You are on page 1of 17

Journal of Service Research http://jsr.sagepub.

com/

The Role of Corporate Image for Quality in the Formation of Attitudinal Service Loyalty
Subhash Jha, George D. Deitz, Emin Babakus and Ugur Yavas
Journal of Service Research 2013 16: 155 originally published online 20 March 2013
DOI: 10.1177/1094670512466441

The online version of this article can be found at:


http://jsr.sagepub.com/content/16/2/155

Published by:

http://www.sagepublications.com

On behalf of:

Center for Excellence in Service, University of Maryland

Additional services and information for Journal of Service Research can be found at:

Email Alerts: http://jsr.sagepub.com/cgi/alerts

Subscriptions: http://jsr.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - Apr 15, 2013

OnlineFirst Version of Record - Mar 20, 2013

What is This?

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Article
Journal of Service Research
16(2) 155-170
The Role of Corporate Image for Quality in ª The Author(s) 2013
Reprints and permission:
sagepub.com/journalsPermissions.nav
the Formation of Attitudinal Service DOI: 10.1177/1094670512466441
jsr.sagepub.com
Loyalty
Subhash Jha1, George D. Deitz2, Emin Babakus2, and Ugur Yavas3

Abstract
Drawing from signaling theory, this study investigates the processes through which corporate image (CI) for quality affects
attitudinal loyalty. The research hypotheses are examined using data from a cross-sectional survey and two scenario-based
experiments. Overall, findings across these three studies suggest that the effects of CI upon loyalty are channeled through cus-
tomer satisfaction and perceived value. The effects of CI on perceived value and loyalty are stronger relative to the effects of
employee interaction quality (IQ) when IQ is measured as an overall evaluation. However, when employee IQ is measured in
reference to a specific service encounter, it becomes a stronger driver of perceived value and loyalty relative to CI. Regardless
of the context of measurement (i.e., overall evaluation vs. a specific service encounter), employee IQ exerts a stronger influence
on customer satisfaction than CI. CI negatively moderates the effect of employee IQ on customer satisfaction and loyalty, render-
ing the effect of employee IQ upon customer evaluations less critical for service providers with stronger CIs. The findings highlight
the relevance and importance of CI as a signal of unobservable quality, which should be measured and closely monitored by man-
agement. Managers should also recognize the central role of customer satisfaction, especially in channeling the effects of IQ upon
loyalty, and therefore, design policies that enhance frontline employee ability and motivation to deliver satisfying customer
experiences.

Keywords
corporate image, employee interaction quality, signaling theory, attitudinal loyalty, common method bias

Introduction In addition, empirical research in this area has adopted a


host of theoretical and methodological perspectives. While
With growing convergence in many markets in terms of the
scholars broadly agree upon the importance of CI, a review
relative performance, price, and availability of competing
of the literature reveals a decided lack of consistency across
offerings, the nurturing of corporate image (CI) as an integral
studies in terms of theoretic rationale and the causal ordering
component of firms’ broader strategic planning and implemen-
of key antecedents and outcomes. For instance, findings from
tation efforts has taken on increased significance. Bontis, Booker, and Serenko (2007) and Walsh et al. (2009)
While an extensive literature has emerged around the image
suggest that satisfaction favorably influences reputation, a term
construct, knowledge of how and when consumers’ perceptions
often used interchangeably with CI. On the other hand,
relating to the image of a given company influence individual
Andreassen (1999) found empirical support for positive rela-
and firm-level outcomes remains in its early stages (Brown and
tionships between CI and satisfaction and loyalty.
Dacin 1997). As noted by Brown et al. (2006), progress has been
blunted by a number of key conceptual, theoretical, and level of
analysis issues arising from the cross-disciplinary nature of the
CI construct. Although a variety of definitions exist, most con- 1
verge on the notion of CI as associations and meanings relating Department of Marketing & Strategy, IBS-Hyderabad, ICFAI Foundation for
Higher Education, Hyderabad, Andhra Pradesh, India
to the firm held in the minds of its stakeholders; thus, there can 2
Fogelman College of Business and Economics, The University of Memphis,
be no unanimously shared CI for any given company. Further, Memphis, TN, USA
3
work by Dowling (1986) and others (e.g., Gürhan-Canli and East Tennessee State University, Johnson City, TN, USA
Batra 2004) has shown that consumers may hold a variety of dif-
ferent associations with respect to a focal firm. That is, CI does Corresponding Author:
George D. Deitz, Fogelman College of Business and Economics, The University
not refer to a generalized view of the firm, but rather relates to of Memphis, 302 Fogelman College Administration Building, Memphis, TN
associations held with respect to a particular referent (e.g., an 38152, USA.
image for trustworthiness, quality, innovation, etc.). Email: gdeitz@memphis.edu

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


156 Journal of Service Research 16(2)

Adding to this confusion, little is understood regarding the data from customers who have an ongoing service history with
impact of CI upon key outcomes in comparison to characteristics a national bank in New Zealand. Confidence in Study 1 results
of the firm’s market offering, such as employee interaction qual- and conclusions are bolstered by a rigorous analytic approach
ity (IQ) (Hartline and Jones 1996), product and service perfor- in which latent factor scores were used to eliminate the effects
mance (Selnes 1993), and service guarantees (Ostrom and of measurement error and common method variance. Next,
Iacobucci 1998). Likewise, more knowledge relating to potential findings from the survey-based research are triangulated and
positive and negative interaction effects between CI and attri- deepened through two follow-up experiments.
butes of the product or service is needed.
The objective of this research is to examine the effects of CI
for quality upon consumers’ cognitive, affective, and beha- Conceptual Development
vioral evaluations of the firm relative to the influence of Signaling theory suggests that information asymmetry exists
employee IQ. While a consensus has yet to emerge regarding between sellers and buyers due primarily to differences in infor-
the dimensionality and measurement of CI (Stern, Zinkhan, and mation availability (Spence 1973). As a result, buyers make pur-
Jaju 2001), the quality dimension is perhaps the most important chase decisions based on sellers’ signals that are broadly
dimension for the success of a service firm (Shapiro 1983). classified as intrinsic and extrinsic cues (Olson and Jacoby
Therefore, we focus on CI for quality which refers to the over- 1972). Intrinsic cues are commonly thought of as built-in char-
all impression of the firm in the minds of its customers as it acteristics of a product or service. Extrinsic cues, on the other
relates to its overall commitment to quality. hand, are related to the offering but are not components parts
Employee IQ refers to customer experiences with frontline of the product or service offering. So, for example, the data plan
employees. This construct encompasses the ‘‘relationship for a smart phone would be an intrinsic cue, whereas the brand
component’’ of the service experience as opposed to the ‘‘core name would represent an extrinsic cue. Customers evaluate
component,’’ which refers to, for instance, the quality of food extrinsic and intrinsic cues concurrently in order to infer the
served in a restaurant or checking account products in a per- overall value of the market offering (Rao and Monroe 1989).
sonal banking context (Iacobucci and Ostrom 1993). Past Based upon this understanding of extrinsic and intrinsic
research suggests that employee IQ is a critical element of cus- cues, we present our conceptual framework in Figure 1. The
tomer assessments of service quality (Hartline and Jones 1996). underlying premise of the model is that buyers and sellers
Moreover, as it is often easier to replicate elements of a firm’s hold differential levels of information during a market inter-
core service offering, employee IQ represents a potential action (Spence 1973). For instance, customers may not be
source of competitive advantage. able to know some aspects of the quality of a service offering
CI for quality differs from employee IQ in that the former because of its experience and credence properties, which are
deals broadly with the company while the latter relates specifi- difficult to judge (Nelson 1970). Therefore, customers look
cally to customers’ service experiences (Brown and Dacin for information that enables them to differentiate between
1997). This is consistent with current views in the literature that high- and low-quality service providers. Firm managers, on
suggest the company and its offerings exist as separate entities in the other hand, seek to transmit signals targeted toward
the minds of customers (Aaker 1996). Further, we note here that selected customer segments in order to overcome information
IQ may be conceptualized as an overall evaluation of an ongoing asymmetries, particularly in instances in which quality is not
service relationship or be tied to a specific service encounter. readily observable (Kirmani and Rao 2000). While managers
Over the course of the three studies presented in this article, may choose from a variety of signals (e.g., warranty, price,
we examine both generalized perceptions of IQ as well as cus- advertisements), we focus upon CI, as an extrinsic cue, and
tomer perceptions relating to a specific service encounter. Our frontline employee IQ, as an intrinsic cue (cf. Andreassen and
expectation is that the relative effects of IQ will be much stron- Lindestad 1998).
ger when the focus is on a specific service encounter. Overall, the present study suggests that customer percep-
The present study contributes to the literature by addressing tions of value and satisfaction are largely shaped through their
three key questions. First, what are the mechanisms through concurrent interpretations of extrinsic and intrinsic cues, which
which consumers’ perceptions of a firm’s CI for quality shape ultimately influence their behavioral intentions (Purohit and
attitudinal loyalty? Second, what are the relative effects of cus- Srivastava 2001; Richardson, Dick, and Jain 1994; Zeithaml
tomer perceived CI, an extrinsic cue, and frontline employee 1988). Thus, it is critical to consider relevant intrinsic and
IQ, an intrinsic cue, in facilitating these outcomes? Finally, is extrinsic cues together in order to identify the processes
the importance of IQ lessened for firms with stronger images through which each effects customer loyalty as well as to deter-
for quality? mine potential synergistic effects.
In the process of addressing these questions, the study
makes several contributions. First, the article builds a parsimo-
Mediating Effects of Perceived Value and Customer
nious theoretical framework based on the tenets of signaling
theory that examines the relative effects of CI and frontline
Satisfaction
employee IQ upon three key customer outcomes: satisfaction, Following Bolton and Lemon (1999), we conceptualize per-
value, and attitudinal loyalty. This model is tested using survey ceived value as a broad construct that encompasses perceptions

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Jha et al. 157

Perceived
Corporate Image Value
(CI) (VAL)

Attitudinal
Loyalty
(LOY)

Interaction Quality Customer


(IQ) Satisfaction
(SAT)

Figure 1. Research model. Dashed arrows signify the role of CI as a moderator of the effects of IQ. Control variables: gender, frequency of
face-to-face interaction, customer tenure.

of quality given price and inputs versus outputs relative to some performance expectations, there is reason to expect a positive
referent, typically a rival offering. When customers perceive association between CI and customer satisfaction.
that the ratio of received benefits to cost of a focal provider’s Providing customers with greater value and higher satisfac-
offering is proportionally greater than the ratio they would have tion is widely recognized as a means of improving customer
received via some alternative, it results in stronger perceptions loyalty (Cronin, Brady, and Hult 2000; Johnson, Hermann, and
of value. Based upon this conceptualization, it is posited that a Huber 2006). Expectancy-disconfirmation theory (Oliver
strong CI for quality may favorably influence perceived value 1980) as well as meta-analytic results (Szymanski and Henard
through one of the three routes. First, research has shown that 2001), lead us to expect positive linkages from CI to customer
CI may enhance perceptions of benefit (e.g., they may view the satisfaction and perceived value to customer satisfaction. Simi-
service as more reliable; e.g., Nguyen and LeBlanc 1998). Sec- larly, there is sufficient theory (Woodruff 1997) and empirical
ond, a favorable CI can serve to minimize consumer percep- evidence to support expectations of a positive relationship
tions of overall costs (e.g., reducing search costs, perceived between CI and perceived value. Indeed, Lai, Griffin, and
risk). For instance, Helm, Eggert, and Garnefeld (2010) found Babin (2009) found that CI affected perceived value, satisfac-
that firm reputation acts as signal that helps reduce buyers’ tion, and loyalty; however, these authors did not examine more
uncertainty pertaining to future service performance. Finally, complex nomological relationships, such as mediation, involv-
the strength of a firm’s CI may also lead to enhanced value per- ing these variables. Based on the preceding discussion, individ-
ceptions by shaping the basis of comparison with a given refer- uals’ perceptions of CI should enhance the perceived value and
ent, or even influencing the choice of referent itself (Woodruff satisfaction with the service offering, which in turn, help shape
1997). behavioral intentions. It is proposed that:
Much of the research relating to customer satisfaction has
centered on the expectancy-disconfirmation paradigm (Oliver Hypothesis 1a and b: The relationship between CI and cus-
1980), in which customers’ a priori expectations provide a tomer loyalty will be mediated by (a) perceived value and
frame of reference for their subsequent evaluative judgments. (b) customer satisfaction.
Thus, satisfaction is a function of consumers’ expectations plus
any perceived disconfirmation of these expectations during the Relative Impact of CI and IQ Upon Customer Attitudes
consumption experience. As CI, which represents consumer Understanding the relative importance of extrinsic and intrinsic
associations relating to the overall quality of the provider, pro- cues upon customer perceptions and behaviors has been an
vides consumers with a reasonable basis for future service issue of keen interest for researchers and practitioners (Olson

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


158 Journal of Service Research 16(2)

and Jacoby 1972; Purohit and Srivastava 2001). The diminished for firms with more favorable CI. Theoretical sup-
accessibility-diagnosticity framework put forward by Feldman port for this negative interaction effect can be drawn from
and Lynch (1988) offers some guidelines for understanding the research relating to anchoring and adjustment processes
relative roles of extrinsic and intrinsic cues upon customer atti- (Tversky and Khaneman 1974). Anchoring and adjustment the-
tudes, intentions, and behaviors. The framework posits that ory suggests that consumers will evince a systematic selection
when diagnostic inputs are accessible, consumers base their bias in the presence of two signals. In this case, we argue that
judgments on memory-based information rather than consumers use CI as an anchor for evaluating the service and
stimulus-based information (Menon, Raghubir, and Schwarz adjust for the effect of IQ in line with a service provider’s
1995). In other words, when two possible cues are present, the image. Since adjustment is often insufficient (Nelson and Sim-
more easily accessible cue should have greater impact upon the mons 2005), we posit that CI guides the effect of IQ in shaping
decision-making process (Feldman and Lynch 1988). subsequent outcomes. For instance, Ostrom and Iacobucci
Similarly, under cue utilization theory (Olson and Jacoby (1998) argued that guarantees work better for a budget hotel
1972), the relative weight of extrinsic and intrinsic cues than for a luxury hotel because image of the luxury hotel pro-
depends on each cues’ predictive value (i.e., the degree to vides sufficient information about service quality for consu-
which consumers associate a given cue with product quality) mers. These authors suggested that the CI provides service
and confidence value (i.e., the degree to which consumers have customers with an implicit idea of receiving good service qual-
confidence in their ability to use and judge that cue accurately). ity during consumption, which discounts the impact of any
Given the risks associated with selecting a poor-quality brand, explicit quality of the service offering (Ostrom and Iacobucci
cues that are highest in these two qualities assume the greatest 1998).
weight in the consumer assessment process. For instance, We reason that in an ongoing service relationship, the
Richardson, Dick, and Jain (1994) found that an extrinsic cue, occurrence of a specific service failure does not immedi-
retail store image, was more useful than an intrinsic cue, taste ately translate into dissatisfaction, but is instead buffered
of actual product ingredients, in explaining variance in consu- by consumers’ favorable image of the organization. In con-
mers’ perceptions of the product quality of retailer store brands. trast, a single episode of excellent IQ may not have a strong
While both extrinsic and intrinsic cues were significant predic- impact on satisfaction or perceived value if CI is poor. This
tors of product quality perceptions, these authors reasoned that argument is in line with the observation that the unique
consumers had greater confidence in their ability to utilize the effect of a cue is diminished when consumers can draw
extrinsic store image cue. inferences about a product from other cues (Biswas 2004).
We argue that in situations where consumers are asked to Hence, we advance:
provide an overall evaluation of an ongoing service relation-
ship (e.g., continuing customer interactions with a particular Hypothesis 3a–c: CI negatively moderates the effect of IQ
bank over a period of time) without reference to a specific ser- on (a) perceived value, (b) customer satisfaction, and
vice encounter, CI provides a holistic summary and is a more (c) loyalty, such that the effects of IQ will be weaker
easily accessible cue relative to IQ. Furthermore, in a setting (stronger) for customers who possess a more (less) favor-
such as retail banking, CI will hold stronger predictive and able CI of the provider.
confidence value for consumers in making judgments relating
to an ongoing service relationship. Based upon this frame- Empirical Studies
work, we posit that CI acts as memory based self-schema,
information that is readily accessible in the overall evaluation Study 1 is based on survey data from individual customers of a
of an ongoing service offering or relationship (Selnes 1993). New Zealand bank. Results from this research were triangu-
Alternatively, due to the intangibility of service offerings lated and extended through two subsequent experiments using
(Zeithaml 1988), it may be relatively more difficult for cus- graduate student data. Study 2 sought to replicate Study 1 find-
tomers to access IQ information and utilize it in subsequent ings in an experimental design utilizing a personal banking sce-
judgments of value, satisfaction, and repurchase intentions. nario. Study 3 sought to examine the same relationships across
This is particularly plausible in evaluations of IQ as a general a more transactional restaurant context. Therefore, Study 1 is
construct, without reference to a particular service encounter. based on an ongoing service relationship, while the experi-
Hence, we offer: ments examine hypothesized relationships using a single ser-
vice encounter.
Hypothesis 2a–c: With respect to an ongoing service rela-
tionship, CI has a relatively stronger impact upon (a) per-
Study 1: Bank Customer Survey
ceived value, (b) customer satisfaction, and (c) customer
loyalty than IQ. Method
Data Collection and Sample Profile. Data were collected via ques-
Moderating Effects of CI tionnaires mailed to customers of a national bank in New Zeal-
For consumers in an ongoing service relationship, we expect and. Respondents were randomly selected and given assurance
that the effect of IQ upon customer outcomes will be of anonymity. Of the 2,500 questionnaires mailed, 872 were

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Jha et al. 159

returned yielding a response rate of 34.9%. Nearly 65% of evaluation of intrinsic cues at the time of purchase may substi-
respondents were female, 60% were younger than 45 years of tute the importance of extrinsic cues after certain period of time
age, and around 85% earned less than NZ$70,000. Approxi- (Selnes 1993). Thus, we control for the length of the bank-
mately 40% had been a customer of the bank for more than 5 customer relationship. Customer tenure with the bank was mea-
years. A comparison of sample characteristics to company sured using a 5-point scale (less than a year ¼ 1; 1–2 years ¼ 2;
records indicated that the respondents were representative of 3–5 years ¼ 3; 6–9 years ¼ 4; more than 9 years ¼ 5). Finally,
the bank’s customer population. it is believed that repeated encounters with frontline employees
enable customers to more easily evaluate intrinsic cues com-
Construct Measurement. CI was measured by 5 items, 3 adapted pared to those who have less frequent employee interaction
from Brady and Cronin (2001) and 2 new items, each using a (Ostrom and Iacobucci 1998). Thus, our third control variable
10-point semantic differential response format. Respondents relates to the frequency of face-to-face customer interaction
were asked to provide a relative evaluation of the bank on these with the bank employees (recorded as more than 4 times/month
items after reading the following introduction: ‘‘Considering ¼ 4, 3–4 times/month ¼ 3, 1–2 times/month ¼ 2, less than
any experiences you may have had with other banks; what you once/month ¼ 1).
may have heard from friends and colleagues about their experi-
ences with various banks; what you may have heard about Validation of Measures. The entire set of items comprising the
banking services from the media, please evaluate XYZ Bank measures of study constructs was initially subjected to an
relative to other banks.’’ An example item reads as follows: exploratory principal component factor analysis using oblique
‘‘Compared to other banks, XYZ Bank is: poor (1)—excellent rotation since the underlying factors were expected to correlate.
(10).’’ IQ was measured using 5 items adapted from Cronin, Listwise deletion of missing responses resulted in a final sam-
Brady, and Hult (2000). Each item was cast using 7-point ple size of 777.1 The analysis initially produced a four-factor
Likert-type responses, ranging from strongly disagree (1) to solution, each factor with eigenvalues larger than 1.0 and in
strongly agree (7). We adapted Fornell et al’s. (1996) 3-item total accounting for 80% of the variance. Satisfaction and loy-
perceived value (VAL) scale with each item reflecting per- alty items loaded heavily on the same factor. A five-factor solu-
ceived value relative to the focal bank’s competitors. A 10- tion was imposed in search of an underlying structure
point response format was used with anchor labels ‘‘XYZ Bank consistent with our theoretical expectations. The five-factor
is much worse’’ (1) and ‘‘XYZ Bank is much better’’ (10). The solution accounted for 84% of the variance, with the first factor
use of comparative scales to measure these constructs was explaining 63% of the variance. A CI scale item, a satisfaction
based on the premise that loyalty develops, in part, as a result scale item, and two loyalty scale items were eliminated as a
of a consumer’s comparative appraisals in relation to a consid- result of high cross-loadings (larger than .45). The remaining
eration set of available brands (Gale 1994). 20 items were reanalyzed and a five-factor solution emerged
Customer satisfaction (SAT) was measured via 5 items. with all items loading heavily (.82 or stronger) on their
Three were drawn from Brady and Cronin (2001) (e.g., intended factor and very small cross-loadings (largest cross-
happy/unhappy) and two (e.g., negative/positive) were added loading was .084). The first factor explained 67% of the var-
for this study. In each case 10-point bipolar scales were iance and the total variance explained by the five factors solu-
employed. Since satisfaction is primarily an affective construct tion was 89%. Table 1 presents scale items as well as the
(Oliver 1993), the selected adjective pairs were emotive in internal consistency (coefficient a) of the measures, which ran-
nature. Customer attitudinal loyalty (LOY) was assessed ged from .94 to .97. The 20 retained items were subjected to
though respondents’ behavioral intentions with regard to the confirmatory factor analysis (CFA) using the sample covar-
focal bank, using an adaptation of 6 items from Brady and Cro- iance matrix as input to LISREL 8.80 (Jöreskog and Sörbom
nin (2001). The items were cast using a 7-point response scale 1996).
ranging from very unlikely (1) to very likely (7). Owing to our single source data, we investigated the poten-
tial for common method variance to influence structural results
Control Variables. In an effort to provide rigorous tests of the pro- using Harman’s one-factor test. We first tested a five-factor
posed theoretical linkages, we treat gender, customer tenure, measurement model, followed by a single-factor model.
and frequency of face-to-face interactions as control variables Results from the w2 difference test indicated that common
in our analysis. Prior research suggests that the relative impor- method bias may not be a serious problem. Although Harman’s
tance of CI and IQ may differ between males and females approach is a useful diagnostic test, it falls short of controlling
(Babakus and Yavas 2008). For instance, it has been argued for the effects of common method variance. Hence, we incor-
that females value social interactions and relationships more porated a common method factor into the measurement model
than males (Iacobucci and Ostrom 1993). By extension, we rea- along with the five factors of substantive interest to the study.
son that females may place greater value on IQ. Gender was We specified the method factor to be uncorrelated with the
recorded as female ¼ 1 and male ¼ 2. other constructs and allowed each item to load on the method
It has also been argued that extrinsic cues will have a greater factor as well as its respective underlying factor (Podsakoff
importance than intrinsic cues in the initial purchase situations et al. 2003). Table 1 presents CFA results with and without the
for consumers (Zeithaml 1988). Moreover, consistent method factor.

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


160 Journal of Service Research 16(2)

Table 1. Confirmatory Factor Analysis—Results Without and With MF.

Scale Items Standardized Loadings t Values

Corporate Image (CI; a ¼ .96; without MF: AVE ¼ .86, F2 ¼ .44–.59; With MF: AVE ¼ .54, F2 ¼ .26–.37)
Please evaluate XYZ Bank relative to other banks (10-point scale)
CI1. Lower quality (1)–Higher quality (10) .90 (.84) * (*)
CI2. Lower standard (1)–Higher standards (10) .95 (.81) 47.6 (17.7)
CI3. Poor (1)–Excellent (10) .94 (.62) 46.5 (10.8)
CI4. Inferior (1)–Superior (10) .93 (.64) 44.7 (12.1)
Interaction Quality (IQ; a ¼ .94; without MF: AVE ¼ .79, F2 ¼ .34–.56; with MF: AVE ¼ .57, F2 ¼ .17–.40)
Employees of XYZ Bank are (strongly disagree ¼ 1 to strongly agree ¼ 7)
IQ1. Dependable .88 (.76) * (*)
IQ2. Competent .90 (.79) 37.4 (29.2)
IQ3. Knowledgeable .87 (.73) 34.5 (26.6)
IQ4. Reliable .90 (.78) 37.6 (29.3)
IQ5. Willing to provide service in a timely manner .85 (.71) 32.8 (25.2)
Perceived value (VAL; a ¼ .96; without MF: AVE ¼ .90, F2 ¼ .34–.59; With MF: AVE ¼ .63, F2 ¼ .17–.34)
Please rate XYZ Bank relative to other banks (much worse ¼ 1 to much better ¼ 10)
VAL1. The quality of service given the fees I pay .94 (.81) * (*)
VAL2. The fees I pay for the quality of service I receive .96(.81) 56.5 (42.1)
VAL3. Overall value from the service I receive .95(.76) 53.1 (37.4)
Customer satisfaction (SAT; a ¼ .97; without MF: AVE ¼ .88, F2 ¼ .42–.65; with MF: AVE ¼ .54, F2 ¼ .34–.47)
Please rate your feelings about your interactions with XYZ Bank (10-point scale)
SAT1. Unhappy (1)–Happy (10) .96 (.75) * (*)
SAT2. Displeased (1)–Pleased (10) .96 (.76) 63.6 (42.4)
SAT3. Terrible (1)–Delighted (10) .91 (.69) 51.2 (33.9)
SAT4. Dissatisfied (1)–Satisfied (10) .94 (.73) 58.0 (38.1)
Customer loyalty (LOY; a ¼ .96; without MF: AVE ¼ .85, F2 ¼ .44–.65; with MF: AVE ¼ .52, F2 ¼ .27–.47)
What is the likelihood that you (very unlikely ¼ 1 to very likely ¼ 7)
LOY1. Would recommend XYZ bank to others? .94 (.73) * (*)
LOY2. Would say good things about XYZ bank to others? .96 (.77) 56.7 (36.7)
LOY3. Encourage friends and relatives to do business with XYZ bank .92 (.72) 48.5 (32.5)
LOY4. Would choose XYZ bank if you had to choose a bank again .88 (.66) 42.1 (27.4)
Model-fit statistics without MF: w2160 ¼ 611.893, p ¼ .00, RMSEA ¼ .060, GFI ¼ .93, CFI ¼ .99, NFI ¼ .99, NNFI ¼ .99
Model-fit statistics with MF: w2140 ¼ 410.43, p ¼ .00, RMSEA ¼ .049, GFI ¼ .95, CFI ¼ 1.00, NFI ¼ .99, NNFI ¼ .99

Note. AVE ¼ average variance extracted; CFI ¼ comparative fit index; GFI ¼ goodness-of-fit index; MF ¼ method factor; NFI ¼ normed fit index; NNFI ¼
nonnormed fit index; RMSEA ¼ root mean square error of approximation.
Results without introducing method factor in the measurement model are presented first followed by results including method factor (in parentheses).
All the loadings are significant at the .01 level in both of the cases.
*Loadings were initially fixed to 1.0.

As shown in Table 1, overall fit statistics indicate that both Results


measurement models fit the data reasonably well and provide
Lance et al. (2010) have recently suggested that the inflationary
sufficient evidence of discriminant and convergent validity.
effects of common method variance on observed relationships are
However, the w2 values, standardized loadings, average var-
effectively washed out by the attenuating effects of measurement
iance extracted (AVE), and shared variance (F2) between
error. By comparing results obtained using estimated latent vari-
constructs diminished noticeably with the inclusion of the
able scores from a measurement model that includes a method
method factor. This suggests that common method variance
factor as well as composite scale scores, we can assess the validity
within the data may potentially inflate measured relationships.
of this argument by assessing differences in hypothesized struc-
In testing the measurement model with the method factor
tural paths. Thus, model testing was performed using both esti-
included, estimated latent variable scores were generated for
mated latent variable scores (free from measurement error and
each underlying construct (Jöreskog 2000). It is important to
common method variance) and composite scale scores (contain-
note that these latent variable scores were freed of both mea-
ing measurement error and common method variance) for each
surement error and common method variance, since they were
construct in the model as input to LISREL 8.80.
generated from a measurement model that explicitly modeled
both sources of bias. Table 2 presents the correlations, means,
and standard deviations of all variables used in the study. Stron-
ger correlations among composite scores relative to correla- Results Based on Latent Variable Scores
tions among estimated latent variable scores provide Hypothesis 1 states that the impact of CI on loyalty is mediated
evidence of common method–induced relationship inflation. by (a) perceived value and (b) customer satisfaction. As

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Jha et al. 161

Table 2. Correlations, Means, and Standard Deviations of Composite Scale Scores and Estimated Latent Variable Scores of Model Constructs
and Control Variables (n ¼ 777; Study 1).a,b,c

Variables CI IQ VAL SAT LOY X1 X2 X3

Corporate image (CI) 1.00 .41 .44 .50 .49 .00 .00 .06
Interaction quality (IQ) .60 1.00 .34 .58 .46 .06 .01 .05
Perceived value (VAL) .64 .53 1.00 .53 .41 .02 .02 .03
Customer satisfaction (SAT) .71 .70 .69 1.00 .64 .00 .02 .03
Attitudinal loyalty (LOY) .69 .62 .62 .77 1.00 .00 .01 .06
Customer gender (X1) .03 .02 .01 .04 .03 1.00 .03 .00
Face-to-face interaction (X2) .03 .03 .01 .06 .03 .02 1.00 .01
Customer tenure (X3) .04 .04 .03 .03 .04 .00 .01 1.00
Composite scale scores:
Mean 8.05 5.68 7.33 8.14 5.70 1.35 1.97 3.77
SD 1.39 1.09 1.80 1.56 1.41 .47 1.03 1.20
Note. Above the diagonal are the correlations based on estimated latent variable scores and below the diagonal are the correlations based on composite scale
scores of perceptual variables.
a
Composite scale scores for each construct were calculated by averaging respective item scores. A higher score indicates a more favorable response. CI, VAL, and
SAT were measured using a 10-point scale. IQ and LOY were measured using a 7-point scale.
b
Estimates of latent variable scores were generated as standardized scores (mean ¼ 0 and SD ¼ 1.0) from confirmatory factor analysis of scale items by including a
common method factor.
c
Customer tenure (for how long a customer has been associated with the bank) was measured using a 5-point scale and face-to-face interaction was measured
using a 4-point scale. Gender was recorded as (1 ¼ female and 2 ¼ male). Correlations > (.08) are significant at the .05 level.

presented in Table 3, CI significantly and positively impacts (as Since the main independent and dependent variables were
indicated by the Z-test statistic values) customer loyalty highly correlated, an alternative approach was used to determine
through three indirect routes (Z ¼ 2.30 for CI!VAL!LOY, the relative influences of CI and IQ on the dependent variables.
Z ¼ 3.73 for CI!SAT!LOY, and Z ¼ 2.44 for We used the relative weight analysis procedure developed by
CI!VAL!SAT!LOY) (Shrout and Bolger 2002).2 In addi- Johnson (2000) to provide an alternative test of Hypothesis
tion, CI has a significant direct effect on loyalty (b ¼ .20, t ¼ 2a–c. This approach determines the percentage of explained var-
6.45). These results show that perceived value and customer iance in the dependent variable that is due to each predictor.3
satisfaction partially mediate the effect of CI on customer loy- The relative weights for CI and IQ for predicting VAL: 65.9%
alty. This finding is consistent with Hypothesis 1a and b. and 32.1%, for SAT: 23.9% and 40.5%, and for LOY: 21.8% and
A number of studies have advanced the opposing causal 16.8%, respectively. These results suggest that CI has a stronger
ordering of the relationships tested by Hypothesis 1a and b impact on VAL and LOY, while IQ has a stronger influence on
(e.g., Bontis, Booker, and Serenko 2007; Hu, Kandampully, SAT. Therefore, Hypothesis 2a–c receives mixed support based
and Juwaheer 2009). Thus, we compared the overall perfor- on the estimated latent variable score data.
mance of our proposed model with the alternative model in The moderation hypothesis advanced in Hypothesis 3a–c was
which customer satisfaction and perceived value were antece- tested by creating an interaction variable using the product
dents to CI. An examination of the overall model-fit statistics of CI and IQ. The moderated model fits the data well,
reveals that our proposed model fits the data, w29 ¼ 11.37, w29 ¼ 12.01, p ¼ .21, RMSEA ¼ .021, goodness-of-fit index (GFI)
p ¼ .25; root mean square error of approximation ¼ 1.00, comparative fit index (CFI) ¼ 1.00, normed fit index (NFI)
(RMSEA) ¼ .018, better than the alternative model, w210 ¼ ¼ .99, nonnormed fit index (NNFI) ¼ .99, and explains 23% of the
152.10, p ¼ .00; RMSEA¼ .136. Since the rival models are not variance in VAL, 51% in SAT, and 47% in LOY. As hypothesized,
nested, a formal hypothesis test cannot be conducted. Nonethe- the interaction variable (CI*IQ) had significant negative impact on
less, all fit statistics favor the proposed model and provide SAT (b ¼ .12, t ¼ 4.69) and LOY (b ¼ .09, t ¼ 3.80), but
further support for Hypothesis 1a and b. not for VAL (b ¼ .02, t ¼ .66). Thus, these data provide partial
Hypothesis 2a–c predicts that in an ongoing service relation- support for Hypothesis 3.
ship, the effects of CI, an extrinsic cue, upon (a) perceived value,
(b) satisfaction, and (c) loyalty will be stronger than the effects of
IQ, an intrinsic cue. In order to test our hypotheses, we sequen- Results Based on Composite Scale Scores
tially imposed equality restrictions on the effects of CI and IQ In order to examine the extent to which the use of latent variable
on VAL (w210 ¼ 14.99), SAT (w210 ¼ 37.55), and LOY (w210 ¼ scores obtained from a method-factor imposed CFA obviates
14.38) and compared each w2 value with the baseline (w29 ¼ concerns over method bias, we examined the same model using
12.01) value for a model without such restrictions. A significant composite scale scores as input. Table 3 shows that overall
w2difference between constrained and unconstrained models was model-fit statistics for the composite score model was more than
found in just one case (SAT). Contrary to our hypotheses, IQ has a adequate, w29 ¼ 12.51, p ¼ .19, RMSEA ¼ .022, GFI ¼ 1.00,
stronger effect on customer satisfaction (SAT). CFI ¼ 1.00, NFI ¼ 1.00, NNFI ¼ 1.00, and was superior to the

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


162 Journal of Service Research 16(2)

Table 3. Tests of Structural Model and Research Hypotheses Using Estimated Latent Variable Scores and Composite Scale Scores (Study 1).a,b

Structural Model Standardized Estimate T Value/Z Valuec R2

Parameter
CI ! VAL .36 (.51) 10.42 (15.18)
IQ ! VAL 19 (.23) 5.58 (6.81)
CI  IQ ! VAL .02 (.04) 0.66 (1.53) .23 (.45)
CI ! SAT .22 (.30) 7.32 (10.34)
IQ ! SAT .38 (.32) 13.51 (12.20)
VAL ! SAT .31 (.32) 10.72 (11.91)
CI  IQ ! SAT .12 (.11) 4.69 (5.52) .51 (.69)
CI ! LOY .20 (.27) 6.45 (8.31)
IQ ! LOY .09 (.08) 2.85 (2.76)
VAL ! LOY .05 (.10) 1.58 (3.26)
SAT ! LOY .44 (.42) 11.70 (11.28)
CI  IQ ! LOY .10 (.11) 3.80 (5.08) .47 (.66)
Indirect effects of CI on LOY:
Route 1: CI!VAL!LOY .02 (.05) 2.30 (3.01)
Route 2: CI!SAT!LOY .10 (.13) 3.73 (5.08)
Route 3: CI!VAL!SAT!LOY .05 (.07) 2.44 (3.73)
Model-fit statistics using estimated latent variable scores:
w29 ¼ 12.01, p ¼ .21, RMSEA ¼ .021, GFI ¼ 1.00, CFI ¼ 1.00, NFI ¼ .99, NNFI ¼ .99
Model-fit statistics using composite scale scores:
w29 ¼ 15.43, p ¼ .08, RMSEA ¼ .030, GFI ¼ 1.00, CFI ¼ 1.00, NFI ¼ 1.00, NNFI ¼ .99
Note. CFI ¼ comparative fit index; GFI ¼ goodness-of-fit index; NFI ¼ normed fit index; NNFI ¼ nonnormed fit index; RMSEA ¼ root mean square error of
approximation.
a
Results using estimated latent variable scores are presented first followed by results based on composite scale scores (in parentheses). T (or Z) values
corresponding to one-tail tests at various significance levels: t > 1.3, p < .10; t > 1.65, p < .05; and t > 2.33, p < .01.
b
Of the three control variables, only customer tenure had a marginally significant positive impact on corporate image when the model was tested using latent
variable scores (g ¼ .06, t ¼ 1.72).
c
Test statistic values for the indirect effects are Z values (based on Sobel’s test) since they are computed based on large sample assumptions (Shrout and Bolger
2002). We also used a bootstrapping procedure (Selig and Preacher 2008) to compute 95% confidence intervals for the indirect effects. The confidence interval
results and Sobel test results are in agreement.

alternative model, w210 ¼ 152.24, p ¼ .00, RMSEA ¼ .136, scale scores for CI and IQ to minimize the effects of multicolli-
GFI ¼ .96, CFI ¼ .96, NFI ¼ .96, NNFI ¼ .88. CI had a positive nearity (Aiken and West 1991). The model fits the data well
indirect effect on loyalty through three indirect routes, Z ¼ 3.01 using composite scale scores (w29 ¼ 15.43, p ¼ .08, RMSEA
for CI to VAL to LOY, Z ¼ 5.08 for CI to SAT to LOY, and ¼ .030, GFI ¼ 1.00, CFI ¼ 1.00, NFI ¼ 1.00, NNFI ¼ .99). The
Z ¼ 3.73 for CI to VAL to SAT to LOY.4 In addition, CI had model explains 45% of the variance in VAL, 69% in SAT, and
a significant direct effect on loyalty (b ¼ .27, t ¼ 8.31). While 66% in LOY when composite scores are used as input. The inter-
the significance of hypothesized paths was consistent with the action variable (CI*IQ) had significant negative impact on SAT
latent variable score model, regression coefficients in the com- (b ¼ .18, t ¼ 5.71) and LOY (b ¼ .16, t ¼ 5.08). How-
posite score model tended to be larger for the direct and indirect ever, the effect of CI*IQ upon VAL was not significant (b ¼ .07,
paths. This finding bolsters the use of method-variance free t ¼ 1.53).5
latent variable scores in testing mediation.
In order to assess Hypothesis 2a–c, we sequentially imposed
equality restrictions on the effects of CI and IQ on VAL (w210 ¼ Discussion
37.03), SAT (w210 ¼ 15.63), and LOY (w210 ¼ 30.20) and Results of Study 1 generally support the hypothesized effects of
compared each w2 value with the baseline (w29 ¼ 15.43) model CI and employee IQ. First, support was found for the hypothesis
free of such restrictions. In two cases (VAL and LOY), we that customer satisfaction and perceived value mediate the effects
found significant w2differences between constrained and of CI on customer loyalty. Second, CI showed stronger effects on
unconstrained models consistent with Hypothesis 2. The rela- perceived value and loyalty relative to the effects of IQ, while,
tive weight analysis was repeated using the composite scores contrary to our hypothesis, IQ had a stronger effect on customer
data. The relative importance weights of CI and IQ in predict- satisfaction than CI. Third, as hypothesized, CI reduced the
ing VAL: 64.3% and 34.5%, for SAT: 31.2% and 31.7%, and effects of employee IQ on customer satisfaction and loyalty.
for LOY: 26.0% and 17.5%. Hence, analysis of composite Results obtained from latent variable scores and composite
scores generated partial support for Hypothesis 2. scale scores are similar, albeit with some noteworthy excep-
Hypothesis 3a–c was tested by creating an interaction vari- tions. First, there is a significant gap between the two sets of
able using the product of CI and IQ after centering the composite results in terms of the proportion of explained variance in the

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Jha et al. 163

dependent constructs. For instance, as shown in Table 3, R2 for MBA students enrolled in a marketing management class (45%
VAL generated by input based on latent variable scores is .23 male; mean age ¼ 23 years; 95% bank account holder) received
while it is .45 when composite scale scores are used as input. one of the four questionnaires that differed only in the CI and
In addition, the magnitudes of indirect effects and the majority IQ manipulations. Participants were told that we were inter-
of direct parameter estimates using composite scores are larger ested in their reactions to a fictitious experience at a bank. After
relative to those estimated by latent variable scores (which are reading the scenario, participants indicated the realism and
free from measurement error and common method variance). comprehensibility of the scenario questions through two, 10-
An unexpected finding in Study 1 related to the relative point semantic differential items (unrealistic-realistic; not easy
influence of CI and IQ on customer satisfaction. Contrary to to understand-easy to understand).
our hypothesis, IQ had a stronger impact on satisfaction than The realism (Ms > 7.83) and comprehensibility (Ms > 8.25)
CI. A potential explanation for this finding may be tied to the ratings were high and equivalent across all four conditions
nature of these constructs. CI is defined as the sum total of (Fs < 1). In addition, the manipulation checks were successful
knowledge of a service provider and relates to schemas held and suggested independence for the CI and IQ manipulations
in customers’ minds. Such cognitive schemas are likely to have (Perdue and Summers 1986).7 Hence, we used them in our
more influence on customer’s cognitive evaluations of per- main study.
ceived value and behavioral intentions. In contrast, IQ appears
to elicit a more emotive response. This seems to hold despite Sample and Procedure. Ninety-one MBA students enrolled in a
the fact that IQ was measured within the context of an ongoing marketing management class (51% male; mean age ¼ 23 years;
service relationship and not in reference to a specific service 98% bank account holder) participated in the main study, a 2
encounter. Satisfaction has been defined primarily an affective (CI: high, low) by 2 (IQ: high, low) between-subjects design.
construct (Oliver 1993); hence, the impact of IQ upon satisfac- They were randomly assigned to one of the four conditions.
tion is relatively stronger compared to CI. Participants were instructed to imagine themselves as the pro-
These results beg several questions. First, what if IQ and CI tagonist of the service encounter described in the scenario.
were measured with a focus on a single service encounter? We After reading the scenario, participants completed a structured
might expect assessments of IQ to be more emotion-laden due questionnaire consisting of questions for realism and compre-
to the immediacy of the encounter which, in turn, would spill hensibility, and the CI and IQ manipulations checks. Next, they
over heavily on all outcome variables, not just satisfaction. If responded to the same customer satisfaction (SAT; a ¼ .96),
so, does this mean that retrieval of CI, which may be the more perceived value (VAL; a ¼ .90), and customer loyalty (LOY;
distal construct under this circumstance, will be slower and sec- a ¼ .90) scales used in Study 1. Finally, participants completed
ondary to the immediate experience of IQ? Second, given our questions related to demographics and relayed their thoughts of
theory and results from Study 1, is it possible that the type of the study’s true purpose.
service might influence the relative roles of CI and IQ? Finally,
to what extent would results of our latent variable score model,
Results
which is free from both measurement error and method bias,
differ from experimental results in which common method var- Manipulation and Realism Checks
iance was controlled by design (i.e., by experimentally manip- A 2 (CI: high vs. low)  2 (IQ: high vs. low) analysis of cov-
ulating the independent variables in order to control for the ariance (ANCOVA) yielded a significant main effect of CI,
effects of common method in their relationship with the depen- F(1, 86) ¼ 102.06, p < .01, controlling for gender. Participants
dent variables)?6 We conducted two experiments in an effort to perceived the overall image of Bank XYZ more favorably
shed light on these questions, the first replicating the personal in the high CI condition than in the low CI condition
banking context of Study 1, and the second using a restaurant (MCI_high ¼ 7.6 vs. MCI_low ¼ 4.8). The same approach was
context. adopted to test the IQ manipulation. The ANCOVA results
produced a significant main effect of IQ, F(1, 86) ¼ 199.42,
p < .01. Participants perceived the IQ of Bank XYZ’s staff
Study 2: Bank Customer Experiment more positively in high IQ condition than in the low IQ con-
Method dition (MIQ_high ¼ 8.1 vs. MIQ_low ¼ 4.4). Thus, manipulations
of CI and IQ worked as intended. Realism (Ms > 7.8) and
We conducted a scenario-based experiment that was a 2 (CI: high, comprehensibility (Ms > 8.1) ratings taken during Study 2 were
low)  2 (IQ: high, low) between-subjects design. Participants high and equivalent all across the four conditions (Fs < 1).
read a scenario about opening a bank account and then respond
to questions. The scenario is provided in the Appendix.
Tests of Hypotheses
Pretests. Several pretests were conducted to develop the CI and To test whether SAT and VAL mediated the relationships
IQ manipulations. Every effort was made to ensure that the between CI and LOY (Hypothesis 1a and b), we added SAT
bank scenario would be as realistic and comprehensible as pos- and VAL as covariates in separate ANCOVA models (Baron
sible to student samples (see Appendix). In the final pretest, 24 and Kenny 1986; Wang 2011). In the first model, we included

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


164 Journal of Service Research 16(2)

Table 4. Effect of CI and IQ on Customer Satisfaction, Perceived Value, and Attitudinal Loyalty (Studies 2 and 3).

MANCOVA Univariate Statistics: F Values and Partial Z2

Wilks’s l F Value Partial Z2 SAT Partial Z2 VAL Partial Z2 LOY Partial Z2

Study 2
Gender 0.985 0.42 .015 0.52 .006 0.00 .000 0.13 .001
CI 0.784 7.70 .216 14.39 .143 2.49 .028 12.71 .129
IQ 0.329 57.13 .671 98.62 .534 56.50 .397 144.92 .628
CI*IQ 0.973 0.76 .027 1.96 .022 0.14 .002 0.06 .001
Study 3
Gender 0.963 1.15 .037 0.77 .008 0.67 .007 0.05 .001
CI 0.728 11.05 .272 25.84 .221 12.85 .124 29.96 .250
IQ 0.356 53.62 .644 153.63 .631 81.54 .473 110.78 .549
CI*IQ 0.761 9.37 .239 25.89 .222 7.87 .080 5.08 .053

Note. CI ¼ corporate image; IQ ¼ interaction quality; MANCOVA ¼ multivariate analysis of covariance.


*F values > 3.6 are significant at the .05 level.

SAT as a covariate between CI and LOY. SAT was significant, LOY, F(1, 86) ¼ .06, p ¼.80; partial Z2 ¼ .00. Thus, in contrast
F(1, 88) ¼ 155.50, p < .01; partial Z2 ¼ .63, but the direct to Study 1 results, the experiment failed to provide support for
CI effect became insignificant, F(1, 88) ¼ .292, p ¼ .59; partial Hypothesis 3a–c. Table 5 provides descriptive statistics for
Z2 ¼ .00, thus indicating full mediation of SAT between CI and study variables from Studies 2 and 3.
LOY. In the second model, we included VAL as a covariate
between CI and LOY. VAL, F(1, 88) ¼ 243.00, p < .01; partial
Z2 ¼ .73, and CI, F(1, 88) ¼ 6. 36, p < .05; partial Z2 ¼ .06,
effects were significant but the effect of CI on LOY,
Discussion
F(1, 89) ¼ 7.28, p < .01; partial Z2 ¼ .07, was reduced with The primary objective of Study 2 was to triangulate results from
inclusion of VAL as a covariate, thus indicating partial media- Study 1’s survey-based examination of an ongoing banking ser-
tion of VAL between CI and LOY. Collectively, these results vice relationship with findings from an experiment in which cus-
provide support for Hypothesis 1a and b. tomers focused upon a single banking service encounter. We
To examine the relative impacts of CI and IQ on SAT, VAL, experimentally manipulated CI and IQ and investigated their
and LOY, we conducted a 2 (CI: high vs. low)  2 (IQ: high vs. effects upon perceived value, customer satisfaction, and cus-
low) multivariate analysis of covariance (MANCOVA) on the tomer loyalty. In using a manipulation to capture our indepen-
scales measuring SAT, VAL, and LOY, controlling for gender. dent variables and survey responses for our dependent
Results showed significant multivariate main effects of CI variables, we have also controlled for method variance.
(Wilks’s l ¼ .784; F ¼ 7.70; p < .01; partial Z2 ¼ .21) and Study 2 results were consistent with Study 1 with respect to
IQ (Wilks’s l ¼ .329, F ¼ 57.13, p < .01; partial Z2 ¼ .67). Hypothesis 1a and b; that is, customer satisfaction and perceived
Gender as a covariate was not found significant. Univariate value partially mediated the relationship between CI and cus-
statistics yielded a significant main effect of CI on SAT, tomer loyalty. As expected, the tentative results from our exam-
F(1, 86) ¼ 14.39, p < .01; partial Z2 ¼ .14, and LOY, F(1, ination of Hypothesis 2a–c in Study 1 regarding the relatively
86) ¼ 12.71, p < .01; partial Z2 ¼ .12, but not on VAL, stronger effects of CI upon customer satisfaction, perceived
F(1, 86) ¼ 2.49, p ¼ .11; partial Z2 ¼ .02. In addition, MAN- value, and customer loyalty when evaluations involved an
COVA results showed a significant main effect for IQ upon ongoing service relationship were not upheld in Study 2, in
SAT, F(1, 86) ¼ 98.62, p < .01; partial Z2 ¼ .53, VAL, which individuals were forced to focus on a particular service
F(1, 86) ¼ 56.55, p <.01; partial Z2 ¼ .39, and LOY, encounter. The mixed results from Study 1 indicated that CI had
F(1, 86) ¼ 144.92, p < .01; partial Z2 ¼ .62. In all the cases (see a greater impact on perceived value and customer loyalty than
Table 4), the effects of IQ on SAT, VAL, and LOY were stron- IQ. However, the magnitude of the relationship between these
ger than that of CI based on F and partial Z2 values. Thus, in variables and satisfaction were reversed. Results from Study 2,
contrast to the partial support for Hypothesis 2a–c in the Study however, showed that IQ had a greater impact on customer satis-
1, Study 2 results provide support for this set of hypotheses. faction, perceived value, and customer loyalty than CI. Thus, in
As shown in Table 4, the interaction of CI*IQ did not have a transaction-specific scenario, IQ should be the dominant vari-
significant multivariate impact (Wilks’s l ¼ .973; F ¼ .76; p ¼ able due to proximity and immediacy. These findings are consis-
.51; partial Z2 ¼ .02) upon our dependent variables. In addi- tent with construal-level theory, which suggests that
tion, univariate statistics did not show significant interaction psychological proximity to an event, which was created using
impact of CI*IQ on SAT, F(1, 86) ¼ 1.96, p ¼ .16; partial service scenario, should result in a stronger impact on outcome
Z2 ¼ .02; VAL, F(1, 86) ¼ .14, p ¼.70; partial Z2 ¼ .00; and variables (Trope, Liberman, and Wakslak 2007).

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Jha et al. 165

Table 5. Descriptive Statistics for Customer Satisfaction (SAT), Perceived Value (VAL), and Attitudinal Loyalty (LOY; Studies 2 and 3).*

Study 2 (Bank)

IQ Low IQ High

Dependent Variable CI Low (n ¼ 23) CI High (n ¼ 22) J CI Low (n ¼ 21) CI High (n ¼ 25) J

SAT 3.59 (0.80) 4.49 (0.10) 0.90 5.49 (0.88) 6.02 (0.60) 0.53
VAL 3.61 (0.83) 4.00 (1.18) 0.39 5.16 (0.89) 5.40 (0.76) 0.24
LOY 2.85 (0.84) 3.61 (1.13) 0.76 5.23 (0.98) 5.88 (0.64) 0.65

Study 3 (Restaurant)

IQ Low IQ High

CI Low (n ¼ 23) CI High (n ¼ 24) J CI Low (n ¼ 24) CI High (n ¼ 25) J

SAT 3.07 (0.81) 4.05 (0.65) 1.58 5.82 (0.89) 5.81 (0.61) 0.01
VAL 3.36 (1.04) 4.47 (0.79) 1.11 5.40 (0.81) 5.55 (0.70) 0.15
LOY 2.98 (1.00) 4.28 (0.97) 1.30 5.38 (0.92) 5.97 (0.72) 0.59
Note. CI ¼ corporate image.
Means are based on 7-point scales. A higher score indicates a more favorable response. All mean differences (J) > (.54) are significant at the .05 level.
*Standard deviations in parentheses.

With respect to Hypothesis 3a–c, we did not find a signifi- participants responded to the questions of realism, comprehen-
cant negative interaction effect of CI*IQ on customer satisfac- sibility, manipulation checks for CI (a ¼ .95) and IQ (a ¼ .94),
tion, perceived value, and customer loyalty as found in Study 1. as well as the SAT (a ¼ .95), VAL (a ¼ .92), and LOY
Although we chose a bank service setting that was relevant for (a ¼ .95) measures. Finally, participants completed questions
students and instructed participants to base their judgment only related to demographics and tried to guess the purpose of the
on information available in the scenario, we cannot rule out the study (at which none succeeded).
possibility that students’ motivation and experiences with per-
sonal banking may differ from the general population. For
instance, students may have less experience with the full range Results
of services offered by a bank (e.g., loans). Thus, in Study 3, we
used a restaurant-based scenario to assess whether the pattern
Manipulation and Realism Checks
of relationships we observed in bank setting recurred in what A 2 (CI: high vs. low)  2 (IQ: high vs. low) ANCOVA yielded
should be a more familiar setting. a significant main effect of CI, F(1, 91) ¼ 100.95, p < .01, con-
trolling for gender. Participants perceived the overall image of
Restaurant XYZ more favorably in high CI condition than that
Study 3: Restaurant Customer Experiment of low CI (MCI_high ¼ 7.6 vs. MCI_low ¼ 4.9). We adopted the
Method similar approach using ANCOVA to test IQ manipulation. The
ANCOVA results produced a significant main effect of IQ,
Pretest. A restaurant scenario was constructed that was as sim- F(1, 91) ¼ 183.45, p < .01. Participants perceived the IQ of Res-
ilar as possible to the bank scenario (see Appendix). We taurant XYZ’s staff more positively in high IQ condition than
followed pretest procedures consistent with Study 2. Twenty- low IQ condition (MIQ_high ¼ 8.2 vs. MIQ_low ¼ 4.8). Thus,
six MBA students enrolled in a marketing management class manipulations of CI and IQ worked as intended. Realism (Ms
(53% male; mean age ¼ 24 years) received one of the four > 8.2) and comprehensibility (Ms > 8.3) ratings taken during
surveys that differed only in CI and IQ manipulations. Realism Study 3 were high and equivalent all across the four conditions
(Ms > 7.2) and comprehensibility (Ms > 7.4) ratings were high (Fs < 1).
and equivalent all across the four conditions (Fs < 1). The
results also indicated that the manipulations were successful.
Tests of Hypotheses
Sample and Procedure. Ninety-six MBA students enrolled in a To examine the mediation effects of SAT and VAL, we fol-
marketing management class (55% male; mean age ¼ 24 years) lowed the same approach used in Study 2. In the first model,
participated in the study. They were randomly assigned to one we included SAT as a covariate between CI and LOY. SAT,
of the four conditions of CI (high vs. low) and IQ (high vs. F(1, 93) ¼ 254.98, p < .01; partial Z2 ¼ .73, and CI,
low). As in Study 2, participants were instructed to imagine F(1, 93) ¼ 4.51, p < .05; partial Z2 ¼ .04, effects were
themselves as the protagonist in the restaurant-based service significant but the effect of CI, F(1, 94) ¼ 13.22, p < .01; partial
encounter described by the scenario. After reading the scenario, Z2 ¼ .12, was reduced with the inclusion of SAT as a covariate,

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


166 Journal of Service Research 16(2)

similar effects on perceived value and loyalty. Thus, the data


provide strong support for Hypothesis 3a–c.

Discussion
The objectives of this study were to replicate and extend the
findings of Study 1 and Study 2 (bank-based studies) an alter-
native restaurant-based service setting. A restaurant service
scenario was designed consistent with the Study 2 bank sce-
nario (see Appendix). Thus, Study 3 allowed us to test the gen-
eralizability of our previous findings relating to Hypothesis 1a
and b, Hypothesis 2a–c, and Hypothesis 3a–c to a service set-
ting distinct from the personal banking context.
In line with findings from Study 1 and Study 2 pertaining to
the mediating effects of customer satisfaction and perceived
value, Study 3 results again suggest that these variables play
a critical role in channeling the effects of CI to attitudinal loy-
alty. With respect to Hypothesis 2a–c and its assessment of the
Figure 2. Corporate image (CI) as a moderator of IQ-SAT rela- relative impact of CI and IQ on customer outcomes, Study 3
tionship: Study 3. results are consistent (contrary) with those from Study 2
(Study 1). That is, in the single encounter restaurant setting,
thus indicating partial mediation of SAT between CI and LOY. IQ had a relatively stronger impact on customer satisfaction,
In the second model, we included VAL as a covariate between CI perceived value, and customer loyalty than that of CI. With
and LOY. VAL, F(1, 93) ¼ 223.00, p < .01; partial Z2 ¼ .70, and respect to the interaction effect proposed in Hypothesis 3a–c,
CI effects were, F(1, 93) ¼ 6. 80, p < .05; partial Z2 ¼ .06, signif- Study 3 results, unlike those from Study 2, were consistent with
icant, but effect of CI on LOY, F(1, 94) ¼ 13.22, p < .01; partial Z2 Study 1. That is, the effects of IQ were weaker for customers
¼ .12, was reduced after including VAL as a covariate, thus indi- who possessed a more favorable CI of the provider.
cating partial mediation of VAL between CI and LOY. Hence,
these results provide support for Hypothesis 1a and b.
To test Hypothesis 2, we examined the relative impacts of General Discussion and Implications
CI and IQ on SAT, VAL, and LOY using 2 (CI: high vs. low) The current work provides worthwhile substantive and methodo-
 2 (IQ: high vs. low) MANCOVA, controlling for gender. logical contributions. Substantive results from Study 1 show that
Results showed significant main impact of CI (Wilks’s l ¼ CI for quality exerts significant direct effects upon attitudinal
.728; F ¼ 11.05; p < .01; partial Z2 ¼ .27) and IQ (Wilks’s loyalty as well as indirect effects through perceived value and
l ¼ .356, F ¼ 53.62, p < .01; partial Z2 ¼ .64). Gender as a customer satisfaction, even after accounting for the influences
covariate was not found significant. Univariate statistics of gender, frequency of face-to-face interaction, and customer
yielded significant main impact of CI on SAT, F(1, 91) ¼ relationship tenure. While various researchers have sought to
25.84, p < .01, partial Z2 ¼ .22; VAL, F(1, 91) ¼ 12.85, p link CI to these outcomes, this study contributes to the literature
< .01, partial Z2 ¼ .12; and LOY, F(1, 91) ¼ 29.75, p < by examining the veracity of a more complex nomological rela-
.01, partial Z2 ¼ .24. In addition, it showed significant main tionship among these variables. This robustness of this partial
impact of IQ on SAT, F(1, 91) ¼ 154.50, p < .01, partial Z2 mediation finding was substantiated across methods as well as
¼ .62; VAL, F(1, 91) ¼ 81.54, p <.01, partial Z2 ¼ .47; and across service contexts over the course of the experiments
LOY, F(1, 91) ¼ 110.78, p < .01, partial Z2 ¼ .54. In all the described by Study 2 and Study 3. Moreover, the fact that CI was
cases as shown in Table 4, the effects of IQ on SAT, VAL, and a consistently significant predictor across outcomes, methods,
LOY are stronger than that of CI based on F and partial Z2 and contexts reinforces the notion that variable represents an
values. Hence, collectively these results do not provide sup- important cue in consumer judgments of service offerings.
port for Hypothesis 2a–c. Study findings with respect to the relative effects of CI for
Table 4 indicates that interaction effect of CI*IQ has signif- quality and employee IQ were mixed, suggesting the nature and
icant impact (Wilks’s l ¼ .766; F ¼ 8.94; p < 0.001; partial strength of these relationships may be somewhat more complex
Z2 ¼ .23). In addition, univariate statistics showed significant than hypothesized. Results from the chi-square difference tests
interaction impact of CI*IQ on SAT, F(1, 91) ¼ 25.84, conducted in Study 1, in which customers were asked questions
p < .01; partial Z2 ¼ .22, VAL, F(1, 91) ¼ 7.87, p <.01; partial with respect to an ongoing service relationship, provided no evi-
Z2 ¼ .080, and LOY, F(1, 91) ¼ 5.08, p <.05; partial Z2 ¼ .05. dence in support of our hypothesis that CI would exert stronger
Figure 2 graphically plots the significant interaction, illustrat- effects. In fact, contrary to our expectation, the employee
ing that the effect of IQ on SAT, VAL, and LOY is weaker IQ-customer satisfaction linkage was stronger in comparison
at higher levels of CI. While not shown here for brevity, IQ has to the CI-customer satisfaction path. Results from the heuristic

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Jha et al. 167

relative weight analysis procedure developed by Johnson (2000), In addition, it is important for the managers to understand that
however, do provide some indication that CI is the more critical IQ has a direct role to play in forming service loyalty. Since CI
factor in driving perceived value and attitudinal loyalty. As and IQ independently affect behavioral intentions, this suggests
expected, we find in each experimental study that when respon- that customers are able to differentiate between CI and IQ as
dents were forced to focus on a particular service experience, extrinsic and intrinsic cues.
employee IQ became the more important factor due to its prox- Second, results from all three studies showed that IQ had a
imity and immediacy. Given differences in results between stronger impact on satisfaction than CI. This finding under-
Study 1, in which IQ was measured as a general construct in scores the importance of maintaining high levels of employee
an ongoing service relationship, compared to Studies 2 and 3, IQ through careful selection, training, supporting, and reward-
in which IQ was experimentally manipulated with reference to ing of frontline employees. Given the more affective nature of
a specific encounter, scholars researching in this area would customer satisfaction, study results suggest that IQ connects
be well advised to exercise great precision in their theorizing. with customers on an emotional level. The centrality of satis-
In addition, we find a significant negative effect for the CI-IQ faction in channeling the effects of image and perceived value
product term, which suggests that the effects of IQ upon perceived upon behavioral intentions places an onus on the design of stra-
value, customer satisfaction, and loyalty are diminished as CI tegic human resource management policies that enhance front-
increases. As depicted in Figures 2, the effects of IQ on satisfac- line employee ability, motivation, and opportunity to deliver
tion, value, and loyalty are weaker for a service provider when highly satisfying customer experiences.
customers hold high levels of CI compared to those holding low Third, this study offers providers an understanding of how
image. This finding highlights the relevance and importance of CI CI for quality and employee IQ interrelate in the formation
as a signal of unobservable quality. That is, customers discount of customer attitudes and intentions. Based upon the significant
the effect of IQ if a service provider has a favorable CI. It also negative effect of product term, however, we conclude that
implies that customers’ subsequent evaluation of multiple cues these variables act interactively such that the effects of one are
in a service environment is biased on the evaluation of the stron- stronger at lower levels of the other. It is interesting that this
gest cue. Our results are consistent with the premise that CI may multiplicative effect did not hold when respondents were
work as a reference point that allows the customers to perform the forced to focus on a single banking encounter in the Study 2
relative diagnosticity of IQ of a service provider. banking experiment, but was supported in the Study 3 restau-
In terms of methodology, we gain insight regarding the rant scenario. Again, differences found across methods and
influence of measurement error and common method variance contexts suggest that this nonlinear relationship may be more
on model estimates. Study results suggest a cautious interpreta- nuanced than what might be apparent at first glance.
tion of recent findings by Lance et al. (2010) regarding the
countervailing effects of measurement unreliability and com-
mon method variance in producing unbiased results. The large
Limitations and Future Research
discrepancies observed between composite score and latent Although this study expands knowledge of the role of CI in
variable score models in terms of the comparative strength of driving service loyalty, key limitations must be observed. For
several path estimates and much higher explained variance in instance, the initial study is based on a cross-sectional sample,
each of our endogenous variables suggest that common method highlighting that the directionality of relationships must be
bias remains a concern when using composite score variables. treated with caution. A true test of the causal structure of our
Further, it is unclear whether such issues become exacerbated behavioral-intention model would require longitudinal data.
in tests of more complex nomological relationships such as While our comparison of the proposed model with an alterna-
mediation and moderation. Study results support the use of esti- tive model sheds some light on the temporal location of CI con-
mated latent variables scores that are free from both measure- structs on service loyalty models, the temporal order can best
ment error and common method variance as a viable response be addressed by longitudinal data. Nevertheless, our study pro-
to the skepticism regarding the use of composite scale scores in vides insights compatible to other studies examining models
model testing (Schumacker 2002). In particular, the approach through cross sectional designs. In addition, over the long run,
demonstrated here is particularly attractive as it makes it much the quality of employee interactions feeds into consumer per-
easier to test the effects of latent variable interactions within ceptions of CI. Thus, findings from this study could be
structural equation models (Jöreskog 2000). advanced by a longitudinal design that investigates study rela-
tionships over time.
Second, we acknowledge that CI is likely to have several
Managerial Implications dimensions but the importance of those dimensions will differ
Study findings provide several key managerial insights. First, across the external stakeholders of a firm. Based upon Sha-
study results reinforce earlier work which supports the role of piro’s (1983) arguments, the overall quality dimension is an
CI in shaping customer attitudes and behavioral intentions. important one for customers. As noted by Dowling (1986) and
CI for quality affected customer satisfaction and perceived Walsh and Beatty (2007), however, there may be other dimen-
value, which should lead to greater sales revenue according sions of CI that play a key role in consumers judgments. Future
to earlier findings (Anderson, Fornell, and Lehmann 1994). studies should measure other dimensions of CI as well as

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


168 Journal of Service Research 16(2)

alternative features of the service offering in order to expand Note: Bold letters (in parentheses) are the manipulations of
our knowledge in this context. Third, we recognize the limita- high (low) corporate image and interaction quality.
tion of relying on behavioral intentions instead of actual beha-
vioral data. Extending our research by measuring actual Declaration of Conflicting Interests
behavior would be a useful direction for future research. The author(s) declared no potential conflicts of interest with respect to
This study examines only two service contexts. To broaden the research, authorship, and/or publication of this article.
the database for further generalizations, testing viability of our
model, relative impacts of extrinsic and intrinsic cues, and Funding
moderating role of CI in other service sectors would be fruitful The author(s) received no financial support for the research, author-
avenues for further research to show if the hypothesized lin- ship, and/or publication of this article.
kages change according to service characteristics. Finally, the
unit of analysis in the present study was the individual con- Notes
sumer in reference to a single service firm without considering 1. An anonymous reviewer raised questions about listwise deletion of
interfirm variations of image. Future studies should consider data. A test of the MCAR (missing completely at random) assump-
multilevel research designs that include service organizations tion (Little 1988) showed that the MCAR hypothesis cannot be
as units of analysis combined with individual-level analysis rejected at the .01 level, w21,655 ¼ 1,758.73, p ¼ .038. In addition,
in an effort to enhance our understanding of the role of CI. 20 cases had more than 10% missing data and none of the variables
had more than 5% missing values. The resulting sample size could
still be considered ‘‘large’’ to alleviate any concerns regarding sta-
Appendix
tistical power. These factors collectively justify listwise deletion
Bank Scenario for Study 2 and the resulting parameter estimates are expected to be unbiased
You are interested in opening a saving account at Bank XYZ. (Allison 2001; Hair et al. 2010).
You consider yourself somewhat familiar with this provider, 2. Sobel’s test (Shrout and Bolger 2002) was used to determine the
having visited the bank several times during the past 6 months. significance of the indirect effects. This test is based on large sam-
The Bank XYZ charges reasonable fees that are comparable to ple assumptions and our sample size (777) is considered large by
other banks. Based upon what you have heard from others and any measure. However, we also used a bootstrapping procedure
your personal experience, you hold a very (a less than) favor- (Selig and Preacher 2008) to compute 95% confidence intervals for
able overall image of the bank. the indirect effects. The confidence interval results and Sobel test
Upon arrival, a clerk escorts you to an account manager’s results were in agreement.
office. The bank account manager introduces himself with a 3. We thank an anonymous reviewer for suggesting this alternative
smile (without smiling) and asks how he can help you today. approach.
The account manager expertly answers (struggles to answer) 4. These indirect effects were also evaluated via bootstrapping and
several questions you have about some of the special types of the resulting 95% confidence intervals were consistent with results
accounts the bank offers. based on Sobel’s test.
After your new account has been opened, the account man- 5. We conducted a series of simple slope tests related to CI and IQ
ager provides you with an information packet, thanks you for interaction, using both composite scale scores and estimated latent
coming and wishes you a pleasant afternoon. variable scores (Aiken and West 1991). For instance, using esti-
mated latent variable scores, the simple slope of IQ in predicting cus-
tomer satisfaction at the unfavorable level of CI (low CI) was .72, t
Restaurant Scenario for Study 3 ¼ 15.80, p < .01, compared to .42, t ¼ 8.38, p < .01, when CI was
You are interested in eating lunch at Restaurant XYZ. You con- favorable (high CI), which clearly show the negative moderator role
sider yourself somewhat familiar with this provider, having of CI. These detailed results are available from the authors.
eaten at the restaurant several times during the past 6 months. 6. We thank two anonymous reviewers and the journal editor for trig-
The Restaurant XYZ charges reasonable prices that are compa- gering these questions.
rable to other restaurants. Based upon what you have heard 7. Detailed statistics are available from the authors but for brevity,
from others and your personal experience, you hold a very they are not included here.
(a less than) favorable overall image of the restaurant.
Upon arrival, the hostess escorts you to your table. Your References
waiter introduces himself with a smile (without smiling) Aaker, David A. (1996), Building Strong Brands. New York: The Free
before taking your drink order. After returning with your Press.
drinks, the waiter expertly answers (struggles to answer) sev- Aiken, Leona S. and Stephen G. West (1991), Multiple Regression:
eral questions you have about some of the special items the res- Testing and Interpreting Interactions. Thousand Oaks, CA: Sage.
taurant is offering. Allison, Paul D. (2001), Missing Data. Thousand Oaks, CA: Sage.
After you have completed your meal, the server provides Anderson, Eugene W., Claes Fornell, and Donald R. Lehman (1994),
you with your bill, thanks you for coming and wishes you a ‘‘Customer Satisfaction, Market Share, and Profitability: Findings
pleasant afternoon. from Sweden,’’ Journal of Marketing, 58 (3), 53-66.

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


Jha et al. 169

Andreassen, Tor Wallin (1999), ‘‘What Drives Customer Loyalty with Gürhan-Canli, Zeynep and Rajeev Batra (2004), ‘‘When Corporate
Complaint Resolution?’’ Journal of Service Research, 1 (4), 324-332. Image Affects Product Evaluations: The Moderating Role of Per-
Andreassen, Tor Wallin and Bodil Lindestad (1998), ‘‘The Effect of ceived Risk,’’ Journal of Marketing Research, 41 (2), 197-205.
Corporate Image in the Formation of Customer Loyalty,’’ Journal Hair, Joseph F., Jr., William C. Black, Barry J. Babin, and Rolph E.
of Service Research, 1 (1), 82-92. Anderson (2010), Multivariate Data Analysis (7th ed.). Upper Sad-
Babakus, Emin and Ugur Yavas (2008), ‘‘Does Customer Sex Influ- dle River, NJ: Pearson.
ence the Relationship between Perceived Quality and Share of Hartline, Michael D. and Keith C. Jones (1996), ‘‘Employee Perfor-
Wallet?’’ Journal of Business Research, 61 (9), 974-981. mance Cues in a Hotel Service Environment: Influence on Per-
Baron, Reuben M. and David A. Kenny (1986), ‘‘The Moderator- ceived Service Quality, Value, and Word-of-Mouth Intentions,’’
Mediator Variable Distinction in Social Psychological Research: Journal of Business Research, 35 (3), 207-215.
Conceptual, Strategic, and Statistical Considerations,’’ Journal of Helm, Sabrina, Andreas Eggert, and Ina Garnefeld (2010), ‘‘Modeling
Personality and Social Psychology, 51 (6), 1173-1182. the Impact of Corporate Reputation on Customer Satisfaction and
Biswas, Dipayan (2004), ‘‘Economics of Information in the Web Loyalty Using Partial Least Squares,’’ in Handbook of Partial Least
Economy: Towards a New Theory?’’ Journal of Business Squares, Vincenzo Esposito Vinci, Wynne W. Chin, Jörg Henseler
Research, 57 (7), 724-733. and Huiwen Wang, eds. Heidelberg, Germany: Springer, 515-534.
Bolton, Ruth N. and Katherine N. Lemon (1999), ‘‘A Dynamic Model Hu, Hsin-Hui, Jay Kandampully, and Thanika D. Juwaheer (2009),
of Customers’ Usage of Services: Usage as an Antecedent and ‘‘Relationships and Impacts of Service Quality, Perceived Value,
Consequence of Satisfaction,’’ Journal of Marketing Research, Customer Satisfaction, and Image: An Empirical Study,’’ The Ser-
36 (2), 171-186. vice Industries Journal, 29 (2), 111-125.
Bontis, Nick, Lorne D. Booker, and Alexander Serenko (2007), ‘‘The Iacobucci, Dawn and Amy Ostrom (1993), ‘‘Gender Differences in the
Mediating Effect of Organizational Reputation on Customer Loy- Impact of Core and Relational Aspects of Services on the Evalua-
alty and Service Recommendation in the Banking Industry,’’ Man- tion of Service Encounters,’’ Journal of Consumer Psychology, 2
agement Decision, 45 (9), 1426-1445. (3), 257-286.
Brady, Michael K. and J. Joseph Cronin Jr. (2001), ‘‘Customer Orien- Johnson, Jeff W. (2000), ‘‘A Heuristic Method for Estimating the
tation: Effects on Customer Service Perceptions and Outcome Relative Weight of Predictor Variables in Multiple Regression,’’
Behaviors,’’ Journal of Service Research, 3 (3), 241-251. Multivariate Behavioral Research, 35 (1), 1-19.
Brown, Tom J., Peter A. Dacin, Michael G. Pratt, and David A. Whet- Jöreskog, Karl G. (2000), ‘‘Latent Variable Scores and Their Uses,’’
ten. (2006). ‘‘Identity, Intended Image, Construed Image, and Rep- (accessed April 11, 2011), [available at http://www.ssicentral.
utation: An Interdisciplinary Framework and Suggested com/lisrel/advancedtopics.html].
Terminology,’’ Journal of the Academy of Marketing Science, 34 Jöreskog, Karl G. and Dag Sörbom (1996), LISREL 8: Structural
(2), 99-106. Equation Modeling with the SIMPLIS Command Language. Hills-
Brown, Tom J. and Peter A. Dacin (1997), ‘‘The Company and the dale, NJ: Lawrence Erlbaum.
Product: Corporate Associations and Consumer Product Kirmani, Amna and Akshay R. Rao (2000), ‘‘No Pain, No Gain: A
Responses,’’ Journal of Marketing, 61 (1), 68-84. Critical Review of the Literature on Signaling Unobservable Prod-
Cronin, J. Joseph Jr., Michael K. Brady, and G. Tomas M. Hult (2000), uct Quality,’’ Journal of Marketing, 64 (2), 66-79.
‘‘Assessing the Effects of Quality, Value, and Customer Satisfac- Lai, Fujun, Mitch Griffin, and Barry J. Babin (2009), ‘‘How Quality,
tion on Consumer Behavioral Intentions in Service Environ- Value, Image, and Satisfaction Create Loyalty at a Chinese Tele-
ments,’’ Journal of Retailing, 76 (2), 193-216. com,’’ Journal of Business Research, 62 (10), 980-986.
Szymanski, David M. and David H. Henard (2001), ‘‘Customer Satis- Lance, Charles E., Bryan Dawson, David Birkelbach, and Brian J. Hoff-
faction: A Meta-Analysis of the Empirical Evidence,’’ Journal of man (2010), ‘‘Method Effects, Measurement Error, and Substantive
the Academy of Marketing Science, 29 (1), 16-35. Conclusions,’’ Organizational Research Methods, 13 (3), 435-455.
Dowling, Grahame R. (1988), ‘‘Measuring Corporate Image: A Little, Roderick J. (1988), ‘‘A Test of Missing Completely at Random
Review of Alternative Approaches,’’ Journal of Business for Multivariate Data with Missing Values,’’ Journal of the Amer-
Research, 17 (1), 27-34. ican Statistical Association, 83 (404), 1198-1202.
Dowling, Grahame R. (1986), ‘‘Managing Your Corporate Images,’’ Menon, Geeta, Priya Raghubir, and Nobert Schwarz (1995), ‘‘Beha-
Industrial Marketing Management, 15 (2), 109-115. vioral Frequency Judgments: An Accessibility-Diagnosticity
Feldman, Jack M. and John G. Lynch (1988), ‘‘Self-Generated Valid- Framework,’’ Journal of Consumer Research, 22 (2), 212-228.
ity and Other Effects of Measurement on Belief, Attitude, Inten- Nelson, Leif D. and Joseph P. Simmons (2005), ‘‘Favored Favorites:
tion, and Behavior,’’ Journal of Applied Psychology, 73 (3), Inequalities in Equivalent Outcomes,’’ Advances in Consumer
421-435. Behavior, 32, 127-128.
Fornell, Claes, Michael D. Johnson, Eugene W. Anderson, Jaesung Nelson, Phillip (1970), ‘‘Information and Consumer Behavior,’’ The
Cha, and Barbara E. Bryant (1996), ‘‘The American Customer Journal of Political Economy, 78 (2), 311-329.
Satisfaction Index: Nature, Purpose, and Findings,’’ Journal of Nguyen, Nha and Gaston LeBlanc (1998), ‘‘The Mediating Role of
Marketing, 60 (4), 7-18. Corporate Image on Customers’ Retention Decisions: An Investi-
Gale, Bradley T. (1994), Managing Customer Value. New York: The gation in Financial Services,’’ International Journal of Bank Mar-
Free Press. keting, 16 (2), 52-65.

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014


170 Journal of Service Research 16(2)

Oliver, Richard L. (1993), ‘‘Cognitive, Affective, and Attribute Bases of the Walsh, Gianfranco, Vincent-Wayne Mitchell, Paul R. Jackson, and
Satisfaction Response,’’ Journal of Consumer Research, 20 (3), 418-430. Sharon E. Beatty (2009), ‘‘Examining the Antecedents and Conse-
Oliver, Richard L. (1980), ‘‘A Cognitive Model of the Antecedents quences of Corporate reputation: A Customer Perspective,’’ British
and Consequences of Satisfaction Decisions,’’ Journal of Market- Journal of Management, 20 (2), 187-203.
ing Research, 17 (4), 460-469. Walsh, Gianfranco and Sharon E. Beatty (2007), ‘‘Customer-Based Cor-
Olson, Jerry C. and Jacob Jacoby (1972), ‘‘Cue Utilization in the Qual- porate Reputation of a Service Firm: Scale Development and Valida-
ity Perception Process’’, in Proceedings of the Third Annual Con- tion,’’ Journal of the Academy of Marketing Science, 35 (1), 127-143.
ference of the Association for Consumer Research, Venkatesan, M. Wang, Xuehua (2011), ‘‘The Effect of Unrelated Supporting Service
, ed. Iowa City, IA: Association for Consumer Research, 167-179. Quality on Consumer Delight, Satisfaction, and Repurchase Inten-
Ostrom, Amy L. and Dawn Iacobucci (1998), ‘‘The Effect of Guaran- tions,’’ Journal of Service Research, 14 (2), 149-163.
tees on Consumers’ Evaluation of Services,’’ Journal of Services Woodruff, Robert B. (1997), ‘‘Customer Value: The Next Source for
Marketing, 12 (5), 362-378. Competitive Advantage,’’ Journal of the Academy of Marketing
Perdue, Barbara C. and John O. Summers (1986), ‘‘Checking the Suc- Science, 25 (2), 139-153.
cess of Manipulations in Marketing Experiments,’’ Journal of Zeithaml, Valarie A. (1988), ‘‘Consumer Perceptions of Price, Qual-
Marketing Research, 23 (November), 317-326. ity, and Value: A Means-End Model and Synthesis of Evidence,’’
Podsakoff, Philip M., Scott B. MacKenzie, Jeong-Yeon Lee, and Nathan Journal of Marketing, 52 (3), 2-22.
P. Podsakoff (2003), ‘‘Common Method Biases in Behavioral
Research: A Critical Review of the Literature and Recommended
Remedies,’’ Journal of Applied Psychology, 88 (5), 879-903. Author Biographies
Purohit, Devavrat and Joydeep Srivastava (2001), ‘‘Effect of Manu-
Subhash Jha is a doctoral candidate at IBS-Hyderabad and was a vis-
facturer Reputation, Retailer Reputation, and Product Warranty
iting researcher at Fogelman College of Business & Economics at the
on Consumer Judgments of Product Quality: A Cue Diagnosticity
University of Memphis from 2010 to 2011. His research interest lies in
Framework,’’ Journal of Consumer Psychology, 10 (3), 123-134.
examining the roles of extrinsic and intrinsic cues in product and ser-
Rao, Akshay R. and Kent B. Monroe (1989), ‘‘The Effect of Price,
vice evaluation. He has published in the Asia Pacific Journal of Mar-
Brand Name, and Store Name on Buyers’ Perceptions of Product
keting and Logistics. His research has also been presented in several
Quality: An Integrative Review,’’ Journal of Marketing Research,
international conferences including those organized by the AMA, Soci-
26 (3), 351-357.
ety for Consumer Psychology, Academy of Marketing Science,
Richardson, Paul S., Alan S. Dick, and Arun K. Jain (1994), ‘‘Extrin-
INFORMS, and Society for Marketing Advances.
sic and Intrinsic Cue Effects on Perceptions of Store Brand Qual-
ity,’’ Journal of Marketing, 58 (4), 28-36.
George D. Deitz is an associate professor of marketing at The Univer-
Schumacker, Randall E. (2002), ‘‘Latent Variable Interaction Model-
sity of Memphis. Current research interests are centered on the area of
ing,’’ Structural Equation Modeling, 9 (1), 40-54.
service innovation, particularly within retail, health care, and financial
Selig, James P. and Kristopher J. Preacher (2008, June). Monte Carlo
service systems. His research has been published in a number of jour-
method for assessing mediation: An interactive tool for creating
nals, including Industrial Marketing Management, Journal of Busi-
confidence intervals for indirect effects [Computer software],
ness Venturing, Journal of Public Policy and Marketing, Journal of
(accessed February 10, 2012), [Available at http://quantpsy.org/].
Business Research, and Journal of Business Logistics.
Selnes, Fred (1993), ‘‘An Examination of the Effect of Product Perfor-
mance on Brand Reputation, Satisfaction, and Loyalty,’’ European Emin Babakus is a Palmer Professor of Marketing at The University
Journal of Marketing, 27 (9), 19-35. of Memphis. His research has been published in the Journal of Service
Shapiro, Carl (1983), ‘‘Premiums for High Quality Products as Returns to Research, Journal of Marketing Research, Journal of the Academy of
Reputations,’’ The Quarterly Journal of Economics, 98 (4), 659-679. Marketing Science, International Journal of Research in Marketing,
Shrout, Patrick E. and Niall Bolger (2002), ‘‘Mediation in Experimen- Journal of Retailing, Decision Sciences, and Journal of Advertising
tal and Nonexperimental Studies: New Procedures and Recom- Research, among others. He serves on the editorial review boards of
mendations,’’ Psychological Methods, 7 (4), 422-445. several journals.
Spence, Michael (1973), ‘‘Job Market Signaling,’’ The Quarterly
Journal of Economics, 87 (3), 355-374. Ugur Yavas is a professor of marketing and Advisory Board Faculty
Stern, Barbara, George M. Zinkhan, and Anupam Jaju (2001), ‘‘Mar- Fellow at East Tennessee State University. He has contributed, among
keting Images: Construct Definition, Measurement Issues, and others, to the Journal of Service Research, Journal of Marketing
Theory Development,’’ Marketing Theory, 1 (2), 201-224. Research, Journal of Retailing, Journal of the Academy of Marketing
Tversky, Amos and Daniel Kahneman (1974), ‘‘Judgment Under Uncer- Science, Journal of Business Research, Journal of Business and Indus-
tainty: Heuristics and Biases,’’ Science, 185 (4157), 1124-1131. trial Marketing, International Journal of Research in Marketing, Eur-
Trope, Yaacov, Nira Liberman, and Cheryl Wakslak (2007), ‘‘Construal opean Journal of Marketing, International Journal of Service Industry
Levels and Psychological Distance: Effects on Representation, Pre- Management, Journal of Consumer Marketing, Journal of World
diction, Evaluation, and Behavior,’’ Journal of Consumer Psychol- Business, Management Decision, Service Industries Journal, Journal
ogy, 17 (2), 83-95. of Services Marketing, and Long Range Planning.

Downloaded from jsr.sagepub.com at University of Stellenbosch on June 9, 2014

You might also like