You are on page 1of 15

Journal of Retailing 82 (3, 2006) 229–243

Determinants of retail patronage: A meta-analytical perspective夽


Yue Pan a,∗ , George M. Zinkhan b,1
a 812 Miriam Hall, Department of Management & Marketing, University of Dayton, Dayton, OH 45469-2271, United States
b Department of Marketing, Terry College of Business, University of Georgiam, Athens, GA 30602, United States

Accepted 11 November 2005

Abstract

The retail patronage idea includes such key concepts as store choice and frequency of visit. In this study, the authors synthesize previous
empirical studies through a formal, critical review of retailing literature. The meta-analysis suggests that various predictors (e.g., service,
product selection, quality) are strongly related to shoppers’ retail choice, whereas others (e.g., store attitude, store image) are important
antecedents of shopping frequency. However, the relationships between the predictors and retail patronage vary according to the study
characteristics (e.g., experimental vs. other designs). The authors offer implications for retailing research and practice.
© 2006 New York University. Published by Elsevier Inc. All rights reserved.

Keywords: Retail patronage; Meta-analysis; Store choice; Frequency of visit

Introduction tude of the effects for the same predictor variables across
studies.
An understanding of patronage behavior is a critical issue Consider “store price level” as an example. Although it is
for retail managers because it enables them to identify and well documented that low prices accelerate retail purchases
target those consumers most likely to purchase. Reflecting (e.g., Tigert 1983; Walters and Rinne 1986), some research
this managerial need, one research stream has focused on has found a positive relationship between monetary price
explaining retail patronage with respect to various elements and perceptions of product quality (e.g., Dodds et al. 1991;
(e.g., store, frequency of visit, store choice). The practical and Rao and Monroe 1989). Shoppers with limited sources of
conceptual importance of this topic has been underscored by diagnostic information tend to make more use of price as
the substantial volume of studies published in leading jour- a quality cue (Rao and Monroe 1988). Following this logic,
nals. Decades of research efforts have produced a rich body some consumers may choose a retailer that offers high-priced
of empirical data associated with a variety of research designs products to enhance their expected quality (Tellis and Gaeth
and study contexts. However, a review of the literature reveals 1990).
marked differences in both the direction and the magni- In addition to conflicting evidence about the direction
of the relationship, the findings pertaining to the strength
of the relationship also vary. For example, though some
夽 This article is based on the first author’s dissertation and was the 2002
researchers report no evidence of a significant relationship
AMS Mary Kay Doctoral Dissertation Competition winner. The authors
acknowledge the helpful input of Don Lehmann, David Henard, Randy
between low-price offerings and retail choice (e.g., Lumpkin
Sparks, and Ashutosh Dixit. They also thank the editors and the anonymous and Burnett 1991–1992), others suggest a significantly posi-
JR reviewers for their valuable comments on previous drafts of this article. tive relationship (e.g., Thelen and Woodside 1997). Because
This work was supported by summer research grants from the University of of heterogeneous findings and diverse study conditions in the
Dayton. extant empirical literature, the relationship between various
∗ Corresponding author. Tel.: +1 937 229 1773; fax: +1 937 229 3788.

E-mail addresses: Yue.Pan@notes.udayton.edu (Y. Pan),


predictors and a shopper’s retail patronage are unclear, which
gzinkhan@terry.uga.edu (G.M. Zinkhan). in turn, complicates our efforts to develop a comprehensive
1 Tel.: +1 706 542 3757; fax: +1 706 542 3738. understanding of what affects shoppers’ decisions to patron-

0022-4359/$ – see front matter © 2006 New York University. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.jretai.2005.11.008
230 Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243

ize a retail outlet. Given this limitation, it is somewhat difficult Relationship between product-relevant factors and retail
to translate academic findings into a form that is useful for patronage
retailing managers.
Despite the importance of retail patronage, no compre- Product quality
hensive work has attempted to assess the general findings Consumers perceive the quality of a product differently
across academic studies. We seek to fill that void by conduct- depending on the store type from which the purchase is made
ing a meta-analysis of empirical findings on the predictors (Darden and Schwinghammer 1985). Furthermore, a con-
of retail patronage. Our dependent variable, retail patronage, sumer’s perception of the quality of a store’s merchandise
includes two dimensions: (1) store choice (i.e., a consumer’s relates to the patronage of that store (Darley and Lim 1993;
choice to patronize a particular store) and (2) frequency of Jacoby and Mazursky 1985; Olshavsky 1985). As an impor-
visit (i.e., how often a shopper patronizes that store). Such a tant component of store evaluation, merchandise quality has
research effort appears useful for at least two reasons. First, a positive link to merchandise value (Grewal et al. 2003).
we seek to reconcile inconsistent findings and establish the In brief, merchandise determines a retailer’s reputation and
generalizability of the relationships between retail patronage influences consumers’ choice at stores.
and its correlates. Second, research on retail patronage has
been conducted in various methodological contexts, yet no H1. Perceived product quality is positively related to retail
attempt has been made to evaluate the robustness of the effects patronage.
across study conditions. Here, we attempt to explain differ-
Price
ences in the results of previous studies by investigating var-
Shopping channels are significantly different in terms of
ious study characteristics that could moderate the effects of
the general price levels for products sold. A higher price rep-
interest.
resents a monetary measure of immediate costs, which leads
In terms of organization, we begin by presenting
to a reduced willingness to buy (Dodds et al. 1991; Walters
antecedents to retail patronage that are frequently reported
and Rinne 1986). Low prices, in the form of either price pro-
in prior research, followed by a discussion of the theoretical
motions or general price levels, can create store traffic and
background of these variables. Then, we discuss the sample
increase category sales. However, consumers’ responsive-
frame for the study. We subsequently present the research
ness to low prices may be heterogeneous. Consumers may
method and data analysis procedure, followed by a discus-
use under uncertainty another choice strategy, what Tellis
sion of the results. We conclude with some implications and
and Gaeth (1990) call “price-seeking,” in which consumers
suggestions for further research.
choose the highest-priced brand to maximize their expected
quality. This theory posits a positive relationship between
monetary price and perceptions of product quality (cf. Dodds
Antecedents of retail patronage: theoretical foundations
et al. 1991; Kerin et al. 1992; Rao and Monroe 1989). How-
ever, the extent to which consumers use price as an indicator
A review of the literature reveals 16 antecedents reported
of quality depends on the availability of alternative diag-
frequently enough (n ≥ 4) to be included in a formal meta-
nostic information (Rao and Monroe 1988), in that when
analysis. To organize the large number of antecedents, we
more extrinsic cues (e.g., brand, store name) are available,
categorize them into three groups: (1) product-relevant fac-
the price–quality relationship weakens (Dodds et al. 1991).
tors, which pertain to product features and attributes, such as
Although price has a positive effect on perceived quality, it
product quality and price; (2) market-relevant factors, which
has a negative effect on perceived value and willingness to
pertain to the retailer of interest, such as the service pro-
buy (Dodds et al. 1991).
vided by the store; and (3) personal factors, which pertain
to consumer characteristics, such as demographics. To check H2. The general price level in a store is negatively related
the coding quality of this categorization, two independent to retail patronage.
investigators who were familiar with retail patronage liter-
ature placed specific predictors within each category and Product selection/assortment
reviewed the final taxonomy for completeness and appropri- Product selection (or assortment) is defined as “the number
ateness of classification. For the 16 items coded, the overall of different items in a merchandise category” (Levy and Weitz
raw intercoder reliability was 93.8%. The coders resolved any 1995, p. 30). As a major retailer descriptor, product selec-
disagreements by discussing the terms until the team arrived tion contributes significantly to the explanation of patronage
at a consensus. The three categories of determinants serve to of alternative retail centers (Arnold et al. 1983; Craig et al.
isolate individual, product, and seller differences and reduce 1984; Koelemeijer and Oppewal 1999; Louviere and Gaeth
the large number of independent variables to a more suc- 1987). The breadth (number of brands) and depth (number
cinct hierarchical model. Note that it is beyond the scope of of stockkeeping units) of an assortment offered in a shop-
this study to build a holistic theory of retail patronage. We ping center helps retailers cater to the heterogeneous tastes
organize the variables around these three categories and test of their patrons (Dhar et al. 2001). Not only can greater vari-
relevant hypotheses using meta-analytical integration. ety help a retailer attract more consumers, it also can entice
Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243 231

them to make purchases while in the retail center. A wide less emphasis on low prices, are more likely to fall into shop-
selection of products also can minimize the perceived costs pers’ consideration set (i.e., retail alternatives a consumer
(e.g., travel time, effort) associated with each shopping trip is aware of and evaluates positively). In Malhotra (1983)
and ease the shopping task (e.g., by enhancing comparison threshold model of store choice, service is one of the five
shopping); in other words, a retailer that offers greater vari- identified salient characteristics (along with variety and selec-
ety in product categories can improve shopping convenience tion, acceptable prices, convenience of location, and physical
and make it easier for consumers to combine their visits to facilities).
different stores (Dellaert et al. 1998). An empirical study by
Stassen et al. (1999) shows that the assortment decision in a H5. Service quality is positively related to retail patronage.
store is as important, if not more important, than other key In particular, higher service quality is associated with higher
variables such as price. retail patronage intentions.

H3. There is a positive correlation between product assort- Friendliness of salespeople


ment and retail patronage. Retail stores offer a chance for human interactions. Some
consumers enjoy talking to and socializing with others dur-
Relationship between market-relevant factors and retail ing a shopping visit and seek a social experience outside the
patronage home (Tauber 1972). These people generally experience a
strong motivation to associate themselves meaningfully with
Convenience groups of “kindred spirits” to reduce feelings of boredom and
A convenience orientation is a key benefit that shoppers loneliness. To cope with and alleviate such feelings of lone-
seek in the modern environment. In this sense, consumers’ liness, people pursue various strategies, including shopping
perceptions of convenience (e.g., opening hours, location, (Rubenstein and Shaver 1980). The desire for human inter-
parking) will have a positive influence on their satisfaction action thus may drive some shoppers to stores in which they
with the service (Berry et al. 2002). Consumers’ perceived find salespeople friendly and communicative.
expenditure of time and effort interacts to influence their
perceptions of service convenience (Berry et al. 2002), and H6. There is a positive correlation between the friendliness
retail facilities can be designed to affect those time and effort of salespeople and shoppers’ retail patronage.
perceptions. For example, a central location can reduce the
transaction costs associated with shopping (e.g., transporta- Store image
tion cost, time spent). The law of retail gravitation (Reilly Here, store image is defined as “the way in which the store
1931) suggests that the potential attraction of a shopping is perceived by shoppers.” Image formations result in predis-
center should be assumed to be inversely proportional to the positions that generally guide patronage (Darley and Lim
driving time from a shopper’s home to the center. The more 1993), including shopping trips, expenditure behavior, and
recent central place theory (Craig et al. 1984) suggests that store loyalty (Arnold et al. 1983; Sirgy and Samli 1985). The
central business districts and regional shopping centers that impressions shoppers form of stores have a significant impact
offer a large agglomeration of goods and services attract cus- on their store patronage. The retail store environment offers a
tomers from greater distances than neighborhood centers that multitude of stimuli that can serve as cues to consumers look-
offer fewer goods and services. Empirical evidence supports ing for information-processing heuristics (Baker et al. 1994).
these theories by showing that easy accessibility has a high These sensory search attributes, particularly visual cues about
correlation with shopping center selection (Bellenger et al. the store, have a significant impact on consumers’ patron-
1977). In addition to a convenient location, other convenience age behavior because consumers tend to make judgments
incentives provided by retailers, such as longer operating about stores on the basis of their subjective impressions (e.g.,
hours or ample parking, can draw patrons to a store (Hansen ambient design, social factors; Baker et al. 1994). Empirical
and Deutscher 1977–1978). evidence suggests that image perceptions account for a very
high proportion of the variance in retail patronage (Finn and
H4. Shopping convenience (opening hours, location, and Louviere 1996; Kasulis and Lusch 1981).
parking) provided by a retailer increases retail patronage.
H7. Store image is positively related to retail patronage.
Service quality
Previous studies have found a direct link between ser- Store atmosphere
vice quality and patronage intentions (e.g., Baker et al. 2002; Unlike store image, which involves shoppers’ percep-
Sirohi and McLaughlin 1998; Zeithaml and Berry 1996). For tions, store atmospherics deal strictly with the physical store
example, Finn and Louviere (1990) find that different apparel attributes. Research on mall shopping has revealed that many
shopper segments tend to choose shopping centers that they consumers are prone to make a decision about where to
associate with different combinations of features. Shopping shop on the basis of their attitude toward the shopping cen-
centers that provide good service and a wide selection, but ter environment (Finn and Louviere 1990, 1996; Gentry
232 Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243

and Burns 1977). For example, recreational shoppers who likely to be of greater importance to them. Therefore, they
enjoy shopping as a leisure activity may shop impulsively might be motivated to shop around and thus be less store
and place higher importance on store décor (Bellenger and loyal than high-income consumers. Levy (1966), however,
Korgaonkar 1980). Lambert (1979) similarly suggests stores states that low-income women may like to go shopping just
should provide rest areas and an appropriate store tempera- to have a reason to get out of the house. Because higher-
ture. Arousal induced by the store environment intensifies income consumers are more likely to visit a mall only when
both pleasure and displeasure, such that time and spend- they have to and lower-income people are more likely to shop
ing behavior increase in pleasant environments and decrease for recreation, there might be a negative correlation between
in unpleasant environments (Donovan et al. 1994). More- income and retail frequency of visit.
over, shoppers’ evaluations of the store’s atmosphere affect
their perceptions of value and their store patronage intentions H10a . Women tend to visit a retail outlet more frequently
(Grewal et al. 2003). than men.

H8. Store atmosphere relates to retail patronage behavior, H10b . Frequent patrons tend to be older than less frequent
such that shoppers’ perceptions of pleasant store atmospher- patrons.
ics lead to greater retail patronage intentions.
H10c . Frequent patrons tend to have lower incomes than
Fast checkout
less frequent patrons.
The time pressures that many people experience are hav-
ing a major effect on consumer behavior; they perceive
their discretionary time available as insufficient to accom- Store/store-type attitude
modate all their desired uses of it. The results are con- According to the theory of planned behavior (Ajzen 1985),
tinual choices among various activities and the pursuit of a behavioral intention (i.e., decision) is partially determined
efficiency-producing behaviors. Retail stores are devoting by the person’s attitudes, which means that consumers’ atti-
more resources to time-saving services, such as fast checkout tudes toward retail stores likely play a key role in their choice
(Lambert 1979). Time savings for consumers are readily rec- of shopping modes. When a consumer holds a general attitude
ognized and therefore likely to influence their retail choice. toward a store (or store type), that attitude is readily acces-
sible and probably will have a direct effect on the person’s
H9. Checkout speed is positively related to retail patronage. store-specific quality perceptions (Bauer and Greyser 1968;
MacKenzie and Lutz 1989), as well as spillover effects on
Relationship between personal factors and retail store patronage through the process of affect transfer (Darley
patronage and Lim 1993; Lutz 1985). Several empirical investigations
provide support for the positive relationship between attitude
Demographic variables and patronage (e.g., Eastlick and Liu 1997; Korgaonkar et al.
A considerable body of empirical research on shopping 1985).
behavior suggests that consumer demographic variables may
be related to retail store patronage (e.g., Bellenger et al. H11. The general attitude toward a store or store type is
1976–1977; Korgaonkar et al. 1985; Samli 1975). However, positively related to retail patronage.
no consensus exists about the relationship between shop-
pers’ demographic profiles and their patronage behavior. Potential moderating effects
For example, in their study of department store shoppers,
Crask and Reynolds (1978) find that frequent patrons tend Sultan et al. (1990) indicate that four broad categories
to be slightly younger, better educated shoppers with higher of characteristics often account for systematic differences
incomes. However, Roy (1994) argues that young people, across correlations: measurement method, research context,
facing greater constraints on their time, may be restrained estimation procedure, and model specification. Because our
from frequently visiting a retailer. Empirical studies of shop- units of analysis are bivariate correlations that are unaf-
per motivations (Westbrook and Black 1985) also identify fected by the estimation procedure or model specification,
a predominantly older age segment that derives satisfaction we seek systematic differences in the study characteristics.
from aspects of shopping, such as negotiation with salespeo- Research on retail patronage has been conducted in a variety
ple and an affiliation with other shoppers. Older shoppers may of contexts with different measurement instruments. Differ-
shop more frequently if they view shopping as a recreational ences in method and research context also could contribute
activity. to the variance in the retail patronage found across studies.
Conflicting views also emerge with respect to income. In our investigation, we examine five potential moderators –
According to Goldman (1977–1978), low-income consumers type of scale used, type of sample, study design, shopping
tend to have lower marginal opportunity costs for their time, mode, and product type – to assess their impact on sample
in that the potential benefits of comparison shopping are homogeneity.
Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243 233

Multi- versus single-item scale ple buy frequently with minimal shopping effort, purchase
Single treatments with low reliability can drastically atten- is often characterized by inertia, which may lead shoppers to
uate effect-size estimates and decrease precision (Fern and ignore or overlook the service quality provided. The relation-
Monroe 1996). The use of multiple-item scales, in contrast, ship between price and patronage also might be moderated by
enhances measurement reliability. Therefore, this practice product type; consumers may choose a high-priced retailer
should provide stronger relationships than single-item mea- to enhance the expected quality of a product when they know
sures can (Peter and Churchill 1986). the price better than they do the quality (Tellis and Gaeth
1990), such as when they shop for specialty goods.
Student versus nonstudent sample
The composition of samples also has the potential to affect
Method and results
the relationships studied. The use of student samples in rela-
tional research can lead to restricted ranges and attenuated Sampling frame
effect sizes (Pedhazur and Schmelkin 1991) because a homo-
geneous sample likely will respond similarly to questionnaire We chose those articles that shared the same criterion
items, which limits the response variation across the range of variable: retail patronage. Our review of empirical work
scale values (Fern and Monroe 1996). This limitation results in this area reveals that this construct is often operational-
in a bias toward stronger effects than would be found among ized as either shoppers’ store choice or shopping frequency.
nonstudent samples. The heterogeneity of the nonstudent For example, some authors focus on shoppers’ store patron-
sample, in contrast, increases error variance, which attenu- age (intention or behavior) choice (e.g., Grewal et al. 2003;
ates the magnitude of effect and possibly underestimates the Kenhove et al. 1999; Woodside and Trappey 1992), whereas
effect size (Fern and Monroe 1996). Moreover, the impact of others examine shoppers’ repeated patronage in a store over
antecedent variables on retail patronage may differ between a period of time (i.e., shopping frequency) (e.g., Darley
student and nonstudent samples because of inherent differ- and Lim 1993; Korgaonkar et al. 1985). Therefore, our
ences in shopping behavior between the two samples. meta-analysis concentrates on these two dimensions of retail
patronage; we define store choice as the likelihood that a
Experimental versus nonexperimental
shopper will patronize a retailer and shopping frequency as
The use of experiments also may represent a poten-
the number of times a shopper patronizes a retailer during a
tial moderator variable. A carefully conceived experimental
given period of time.
design enables a researcher to assign subjects randomly to
We identified the studies through the following search
conditions and exercise more control over the variables under
procedure: (1) an initial keyword search in business and non-
consideration, which in turn generates less error variance in
business databases, including ABI/Inform, PsycInfo, Socio-
the denominator of the effect size and produces larger effect
logical Abstracts, Social Science Citation Index, and Social
sizes. In addition, the exclusion of potential confounds may
Sciences Abstracts; (2) a search through listings of con-
lead to seemingly stronger relationships in an experimental
ference proceedings in the “Papers First” database and the
setting compared with a nonexperimental one.
Advances in Consumer Research online proceedings; (3) an
Store versus nonstore shopping mode interactive search of the references from relevant articles
The shopping mode likely will moderate the relation- identified from the keyword search until no new references
ship between retail patronage and its correlates. That is, could be identified; and (4) letters sent to 165 authors known
store patronage in a traditional bricks-and-mortar store for- for their work on retail patronage, in which we requested their
mat might differ from nonstore retailing in various ways. published or unpublished studies on this topic.
For example, a consumer who chooses a traditional retailer Our search did not include those studies that examined
over an e-tailer (or catalog) may attach more importance to patronage intentions for a specific product (e.g., product price
customer service, and customer service is more accessible affects a shopper’s intention to buy the product; Dodds et al.
in a store (e.g., helpful salespeople, easy return of defec- 1991; Grewal et al. 1998), because our dependent variable is
tive products). Moreover, the physical facilities (e.g., parking the patronage choice/intentions toward a retailer. Of the 165
facilities, convenient location, hours of operation) provided authors we contacted, 31 responded to our inquiry, and these
by a bricks-and-mortar retailer should not carry equal weight responses resulted in one usable study. In total, we identi-
for a nonstore retailer. fied 80 studies that reported one or more antecedents of retail
patronage. Among the 80 studies, 35 examine idiosyncratic
Type of good relationships that pertain to a specific shopping mode (e.g.,
Shopping patterns may vary for different product types. Internet experience as an antecedent to online shopping inten-
For example, for specialty goods that demand a considerable tion, Brengman and Geuens 2002) and are excluded from
amount of shopping effort, consumers are more likely to rely further discretion. The remaining 45 studies, which we report
on salespeople’s expertise to help them compare among alter- in Appendix, represent empirical work in business, psycho-
natives, which would increase the effect of service quality on logical, and sociological literature and, we believe, a fairly
retail patronage. In contrast, for convenience goods that peo- well rounded set of studies on this topic.
234 Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243

Table 1
Effects reported in the studies
Predictor Number of effects reported Range of reported effects (r) Cumulative n
Product-relevant factors
Low pricea 14b /19c −.01 to .702 4302
Qualitya 16/18 −.05 to .90 4443
Selectiona 14/14 .102 to .92 3272
Market-relevant factors
Convenient parking facilitiesa 8/9 .01 to .53 449
Convenient locationa 21/22 −.05 to .76 934
Convenient opening hoursa 8/9 .06 to .56 449
Friendliness of salespeoplea 11/13 −.02 to .62 1206
Servicea 14/17 .045 to .95 3802
Fast checkouta 9/11 .09 to .712 643
Store atmospherea 3/5 .016 to .55 970
Store imaged 12/20 .04 to .468 1445
Personal factors
Store/store-type attituded 13/13 .111 to .45 1271
Genderd 4/4 .139 to .19 1393
Incomed 2/11 −.13 to .198 2514
Aged 5/11 −.104 to .12 2275
a Predictor variables for store choice.
b Number of statistically significant (at α = .05) effects reported.
c Total number of effects reported.
d Predictor variables for frequency of visit.

In Table 1, we provide a complete taxonomy of predictors the same relationship. For correlational relationships, this
of retail patronage and the range of effects (r) reported in the process typically involves transforming the raw Pearson cor-
original studies. It reveals marked differences in the direction, relation coefficients (r) into an associated Z statistic and
magnitude, and statistical significance of the effect sizes for then transforming it back to r. In the studies examined here,
the same pairwise relationships across studies. Take “store 199 raw test statistics yielded 150r that we coded into our
price level” as an example. The range of reported effects for database.
this relationship across all studies spans −.01 to .702. Among To avoid the possibility that smaller or less representa-
the 19 studies that examine this relationship, 14 (74%) show tive samples are overrepresented in the analysis, we weight
statistically significant results. Only 2 (10.5%) report rela- the effect size r using the relevant sample size information.
tively large effect sizes (r > .5), whereas a few more (5) report In addition to considering sampling errors, we also correct
relatively small effect sizes (r < .1). We note that these appar- the effect sizes for attenuation due to measurement errors
ent inconsistencies might be due to nonsubstantive factors if scale reliability measures are available for the predictor
(e.g., sampling error, measurement error, deviation from con- and/or criterion variables. File drawer N indicates the num-
struct validity). ber of studies that confirm the null hypothesis that would
be needed to reverse a conclusion that a significant rela-
Computation and coding tionship exists, which we estimated for a significance level
of .05. For computational details, see Hunter and Schmidt
We define the effect-size estimate as the degree to which (1990).
the predictor–/criterion–variable relationship appears in the In Tables 2 and 3, we summarize the number of studies,
population of retail patronage research. Although correla- weighted mean observed correlation (sample-size adjusted
tions are the most common metrics used in our studies, mean), weighted mean correlation corrected for attenuation
many also report F, t, and chi-square statistics. To exam- from measurement error (reliability adjusted mean), total
ine the strength of the relationship, we convert various variance, sampling error variance, reliability variation vari-
summary statistics (e.g., F, t, Z, chi-square) to the com- ance, remaining variance, 95% confidence interval for each
mon correlation coefficient metric, r, following the formu- pairwise relationship represented by multiple study effects,
las suggested by Hays (1973), Kendall and Stuart (1967), file drawer N, and results of homogeneity tests. Conven-
Rosenthal (1991), and Wolf (1986). Furthermore, some stud- tional, appropriate magnitudes of r that correspond to small,
ies report several effect-size estimates for one predictor using medium, and large effect sizes are .1, .3, and .5, respectively.
the same subjects. To obtain a single result for the mul- Following this guideline, we find that a group of predictors
tiple correlated results from a single study, Wolf (1986) (product quality, service quality, wide selection) provides
suggests using the average of the statistics that examine relatively large effect sizes for explaining consumers’ retail
Table 2
Main effects of retail choice
Independent variables Ka Weighted ṙ Weighted ṙ 95% Confidence interval 95% Confidence interval Total Sampling Reliability Remaining File drawer Q (df)
(observed) (corrected) (observed) (corrected) variance error variation varianceb N (p = .05)
variance variance
Convenient parking facilities 9 .266 n.a.c .0055 .5265 n.a. .015 .018 n.a.d 39 7.53 (8)
Friendliness of salespeople 12 .273 .286 .0912 .4548 .0946 .4774 .027 .008 .00002 .019 (70.6) 67 37.80e (11)
Service quality 15 .637 .649 .5637 .7103 .5736 .7244 .228 .003 .00003 .225 (98.7) 75 1278.75e (14)
Low price 17 .364 .389 .2569 .4711 .2773 .5007 .044 .003 .00003 .04 (92.7) 94 234.25e (16)
Good quality 17 .579 .595 .4983 .6597 .5117 .6783 .092 .003 .00055 .089 (96.7) 138 710.66e (16)
Store atmosphere 4 .395 .419 .2886 .5014 .3003 .5377 .052 .003 .00011 .049 (94.4) 35 63.2e (3)
Fast checkout 11 .382 n.a. .1612 .6028 n.a. .042 .014 n.a. .028 (67.1) 61 37.66e (10)
Wide selection 12 .696 n.a. .6347 .7573 n.a. .171 .003 n.a. .168 (98.5) 86 950.92e (11)

Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243


Convenient location 12 .386 n.a. .1957 .5763 n.a. .037 .009 n.a. .028 (75.7) 83 54.51e (11)
Convenient opening hours 9 .363 n.a. .1196 .6064 n.a. .027 .017 n.a. .01 (37.4) 45 13.11 (8)
a The number of effect sizes combined.
b The percentage of total variance remaining is in parentheses.
c n.a. indicates insufficient reliability information is available to correct the study effects for measurement error.
d Reliability variation variances are not calculated as predictor or criterion reliability estimates are not available to adjust the mean correlation for the differences in scale reliabilities.
e Significant at p < .05.

Table 3
Main effects of frequency of visit of a particular store
Independent variables Ka Weighted ṙ Weighted ṙ 95% Confidence interval 95% Confidence interval Total Sampling Reliability Remaining File drawer Q (df)
(observed) (corrected) (observed) (corrected) variance error variation varianceb N (p = .05)
variance variance
Age 6 .073 n.a.c −.0273 .1733 n.a. .0052 .0026 n.c.d .0026 (49.6) n.c.e 11.53f (5)
Store/store-type attitude 5 .292 .334 .1793 .4047 .2021 .4659 .0112 .0033 .00032 .0076 (67.6) 30 16.94f (4)
Store image 9 .157 .170 .0057 .3083 .0096 .3304 .0094 .0058 .00001 .0035 (37.9) 26 12.64 (8)
Genderg 4 .154 n.a. .0513 .2567 n.a. .0004 .0027 n.c. 9 .45 (3)
Income 6 .057 n.a. −.0386 .1526 n.a. .0099 .0024 n.c. .008 (75.9) n.c. 23.22f (5)
a The number of effect sizes combined.
b The percentage of total variance remaining is in parentheses.
c n.a. indicates insufficient reliability information is available to correct the study effects for measurement error.
d Reliability variation variances are not calculated as predictor or criterion reliability estimates are not available to adjust the mean correlation for the differences in scale reliabilities.
e n.c. means the file drawer N was not calculated, because the 95% confidence interval contained 0.
f Significant at p < .05.
g Coded as (1) male, (2) female.

235
236 Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243

choices. Other predictors (store attitude) are important pre- between the given variance of the observed correlations and
dictors of the frequency of patrons’ visits to a particular store. the statistically given sampling error variance, in which the
In this sense, we find evidence that retailing literature gener- variance of the observed correlations is a sample estimate.
ates robust results and has been successful in identifying key Therefore, unless the number of studies is infinite, there will
variables that managers should emphasize in their planning be some error in the empirical estimate. In our case, sampling
process. error caused the variance of the observed correlations to dif-
Tables 2 and 3 also show that the corrected mean corre- fer slightly from the expected value, and that error caused the
lations for 13 of the antecedents are significant at the .05 smaller estimated total variance.
level. The high file drawer numbers for publication bias also
indicate that these results are significant beyond chance. For Multivariate analysis of the antecedents of retail
example, to bring the significant effect of service on store patronage
choice down to the level of just significant at α = .05, 75 null
results would need to be uncovered and included in our analy- In addition to a bivariate analysis of the correlations, a
sis. The effect sizes for most demographic variables, however, multivariate analysis can further our understanding because
are relatively small. The mean correlations for age (r = .073) we simultaneously estimate the relative impact of the inde-
and income (r = .057) are not statistically significant at con- pendent variables on the dependent variables. We estimate
ventional probability levels (p > .05), which means that the an ordinary least squares regression model of retail patron-
cumulative evidence indicates their effects on retail patron- age, whose matrix of corrected correlations we constructed
age do not generalize across studies. As Tables 2 and 3 also from the available data. Because few studies report correla-
suggest, reliability differences and sampling error account tional data about the interrelationships among the predictor
for only a small proportion of the variation across stud- variables, our correlation matrix contains data for only a sub-
ies, which leaves much room for moderator variables to set of the predictors. Table 4 contains the findings from our
operate. estimation of the regression models for retail choice and fre-
Our bivariate analysis confirms all our main effects pre- quency of visit.
dictions except for two hypotheses: H10b (age) and H10c As we show in Table 4, the relatively parsimonious model
(income). Shoppers’ retail choice is influenced by the fol- (Model 1) accounts for a majority of the variance (61.8%)
lowing factors (in order of importance): assortment, service, in retail choice. Product quality (β = .502, p < .001) and low
product quality, store atmosphere, store location, price level, price level (β = .44, p < .001) have the largest effects on shop-
checkout speed, hours of operation, friendliness of salespeo- pers’ intentions to patronize a retailer. Moreover, the effects of
ple, and parking facilities. The frequency of visits to a store the friendliness of salespeople and store atmosphere, though
is subject to the influence of store attitude, store image, and relatively small in magnitude, are significant in the regression
gender. model.
Note that for two variables (i.e., convenient parking facil- In Model 2, we identify store attitude as a dominant predic-
ities and gender), the total variances are smaller than the tor of shopping frequency (β = .377, p < .001), and the regres-
sampling error variances, a situation for which Hunter and sion coefficients for gender (β = .13, p < .001) and income
Schmidt (1990) provide an explanation. We compute the esti- (β = .188, p < .001), though relatively modest in size, also
mated variance of population correlations as the difference emerge as important drivers of shopping frequency. Age is

Table 4
Multiple regression models for selected predictors of retail patronage
Predictor Standardized coefficient Standardized coefficient
(standard error): Model 1a (standard error): Model 2b
Dependent variable: Dependent variable:
retail choice frequency of visit
Friendliness of salespeople .106 (.02)c –
Low price .44 (.019)c –
Good quality .502 (.02)c –
Store atmosphere .281 (.02)c –
Age – .019 (.033)
Store/store-type attitude – .377 (.034)c
Gender – .130 (.033)c
Income – .188 (.035)c
R2 (adjusted R2 ) .618 (.617) .158 (.154)
F (p value) 424.23 (<.001) 37.86 (<.001)
a Statistical significance is based on the median sample size of 1054, on which the individual correlations are based.
b Statistical significance is based on the median sample size of 812, on which the individual correlations are based.
c p < .001.
Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243 237

Table 5
Comparison of complete set and homogeneous subsets
Relationship Complete set Homogeneous subset
K Cumulative N Weighted ṙ Weighted ṙ K Cumulative N Weighted ṙ Weighted ṙ
(observed) (corrected) (observed) (corrected)
Friendliness of salespeople 12 1206 .273 .286 9 968 .212 .221
Service 15 3802 .637 .649 12 1114 .175 n.a.
Low price 17 4302 .364 .389 9 1287 .073 .089
Good quality 17 4443 .579 .595 11 1194 .401 .428
Fast checkout 11 643 .382 n.a.a 9 449 .271 n.a.
Wide selection 12 3272 .696 n.a. 10 759 .344 n.a.
Convenient location 12 934 .386 n.a. 9 599 .271 n.a.
Store atmosphere 4 970 .395 .419 3 679 .543 .640
Store/store-type attitude 5 1271 .292 .334 3 682 .216 .237
Age 6 2275 .073 n.a. 5 1995 .098 n.a.
Income 6 2514 .057 n.a. 5 1933 .015 n.a.
a n.a. indicates insufficient reliability information is available to correct the study effects for measurement error.

the only predictor variable that does not capture much of the data indicate significant variation in the pairwise relation-
variance in the bivariate and multivariate analyses. ships.
Because the heterogeneity of study effects suggests the
Homogeneity tests potential presence of moderator variables, we conducted sep-
arate meta-analytic syntheses for different subsets of studies
For each pairwise relationship, we conduct a homogeneity by examining a series of potential moderators. Then, we con-
test in which we successively delete the study effects to iden- ducted moderating analyses only for those relationships for
tify any outliers, following the procedure recommended by which at least 10 study effects were available.
Hedges and Olkin (1985). We compute the test statistic, Q,
for each pairwise relationship on Fisher’s z-transformations
of the correlation coefficients according to the formula Potential moderating effects of study characteristics:
tests of moderators
k

Q= (ni − 3)(zi − z+ )2 , (1) We partitioned the study effects into subgroups on the
i=1 basis of the values of their moderator variables, then com-
where z+ is the mean of the weighted z-transformed corre- pared the weighted mean correlations of the subgroups. In
lations (weights are the reciprocals of the variances, to give Table 6, we report the results of the moderator analyses,
greater weight to more precise estimates), and ni is the sample including the means and significance of each moderator for
size of a study. This statistic has an approximate chi-square both the corrected and uncorrected correlations. The aggre-
distribution with k − 1 degrees of freedom, where k is the gated data provide some interesting insights regarding the
number of included studies. relative strength of the effect sizes in specific study con-
Table 2 includes the results of our test for homogeneity ditions. The five potential moderators (i.e., study design,
of effects across all studies. If the results are heterogeneous, sample type, scale type, shopping mode, product type) appear
moderator variables exist. The homogeneity tests reveal the to have a pervasive influence for all but one pairwise relation-
overall consistency of results for four relationships (i.e., con- ship (fast checkout).
venient parking facilities → store choice, convenient opening As we reveal in Table 6, measurement and contextual
hours → store choice, store image → frequency of visit, gen- factors can account for a statistically significant amount of
der → frequency of visit), which signals the robustness and the variance. As we expected, shopping mode moderates the
generalizability of these relationships across study contexts. effects of several variables. For example, the effects of ser-
We also find significant heterogeneity among the effect sizes vice, quality, and selection on retail choice are more manifest
obtained from the independent studies for the other relation- for store than for nonstore patronage. Product type also mod-
ships. erates the effect of various variables on retail choice, such
To identify outliers, we deleted that study with the largest that the impact of service, quality, and selection on patronage
weighted deviation from the mean effect size, recalculated intention is significantly greater for specialty goods than for
the mean effect size, and repeated the homogeneity test. We convenience goods. This unsurprising result suggests that,
continued until we arrived at a reduced set of homogeneous when purchasing a specialty good (as opposed to a conve-
studies, the results of which we report in Table 5. For most nience good), shoppers are more highly involved with the
pairwise relationships, we must delete a small number of purchase and willing to spend more time comparing alter-
study effects (n ≤ 3) to achieve overall homogeneity. These native stores on criteria such as service and quality. A few
238 Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243

Table 6
Subgroup means by moderator variables
Experimental vs. Single vs. multiple Student vs. Store vs. nonstore Convenience vs.
nonexperimental scales nonstudent subjects shopping mode nonconvenience
goods
Observed ṙ, corrected ṙ, Observed ṙ, corrected ṙ, Observed ṙ, corrected Observed ṙ, corrected ṙ, Observed ṙ, corrected
N N ṙ, N N ṙ, N
Friendliness of .251 vs. .286, .284 vs. .386 vs. .205a , .386 vs. .251 vs. .286, .284 vs. n.a., n.a., 12 vs. 0 0.251 vs. 0.286, 0.284
salespeople .286, 2 vs. 10 .217a , 9 vs. 3 .286, 2 vs. 10 vs. 0.286, 2 vs. 10
Service .464 vs. .665a , .526 vs. .705 vs. .360a , .705 vs. .503 vs. .655a , .591 .683 vs. .210a , .699 vs. 0.158 vs. 0.752a ,
.665a , 3 vs. 12 .378a , 12 vs. 3 vs. .655a , 2 vs. 13 .210a , 13 vs. 2 0.158 vs. 0.774a , 10
vs. 5
Low price .031 vs. .450a , .037 vs. .483 vs. .048a , .483 vs. .028 vs. .427a , .036 .365 vs. .348, .392 vs. 0.358 vs. 0.365, 0.358
.450a , 4 vs. 13 .058a , 12 vs. 5 vs. .429a , 3 vs. 14 .348, 16 vs. 1 vs. 0.397, 11 vs. 6
Good quality .444 vs. .604a , .528 vs. .604 vs. .444a , .604 vs. .444 vs. .604a , .528 .674 vs. .206a , .700 vs. 0.487 vs. 0.595a ,
.604a , 3 vs. 14 .528a , 14 vs. 3 vs. .604a , 3 vs. 14 .206a , 15 vs. 2 0.487 vs. 0.615a , 11
vs. 6
Fast checkout n.a.b , n.a., 0 vs. 11 n.a., n.a., 11 vs. 0 n.a., n.a., 0 vs. 11 n.a., n.a., 0 vs. 11 n.a., n.a., 11 vs. 0
Wide selection n.a., n.a., 0 vs. 12 .754 vs. .103a , n.a., 11 n.a., n.a., 0 vs. 12 .737 vs. 307a , n.a., 11 0.265 vs. 0.822a , n.a.,
vs. 1 vs. 1 10 vs. 2
Convenient location n.a., n.a., 0 vs. 12 .464 vs. .216a , n.a., 11 n.a., n.a., 0 vs. 12 n.a., n.a., 12 vs. 0 n.a., n.a., 12 vs. 0
vs. 1
a Significant at p < .05.
b n.a. means it has not been calculated because the moderator does not exist for this category or the attenuation factor is not available.

other patterns are evident across the data, including the differ- patronage are explained by very different sets of predictors.
ence in effect sizes for experimental versus nonexperimental Of the three categories, personal factors (e.g., demographics,
designs. In contrast to our expectation, correlations are lower attitude toward a store) seem to be the dominant predictors of
with experiments. Furthermore, the impact of sample type on shopping frequencies, whereas market- and product-relevant
the effect sizes is directionally inconsistent with our expecta- variables are more likely to influence shoppers’ decisions to
tions; student subjects appear to create a downward bias in the patronize a particular store, given that a variety of stores are
strength of the studied relationships. We also expected rela- available. These findings suggest that retailers have various
tionships to be weaker in studies that used single scales, but tools at hand (e.g., greater assortment, low prices) to influence
this expectation is not supported. We discuss the implications shoppers’ intention to patronize their stores. However, shop-
of these findings subsequently. ping frequency, over which retailers have much less control,
largely depends on a consumer’s will. This conclusion should
be interpreted with caution though. In our analysis of shop-
Discussion and conclusion ping frequencies, some factors (e.g., assortment, quality) that
occur in some studies have not been investigated frequently
Using a meta-analytical approach, our study provides enough to be integrated into our meta-analysis. Therefore,
insights into retail patronage, synthesizes empirical works we cannot rule out the possibility that some other product-
from retailing literature, and provides a quantitative summary and market-relevant factors may have significant influence
of the cumulative findings. As we show in Tables 2 and 3, we on shopping frequencies.
find a relatively strong relationship between shoppers’ store Our findings also reveal the existence of methodological
choice and several important predictors. Of the three cate- artifacts. As we indicate in Table 6, the study designs and
gories of predictor variables, selection has the highest average research methods used can influence the effects of some key
correlation with store choice, followed by service, quality, retailing predictors. Take location as an example. Theoret-
store atmosphere, low price levels, convenient location, fast ically, location is a key variable for predicting the success
checkout, convenient opening hours, friendliness of sales- of traditional retailers, but our results only detect a medium
people, and convenient parking facilities. Other antecedent effect (r = .386). A post hoc analysis reveals that most retail-
variables (e.g., store image, store attitude, gender) are impor- ing studies are not really designed to allow location to display
tant predictors of shopping frequencies, though the effect its key effect (e.g., much of the location effect may occur
sizes of some variables (e.g., age, income) are not significant. below the conscious level). In Table 6, we also reveal that in an
Gender is the only successful demographic variable, which artificial research context (e.g., student sample in experimen-
suggests that women tend to be more frequent shoppers than tal settings), the impact of predictors on patronage behavior
men. is less obvious than in nonexperimental settings. In an exper-
A second insight relates to the three categories of predictor imental setting, researchers often ask subjects to evaluate a
variables that explain patronage. The two dimensions of retail retailer on the basis of a brief description and then indicate
Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243 239

their behavioral intentions. These behavioral accounts are variations in the observed effect sizes beyond that produced
often speculative and perhaps constructed on the spot rather by sampling error. If the true effect size is homogeneous
than being true indicators of the relevant behavior. The use across studies, the variation in reliability would produce a
of these student samples also presents a major challenge to false impression of heterogeneity (Hunter and Schmidt 1990).
the generalizability of retailing research findings. Students In our meta-analysis, we corrected for this attenuation by
seem less interested in retailers’ strategies to attract patrons weighting each study with reliability measures, but many
(e.g., good service, high product quality) than general con- studies (especially those published a long time ago) did not
sumers, and their shopping behavior may be more influenced report reliability measures, which could lead to systematic
by factors such as peer recommendation and brand. reductions in the mean effect sizes.
In addition, single scales produce higher correlations than Homogeneity tests reveal that the effects of convenient
multiple scales in many cases. Our post hoc review reveals parking facilities and hours of operation on retail choice and
varied and sometimes weak measures of key variables and the effects of gender and store image on shopping frequency
thereby highlights the need to identify superior measurement are relatively robust across study contexts. For other rela-
approaches in future studies. The way in which retail patron- tionships (e.g., fast checkout, selection, store atmosphere),
age is measured can lead to different estimates of relationship we must delete one or two outliers to achieve overall homo-
strength; for example, research often measures patronage as geneity. Homogeneity tests, however, indicate substantial
recalled behavior (e.g., “Name the regular store in which you variation in the relationships between shoppers’ perceptions
concentrate the majority of your purchases”) or anticipated of low price and good product quality and their retail choice.
future behavior (e.g., “The likelihood that I would shop in this These study effects are significantly affected by the method-
store is very high”). Neither of these approaches provides a ological choices and study contexts.
good estimate of shoppers’ actual patronage behavior. An
understanding of these effects can shed more light on the
robustness of findings in retailing research. Implications and directions for further research
The findings of our study thus call for different theories
for studying retail patronage. For traditional retailers, fac- Our study has implications for both academicians and
tors such as physical location, parking facilities, checkout retail managers. We gauge the current level of knowledge
speed, and store atmosphere can make or break a store. How- about retail patronage with our critical review of the empiri-
ever, with the evolution of nonstore retailing formats (e.g., cal studies on this topic and a meta-analysis of these studies.
e-tailing), these traditionally important predictors of retail By synthesizing the traditional retailing literature in a formal
patronage may become less crucial or even obsolete. Our find- way, our study offers greater understanding of the general
ings provide some indication of this trend; we find that the strength and variability of the relationships and the condi-
effects of various factors on patronage (e.g., service, quality, tions that moderate those relationships. It also tests a broad
selection) are less obvious in nonstore than in store shopping theoretical typology and a wide range of variables that may
modes. Shoppers may attach more importance to other fac- explain retail patronage behavior.
tors (e.g., return policy, company reputation) in their decision We therefore tap into areas that are of great interest to retail
to patronize a nonstore retailer, so the evolution of nonstore managers. Retailers have tracked, scanned, monitored, and
retailing may signal the need to develop an updated set of followed consumers’ shopping behavior for decades. Ques-
retailing theories. tions such as how shoppers choose a particular store, how
We observe a pattern of strong relationships in our meta- often they visit, why they visit, and who visits have been
analysis, but their strength may be inflated by a potential popular inquiries. However, inconsistent research findings
publication bias. Despite our effort to locate unpublished make it problematic to communicate with retailers who want
studies, we did not find many, which could influence the to apply this knowledge. We thus provide basic answers to
effect-size estimates. The file drawer Ns calculated in our some key questions. A retailer can enhance consumer patron-
study provide some estimates of the number of studies with age behavior by identifying and implementing an appropriate
nonsignificant results that would be required to nullify the sig- marketing strategy, which must start with a good under-
nificant results, such as the 94 nonsignificant results studies standing of the many factors and dimensions that influence
needed to nullify the “low price” result at the .05 signifi- shoppers’ choice behavior. For example, to increase initial
cance level. Given the relatively limited number of studies patronage, consumer promotions should focus on store- and
we found that report this variable (17), the possibility of find- product-specific elements (e.g., wide assortment, premium
ing an additional 94 studies is rather low. All the significant service, pleasant in-store decor). Managers also must rec-
relationships (except that between gender and frequency of ognize that shopping frequencies tend to be associated with
visit) require many zero-effect studies to lead to conclusions shopper characteristics, so they should tailor their marketing
of nonsignificance. Therefore, we have confidence in these communications to frequent shoppers, which will increase
significant results. the likelihood that they experience positive returns from their
Measurement error in the variables reduces effect-size promotional efforts. Examining the determinants of store
estimates, and variations in reliability across studies cause patronage thus enables managers to evaluate and understand
240 Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243

the extent to which their own perceptions of the store and its small number of studies (n < 10) available, we cannot under-
offerings are congruent with shoppers’ perceptions. take a moderator analysis to uncover why such a disparate
We conclude that differences in study and sample charac- pattern occurs. Nonetheless, we take a closer look at the stud-
teristics contribute to the variances in store patronage found ies’ characteristics to pinpoint some potential explanations.
across prior studies. For example, the use of student sub- Of the two studies that show significant effects for income,
jects remains a somewhat controversial issue in marketing one investigates discount stores and finds a negative relation
and retailing research. It is argued that student subjects may between income and shopping frequency (Korgaonkar et al.
be appropriate for tests of theory. Here, however, we find that 1985), and the other studies the frequency of catalog orders
the use of such subjects can deflate estimated effect sizes. In and finds a positive relationship (Lumpkin and Hawes 1985).
the same way, experimental designs may introduce an ele- For age, the only negative effect reported appears in a study
ment of artificiality that can distort results. On the one hand, analyzing convenience food stores (Darden and Lumpkin
such method effects are well known; on the other, researchers 1984). We suggest that additional research should explore
sometimes forget about these crucial differences when trying the effect of income and age on retail patronage across var-
to make sense out of cumulative findings (e.g., in research ious store types and product types (e.g., convenience versus
reports and academic literature). nonconvenience goods).
In terms of main effects, we find that the following vari- Because main effects can be relatively uninteresting in
ables are especially important for explaining retail choice: marketing research, we must go beyond those main effects
wide selection, service, and product quality. Store attitude and explore the interaction effects. We provide evidence that
is a more important predictor than image or demographic as yet unnamed moderators affect key retailing relationships
variables for explaining shopping frequencies. But all of the (e.g., product quality → retail choice). Future researchers
variables in Tables 2 and 3 have been shown to be important therefore must explore if and to what extent those moder-
in some studies, so our analysis reveals a way to rank-order ators account for heterogeneity in studies, which could lead
these predictors. to greater confidence regarding the generalizability of the
The study of demographic variables represents pioneer- relationships.
ing work on store patronage; we find that the majority of Our study includes only those antecedent factors that have
demographic variables do not have a strong effect, except for been frequently (n ≥ 4) investigated in retailing literature.
gender, which remains a key predictor of shopping frequency. Among the predictors that did not meet this threshold are
With respect to the effects of age and income, our null results task definition (i.e., the purpose of a shopping trip), intended
may be due to two factors: First, only a small proportion of the purchase quantities, and personality traits (e.g., risk percep-
studies finds a significant effect, and second, there appears to tion, self-confidence, locus of control). These factors there-
be a bimodal distribution around 0, in that some studies report fore may have not received sufficient attention from retailing
a positive effect and others report a negative one. Given the researchers.
Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243 241

Appendix A

Selected empirical studies on retail patronage


242 Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243

References Gentry, J.W. and A.C. Burns (1977). “How ‘important’ are evaluative criteria
in shopping center patronage?,” Journal of Retailing, 53 (4), 73–86, 94.
Ajzen, I. (1985). “From intentions to actions: a theory of planned behavior,” Goldman, Arieh (1977). “The shopping style explanation for store loyalty,”
in Action Control: From Cognition to Behavior, J. Kuhl and J. Beckmann Journal of Retailing, 53 (4), 33–46, 94.
eds. New York: Springer-Verlag, 11–39. Grewal, Dhruv, Kent Monroe and R. Krishnan (1998). “The effects of price-
Arnold, Stephen J., Tae H. Oum and Douglas J. Tigert (May, 1983). comparison advertising on buyers’ perceptions of acquisition value,
“Determining attributes in retail patronage: seasonal, temporal, regional, transaction value, and behavioral intentions,” Journal of Marketing, 62
and international comparisons,” Journal of Marketing Research, 20, (2), 46–59.
149–157. Grewal, Dhruv, Julie Baker, Michael Levy and Glenn Voss (2003). “The
Baker, Julie, Dhruv Grewal and A. Parasuraman (1994). “The influence of effects of wait expectations and store atmosphere evaluations on patron-
store environment on quality inferences and store image,” Journal of the age intentions in service-intensive retail stores,” Journal of Retailing, 79,
Academy of Marketing Science, 22 (4.), 328–339. 259–268.
Baker, Julie, A. Parasuraman, Dhruv Grewal and Glenn B Voss (2002). “The Hansen, Robert and Terry Deutscher (1977). “An empirical investigation of
influence of multiple store environment cues on perceived merchandise attribute importance in retail store selection,” Journal of Retailing, 53
value and patronage intentions,” Journal of Marketing, 66 (2.), 120–141. (4), 59–72, 95.
Bauer, R.A. and Stephen A. Greyser (1968). Advertising in America: The Hays, W.L. (1973). Statistics for the Social Sciences. New York: Holt, Rin-
Consumer View. Boston, MA: Harvard University. chart & Winston.
Bellenger, D.N. and P.K. Korgaonkar (1980). “Profiling the recreational Hedges, L.V. and Ingram Olkin (1985). Statistical Methods for Meta-
shopper,” Journal of Retailing, 56 (3), 77–91. Analysis. Orlando, FL: Academic Press.
Bellenger, D.N., Dan Robertson and Elizabeth Hirschmann (1976). “Age Hunter, John E. and Frank L. Schmidt (1990). Methods of Meta-Analysis:
and education as key correlates of store selection for female shoppers,” Correcting Error and Bias in Research Findings. Newbury Park, CA:
Journal of Retailing, 52 (4), 71–78. Sage Publications.
Bellenger, D.N., D.H. Robertson and B.A. Greenberg (1977). “Shopping Jacoby, Jacob and David Mazursky (1985). “The impact of linking brand
center patronage motives,” Journal of Retailing, 53 (2), 29–38. and retailer images on perceptions of quality,” in Perceived Quality:
Berry, Leonard, Kathleen Seiders and Dhruv Grewal (July, 2002). “Under- How Consumers View New Stores and Merchandise, Jacob Jacoby and
standing service convenience,” Journal of Marketing, 66, 1–17. C. Olson Jerry eds. Lexington, MA: D.C. Heath and Company, 155–159.
Brengman, Malaika and Maggie Geuens (2002). “Profiling Internet users Kasulis, Jack and Robert Lusch (1981). “Validating the retail store image
based on their propensity to adopt online shopping,” Asia Pacific concept,” Journal of the Academy of Marketing Science, 9 (4), 419–
Advances in Consumer Research, 5, 30–39. 435.
Craig, S., A. Ghosh and S. McLafferty (1984). “Models of retail location Kendall, M.G. and A. Stuart (1967). The Advanced Theory of Statistics sec-
process: a review,” Journal of Retailing, 60 (1), 5–36. ond ed. New York: Hafner.
Crask, Melvin and Fred Reynolds (1978). “An indepth profile of the depart- Kenhove, Patrick Van, Kristof De Wulf and Walter Van Waterschoot (1999).
ment store shopper,” Journal of Retailing, 54 (Summer), 23–32. “The impact of task definition on store-attribute saliences and store
Darden, William R. and James R. Lumpkin (1984). “Psychographic and choice,” Journal of Retailing, 75 (1), 125–137.
demographic profile of convenience food store users: why people con- Kerin, Roger, Ambuj Jain and Daniel J. Howard (1992). “Store shopping
venience shop,” Review of Business & Economic Research, 19 (2), 68–80. experience and consumer price-quality-value perceptions,” Journal of
Darden, William R. and JoAnn Schwinghammer (1985). “The influence of Retailing, 68 (4), 376–397.
social characteristics on perceived quality in Patronage Choice Behav- Koelemeijer, Kitty and Harmen Oppewal (1999). “Assessing the effects of
ior,” in Perceived Quality: How Consumers View New Stores and Mer- assortment and ambience: a choice experimental approach,” Journal of
chandise, Jacoby Jacob and C. Olson Jerry eds. Lexington, MA: D.C. Retailing, 75 (3), 319–345.
Heath and Company, 161–172. Korgaonkar, Pradeep K., Daulat Lund and Barbara Price (1985). “A
Darley, William K. and Jeen-Su Lim (1993). “Store-choice behavior for structural-equation approach toward examination of store attitude and
pre-owned merchandise,” Journal of Business Research, 27, 17–31. store patronage behavior,” Journal of Retailing, 61 (2), 39–60.
Dellaert, Benedict, Theo Arentze and Michel Bierlaire (1998). “Investigat- Lambert, Z.V. (1979). “An investigation of older consumers: unmet needs
ing consumers’ tendency to combine multiple shopping purposes and and wants at the retail level,” Journal of Retailing, 55 (4), 37–57.
destinations,” Journal of Marketing Research, 3 (2), 177–188. Levy, Sidney I. (1966). “Social class and consumer behavior,” in On Know-
Dhar, S.K., S.J. Hoch and Nirmalya Kumar (2001). “Effective category man- ing the Consumer, J. W. Newman ed. New York: John Wiley & Sons,
agement depends on the role of the category,” Journal of Retailing, 77 146–160.
(2), 165–184. Levy, Michael and Barton A. Weitz (1995). Retailing management. Chicago:
Dodds, William B., Kent B. Monroe and Dhruv Grewal (August, 1991). Richard D. Irwin.
“The effects of price, brand, and store information on buyers’ product Louviere, J.J. and G.J. Gaeth (1987). “Decomposing the determinants of
evaluations,” Journal of Marketing Research, 28, 307–319. retail facility choice using the method of hierarchical information inte-
Donovan, Robert J., John R. Rossiter, G. Marcoolyn and A. Nesdale (1994). gration: a supermarket illustration,” Journal of Retailing, 63 (1), 25–48.
“Store atmosphere and purchasing behavior,” Journal of Retailing, 70 Lumpkin, James R. and John J. Burnett (1991). “Identifying determinants of
(3), 283–294. store type choice of the mature consumer,” Journal of Applied Business
Eastlick, Mary Ann and Mengmeng Liu (1997). “The influence of store Research, 8 (1), 89–102.
attitudes and other nonstore shopping patterns on patronage of television Lumpkin, James R. and J.M. Hawes (1985). “Retailing without stores: an
shopping programs,” Journal of Direct Marketing, 11 (3), 14–24. examination of catalog shoppers,” Journal of Business Research, 13,
Fern, F. Edward and Kent B. Monroe (September, 1996). “Effect size esti- 139–151.
mates: issues and problems in interpretation,” Journal of Consumer Lutz, Richard J. (1985). “Affective and cognitive antecedents of attitude
Research, 23, 89–115. toward the ad: a conceptual framework,” in Psychological Processes
Finn, A. and J. Louviere (1990). “Shopping center patronage models: fash- and Advertising Effects: Theory, Research and Application, L. F. Alwitt
ioning a consideration set segmentation solution,” Journal of Business and A. A. Mitchell eds. Hillsdale, NJ: Lawrence Erlbaum Associates,
Research, 21 (3), 159–175. 45–63.
Finn, A. and J. Louviere (March, 1996). “Shopping center image, con- MacKenzie, Scott B. and Richard L. Lutz (April, 1989). “An empirical
sideration, and choice: anchor store contribution,” Journal of Business examination of the structural antecedents of attitude toward the ad in
Research, 35, 241–251. an advertising pretesting context,” Journal of Marketing, 53, 48–65.
Y. Pan, G.M. Zinkhan / Journal of Retailing 82 (3, 2006) 229–243 243

Malhotra, N.K. (1983). “A threshold model of store choice,” Journal of Sirohi, Niren and Edward McLaughlin (1998). “A model of consumer per-
Retailing, 59 (2), 3–21. ceptions and store loyalty intentions for a supermarket retailer,” Journal
Olshavsky, Richard W. (1985). “Perceived quality in consumer decision of Retailing, 74 (2), 223–245.
making: an integrated theoretical perspective,” in Perceived Quality: Stassen, Robert, John Mittelstaedt and Robert Mittelstaedt (1999). “Assort-
How Consumers View Stores and Merchandise, Jacoby Jacob and ment overlap: its effect on shopping patterns in a retail market when the
C. Olson Jerry eds. Lexington, MA: D.C. Heath and Company, 3– distributions of prices and goods are known,” Journal of Retailing, 75
29. (3), 371–386.
Pedhazur, Elazar and Liora Schmelkin (1991). Measurement, Design, and Sultan, Fareena, John Farley and Donald Lehmann (February, 1990). “A
Analysis: An Integrated Approach. Hillsdale, NJ: Lawrence Erlbaum meta-analysis of applications of diffusion models,” Journal of Marketing
Associates. Research, 27, 70–77.
Peter, J.Paul and Gilbert Churchill (February, 1986). “Relationships Tauber, E.M. (October, 1972). “Why do people shop?,” Journal of Marketing,
among research design choices and psychometric properties of rat- 36, 46–49.
ing scales: a meta-analysis,” Journal of Marketing Research, 23, 1– Tellis, Gerard and Gary Gaeth (1990). “Best value, price-seeking, and price
10. aversion: the impact of information and learning on consumer choices,”
Rao, Akshay R. and Kent B. Monroe (1988). “The moderating effect of Journal of Marketing, 54 (2), 34–45.
price knowledge on cue utilization in product evaluations,” Journal of Thelen, Eva M. and Arch G. Woodside (1997). “What evokes the brand or
Consumer Research, 15 (2), 253–264. store? Consumer research on accessibility theory applied to modeling
Rao, Akshay R. and Kent B. Monroe (1989). “The effect of price, brand primary choice,” International Journal of Research in Marketing, 14,
name, and store name on buyers’ subjective product assessments: 125–145.
an integrative review,” Journal of Marketing Research, 26 (3), 351– Tigert, Douglas J. (1983). “Pushing the hot buttons for a successful retailing
357. strategy,” in Patronage Behavior and Retail Management, R. Darden
Reilly, William J. (1931). The Law of Retail Gravitation. New York: Author. William and F. Lusch Robert eds. New York: North-Holland.
Rosenthal, Robert (1991). Meta-Analytic Procedures for Social Research. Walters, Rockney G. and Heikki J. Rinne (1986). “An empirical investigation
Newbury Park, CA: Sage Publications. into the impact of price promotions on retail store performance,” Journal
Roy, Abhik (1994). “Correlates of mall visit frequency,” Journal of Retailing, of Retailing, 62 (Fall), 237–266.
70 (2), 139–161. Westbrook, R.A. and W.C. Black (1985). “A motivation-based shopper typol-
Rubenstein, C. and P. Shaver (1980). “Loneliness in two northeastern cities,” ogy,” Journal of Retailing, 61 (1), 78–103.
in The Anatomy of Loneliness, New York: International Universities Wolf, Fredric M. (1986). Meta-Analysis: Quantitative Methods for Research
Press. Synthesis. Beverly Hills, CA: Sage Publications.
Samli, A. Coskun (1975). “Use of segmentation index to measure store Woodside, Arch G. and Randolph J. Trappey (1992). “Finding out why
loyalty,” Journal of Retailing, 51 (1), 51–113. customers shop your store and buy your brand: automatic cognitive pro-
Sirgy, M. Joseph and A. Coskun Samli (1985). “A path analytic model of cessing models of primary choice,” Journal of Advertising Research, 32
store loyalty involving self-concept, store image, geographic loyalty, and (6), 59–78.
socioeconomic status,” Journal of the Academy of Marketing Science, Zeithaml, V.A. and Leonard Berry (1996). “The behavioral consequences of
13 (3), 265–291. service quality,” Journal of Marketing, 60 (2), 31–46.

You might also like