Journal of Retailing 84 (3, 2008) 243255

A Meta-Analysis of Relationships Linking Employee Satisfaction

to Customer Responses
Steven P. Brown , Son K. Lam 1
C.T. Bauer College of Business, University of Houston, Houston, TX 77204-6285, United States

Abstract
The authors report a meta-analysis of relationships linking employee job satisfaction to customer satisfaction and perceived service quality in
studies that correlate employee data with customer data. Overall, both relationships are positive and statistically and substantively significant.
Moderator analyses show that service industry characteristics (relationship versus encounter, personal versus non-personal) and methodological
characteristics (aggregated versus individual level of analysis, concurrent versus predictive survey) moderate these relationships. Estimation of
a path analytic model using the aggregated data shows that customer-perceived service quality completely mediates the relationship between
employee job satisfaction and customer satisfaction. Implications for research and practice are discussed.
2008 New York University. Published by Elsevier Inc. All rights reserved.

Keywords: Employee satisfaction; Customer satisfaction; Service quality; Service-profit chain; Meta-analysis

In retail and other service environments, the interaction Such dyadic data, while difficult and costly to obtain, is prefer-
between customers and front-line employees is likely to able to estimating relationships on non-dyadic data (e.g., by
importantly affect customer perceptions of the shopping and asking employees to rate customer satisfaction or customers
consumption experience. High quality interactions with service to rate employee satisfaction). Thus, we limit our review to
providers often result in customer satisfaction despite problems studies that have used dyadic data to estimate linkages between
with other aspects of service delivery, whereas dissatisfying employee job satisfaction and customer responses.2
experiences with service providers have the potential to ruin oth- The accumulated research includes findings of a positive
erwise pleasant service encounters. Service providers often are (e.g., Payne and Webber 2006; Homburg and Stock 2004),
the organization in customers perception, and their attitudes, negative (e.g., Silvestro and Cross 2000), and non-significant
expressed verbally and non-verbally during customer interac- relationship (e.g., Adcock 1999; Brown and Mitchell 1993),
tions, may leave a lasting impression that affects satisfaction, leaving in doubt the issues of its strength and substan-
repeat buying, and financial performance (Magi 2003; Schneider tive significance. Moreover, little is known about contextual
and Bowen 1995). The financial significance of these interac- and methodological influences that moderate the relationship.
tions may not be appreciated by front-line employees, or even Understanding the strength and context dependency of this rela-
by management. tionship may provide insights into the degree to which employee
Commensurate with the importance of these issues, a robust satisfaction influences customer experiences and how effect
literature has developed regarding interactions between front- sizes vary across service contexts and research designs. We
line employees and customers. In particular, researchers have assess the strength and context dependency of relationships
focused considerable effort on investigating the relationship
between employee job satisfaction and customer satisfaction,
2 Other meta-analyses (e.g., Brown et al., 2002; Carrillat et al., 2007) address
using data linking employee responses with customer responses.
the relationship between service quality and customer satisfaction across studies
that use only customer data. Because our focus is on linkages between employee
job satisfaction and these outcomes, we limit our review to studies that use dyadic
Corresponding author. Tel.: +1 713 743 4560. data. The weighted average effect size between service quality and customer sat-
E-mail addresses: steve.brown@mail.uh.edu (S.P. Brown), slam5@uh.edu isfaction that we obtain based on dyadic studies is very similar to those obtained
(S.K. Lam). by Brown et al. and Carrillat, Jaramillo, and Mulki based on their reviews of
1 Tel.: +1 713 743 4577. customer data only studies.

0022-4359/$ see front matter 2008 New York University. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.jretai.2008.06.001
244 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255

linking employee job satisfaction to customer-perceived service assess whether coded study characteristics (e.g., study contexts,
quality and satisfaction in a meta-analysis of studies that corre- methodological choices made by researchers) can account for
late employee satisfaction data with customer response data. between-study variance in effect sizes, and (3) to test whether
Such insights may provide guidance for managers contem- customer-perceived service quality mediates the relationship
plating actions that affect employee and customer satisfaction. between employee satisfaction and customer satisfaction based
For example, Walmarts recent decision to employ computer- on the aggregated data.
ized scheduling to staff its stores dynamically according to the
number of shoppers in the store is predicated on the assump- Previous research on the relationships linking ES to CS
tion that customer satisfaction will increase as a result of having and SQ
the optimal number of employees on hand, while, at the same
time, the move may decrease employee satisfaction by mak- Each of the three focal constructs has a long history of
ing work schedules and pay less predictable (Maher 2007). A research. Employee job satisfaction (hereafter ES) has been
strong employee satisfactioncustomer satisfaction relationship defined as a pleasurable or positive emotional state resulting
would suggest that such moves may not generate the desired from the appraisal of ones job or job experiences (Locke 1976,
results. Having a precise understanding of the strength of the p. 1300). It has been operationalized as (1) a global reaction to
employee satisfactioncustomer satisfaction relationship will overall job circumstances, and (2) perceptions of and affective
furnish guidance for managers facing such trade-offs. reactions to different job factors (e.g., work, supervisor, pay, co-
The analysis will also indicate how employeecustomer workers, etc.; Brown and Peterson 1993). Customer-perceived
linkages differ across industry contexts (e.g., personal versus service quality (hereafter SQ) has been defined conceptually as
non-personal services, encounter versus relationship business, customers perceptions of the difference between prior expecta-
and B2B versus B2C contexts), highlighting conditions in which tions and delivered service on different aspects of the service
customer responses are especially sensitive to employee satis- experience (e.g., Gronroos 1982; Parasuraman et al. 1988).
faction. Such findings should provide guidance for managers in Customer satisfaction (hereafter CS) represents a fulfillment
formulating human resource policy and managing the front-line response that involves an assessment that a product or service
workforce in a manner commensurate with customer rela- provides a pleasurable level of consumption-related fulfill-
tionship objectives in different business settings (Kumar and ment (Oliver 1997, p. 13).
Shah 2004). The meta-analysis is also likely to provide useful Both SQ and CS have been defined as perceived perfor-
insights for researchers concerning the effects of methodolog- mance relative to expectations, and early in the development
ical decisions by indicating research design factors (e.g., types of SQ research, debate ensued over whether they represent dis-
of measures, levels of analysis, sampling frames, and timing of tinct constructs (e.g., Bitner and Hubbert 1994; Dabholkar 1993;
data collection) that are systematically related to effect sizes. Iacobucci et al. 1994). A dominant view has emerged that they
Our study addresses the following objectives: (1) to ascer- represent distinct but closely related constructs, with SQ rep-
tain whether extant studies converge on a population parameter, resenting a cognitive representation of service delivered and
(2) if they do not converge on a population parameter, to CS representing the resulting affective response (Brady et al.

Table 1
Comparing and contrasting three conceptual models
Framework Level of analysis Temporal horizon Causal mechanism Conceptual scope Representative studies

Emotional contagion Individual Transaction Transfer through mimicry Person-to-person Homburg and Stock
of emotional expression interactions (2004) and Wangenheim
and subjective experience et al. (2007)
Serviceprofit chain Empirical research Relationship Employee satisfaction Quality and value Jones et al. (2003),
calibrated at individual increases productivity and delivered in one-to-one Silvestro and Cross
level, but framework is value delivered, which relationship (2000), Brown and Chin
conceptually applicable to enhances customer (2004), Hartline and
business-unit level as well perceived service quality Ferrell (1996), Bernhardt
and customer satisfaction et al. (2000), Koys (2001),
Netemeyer et al. (2005),
Payne and Webber
(2006), Schlesinger and
Zornitsky (1991), Tornow
and Wiley (1991), and
Keiningham et al. (2006)
Service climate Empirical research Relationship Organizational support Relationship quality and Pedrick (1998), Schneider
calibrated at SBU level, and facilitation of service satisfaction in et al. (1980), Yoon and
but framework is effort fosters quality organizational context Suh (2003), Brown and
conceptually applicable to service and customer Mitchell (1993), Liao and
individual level as well satisfaction Chuang (2004), and Yoon
et al. (2001)
 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255 245

2005; Oliver 1997). As such, the linkages of ES to both SQ ceptions of SQ and, ultimately, CS. Morris and Feldman (1996)
and CS are likely to be strong and based on similar underlying build on the work of Hochschild (1983) to suggest that ES facil-
processes. itates emotional labor, which contributes to perceptions of SQ
The view that ES positively affects SQ and CS has been and CS. Similarly, Ulrich (1992) and Ulrich et al. (1991) main-
grounded primarily on three conceptual models: emotional con- tain that satisfied employees develop commitment to a service
tagion (Hatfield et al. 1993), the serviceprofit chain (Heskett ethic, which leads to an understanding of customer expecta-
et al. 1997), and service climate (Schneider 1990). Table 1 tions and efforts and ability to satisfy them. These perspectives
compares and contrasts these conceptual models in terms of suggest that SQ mediates the ESCS relationship, in a man-
the levels of analysis they address, their temporal horizons, ner consistent with the serviceprofit chain and service climate
causal mechanisms, and conceptual scope. It is important to perspectives.
note that their explanations of EScustomer response relation-
ships are not mutually exclusive. In fact, they can perhaps best be Moderators of the ESCS, ESSQ relationships
viewed as nested explanations, ranging from a micro-focus on
affect transfer in the context of one-time person-to-person inter- Meta-analysis makes possible a systematic examination of
actions (emotional contagion) to delivery of superior service the ESSQ and ESCS relationships across a diverse set of stud-
and value by business units in long-term customer relationships ies and research contexts. In this section, we present rationales
(serviceprofit chain, service climate). for possible moderators of these relationships. We investigate
The emotional contagion framework, which has frequently four groups of moderators: service characteristics, research con-
been invoked to explain correspondence between employee texts, study methods, and publication status.
affect and customer response (e.g., Pugh 2001; Tsai and Huang
2001, 2002), focuses on simple affect transfer in the context of
employeecustomer interactions in limited service encounters. Service characteristics
It holds that affect transfer alone (i.e., exclusive of quality or
value considerations) is sufficient to account for EScustomer The nature of interactions between employees and customers
response relationships. differs across service contexts. Lovelock (1983) and Gutek et al.
In contrast, the serviceprofit chain framework holds that (1999) offer particularly useful classifications of services, which
satisfied employees are more productive and provide better SQ we use to develop hypotheses and a coding scheme for moderator
and value than less satisfied employees, leading to higher CS. analyses.
While this view is not incompatible with emotional contagion, its
scope is greater, as it suggests that cumulative perceptions of ser- Personal versus non-personal services
vice quality and value are foundations of EScustomer response According to Lovelocks (1983) classification, whether a ser-
relationships. Because the effects posited by the serviceprofit vice is performed on a person (e.g., medical services, fitness)
chain are not limited to affect transfer within specific ser- or on possessions (e.g., television repair) constitutes a major
vice encounters, this model is also better suited to relationship distinction. Personal services involve up close interactions
marketing contexts than emotional contagion is. Although the between service providers and customers. Such services are
serviceprofit chain is potentially relevant to analysis of busi- not amenable to a service factory approach, in which cus-
ness unit-level processes, empirical research grounded on this tomer interactions are limited to transactional elements and
model has typically been conducted at the individual level of service production takes place outside of the presence of the
analysis (e.g., Schlesinger and Zornitsky 1991; Ulrich et al. customer (Chase and Garvin 1989). The service provider and
1991). customer engage in close interpersonal contact throughout the
Finally, the service climate framework (e.g., Schneider et al. encounter, which provides opportunity for affective cues to be
1992, 1998) posits that employee perceptions of management communicated. In contrast, in services performed on possessions
support and facilitation of service performance leads to favor- personal contact between service providers and customers is
able customer reactions. This view is similar to the serviceprofit more limited in scope and duration. Under these circumstances,
chain, but it is more explicit regarding how specific employee service providers may be able to conceal true emotions that are
perceptions of managerial policies and the organizational envi- counter-normative for the situation through emotional labor
ronment contribute to favorable customer responses. Like the (Hochschild 1983).
serviceprofit chain model, the effects of service climate can Similarly, in personal services, customers can closely observe
potentially be assessed at either the individual or business-unit the processes and outcomes of service delivery, whereas in ser-
levels of analysis, although empirical research has most often vices performed on possessions delivery processes often take
been focused on the business-unit level (e.g., Schneider et al. place in a service factory that is invisible to the customer, while
1980, 1998). customer interactions occur in a front office or lobby, separate
Other authors have elaborated behavioral processes underly- from the space where the core service is performed. In many
ing the chain of effects linking ES to SQ and CS. For example, cases, it may also be more difficult for customers to evaluate the
Wilson and Frimpong (2004) argue, based on the work of Isen outcomes of non-personal as opposed to personal services (e.g.,
(Isen and Levin 1972; Isen et al. 1978), that employees who are of preventive car maintenance versus a haircut). This reasoning
in a good mood engage in helping behaviors that result in per- leads to the following hypotheses:
246 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255

H1. The (a) ESCS and (b) ESSQ relationships are stronger to-consumer (e.g., Keiningham et al. 2006; Koys 2001) contexts.
in personal services than in services performed on possessions. Because retailers must be concerned with distribution channel
as well as customer service issues, both contexts have direct
Encounter versus relationship businesses implications for retailing, and the relative strength of these rela-
Gutek et al. (1999) categorize service business models into tionships at multiple points in the value chain should be of
relationship and encounter businesses. The defining charac- interest (e.g., Ping 2003).
teristic of relationship businesses is that customers seek out Although it has been argued that organizational buying tends
the same provider for each service encounter (e.g., hairstylist, to be more objective and impersonal than consumer buying (e.g.,
doctor, or dentist). In contrast, in encounter businesses cus- Robinson et al. 1967), recent research also indicates that personal
tomers interact with a different service provider each time (e.g., relational dynamics play an important role in B2B supply rela-
airlines, fast food service). Relationship businesses give employ- tionships (e.g., Heide and Wathne, 2006; Wathne et al. 2001).
ees time and opportunity to become personally acquainted Recent research also indicates that most distribution channel
with customers, understand their needs and expectations, and relationships are characterized by long-term relational focus
develop personal as well as commercial bonds. In contrast, (e.g., Palmatier et al. 2006). This research suggests that link-
transactions in encounter businesses tend to be brief and ages between employees and customers in B2B settings may
impersonal. In fact, Gutek and Welsh (2000) observe that not be any less personal or subject to processes that lead ES to
customers often react negatively to attempts to turn service be related to SQ and CS than they would be in B2C contexts.
encounters into pseudo-relationships in encounter businesses Positive linkages of ES with SQ and CS have been found in both
by being more personal with customers than the situation contexts (e.g., Brown and Chin 2004; Yoon et al. 2001). Because
warrants. affect transfer mechanisms and service quality outcomes are
Because relationship businesses involve sustained personal important in both B2B and B2C contexts, and neither is system-
contact over time, it is likely that SQ and CS will be more closely atically more or less personal than the other, we assess ESSQ
related to ES in them than in encounter businesses, in which con- and ESCS relationships in B2B and B2C contexts without
tact is brief and superficial. Sustained contact provides greater directional hypotheses.
opportunity for diffusion of affect. It is also more difficult for
service providers to disguise counter-normative affective states Method factors
through emotional labor in relationship (compared to encounter)
businesses, which is likely to lead to greater correspondence Studying employeecustomer relationships involves mak-
between their experienced affective states and those of cus- ing methodological decisions and trade-offs that can affect
tomers. Customers are also more likely to empathize with the results and conclusions. We investigate the moderating effects
feelings of service providers in relationship businesses than in of three method factors, including: levels of analysis (individ-
encounter businesses, also leading to greater correspondence ual employee versus organizational unit), timing sequence of
between their affective states. surveys (employees first or concurrent) and measurement scales
Social exchange theory would suggest that long-term rela- (global ES versus ES by job facet).
tionships with valued customers are likely to motivate satisfied
employees to perform well in serving them (Blau 1964). Rela- Levels of analysis
tionship customers are likely to perceive service performance As previously noted, different streams of research investi-
by satisfied service providers as being responsive and empa- gate relationships linking ES to CS and SQ at different levels
thetic, in addition to being technically competent. In contrast, in of analysis. For example, research testing specific linkages in
encounter businesses, at best, customers are likely to perceive the serviceprofit chain (e.g., Schlesinger and Zornitsky 1991)
service performance as technically competent and cheerful. As a typically views the relationships from the individual level of
result, service providers have greater opportunity to distinguish analysis, whereas research focusing on service climate issues
themselves through quality service in relationship businesses; (e.g., Schneider et al. 1980, 1998) typically assesses them at the
thus, customer perceptions are likely to have a higher ceiling organizational level (e.g., at the store level).
and greater variation than in encounter businesses. Richer and Decisions regarding levels of analysis can influence effect
more variable perceptions of service quality are likely to result sizes. For example, several studies (e.g., Ostroff and Harrison
in a stronger ESSQ relationship in relationship (as compared 1992, 1999; Gully et al. 1995; Judge et al. 2001) have indi-
to encounter) businesses. cated a stronger job satisfactionwork performance relationship
at the organizational level than at the individual level. Aggre-
H2. The (a) ESCS and (b) ESSQ relationships are stronger gating data to the unit or organizational level allows researchers
in relationship businesses than in encounter businesses. to avoid individual bias in linking employee ratings to orga-
nizational outcomes (Schneider et al. 2005). Also, employees
Business-to-business (B2B) versus business-to-consumer working together in a collaborative and supportive environment
(B2C) may create what Glick and Roberts (1984) and Ostroff and
Harrison (1993) refer to as interdependence effects, meaning
Research linking ES to CS and SQ has been conducted in both that satisfied employees realize synergies by feeding off of one
business-to-business (e.g., Brown and Chin 2004) and business- anothers positive affect in customer-focused efforts. Sociolo-
 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255 247

gists (e.g., Hannan 1971) have observed that aggregation of data study designs.4 On this basis, we venture the tentative hypothesis
to larger units of analysis tends to produce stronger effect sizes. that:
On this basis, we hypothesize that:
H5. The (a) ESCS and (b) ESSQ relationships are stronger in
H3. The (a) ESCS and (b) ESSQ relationships are stronger studies using predictive designs than in studies using concurrent
in studies conducted at the organizational level than in studies designs.
conducted at the individual level.

Sampling frames Measurement scales

Studies investigating the relationships linking ES to CS and ES has been measured as global satisfaction (e.g., Homburg
SQ differ in the types of sampling frames they use. Some stud- and Stock 2004) and satisfaction with various job facets (e.g.,
ies (e.g., Schneider et al. 1980, 1998) draw data from a single work, pay, supervisor, co-workers, etc.; e.g., Hartline and Ferrell
organization, whereas others (e.g., Hartline and Ferrell 1996) 1996). Of these, global measurement of satisfaction comes clos-
use data from multiple organizations. Following the recommen- est to capturing the conceptualizations of simple positive affect
dation of Ostroff and Harrison (1999), we partition the primary underlying the ESCS and ESSQ relationships. These con-
studies on this basis and test for moderating effects of sampling ceptualizations, grounded on affect transfer and the positive
frame. relationship between positive affect and helping behavior, are
Studies using data from multiple organizations match more consistent with global measurement of simple job-related
employee observations to customer organizations either by affect than with extensive measurement of multi-dimensional
asking employees to nominate customers or vice versa (e.g., job perceptions. That is, if affect is transmitted from employees
Homburg and Stock 2004; Payne and Webber 2006). As a result, to customers, it is more likely to be in the form of simple feelings
effect sizes in these studies tend to be estimated on relatively few related to ones work than as reactions to detailed perceptions of
customer observations per sampling unit. In contrast, studies that a myriad of job facets. Moreover, some job facets are likely to be
use data from a single organization often use large-scale cus- more weakly correlated with SQ and CS than others, leading to
tomer data bases for sampling frames and estimate effect sizes attenuation of the overall relationship when facets are averaged
based on larger samples per unit of analysis (e.g., Bernhardt into a global assessment of ES. These considerations lead to the
et al. 2000; Koys 2001).3 Greater aggregation of customer hypothesis that:
observations reflects more comprehensive experience with the
H6. The (a) ESCS and (b) ESSQ relationships are stronger
service provider, and larger samples per unit of analysis result
in studies that measure ES globally than in those that measure
in less sampling error and more reliable measurement, which
ES by facet.
should result in stronger correlations. Therefore, we hypothesize
that:
Publication status
H4. The (a) ESCS and (b) ESSQ relationships are stronger It has often been argued that published studies are likely
in studies that use data from a single organization than in those to report larger effect sizes than unpublished studies (e.g.,
that use data from multiple organizations. Rosenthal 1984; Rust et al. 1990). The rationale underly-
ing this argument is that the review process tends to be
Timing of data collection biased against non-significant effects, which thus have a greater
Studies in the literature we review vary in the temporal likelihood of going unpublished. An important implication
sequence of data collection. Some, which we label predictive for researchers conducting meta-analyses is that every effort
(e.g., Brown and Chin 2004), obtained employee data first and should be made to collect unpublished as well as published
then matched employee responses with subsequently collected studies to reduce publication bias and increase the represen-
customer data. Others, which we refer to as concurrent (e.g., tativeness of the effect sizes being aggregated. Our sample
Bernhardt et al. 2000), obtained data from employees and cus- includes a number of unpublished studies, which we compare
tomers at the same time. against those published in the literature under the hypothesis
Conceptual models in the primary studies uniformly repre- that:
sent ES as causally antecedent to customer responses. Although
H7. The (a) ESCS and (b) ESSQ relationships are stronger
concurrent data collection may also capture covariation between
for published studies than for unpublished studies.
employee and customer responses, to the extent that the concep-
tual models are accurate, customer sensitivity to changes in ES
may have a tendency to produce larger effect sizes in predictive 4 We note, however, that Judge et al. (2001) and Harrison et al. (2006) found
largest effect sizes in concurrent study designs. A difference between the litera-
ture we review here and those reviewed in those studies is that all of the studies
3 For example, in our sample the number of customers per sampling unit in in our review collect data from multiple sources. In other words, none of the
studies using the sampling frame of a single organization ranged from 30 to relationships we study are subject to common-method bias, which is frequently
54,000, whereas the number of customers per sampling unit in studies using a present in the concurrent study designs reviewed by Judge et al. (2001) and
sampling frame consisting of multiple organizations ranged from 1 to 16. Harrison et al. (2006) and may have inflated effect sizes in those studies.
248 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255

Method computed sample-size-weighted mean effect sizes and variances

over all studies reporting a particular relationship.
Identication of relevant studies As per the recommendation of Hedges and Olkin (1985), we
conducted homogeneity analyses using Q-tests to assess whether
We searched EBSCO, ABI/Inform, PsycINFO, and Disser- the accumulated study effects converge on a population value.
tation Abstracts database using key words such as employee The Q-statistic consists of the sample-size-weighted between-
satisfactioncustomer satisfaction link, employee satisfaction study variance, which is distributed chi square and constitutes
and service quality, serviceprofit chain, etc. We also reviewed a powerful test of effect size homogeneity. Non-significant Q-
references from studies we identified as relevant to the review tests indicate that the accumulated study effects converge on a
to find additional studies that might be relevant. In addition, we population mean, whereas significant values of Q indicate signif-
communicated with researchers known to be working in the field icant between-study variance and a need to conduct moderator
and posted notes on relevant bulletin boards to identify unpub- analyses to attempt to account for it (Hedges and Olkin 1985).
lished papers and data sets that might be included. In addition, we Studies were coded for contextual and methodological char-
conducted issue-by-issue searches of a number of journals from acteristics independently by two coders. The coders agreed on 92
1995 to the present, including Journal of Marketing, Journal percent of the codings, indicating high reliability. Disagreements
of Retailing, Journal of Service Research, Journal of Applied were resolved by discussion.
Psychology, Personnel Psychology, Academy of Management Hedges and Olkin (1985) recommend checking for outliers
Journal, International Journal of Service Industry Management, prior to conducting moderator analyses to determine whether
Service Industries Journal, Journal of Vocational Behavior, and removing a small number of outlying effect sizes results in a
Journal of Occupational Psychology. homogeneous set of effects. Following this recommendation,
we plotted effect sizes and within-study confidence intervals to
Criteria for inclusion identify highly discrepant observations (e.g., greater than three
standard deviations from the mean). Using these procedures, we
Because our conceptual focus is on linkages between ES and identified four ESCS effect sizes as outliers. Table 3 reports
customer reactions, we included only studies that report dyadic both the analysis of all effects and of the reduced set of effects
data linking employee-rated job satisfaction to customer satis- with outliers deleted.7 We did not identify any outliers in the
faction and service quality or both. Brown et al. (2002) reviewed ESSQ analysis.
a larger number of studies of the SQCS relationship that use As per the recommendation of Hedges and Olkin (1985), we
non-dyadic data. We include an estimate of the SQCS relation- conducted moderator analyses on the reduced set of study effects
ship in this study only to assess the role of SQ in mediating the (i.e., after removing the outliers). We conducted the analyses
ESCS relationship. Our estimate of this relationship, based on using the Comprehensive Meta-Analysis program (Borenstein
studies with dyadic data, is consistent with that of Brown et al. and Rothestien 2000), which incorporates the Hedges and Olkin
(2002). (1985) algorithms.
After attempting to obtain usable study effects by contacting
authors, we excluded studies for which we were unable to obtain
zero-order correlations (or statistics that could be converted to r). Results
We also excluded studies that appeared to report results using the
same data as studies that were already included (e.g., Homburg Results of the overall analyses of the ESCS and ESSQ rela-
and Stock 2004, 2005; Schneider et al. 1998, 1992; Schneider tionships are presented in Table 2. The overall analyses indicate
and Bowen 1995). As a result of these procedures, the overall attenuation-corrected means of r = 0.23 for the ESCS relation-
analysis included 36 correlations reported in 28 studies, with a ship (r = 0.25 after removal of four outliers) and r = 0.29 for
cumulative sample size of 6,680.5 the ESSQ relationship. Neither of the 95 percent confidence
intervals includes zero, indicating that both differ significantly
Procedures from zero. The Q-statistic was statistically significant for both
the ESCS and ESSQ relationships, indicating that moderator
We conducted the analyses using procedures recommended analyses are warranted to attempt to account for between-
by Hedges and Olkin (1985). We first corrected the zero-order study variance. Therefore, we tested the hypothesized moderator
correlations for attenuation from measurement error, imputing effects of service characteristics, research contexts, method char-
weighted-average reliability estimates based on all studies that acteristics, and publication status.
do report reliabilities for those that do not report them.6 We then

latent constructs rather than the relationship between imperfect measures of con-
5 The list of included studies and their coded study characteristics can be structs. Therefore, following standard practice, we correct the study effects for
obtained from the first author. measurement unreliability.
6 Although Spector and Levine (1987) indicate that artifact correction can 7 We checked the studies producing these outliers for characteristics that might

result in some bias in estimates of between-study variance, correction for attenua- account for their seemingly anomalous observations in the overall distribu-
tion for measurement unreliability is standard meta-analytic practice. Correction tion of effects, but were unable to identify any that might constitute plausible
for measurement unreliability results in estimates of the relationship between explanations.
 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255 249

Moderator analyses

Failsafe, N
Service characteristics

971
1,188

361
Results of the service characteristics moderator analyses are
presented in Table 3. We report the results of the hypothesis tests
Q-Test (d.f.)***
in the following sections.

180.5 (21)
136.8 (17)

25.2 (8)
Personal versus non-personal. As predicted in H1a and H1b,
both the ESCS and ESSQ relationships were stronger in per-
sonal services than in services on possessions. Therefore, these
0.21 < r < 0.26
0.22 < r < 0.27

0.26 < r < 0.33

hypotheses are supported.
(corrected)
confidence
95 percent

interval

Relationship versus encounter businesses. Hypotheses H2a and

H2b, concerning the effects of relationship versus encounter
businesses, were not supported. The results showed that the
S.D. (corrected)

strength of the ESCS relationship did not differ signifi-

cantly across these types of businesses, whereas, surprisingly,
the ESSQ relationship was stronger in studies conducted in
0.14
0.13

0.09

encounter businesses. We discuss this unexpected but interesting

finding in the discussion section.
0.17 < r < 0.22
0.18 < r < 0.23

0.20 < r < 0.27

Research contexts. The results indicated no significant differ-

confidence

(observed)
95 percent

interval

ence in either the ESCS or ESSQ relationship between B2C

and B2B contexts. Although the relationships are directionally
stronger in B2C contexts, the differences were not statistically
corrected, r a

significant.
Weighted

0.23
0.25

0.29

Method factors
Note: ES: employee job satisfaction; CS: customer satisfaction; SQ: service quality. Fixed effects model.

Levels of analysis. Hypotheses H3a and H3b predicted stronger

relationships in studies conducted at the business-unit than at the
observed, r
Weighted

individual level of analysis. Consistent with these predictions,

0.19
0.21

0.23

the correlations between ESCS and ESSQ are both stronger

in studies conducted at the business-unit than at the individual
corrected, r a

level of analysis. These results are presented in Table 4.

Unweighted

0.16
0.26

0.30

Sampling frame. Contrary to the predictions of hypotheses H4a

and H4b, effect sizes did not differ significantly between studies
Unweighted

using data from single firms and those using data from multiple
observed, r

firms. Thus, these hypotheses are not supported.

0.14
0.22

0.25

Data collection sequences. There was no significant difference

Cumulative N

in either relationship between predictive and concurrent research

designs. Therefore, H5a and H5b are not supported.
Strength and variability of pair-wise relationship

5,606
5,391

2,771

ES measurement. Hypotheses H6a and H6b predicted stronger

a Corrected for measurement error.

effect sizes in studies that use global ES measurement scales

study effects, k

than in those that measure ES by facet. Consistent with these

Number of

hypotheses, the results indicate stronger ESCS and ESSQ rela-

*** Significant at p < 0.00.

tionships in studies that use global satisfaction scales compared

22
18

to those that measure satisfaction by job facet.

ESCS (outliers

Publication status. As reported in Table 4, consistent with H7a

Relationship

ESCS (all)

removed)

and H7b, published studies report stronger effect sizes for both
Table 2

ESSQ

the ESCS and ESSQ relationships than unpublished studies

do.
250 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255

Table 3
Moderator analyses service characteristics
Relationship and moderator Number of Cumulative N Observed, r Corrected, r a 95 percent 95 percent QB (d.f. = 1)
study effects, confidence interval confidence interval
k (observed) (corrected)

Personal versus non-personal

ES-CS 30.1 ***
Personal 6 793 0.34 0.41 0.28 < r < 0.40 0.35 < r < 0.47
Non-personal 11 4,570 0.18 0.22 0.15 < r < 0.21 0.19 < r < 0.24
ESSQ 5.7**
Personal 3 310 0.33 0.40 0.23 < r < 0.43 0.31 < r < 0.49
Non-personal 6 2,461 0.22 0.28 0.18 < r < 0.26 0.24 < r < 0.31

Relationship versus encounter

ESCS 0.6
Relationship 9 953 0.19 0.22 0.13 < r < 0.25 0.16 < r < 0.28
Encounter 8 4,410 0.21 0.25 0.18 < r < 0.24 0.22 < r < 0.28
ESSQ 4.8**
Relationship 2 299 0.13 0.17 0.02 < r < 0.24 0.06 < r < 0.28
Encounter 6 2,431 0.24 0.30 0.20 < r < 0.28 0.26 < r < 0.34

B2B versus B2C

ESCS 1.0
B2B 4 574 0.18 0.21 0.10 < r < 0.26 0.13 < r < 0.29
B2C 13 4,789 0.21 0.25 0.18 < r < 0.23 0.22 < r < 0.28
ESSQ 1.4
B2B 4 413 0.19 0.24 0.10 < r < 0.28 0.14 < r < 0.33
B2C 5 2,358 0.24 0.30 0.20 < r < 0.28 0.26 < r < 0.34

Note: ES: employee job satisfaction; CS: customer satisfaction; SQ: service quality. Fixed effects model. *Significant at p < 0.10, ** at p < 0.05, *** at p < 0.01.
Significance of Q-statistics indicates rejection of the homogeneity hypothesis. QB = between-group homogeneity statistic.
a Corrected for measurement error.

square root of the sample-size-weighted reliability and random

measurement errors to 1 . The resulting model is just iden-
tified, estimating five structural parameters with five degrees of
freedom, resulting in an artifactually perfect fit to the data.8 As
per the recommendation of Viswesvaran and Ones (1995), we
use the harmonic mean sample size across the three relation-
ships in the analysis. The results indicate a significant effect of
ES on SQ and a very strong effect of SQ on CS, but a non-
significant direct effect of ES on CS. Results of the Sobel test
(12.5, p < 0.001) support the conclusion that the ESCS relation-
ship is completely mediated by SQ. This finding is consistent
with the serviceprofit chain framework.

Fig. 1. Path analysis. Discussion

Does SQ mediate the ESCS relationship? The meta-analysis indicates that the central links in the
We addressed the question of whether SQ mediates the serviceprofit chain, consisting of the relationships of ES with
ESCS relationship by estimating a path analytic model using SQ and CS, are supported by significant positive effects. How-
AMOS with the correlation matrix developed from the meta- ever, the analysis also shows significant variability in the strength
analyses of the ESSQ, ESCS, and SQCS relationships as of these relationships over service contexts and method factors.
input data. Estimating path analytic models using meta-analytic We found a portion of this variance predictable from coded
data has become a common way of testing theoretical propo- study characteristics, although a substantial portion remains
sitions using aggregated data (e.g., Brown and Peterson, 1993; unexplained.
Geyskens et al., 1999; Viswesvaran and Ones 1995). The struc-
tural model we estimated is represented in Fig. 1. It posits ES 8 Despite the limitation of not being able to rigorously assess the fit of the
as antecedent to both SQ and CS, and SQ as antecedent to CS. structural model to the data, we proceed with the analysis in the interest of
We used the observed correlations in the analysis and fixed the addressing the conceptual issue of the mediating effect of SQ on the ESCS
paths linking observed measures to the latent constructs to the relationship in the best manner possible given the data limitations.
 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255 251

Table 4
Moderator analyses method characteristics
Relationship and moderator Number of Cumulative N Observed r Corrected r a 95 percent 95 percent QB (d.f.)
study effects, k confidence interval confidence interval
(observed) (corrected)

Level of analysis
ESCS 18.9 (1)***
Organizational 8 1,359 0.28 0.34 0.23 < r < 0.33 0.29 < r < 0.39
Individual 9 4,004 0.18 0.21 0.15 < r < 0.21 0.18 < r < 0.24
ESSQ 5.1 (1)**
Organizational 2 64 0.45 0.53 0.22 < r < 0.63 0.32 < r < 0.69
Individual 7 2,707 0.23 0.28 0.19 < r < 0.26 0.25 < r < 0.32

Sampling frames
ESCS 1.9 (1)
Within a firm 12 4,854 0.21 0.25 0.18 < r < 0.23 0.22 < r < 0.28
Multiple firms 5 509 0.17 0.19 0.08 < r < 0.25 0.10 < r < 0.27
ESSQ 0.4 (1)
Within a firm 6 2,276 0.24 0.30 0.20 < r < 0.28 0.26 < r < 0.33
Multiple firms 3 495 0.21 0.27 0.12 < r < 0.29 0.18 < r < 0.35

Timing of data collection

ESCS (without Koys 2001) 2.1 (1)
Predictive 5 697 0.18 0.21 0.11 < r < 0.25 0.13 < r < 0.28
Concurrent 7 3,966 .22 0.26 0.19 < r < 0.25 0.23 < r < 0.29
ESSQ 0.01(1)
Predictive 4 794 0.24 0.29 0.17 < r < 0.30 0.23 < r < 0.36
Concurrent 5 1,977 0.23 0.29 0.19 < r < 0.27 0.25 < r < 0.33

ES measurement scales
ESCS 6.0 (1)***
Overall 10 4,587 0.22 0.26 0.19 < r < 0.24 0.23 < r < 0.29
Facet composite 7 776 0.14 0.17 0.07 < r < 0.21 0.10 < r < 0.24
ESSQ 2.7 (1)*
Overall 2 1,682 0.25 0.30 0.21 < r < 0.30 0.27 < r < 0.36
Facet composite 7 1,089 0.20 0.24 0.15 < r < 0.26 0.20 < r < 0.31

Published versus unpublished

ESCS 4.4 (1)**
Published 13 4,830 0.21 0.26 0.19 < r < 0.24 0.23 < r < 0.28
Unpublished 4 733 0.15 0.18 0.08 < r < 0.22 0.11 < r < 0.25
ESSQ 3.9 (1)**
Published 6 2,454 0.24 0.30 0.21 < r < 0.28 0.27 < r < 0.34
Unpublished 3 317 0.15 0.19 0.04 < r < 0.26 0.08 < r < 0.29

Note: ES: employee job satisfaction; CS: customer satisfaction; SQ: service quality. Fixed effects model. *Significant at p < 0.10, ** at p < 0.05, *** at p < 0.01.
Significance of Q-statistics indicates rejection of the homogeneity hypothesis. QB = between-group homogeneity statistic.
a Corrected for measurement error.

Although the focal relationships are positive and statistically about substantial increases in customer retention, loyalty, and
significant, the effect sizes are relatively modest. Overall, ES equity, as well as in profitability and competitive advantage
accounts for between four and six percent of total variance in CS (Reicheld, 1996; Zeithaml 2000). Calibration of these effect
and between six and ten percent in SQ (i.e., based on confidence sizes across the accumulated research increases scholars
intervals around the attenuation-corrected correlations). and managers ability to assess the leverage that ES has on
In interpreting these effects, however, we believe it is wise downstream links in the serviceprofit chain (e.g., Rucci et al.
for both researchers and managers to regard them as being 1998).
substantively as well as statistically significant (Rosenthal et al. The meta-analysis has also indicated several contextual fac-
2000). In a sense, it is remarkable that ES explains that much tors that moderate the ESCS and ESSQ relationships. For
variance in aggregated ratings of CS and SQ, which are also example, both relationships are substantially stronger (a) in ser-
likely to be affected also by a number of other factors (e.g., vices performed on people (as compared to services performed
Oliver 1997; Szymanski and Henard 2001). A modest lift in on peoples possessions), (b) in studies that are conducted at the
CS and SQ based on dealings with employees is likely to bring organizational (as compared to the individual) level of analysis,
252 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255

and (c) when ES is measured globally (as compared to when it Path analysis using the aggregated data indicated that SQ
is measured by facet). completely mediates the relationship between ES and CS. This
Stronger relationships in personal services (compared to ser- finding is consistent with the serviceprofit chain; that is, ES
vices performed on possessions) reflect greater opportunity for leads directly to providing better service, which in turn enhances
affect transfer and observation of service delivery processes CS. Although this finding does not rule out the emotional con-
and outcomes in contexts where service providers have close tagion explanation as a factor in the ESCS relationship, it does
contact with customers throughout the service delivery pro- strongly suggest that there is more to the story than simple affect
cess. Although meaningful effects of ES on customer reactions transferand that enhancement of SQ and perceptions of value
also occur in services performed on possessions, they are much contribute to CS beyond sharing of simple affect. It should also
greater in personal services. These results strongly suggest the be noted that SQ and CS are very strongly related, empirically
need for special care in fostering job satisfaction among service as well as conceptually (e.g., Brown et al., 2002; Iacobucci et
providers who perform such personal services, as their job- al. 1995).
related affect leaves a strong imprint on customer perceptions
and satisfaction. Managerial implications
The finding of stronger relationships in studies conducted at
the organizational (compared to the individual) level of analysis The results suggest that retail managers would be ill advised
appears to reflect greater measurement reliability and variability to take actions that are likely to detract from the job satisfaction
of responses when many responses are aggregated to a super- of customer contact personnel. Doing so appears likely to detract
ordinate level of analysis (Hannan 1971; Ostroff and Harrison from customer satisfaction and perceptions of service quality.
1993). It is also consistent with other meta-analyses (e.g., Gully For example, Walmarts initiative to staff its stores dynamically
et al. 1995; Judge et al. 2001) that indicate stronger correlations according to the amount of customer traffic, leading to employee
at the organizational than at the individual level of analysis. The uncertainty regarding their work schedules and pay, may have
finding suggests that the positive effects of overall workforce the unintended effects of decreasing employee job satisfaction
morale on customer responses exceed those of individual-level and, to a lesser extent, customer responses. Our study calibrates
job satisfaction, and that analysis at the individual level tends the magnitude of the employeecustomer linkages so that man-
to underestimate the cumulative effect of employee job satisfac- agers can more easily estimate the effects that are likely to occur
tion. This is consistent with the notion of interdependence effects in their businesses.
(Glick and Roberts 1984; Ostroff and Harrison 1993), whereby In considering which stakeholders interests to favor in situa-
employees feed off of one anothers positive affect, creating syn- tions in which trade-offs are unavoidable, managers should take
ergy effects. This suggests that a strong climate for service (e.g., careful note of the effects of employee satisfaction on customer
Schneider et al. 1998) is a powerful tool for creating favorable responses. For example, managers may have a tendency to
customer reactions. conclude that customers are the focus of all marketing activities
The finding that employeecustomer linkages are stronger and, as such, their interests should be the primary focus of
when job satisfaction is measured globally rather than by facet organizational activities. While from a marketing perspective
is consistent with conceptualizations of these relationships as such customer orientation is desirable, it should also be noted
being grounded on simple affect transfer mechanisms and (as has been done, for example, by former Southwest Airlines
research showing that positive mood is associated with help- CEO Herb Kelleher) that customers are not always right in
ing behavior (Isen et al. 1978). Job satisfaction represents an fact, they are often wrong and when they are, their actions
overall attitude, consisting of both cognitive appraisals of spe- can be demoralizing to front-line service providers (Menon and
cific job elements and feelings about ones overall situation at Dube 2004; Schneider and Bowen 1995). When push comes to
work. Our findings suggest that simple job-related affect (as shove on the front line, managers should realize that satisfied,
captured in simple global measures) is more strongly related to motivated, and committed front-line employees constitute a
customer-perceived service quality and satisfaction than cogni- powerful engine for delivery of service quality and customer
tive appraisals of job facets are. satisfaction, and that such a workforce can be the firms most
Unexpectedly, we found the ESSQ linkage stronger in valuable asset in developing a loyal customer base (Pfeffer
encounter than in relationship businesses. This finding ran 1998). Abusive customers are not likely worth the strain they
counter to our reasoning that relationship businesses offer place on front-line customers and service delivery systems.
greater opportunity for affect transfer and development of Many kinds of managerial actions can potentially enhance
empathy and understanding than encounter businesses do. In employee satisfaction and workforce morale (e.g., Huselid 1995;
retrospect, it appears possible that, consistent with the reason- Pfeffer 1998). Managerial policies and actions that clarify role
ing presented above, this relationship is based on the simplest of expectations and provide material and psychological support
affect transfer mechanisms and that ES has greatest effect on SQ for employees are of primary importance. Fostering an orga-
when there is little or no prior acquaintance and no other con- nizational environment in which service quality and customer
siderations beside the affective tone of a simple exchange taking satisfaction are primary goals, and in which employee efforts to
place between the service provider and customer. In other words, provide solutions for customers are supported, recognized, and
ES is likely to matter most when few, if any, other forces are at rewarded is important (Schneider and Bowen 1995). Empow-
play. ering employees to do what is necessary to satisfy customers
 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255 253

without fear of sanction enhances both employee satisfaction within organizational units, which, as Ostroff and Harrison
and customer responses. Avoiding the temptation to sacrifice (1993) has shown analytically, leads to stronger correlations
employee interests (e.g., laying off loyal employees) in the inter- between the same constructs at the organizational than at the
est of meeting short-term financial objectives is a difficult but individual level. Studies at the different levels of analysis essen-
very important mandate for managers, as employee uncertainty tially address different objectivesthose at the organizational
about employment status leads to paranoia and distraction from level focusing on organizational drivers (e.g., service climate;
job focus (Huselid 1995; Pfeffer 1998). Schneider et al. 1998), and those at the individual level analyzing
The finding that relationships did not differ between B2B and the interpersonal dynamics underlying the relationships (e.g.,
B2C contexts suggests that employee satisfaction and customer Pugh 2001). Both types have important contributions to make,
linkages are important throughout the value chain, affecting sup- but when the objective is assessing the organizational impact of
plier as well as end-user relationships. This implies that efforts employee satisfaction on customer response and organizational
to enhance employee job satisfaction should be pervasive and outcomes, metrics calibrated at the individual level of analysis
not focus narrowly on front-line service providers. All internal are likely to understate the true effects.
and external relationships are likely to benefit from improved We have noted practical situations in which organizations
employee satisfaction. may take actions in the interest of improving customer satis-
faction that may at the same time have the effect of decreasing
Theoretical and research implications employee job satisfaction (Maher 2007). This type of situation
presents an interesting and potentially researchable dilemma
Our finding that employeecustomer linkages are stronger concerning trade-offs between employee and customer satisfac-
in studies that measure job satisfaction globally than in studies tion. It will be interesting for researchers to consider the total
that measure it by facet suggests that simple job-related affect, effects of actions intended to enhance customer satisfaction that
rather than elaborate, multi-faceted job perceptions, are primar- have the (perhaps unintended) effect of decreasing employee
ily responsible for employeecustomer linkages. Although this satisfaction, or actions intended to enhance employee satis-
interpretation is consistent with existing theoretical accounts of faction that have the unintended consequence of decreasing
these relationships, the finding also suggests that any employee customer satisfaction. While such situations must certainly exist
experience of positive affect can potentially generate positive in practice, they have not been systematically research to date.
customer responses and that, in a sense, job satisfaction serves Conclusions
as a proxy for positive affect experienced on the job. For exam-
ple, Pugh (2001) and Tsai and Huang (2001, 2002) show that The meta-analysis indicates statistically significant and sub-
service providers situational (or state, as opposed to trait) affect stantively important relationships linking employee satisfaction
produces effects on customer responses that are similar to the to customer satisfaction and perceived service quality. We found
effects of employee satisfaction. In future research, finer-grained these relationships to be systematically moderated by (a) per-
analyses of employee affect may be able to identify the mini- sonal versus non-personal service contexts, (b) organizational
mal conditions necessary for employeecustomer affect transfer versus individual level of analysis, and (c) global versus by-
to occur and the extent to which the cognitive (versus affec- facet measurement of employee satisfaction. We also found
tive) components of job attitudes and perceptions account for that perceived service quality completely mediates the job
employeecustomer linkages. satisfactioncustomer satisfaction relationship in a manner con-
Our results indicate that SQ completely mediates the ESCS sistent with the serviceprofit chain conceptualization, which
relationship. This finding suggests that perceptual and cogni- suggests that emotional contagion alone does not provide a suf-
tive mechanisms play an important role in customer responses, ficient conceptual account of the relationship. The results show
but it does not rule out the possibility that emotional conta- employee satisfaction to be a consistently important driver of
gion also plays a role in affect transfer between employees and customer responses and indicate service and research contexts in
customers. Research under tightly controlled conditions could which these effects are heightened. Even so, carefully designed
potentially provide further insights into the relative contribu- research is needed to provide a more precise conceptual account
tions of emotional contagion and service quality to customer of these relationships, as boundary conditions remain to be
satisfaction judgments. Such research might help to further established for the affect transfer and service quality mediation
discriminate between these different conceptual accounts of explanations of the relationships.
employeecustomer linkages and indicate boundary conditions
on each theoretical explanation. References
Our finding of stronger effect sizes at the organizational *Denotes studies that were used in the meta-analysis.
(compared to the individual) level of analysis suggests that *Adcock, Brian P. (1999), The Role of Organizational Citizenship Behavior in
the organizational impact of employeecustomer linkages may the Linkage of Employee and Customer Satisfaction, Ph.D. Dissertation,
be underestimated when effects are estimated at the individual University of South Florida.
level. One reason for this is that greater aggregation of responses *Bernhardt, Kenneth L., Donthu Naveen and Pamela A. Kennett (2000), A
Longitudinal Analysis of Satisfaction and Profitability, Journal of Business
results in more reliable measurement of employeecustomer Research, 47 1617.
dynamics at the organizational level. Another is that greater Bitner, Mary J. and Amy R. Hubbert (1994), Encounter Satisfaction Versus
aggregation typically results in greater variability of responses Overall Satisfaction Versus Quality, in Service Quality: New Directions in
254 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255

Theory and Practice, Rust Roland. T. and OLiver Richard. L., eds. Thousand Hochschild, Arlie R. (1983), The Managed Heart, Berkeley, CA: University
Oaks, CA: Sage. of California Press.
Blau, Peter M. (1964), Exchange and Power in Social Life, New York, NY: *Homburg, Christian and Ruth M. Stock (2004), The Link between Salespeo-
Wiley. ples Job Satisfaction and Customer Satisfaction in a Business-to-Business
Borenstein, Michael and Hannah Rothestien (2000), Comprehensive Meta Context: A Dyadic Analysis, Journal of the Academy of Marketing Science,
Analysis: A Computer Program for Research Synthesis, Englewood, NJ: 32 (2), 14458.
Biostat. and (2005), Exploring the Con-
Brady, Michael K., Gary A. Knight, J. Joseph Cronin, G. Thomas Hult and Bruce ditions Under Which Salesperson Work Satisfaction Can lead to Customer
D. Keillor (2005), Removing the Contextual Lens: A Multinational, Multi- Satisfaction, Psychology & Marketing, 22 (5), 393420.
Setting Comparison of Service Evaluation Models, Journal of Retailing, 81 Huselid, Mark A. (1995), The Impact of Human Resource Management Prac-
(3), 21530. tices on Turnover, Productivity, and Corporate Financial Performance,
*Brown, Steven P. and Wynee W. Chin (2004), Satisfying and Retaining Cus- Academy of Management Journal, 38 64572.
tomers through Independent Service Representatives, Decision Sciences, Iacobucci, Dawn, Kent A. Grayson and Amy L. Ostrom (1994), The Calculus
35 (3), 52750. of Service Quality and Customer Satisfaction: Theoretical and Empirical
*Brown, Karen A. and Terence R. Mitchell (1993), Organizational Obstacles: Differentiation and Integration, Greenwich, CT: JAI Press, 167.
Links with Financial Performance, Customer Satisfaction, and Job Satisfac- , Amy Ostrom and Kent Grayson (1995), Distinguishing
tion in a Service Environment, Human Relations, 46 (6), 72557. Service Quality and Customer Satisfaction: The Voice of the Consumer,
*Brown, Steven P. and Robert A. Peterson (1993), Antecedents and Conse- Journal of Consumer Psychology, 4 (3), 277303.
quences of Salesperson Job Satisfaction: Meta-Analysis and Assessment of Isen, Alice M. and Paula F. Levin (1972), Effect of Feeling Good on Helping:
Causal Effects, Journal of Marketing Research, 30 (1), 6377. Cookies and Kindness, Journal of Personality and Social Psychology, 21
*Brown, Steven P., Amy L. Ostrom, Stephen W. Brown, (2002), A Meta- 3448.
Analysis of Service Quality Relationships, Working paper presented at Isen, Alice M., Thomas E. Shalker, Margaret Clark and Lynn Karp (1978),
Frontiers in Service, Maastricht, Netherlands. Affect, Accessibility of Matierial in Memory, and Behavior: A Cognitive
Carrillat, Francois A., Jaramillo Fernando and Jay P. Mulki (2007), The Valid- Loop?, Journal of Personality and Social Psychology, 36 (1), 113.
ity of the SERVQUAL and SERVPERF Scales, International Journal of *Jones, Eli, Busch Paul and Peter Dacin (2003), Firm Market Orientation
Service Industry Management, 18 (5), 47290. and Salesperson Customer Orientation: Interpersonal and Intrapersonal
Chase, Richard B. and David Garvin (1989), The Service Factory, Harvard Influences on Customer Service and Retention in Business-to-Business
Business Review, 67 (4), 619. BuyerSeller Relationships, Journal of Business Research, 56 (4), 32340.
Dabholkar, Parthiba (1993), Customer Satisfaction and Service Quality: Two Judge, Timothy A., Carl J. Thoresen, Joyce E. Bono and Gregory K. Patton
Constructs Or One? in AMA Educators Proceedings: Enhancing Knowledge (2001), The Job SatisfactionJob Performance Relationship: A Qualitative
Development In Marketing, vol.4, David W. Cravens and Dickson Peter. R., and Quantitative Review, Psychological Bulletin, 127 (3), 376407.
eds. Chicago: American Marketing Association, 108. *Keiningham, Timothy L., Aksoy Lerzan, Robert M. Daly, Kathy Perrier and
Geyskens, Inge, Jan-Benedict E.M. Steenkamp and Nirmalya Kumar (1999), A Aantoine Solom (2006), Reexamining the Link between Employee Sat-
Meta-Analysis of Satisfaction in Marketing Channel Relationships, Journal isfaction and Store Performance in a Retail Environment, International
of Marketing Research, 36 (2), 22338. Journal of Service Industry Management, 18 (1), 517.
Glick, Wilma H. and Karlene H. Roberts (1984), Hypothesized Interdepen- *Koys, Daniel J. (2001), The Effects of Employee Satisfaction, Organizational
dence, Assumed Independence, Academy of Management Review, 9 (4), Citizenship Behavior, and Turnover on Organizational Effectiveness: A Unit-
72235. Level, Longitudinal Study, Personnel Psychology, 54 (1), 10114.
Gronroos, Christian (1982), Strategic Management and Marketing in the Ser- Kumar, V. and Denish Shah (2004), Building and Sustaining Protable Cus-
vice Sector, Helsingfors, Finland: Swedish School of Economics and tomer Loyalty for the 21st Century, Journal of Retailing, 31730.
Business Administration. *Liao, Hui and Aichia Chuang (2004), A Multilevel Investigation of Fac-
Gully, Stanley M., J.D. Dennie and David J. Whitney (1995), A Meta-Analysis tors Influencing Employee Service Performance and Customer Outcomes,
of Cohesion and Performance: Effects of Level of Analysis and Task Inter- Academy of Management Journal, 47 (1), 4158.
dependence, Small Group Research, 26 (4), 497520. Locke, Edwin A. (1976), The Nature and Causes of Job Satisfaction, in Hand-
Gutek, Barbara A. and Theresa Welsh (2000), The Brave New Service Strat- book of Industrial and Organizational Psychology, Dunnatte Marvin ed.
egy: Aligning Customer Relationships, Market Strategies, and Business Chicago: Rand McNally, 1297350.
Structures, New York, NY: AMACOM. Lovelock, Christopher H. (1983), Classifying Services to Gain Strategic Mar-
Gutek, Barbara A., Anita D. Bhappu, Mathew A. Liao-Troth and Bennett Cherry keting Insights, Journal of Marketing, 47 (3), 920.
(1999), Distinguishing between Service Relationships and Encounters, Magi, Anne W. (2003), Share of Wallet in Retailing: The Effects of Cus-
Journal of Applied Psychology, 84 (2), 21833. tomer Satisfaction, Loyalty Cards, and Shopper Characteristics, Journal
Hannan, Michael T. (1971), Aggregation and Disaggregation in Sociology, of Retailing, 79 (2), 10720.
Lexington, MA: Lexington Books. Maher, Kris (2007), Wal-Mart Seeks New Flexibility in Worker Shifts, Wall
Harrison, David A., Daniel A. Newman and Philip L. Roth (2006), How Street Journal, 249 (2), 1.
Important Are Job Attitudes? Meta-Analytic Comparisons of Integrative Menon, Kalyani and Laurette Dube (2004), Service Provider Responses to
Behavioral Outcomes and Time Sequences, Academy of Management Jour- Anxious and Angry Customers: Different Challenges, Different Payoffs,
nal, 49 (2), 3052. Journal of Retailing, 80 (3), 22938.
*Hartline, Michael D. and O.C. Ferrell (1996), The Management of Customer- Morris, J. Andrew and Daniel C. Feldman (1996), The Dimensions,
Contact Service Employees: An Empirical Investigation, Journal of Antecedents, and Consequences of Emotional Labor, Academy of Man-
Marketing, 60 (4), 5270. agement Review, 21 (4), 9861010.
Hatfield, Elaine, John T. Cacioppo and Richard L. Rapson (1993), Emotional *Netemeyer, Richard G., James G. Maxham III and Chris Pullig (2005), Con-
Contagion, Current Directions in Psychological Science, 2 (3), 969. flicts in the WorkFamily Interface: Links to Job Stress, Customer Service
Hedges, Larry and Olkin S Ingram (1985), Statistical Methods for Meta- Employee Performance, and Customer Purchase Intent, Journal of Market-
Analysis, New York, NY: Academic Press. ing, 69 (2), 13043.
Heide, Jan and Kenneth H. Wathne (2006), Friends, Businesspeople, and Rela- Oliver, Richard L. (1997), Satisfaction: A Behavioral Perspective on the Con-
tionship Roles: A Conceptual Framework and a Research Agenda, Journal sumer, New York, NY: McGraw-Hill.
of Marketing, 70 (3), 90103. Ostroff, Cheri and David A. Harrison (1992), The Relationship between Sat-
Heskett, James L., W. Earl Sasser Jr. and Leonard A. Schlesinger (1997), The isfaction, Attitudes, and Performance: An Organizational Level Analysis,
Service Profit Chain, New York, NY: The Free Press. Journal of Applied Psychology, 77 (6), 96374.
 S.P. Brown, S.K. Lam / Journal of Retailing 84 (3, 2008) 243255 255

and (1993), Comparing Corre- Szymanski, David M. and David H. Henard (2001), Customer Satisfaction:
lations Based on Individual-Level and Aggregated Data, Journal of Applied A Meta-Analysis of the Empirical Evidence, Journal of the Academy of
Psychology, 78 (4), 56982. Marketing Science, 29 (1), 1635.
and (1999), Meta-Analysis, *Tornow, Walter W. and Jack W. Wiley (1991), Service Quality and Manage-
Level of Analysis, and Best Estimates of Population Correlations: Cautions ment Practices: A Look at Employee Attitudes, Customer Satisfaction, and
for Interpreting Meta-Analytic Results in Organizational Behavior, Journal Bottom-Line Consequences, Human Resource Planning, 14 (2), 1051.
of Applied Psychology, 84 (2), 2607. Tsai, Wei Chi and Yin-Mei Huang (2001), Determinants and Consequences of
Palmatier, Robert W., Rajiv P. Dant, Dhruv Grewal and Kenneth R. Evans (2006), Employee Displayed Positive Emotions, Journal of Management, 27 (4),
Factors Influencing the Effectiveness of Relationship Marketing: A Meta- 497512.
Analysis, Journal of Marketing, 70 (4), 13653. and (2002), Mechanisms Link-
Parasuraman, A., Valarie A. Zeithaml and Leonard L. Berry (1988), ing Employee Affective Delivery and Customer Behavioral Intentions,
SERVQUAL: A Multiple-Item Scale for Measuring Customer Perceptions Journal of Applied Psychology, 87 (5), 10018.
of Service Quality, Journal of Retailing, 64 (1), 1240. Ulrich, Dave (1992), Strategic and Human Resource Planning: Linking Cus-
*Payne, Stephanie C. and Sheila S. Webber (2006), Effects of Service Provider tomers and Employees, Strategic and Human Resource Planning, 15 (2),
Attitudes and Employment Status on Citizenship Behaviors and Customers 4762.
Attitudes and Loyalty Behavior, Journal of Applied Psychology, 91 (2), Ulrich, Dave, Halbrook Richard, Dave Meder, Mark Stuchlik and Steve Thorpe
36578. (1991), Employee and Customer Attachment: Synergies for Competitive
*Pedrick, Dennis Lee (1998), The Influence of Organizational Climate on Advantage, Human Resource Planning, 14 (2), 89103.
Employee, Customer, and Firm Performance, Ph.D. Dissertation, Univer- Viswesvaran, Chockalingam and Deniz S. Ones (1995), Theory Testing: Com-
sity of Memphis. bining Psychometric Meta-Analysis and Structural Equations Modeling,
Pfeffer, Jeffrey (1998), The Human Equation: Building Profits by Putting Peo- Personnel Psychology, 48 (4), 8658.
ple First, Boston, MA: Harvard Business School Publishing. *Wangenheim, Florian V., Heiner Eanschitzky and Maren Wunderlich (2007),
Ping, Robert A. Jr. (2003), Antecedents of Satisfaction in a Marketing Channel, Does the EmployeeCustomer Satisfaction Link Hold for All Employee
Journal of Retailing, 79 (4), 24958. Groups?, Journal of Business Research, 60 (7), 6907.
Pugh, S. Douglas (2001), Service With a Smile: Emotional Contagion in the Wathne, Kenneth H., Harold Biong and Jan B. Heide (2001), Choice of Sup-
Service Encounter, Academy of Management Journal, 44 (5), 101827. plier in Embedded Markets: Relationship and Marketing Program Effects,
Reicheld, Frederick F. (1996), The Loyalty Effect: The Hidden Force Behind Journal of Marketing, 65 (2), 5466.
Growth, Profits, and Lasting Value, Boston, MA: Harvard Business School Wilson, Alan and Jacob Frimpong (2004), A Reconceptualisation of the
Publishing. SatisfactionService Performance Thesis, Journal of Services Marketing,
Robinson, Patrick J., Charles W. Caris and Yoram Wind (1967), Industrial 18 (6), 4718.
Buying and Creative Marketing, Boston, MA: Allyn and Bacon. *Yoon, Mahn H. and Jacob Suh (2003), Organizational Citizenship Behav-
Rosenthal, Robert (1984), Meta-Analytic Procedures for Social Research, iors and Service Quality as External Effectiveness of Contact Employees,
Beverly Hills, CA: Sage. Journal of Business Research, 56 (8), 597611.
Rosenthal, Robert, Ralph L. Rosnow and Donald B. Rubin (2000), Contrasts *Yoon, Mahn H., Sharon E. Beatty and Jaebeom Suh (2001), The Effect of Work
and Effect Sizes in Behavioral Research, New York, NY: Cambridge Uni- Climate on Critical Employee and Customer Outcomes: An Employee-Level
versity Press. Analysis, International Journal of Service Industry Management, 12 (5),
Rucci, Anthony J., Steven P. Kirn and Richard T. Quinn (1998), The 50021.
EmployeeCustomer Profit Chain at Sears, Harvard Business Review, 76 Zeithaml, Valarie A. (2000), Service Quality, Profitability, and the Economic
(1), 8297. Worth of Customers: What We Know and What We Need to Learn, Journal
Rust, Rolant T., Donald R. Lehmann and John U. Farley (1990), Estimating of the Academy of Marketing Science, 28 (1), 6785.
Publication Bias in Meta-Analysis, Journal of Marketing Research, 27 (2),
Further reading
2207.
*Schlesinger, Leonard A. and Jeffrey Zornitsky (1991), Job Satisfaction, *Banker, Rajiv D., Constantine Konstans, Raj Mashruwala (2000), A Contex-
Service Capability, and Customer Satisfaction: An Examination of Link- tual Study of Links between Employee Satisfaction, Employee Turnover,
ages and Management Implications, Human Resource Planning, 14 (2), Customer Satisfaction and Financial Performance, (accessed October 30,
1419. 2006), [available at www.agsm.ucr.edu/faculty/pages/banker/PAPER8.pdf].
Schneider, Benjamin (1990), Organizational Climate and Culture, San Fran- *Burleson, Andrea (1997), The Service Profit Chain: An Investigation of the
cisco, CA: Jossey-Bass. Linkages Between Employees, Customers, and Organizational Profitability,
Schneider, Benjamin and David E. Bowen (1995), Winning the Services Game, Ph.D. Dissertation, DePaul University.
Boston, MA: Harvard Business School Press. *Galer, Michael M. (2000), The Effect of Market Orientation on Customer Sat-
*Schneider, Benjamin, John J. Parkington and Virginia M. Buxton (1980), isfaction and Employee Satisfaction on Service Quality, Ph.D. Dissertation,
Employee and Customer Perceptions of Service in Banks, Administrative Nova Southeastern University.
Science Quarterly, 25 (2), 25267. *Hagan, Christine M. (1999), The Relationship Between Employee Job Satis-
Schneider, Benjamin, Jill K. Wheeler and Jonathan F. Cox (1992), A Passion faction and Key Customer Outcomes: A Study Among Temporary Workers,
for Service: Using Content Analysis to Explicate Service Climate Theme, Ph.D. Dissertation, Florida Atlantic University.
Journal of Applied Psychology, 77 (5), 70516. *Kitzman, Brian C. (1999), The Effects of the Type of Service Interaction
, Susan S. White and Michelle C. Paul (1998), Linking on the Relationship Between Employee and Customer Attitudes, Ph.D.
Service Climate and Customer Perceptions of Service Quality: Test of a Dissertation, DePaul University.
Causal Model, Journal of Applied Psychology, 83 (2), 15063. *Moshavi, Dan and James R. Terborg (2002), The Job Satisfaction and Per-
Schneider, Benjamin, Mark G. Ehrhat, David M. Mayer, Jessica L. Saltz and formance of Contingent and Regular Customer Service Representatives:
Kathryn Niles-Jolly (2005), Understanding OrganizationCustomer Links A Human Capital Perspective, International Journal of Service Industry
in Service Settings, Academy of Management Journal, 48 (6), 101732. Management, 13 (4), 33347.
*Silvestro, Rhian and Stuar Cross (2000), Applying the Service Profit Chain in *Sepulveda-Martinez, Cesar J. (2001), Relationship of Organizational Citizen-
a Retail Environment, International Journal of Service Industry Manage- ship Behaviors and Customer Orientation to Service Quality and Customer
ment, 11 (3), 24468. Satisfaction, Ph.D. Dissertation, Tulane University.
Spector, Paul E. and Edward L. Levine (1987), Meta-Analysis for Integrating *Subramony, Mahesh, Terry A. Beehr and C. Merle Johnson (2004), Employee
Study Outcomes: A MonteCarlo Study of Its Susceptibility to Type I and and Customer Perceptions of Service Quality in an Indian Firm, Applied
Type II Errors, Journal of Applied Psychology, 72 (1), 39. Psychology, An International Review, 53 (2), 31127.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.