economic Psychology, 20,
285314.
Srinivasan, Narasimhan and Brian T. Ratchford (1991), An Empirical Test of a
Model of External Search for Automobiles, Journal of Consumer Research,
18 (2), 23342.
Stahl, Dale O. II (1989), Oligopolistic Pricing with Sequential Consumer
Search, The American Economic Review, 79 (4), 70012.
Staal, Mark A. (2004), Stress, Cognition, and Human Performance: A Literature Review and Conceptual Framework [NASA Technical Memorandum
212824], Moffett Field, CA: NASA-Ames Research Center.
Stigler, George (1961), Economics of Information, Journal of Political Economy, 69 (3), 21325.
Stone, Robert N. (1984), The Marketing Characteristics of Involvement,
Advances in Consumer Research, 11 (1), 2105.
Sultan, Fareena, John U. Farley and Donald R. Lehmann (1990), A MetaAnalysis of Applications of Diffusion Models, Journal of Marketing
Research, 27 (February), 707.
Szymanski, David M. and David H. Henard (2001), Customer Satisfaction:
A Meta-Analysis of the Empirical Evidence, Journal of the Academy of
Marketing Science, 29 (1), 1635.
Tellis, Gerald J. and Birger Wernerfelt (1987), Competitive Price and Quality under Asymmetric Information, Marketing Science, 6 (Summer),
24053.
Urbany, Joel E. (1986), An Experimental Examination of the Economics of
Information, Journal of Consumer Research, 13 (2), 25771.

254

M. Maity et al. / Journal of Retailing 90 (2, 2014) 233254

Urbany, Joel E., Peter A. Dickson and Rosemary Kalapurakal (1996), Price
Search in the Retail Grocery Market, Journal of Marketing, 60 (2),
91104.
Urbany, Joel E., Peter R. Dickson and William L. Wilkie (1989), Buyer Uncertainty and Information Search, Journal of Consumer Research, 16 (2),
20814.
Varian, Hal R. (1980), A Model of Sales, The American Economic Review, 70
(4), 6519.
Viswesvaran, Chockalingam and Juan I. Sanchez (1998), Moderator
Search in Meta-Analysis: A Review and Cautionary Note on Existing Approaches, Educational and Psychological Measurement, 58 (1),
7787.
Vriend, Nicolaas J. (1996), Rational Behavior and Economic Theory, Journal
of Economic Behavior and Organization, 29, 26385.

Weiss, Allen M. and Jan B. Heide (1993), The Nature of Organizational


Search in High Technology Markets, Journal of Marketing Research, 30
(2), 22033.
Wolf, Fredric M. (1986), Meta-Analysis: Quantitative Methods for Research
Synthesis, Beverly Hills, CA: Sage Publications, Inc.
Zauberman, Gal (2003), The Intertemporal Dynamics of Consumer Lock-In,
Journal of Consumer Research, 30 (3), 40519.
Zeithaml, Valarie A. (1981), How Consumer Evaluation Processes Differ
Between Goods and Services, in Marketing of Services (unpublished doctoral dissertation), Donnelly James H. and George William R., eds. Chicago,
IL: American Marketing Association, 18690.

Potrebbero piacerti anche