SERVPERF Versus SERVQUAL: Reconciling Performance-Based and Perceptions-Minus-Expectations Measurement of Service Quality

Abstract

The authors respond to concerns raised by Parasuraman, Zeithaml, and Berry (1994) about the relative efficacy of performance-based and perceptions-minus-expectations measures of service quality. They demonstrate that the major concerns voiced by these authors are supported neither by a critical review of their discussion nor the emerging literature. Several research issues relative to service quality measurement and strategic decision making also are identified.

Full text

J.
Joseph Cronin, Jr. & Steven A. Taylor
SERVPERF Versus SERVQUAL:
Reconciling Performance-Based and
Perceptions-Minus-Expectations
Measurement of Service Quality
The authors respond to concerns raised by Parasuraman, Zeithaml, and Berry (1994) about the relative efficacy of
performance-based and perceptions-minus-expectations measures of service quality. They demonstrate that the
major concerns voiced by these authors are supported neither by a critical review of their discussion nor the emerg-
ing literature. Several research issues relative to service quality measurement and strategic decision making also
are identified.
P
araraman, Zeithaml, and Beny (1994), hereafter referred
to as PZB (1994), present a critique of the perfor-
mance-based measure of service quality (SERVPERF) iden-
tified by Cronin and Taylor (1992). We begin our response
by first expressing our appreciation for the opportunity to
continue the discussion of how best to conceptualize and
operationalize the service quality construct. We further con-
cur with PZB (1994) that many of the issues discussed
herein remain unresolved and similarly call for further re-
search into the important areas of service quality and con-
sumer satisfaction. However, the following discussion dem-
onstrates that the criticisms PZB (1994) make of the
SERVPERF model appear largely to lack substance.
Before we address PZB's (1994) comments on a point-
by-point basis, it is appropriate to address two general
points they raise. First, Cronin and Taylor (1992) do not
"conclude that it is unnecessary to measure customer expec-
tations in service quality research" (p. 111). Rather, our re-
sults suggest that the perfomiance-minus-expectations is an
inappropriate basis for use in the measurement of service
quality. The reported results in no way contradict the impor-
tance of the unique effect that expectations can have on con-
sumers' perceptions of service quality (cf. Boulding et al.
1993):
Second, our results do not actually suggest "that ser-
vice quality fails to affect purchase intentions." In fact, a
close examination of Table 5 in Cronin and Taylor (1992)
indicates that in the SERVQUAL Model (Model 1), service
quality has a statistically significant effect (p < .05) for two
of the four industries (pest control and fast food). In the
SERVPERF Model (Model 2), the effect is statistically sig-
nificant (p < .05) in three industries (banking, pest control.
J.
Joseph Cronin, Jr.,
is an
Associate Professor, Florida State University.
Steven .^. Taylor
is an
Assistant Professor, lillnois State University.
Joumal cf Marketing
Vol.
58 (January 1994), 125-131
and fast food). However, as we discuss subsequently, our re-
sults do suggest "that consumer satisfaction exerts a
stronger influence on purchase intentions than does service
quality" (Cronin and Taylor 1992, p. 65). We now tum to
a consideration of the specific issues raised by PZB (1994).
Conceptual Issues
Perceptions Expectations Conceptuaiization
The first issue raised by PZB (1994) involves the appropri-
ateness of the perceptions-expectations gap conceptualiza-
tion, which is the basis of the SERVQUAL scale. PZB
(1994) state that their focus groups captured not only the at-
tributes of service quality, but also the underlying psycholog-
ical process by which consumers form service quality judg-
ments. Essentially, on the basis of their focus group find-
ings,
PZB (1985, 1988) conclude that service quality judg-
ments comprise of five underlying attributes that consum-
ers evaluate on the basis of the expectancy-disconfirmation
paradigm (Oliver 1980).
PZB (1994) imply that the literature we cite in the devel-
opment of our research hypotheses is unconvincing and re-
state their arguments for conceptualizing service quality per-
ceptions as being based on the expectancy-disconfirmation
process. However, the SERVPERF conceptualization repre-
sents just one of a number of recent challenges to the
SERVQUAL-based normal science exemplar of service qual-
ity (cf. Babakus and Boiler 1992; Babakus and Mangold
1992;
Boulding et al. 1993; Carman 1990; Oliver 1993).
We refrain from defending our original literature support
for testing altematives to the SERVQUAL paradigm and in-
stead direct our discussion toward the more direct question
of the evidence that has emerged since the publication of
the SERVPERF model. It is important to note that the
emerging literature largely has supported the emerging per-
SERVPERF Versus SERVQUAL /125

formance-based paradigm over the disconfirtnation-based
SERVQUAL paradigm.
In perhaps the most telling evidence to date, one of the
original coauthors of SERVQUAL recently reports results
(Boulding, Kaka, Staelin, and Zeithaml 1993, p. 24) that ap-
pear to support the conclusions of Cronin and Taylor
(1992) over that of PZB (1994):
Our results are incompatible with both the one-dimen-
sional view of expectations and the gap formation for ser-
vice quality
[italics added]. Instead, we find that service
quality is directly influenced only by perceptions [of
performance].
Other recent studies bear out these conclusions. Peter,
Churchill, and Brown (1992) present a compelling argu-
ment of why difference scores (such as those employed by
the SERVQUAL scale operationalization) should be
avoided. Brown, Churchill, and Peter (1992) specifically ex-
tend these arguments to an investigation of the
SERVQUAL scale and conclude that there are serious prob-
lems in conceptualizing service quality as a difference score
(see Brown, Churchill, and Peter 1992 for a comprehensive
discussion of their criticisms). Babakus and Boiler (1992)
and Babakus and Mangold (1992) also report results support-
ing the use of performance-based measures of service qual-
ity over gap measures.
Revisiting the conceptual foundations of the
SERVQUAL model provides some insight into the conflict-
ing results. Origins of the "gap" model can be found in the
early writings on disconfirmation in the consumer satisfac-
tion literature (Oliver 1977, 1980a, 1981). Oliver proposes
that consumers make 1
'
'better-than-expected/worse-than-
expected'' (disconfirmation) judgments on the basis of a
comparison of product performance to expectations in the de-
termination of consumer satisfaction. Though conceptually
consumers can make arithmetic or calculated comparisons
between expectations and performance (as in a car's gas
mileage), Oliver
(1981;
Oliver and Bearden 1985) argues
that they may not because of measurement failure or effort
or because the relevant performance dimensions cannot be
quantified (e.g., aesthetics, pleasure). Thus, the perceived
summary disconfirmation judgment is sufficient as a causal
agent for satisfaction.
In situations in which expectation and performance data
are available, at least to the researcher, it is possible to infer
the consumer's disconfirmation through arithmetic means.
Oliver (1980b), Oliver and Bearden (1985), and Swan and
Trawick (1981) have tested variations of this inferential
measure in models including a summary measure and have
shown that the summary measure mediates the effect of the
inferential measure on satisfaction. This would appear to be
a reasonable finding because consumer perceptions, not cal-
culations, govern behavior. In this context, the
SERVQUAL gap measure is analogous to the inferential dis-
confirmation measure, and is thus an incomplete form of
the summary comparative judgment consumers might use
in quality decisions.
Thus,
PZB's (1985, 1988) generalization of the satisfac-
tion paradigm to service quality evaluations based on the
qualitative evidence apparent in their focus group interpre-
tations could refiect the general ambiguity inherent in the ser
vice quality literature regarding the distinction between ser-
vice quality and consumer satisfaction, rather than clear sup-
port for the efficacy of the PZB gap model. The ambiguity
inherent in the service quality literature relative to these two
closely related constructs also could refiect a similar ambi-
guity in consumers' and managers' understanding of the dif-
ference between service quality and consumer satisfaction—
ambiguity that possibly is refiected in the focus groups con-
ducted by PZB (1985). However, as is depicted in the pre-
vious discussion, this same ambiquity is not evident in the
consumer satisfaction literature.
Given the arguments presented here, along with the
growing body of literature criticizing the five-dimension
conceptualization (cf. Babakus and Boiler 1992; Babakus
and Mangold 1992; Brown, Churchill, and Peter 1992; Ol-
iver 1993; Patterson and Johnson 1993; Peter, Churchill,
and Brown 1992), the assertion that "little if any theoretical
or empirical evidence supports the relevance of the expecta-
tions-performance gap as the basis for measuring service
quality" (Cronin and Taylor 1992, p. 56) appears
creditable.
Two minor points relative to this discussion also should
be addressed. PZB (1994) question our use of Bolton and
Drew (1991) to support our performance-based measure-
ment of service quality. Slightly farther down the page
from the quote offered by PZB (1994) is a second: "(A) cus-
tomer'
s
assessment of overall service quality is also directly
affected by perceptions of performance levels" (p. 383). In
addition, PZB (1994) question our use of Mazis, Ahtola,
and Klippel (1975) on the basis of the fact that the paper
dealt with neither service quality nor tested performance-
based measures incorporating expectations. A thorough read-
ing of this article, however, does suggest that the inclusion
of importance weights does not enhance the predictive abil-
ity of attitude models. Because the use of such weights was
examined in our study, the reference does not appear
inappropriate.
Attitude Formation Versus Attitude Measurement
The second conceptual issue raised by PZB (1994) is
whether service quality measurement is properly associated
with attitude formation or attitude measurement. We recog-
nize that the SERVQUAL and SERVPERF scales are tools
designed for the measurement of service quality; that is, the
measurement of a specific long-term attitude at a single
point in time. Indeed, much of the conceptual support for
using performance-based measures of service quality atti-
tudes over disconfirmation-based measures is derived from
this distinction.
Specifically, a review of the existing literature identifies
an apparent consensus regarding a fundamental distinction
between service quality and consumer satisfaction: Service
quality is a long-term attitude, whereas consumer satisfac-
tion is a transitory judgment made on the basis of a specific
service encounter (cf. Bitner 1990; Bolton and Drew 1991;
Cronin and Taylor 1992; Oliver 1993; Patterson and
Johnson 1993). This distinction is refiected in the concep-
tual domains of the relevant constructs. Service quality per-
126
/
Journal of Marketing, January 1994

ceptions reflect a consumer's evaluative perceptions of a ser-
vice encounter at a specific point in time. In contrast, con-
sumer satisfaction judgements are experiential in nature, in-
volving both an end state and a process and reflecting both
emotional and cognitive elements (Oliver 1993). Satisfac-
tion judgments are believed to degenerate into overall ser-
vice quality judgments over time (Cronin and Taylor 1992,
Oliver 1993). Expectancy-disconfirmation
judgments,
how-
ever, (1) are distinct from both consumer satisfaction judg-
ments and service quality perceptions, (2) involve calcu-
lated and subjective forms, and (3) can involve a number of
referents. (See Oliver 1993 for a comprehensive discussion
of the differences between service quality perceptions, con-
sumer satisfaction judgements, and the role of expectancy-
disconfirmation in consumer decision making processes.)
In short, recent conceptud advances suggest that the dis-
confirmation-based SERVQUAL scale is measuring neither
service quality nor consumer satisfaction. Rather, the
SERVQUAL scale appears at best an operationalization of
only one of the many forms of expectancy-disconfirmation
(cf. Boulding, et al. 1993; Oliver 1993; Zeithaml, Berry,
and Parasuraman 1993). In Cronin and Taylor (1992), it is
suggested that performance-based measures better refiect
long-term service quality attitudes in cross-sectional stud-
ies.
We stand by our original position because disconfirma-
tion and consumer satisfaction judgments are both process
constmcts that share a similar reliance on the consumer ex-
periencing a service encounter, whereas performance percep-
tions are not constrained to actual consumer experiences.
Such a distinction is consistent with Patterson and
Johnson's (1993) comparison of consumer satisfaction and
service quality paradigms.
The Reiationship Between CS and SQ and
Comparison Standards for Expectancy
Disconfirmation
The final conceptual arguments raised by PZB (1994) con-
cem the closely related issues of the (1) causal relationship
between service quality and consumer satisfaction and (2)
the appropriate comparison standard for expectancy-discon-
firmation. We begin first by apologizing for apparently mis-
interpreting the nature of the relationship between service
quality and consumer satisfaction implicit in the
SERVQUAL paradigm as being consistent with our empiri-
cal results.
Second, PZB (1994) argue that Cronin and Taylor
(1992) imply that the debate over the appropriateness of the
various comparison norms used in the measurement of
expectations is resolved. We recognize the unresolved na-
ture of the debate over the appropriate comparison stan-
dards against which perceptions are compared in expec-
tancy-disconfirmation measures (cf. Boulding et al. 1993;
Gardial, Clemons et al. 1993; Gardial, Woodruff et al.
1993).
The intent of the quote cited by PZB (1994) was to
suggest that the distinction between service quality and con-
sumer satisfaction in their work was not clear. Though the
research of Boulding et al. (1993) and Gardial, Woodruff et
al.
(1993) has furthered our understanding of appropriate
comparison standards, we concur with PZB (1994) that fur-
ther research into this area is warranted. In short, PZB's
(1994) interpretation of
the
finality of our arguments involv-
ing comparison standards is unintended and we thank the au-
thors for allowing us the opportunity to prevent further mis-
interpretation of our original work.
In sum, the weight of the emerging literature supports
the efficacy of performance-based measures of service qual-
ity in general and the SERVPERF scale specifically as the
appropriate exemplar of service quality operationalization.
The next section addresses PZB's (1994) criticisms of
Cronin and Taylor's (1992) empirical analyses.
Empirical Issues
Dimensions of Service Quaiity/Factor
Anaiysis Procedures
PZB (1994) correctly question whether we allowed the five
dimensions of service quality to be intercorrelated. We did
neglect to correctly depict the correlations in Figure
1
of the
original article (Cronin and Taylor 1992). However, the ac-
tual analysis did correctly account for the intercorrelations
as conceptualized in Figure 1 of this article (see Table 1 for
the intercorrelations).
Further evidence that the correlations between the dimen-
sions were accounted for is evident in the degrees of fi^e-
dom reported in Table 1 of the 1992 article (p. 61). We in-
cluded 22 variables in the confirmatory factor analysis, re-
sulting in 253 parameters. In the TD matrix, all 22 diagonal
elements were freed, in the LX matrix the first element for
each dimension was fixed at 1 and the remaining 17 were
freed, and in the PHI matrix the diagonal elements (the var-
iances of the latent variables) were fixed at 1 and the off-
diagonal elements were freed, thus resulting in 204 degrees
of freedom as reported in Table 1 of Cronin and Taylor
(1992).
Because the off-diagonal elements of the PHI ma-
trix are actually covariances of the five latent variables, free-
ing those elements is equivalent to allowing the five dimen-
sions to be intercorrelated. In addition, when the dimen-
sions are considered with the diagonal elements of the PHI
matrix freed, there is no appreciable change in the results
(see Table 2).
PZB (1994) raise two additional questions relative to
the factor analyses presented in Cronin and Taylor (1992).
First, they suggest that because the variance captured by the
SERVQUAL item "is much lower than that for the
SERVPERF items ... the former could be a richer con-
struct" (p. 113). This interpretation is unusual and problema-
tic.
Second, in contrast to their earlier position that the di-
mensions of service quality should reflect their inherent
intercorrelation, PZB (1994) suggest that the rotated factor
loadings identified in Table 2 of Cronin and Taylor (1992)
could infiate the variance explained by a factor because an
oblique rotation could attribute variance to one factor that
also is explained by one or more additional factors. Why
this issue is raised by PZB (1994) is puzzling, given that
they have consistently suggested that the dimensions of ser-
vice quality are intercorrelated (PZB 1988, 1991, 1994) and
that the factor analysis procedure utilized in Cronin and Tay-
lor (1992) is identical to that used in PZB (1988).
SERVPERF Versus SERVQUAL /127

FIGURE 1
Service Quality as Conceptualized by Parasuraman, Zeithanni, and Berry (1988) and
Exannined by Cronin and Taylor (1992)
SERVICE
QUALITY
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22
TABLE 1
The Intercorrelations of the Five-F«;tor
Conceptualization of Service Quaiity
Tangibles 1.000 703 ^458
iil4
.478
Reliability .457 1.000 .661 .744 .533
Responsiveness .240 .680 1.000 .644 .687
Assurances .523 .854 .566 1.000 .559
Empathy .256 .854 .772 .539 1.000
Note:
Cell entries below the diagonal are for banks
and
those
above are for pest control.
Tangibles
Reliability1.000 .316
.638 1.000.197
.686
Responsiveness .319
Assurances .589
Empathy .395
.495 1.000
.846
.490
.282 .165
.525 .449
.620 .686
.503 1.000 .575
.573 .645 1.000
Note:
Cell entries below the diagonal are for dry cleaners
and
those
above are for fast
food.
Cronin and Taylor (1992) actually suggest that both
SERVQUAL and SERVPERF can be treated unidimension-
ally after a five-factor solution failed to fit the SERVQUAL
items in any of the four industries. Here we confess that
some of Cronin and Taylor's (1992) wording is potentially
confusing, though the treatment of the SERVQUAL items
is quite consistent with the extant literature. For example.
Carman (1990) also demonstrates that the SERVQUAL
scale fails to exhibit the predicted five-factor structure
TABLE 2
Confimutory Factor Analysis Paran^er
Estimates For Five-Factor {hntercorreiirited)
Conceptuaiization of Service Quaiity
Parameter
Chi Square
d.f.
P
Banks
333.26
199.000
.867
.831
.204
Pest
Control
464.08
199.000
.810
.758
.247
Dry
Cleaning
395.01
199.000
.828
.781
.257
Fast
Food
346.10
199.000
.862
.824
.255
aGoodness of fit
>>Adjusted
goodness of fit
<=Root mean
square residual
when tested for a tire store, placement center, and dental
clinic.
Cronin and Taylor (1992) in effect use the two sets of
items (SERVQUAL and SERVPERF) as two indices rather
than as two factor-based scales after failing to identify a con-
sistent factor structure across the four industries examined
in the study. The difference is that an index is an exact lin-
ear combination of observed items. The dimensionality of
the items of an index used as an observed variable is not rel-
evant. The point of the comparison of the two models is to
determine which of the indices is the superior measure of ser-
vice quality. The conclusions drawn in the research are not
128
/
Joumai of Hbiriwting, January 1994

invalidated by the psychometric properties of the items that
comprise the indices. If in fact the superiority of the
SERVPERF approach pertains in some way to those psy-
chometric properties of the measure, the fact remains—it is
a better measure of service quality.
Vaiidity
PZB (1994) assert that Cronin and Taylor (1992) claim that
the performance-only SERVPERF measures have greater va-
lidity because of their content and discriminant validity,
and they present a lengthy comparison of the convergent
and discriminant characteristics of the SERVQUAL and
SERVPERF measures. Though we appreciate the demonstra-
tion of the superior convergent and discriminant power of
the SERVPERF measures, reB (1994) have not accurately
depicted our basis for attributing greater validity to
SERVPERF. We assert that SERVPERF has greater con-
struct validity based on our review of the literature and the
fact that the SERVPERF measures also exhibit convergent
and discriminant validity. Nowhere in Cronin and Taylor
(1992) is there any consideration or comparison of the con-
vergent or discriminant validity of the SERVQUAL meas-
ure.
Our implication is simply that, on the basis of a review
of the extant literature, SERVQUAL fails to exhibit con-
struct validity.
Regression Anaiyses
PZB (1994) suggest that the dependent measure utilized in
Cronin and Taylor's (1992) regression analyses is perfor-
mance based. The exact wording of the question is:
'
'The
quality of XYZ's services is (very poor (1) to excellent
(7))"
(Cronin and Taylor 1992, p. 67). PZB's argument
that "shared method variance" can account for the greater
explanatory power of the SERVPERF measures thus is du-
bious because (1) Likert scale items were used for the
SERVQUAL and SERVPERF measures, whereas a seman-
tic differential item was utilized for the overall service qual-
ity measure (i.e., different measurement methods) and (2) it
is not obvious that the dependent measure is "performance
based."
PZB (1994) also suggest that the overall pattern of re-
gression coefficients in Cronin and Taylor's (1992) Table 4
support their five-factor dimensionality. The interpretation
offered is suspect for several reasons. First, it is inappro-
priate to assign the 22 individual items to the five factors
when the bulk of empirical evidence presented in the extant
literature rejects the five-factor structure and further sug-
gests that the five dimensions are intercorrelated (cf.
Babakus and Boiler 1992; Babakus and Mangold 1992; Car-
man 1990; Cronin and Taylor 1992; Parasuraman, Zei-
thaml, and Berry 1988, 1991). Second, the importance
weighted measures should not be included in the compari-
son because one cannot disaggregate the importance, perfor-
mance, and interaction effects. When only the unweighted
measures are considered, the pattern of significant regres-
sion coefficients reported by PZB (1994) changes dramati-
cally as is shown following:
Segment 1
Segment 2
Segment 3
Segment 4
Segment 5
SERVQUAL SERVPERF
Tangibles 2/16 2/16
Reliability 7/20 6/20
Responsiveness 2/16 2/16
Assurances 3/16 5/16
Empathy 4/20 2/20
A final point made by PZB (1994) involves the deriva-
tion of the importance weights used in the regression analy-
sis.
PZB contend first that the weights should be calculated
for the dimensions and not for the individual measures (i.e.,
5 "weights" as opposed to 22). Our contention is that the
five-factor dimensionality is problematic; therefore, interpret-
ability is enhanced by asking respondents to assign weights
to each measure.
PZB (1994, p. 115) also contend that asking respon-
dents to indicate the importance of individual items is "not
meaningful because a primary purpose of regression analy-
sis is to derive the importance weights indirectly." Multiple
regression calculates beta coefficients on the basis of the
ability of an independent variable to predict changes in a de-
pendent variable. Mathematically (i.e. indirectly) calculat-
ing the ability of a predictor to explain variation in a depend-
ent measure simply is not the same as directly asking con-
sumers to indicate their perception of the importance of a
specific aspect of service. In addition, the importance
weights failed to contribute significantly to the predictive
ability of either the SERVQUAL or SERVPERF measures,
so the efficacy of the issue is especially dubious.
Relationships Among Service Quality, Consumer
Satisfaction, and Purchase Intentions
Initially, the point should be made that the purpose of
Cronin and Taylor (1992) was to test a performance-based
alternative to SERVQUAL's gap formulation. Our consider-
ation of the relationships between overall service quality,
consumer satisfaction, and purchase intentions was under-
taken to emphasize the importance of the measurement is-
sues relative to service quality. We are in agreement with
PZB's (1994) suggestion that the directionality of the ser-
vice quality/satisfaction relationship is still in question and
that future studies of these relationships should incorporate
multi-item measures.
PZB (1994) also correctly indicate that we suggest that
"only" consumer satisfaction has a significant effect on pur-
chase intentions. However a review of the t-values pre-
sented in Table 5 suggests that overall service quality has a
statistically significant (p < .05) effect on purchase inten-
tions in two industries (pest control and fast food) for
Model 1 (SERVQUAL) and in three industries (banking,
pest control, and dry cleaning) for Model 2 (SERVPERF).
In contrast, consumer satisfaction has a statistically signifi-
cant (p < .01) effect for either model in all four industries.
Thus,
the correct interpretation of our results is that con-
sumer satisfaction appears to have an effect on purchase in-
tentions, which is more frequently statistically significant,
and the effect tends to achieve a greater level of statistical
significance when both constructs have a significant effect
on purchase intentions. Consumer satisfaction, thereby, ap-
pears to be a "richer" construct for use in predicting pur-
SERVPERF Versus SERVQUAL /129

chase intentions. This should hardly be surprising given
that it has long been assumed that consumers do not always
purchase the highest' 'quality'' product due to cost, budget,
availability, and other constraints.
We also take exception to PZB's (1994) interpretation
of the methodological issues related to our examination of
the structural models identified in Figure 2 of Cronin and
Taylor (1992). First, PZB (1994) correctly point out that a
one degree of freedom model fits better, by definition, than
a two degree of freedom model, and so on. However, their
logic misses the point of our analysis, which is to compare
two models that are identical except that SERVQUAL is
used in one and SERVPERF in the other. The degrees of
freedom are not a relevant issue because we are comparing
apples to apples (i.e., a one degree of freedom model to an
identical one degree of freedom model, which substitutes
22 SERVPERF measures for 22 SERVQUAL items). They
also raise the "shared-method variance" argument here
again, which, as we indicate previously, is simply not sup-
ported by a close examination of the research methods
utilized.
Practical issues
We also are pleased to support an emphasis on practical is-
sues involving the measurement of service quality. PZB
(1994) sjjecifically suggest that disconfirmation measures
(1) provide richer information than do the performance-
based measures we propose and (2) have a greater diagnos-
tic value for managers. Because the preceding discussion of
the recent literature identifies that the SERVQUAL opera-
tionalization of service quality (performance-expectations)
appears largely unsupported on both conceptual and empir-
ical grounds (cf. Babakus and Boiler 1992; Babakus and
Mangold 1992; Brown, Churchill, and Peter 1992; Oliver
1993;
Patterson and Johnson 1993; Peter, Churchill, and
Brown 1992), we suggest that the more relevant question in
terms of practical considerations is whether the
SERVPERF scale provides a reliable and valid scale for
operationalizing the service quality construct.
As previously discussed, the performance-based meas-
ures of service quality captured by the SERVPERF scale
can provide a longitudinal index of the service quality per-
ceptions of a service firm's constituencies. In other words,
the SERVPERF scale can provide managers with a
summed overall service quality score that can be plotted rel-
ative to time and specific consumer subgroups (e.g., demo-
graphic subcategories, individual constituencies). As such,
the SERVPERF scale provides a useful tool for measuring
overall service quality attitudes by service managers. How-
ever, we suggest that great care should be exercised by man-
agers of service firms in attempts to derive more specific in-
formation from data derived using the SERVPERF scale
for strategic decision-making.
Specifically, service quality measures clearly exhibit a fac-
tor structure that varies across service industries (cf.
Babakus and Boiler 1992; Brown, Churchill, and Peter
1992;
Cronin and Taylor 1992). It was this finding that led
to our original argument for using the SERVQUAL and
SERVPERF scales as indexes to allow the comparison of re-
sults across altemative models and service industries. We
have unpublished results that further demonstrate that the
SERVPERF scale exhibits the same factor instability inher-
ent in the SERVQUAL difference scores across service in-
dustries. Consequently, we caution marketers to assess the
factor structure implicit in data sets derived from
SERVPERF measures to ensure that the hypothesized five-
factor structure identified by PZB (1988) can be replicated
specific to their own research setting. If not, appropriate
steps should be taken to ensure that only information im-
plicit in their own specific data set is used for strategic
decision making.
On the basis of the preceding discussion, we remain un-
convinced that including consumer expectations in meas-
ures of service quality is a position managers should sup-
port. However, this is not to say that these same measures
cannot impart valuable information for managers if their
unique effect on purchase behaviors and performance percep-
tions are conceptualized properly (cf. Boulding et al. 1993).
We also suggest that one currently unexplored area for gain-
ing additional information from the SERVPERF scale
could be in the use of performance-importance maps
(Hawes and Rao 1985). That is, maps can be developed for
specific data sets that plot consumers' perceptions of the im-
portance of individual scale items relative to perceptions of
service firm performance for each performance attribute.
Summary
The preceding discussion addresses the concems raised by
PZB (1994) on a point-by-point basis. The conclusion of
this discussion is that (1) the criticisms identified by PZB
(1994) appear related more to issues of interpretation than is-
sues of substance and (2) the emerging literature clearly sup-
ports the original conclusions of Cronin and Taylor (1992)
over the PZB (1994) defense of the SERVQUAL exemplar.
We acknowledge that the ground breaking work of PZB
(1985,
1988, 1991, 1994) has made a significant contribu-
tion to service quality research. We also second their call
for continued study of research models that integrate ser-
vice quality, consumer satisfaction, and, we would add, ser-
vice value.
PZB (1994) present a research agenda to aid in meeting
these research goals. Though we do not directly address
their proposed goals, we would ask marketing researchers
and practitioners to consider one key point in further inves-
tigations of service quality and its relationship to consumer
satisfaction, service value, and purchase behavior/inten-
tions.
PZB (1994) suggest that the service quality and con-
sumer satisfaction constructs can be examined at both the
transaction-specific and global levels of analysis.
Though such an extension in the literature could allow
the reconciliation of their paradigm with emerging evi-
dence, we suggest it ultimately will be at the expense of clar-
ity of understanding, the discriminant validity of the rele-
vant measures, and the predictive ability of models of con-
sumer decision making. We remain troubled that expec-
tancy-disconfirmation, though originally developed to ex-
plain consumer satisfaction judgments (Oliver 1980), has
been extended to both represent a unique construct (Bolton
130
/
Joumal
of
Marketing, January 1994

and Drew 1991, Qliver 1993) and measure service quality
perceptions (PZB 1985, 1988).
The point is that it may be time for service marketers to
investigate the possibility of a commensurable agreement
as to the domains of service quality, consumer satisfaction,
expectancy disconfirmation, and service value. We suggest
that a useful starting point, consistent with what is currently
known, would be to restrict the domain of service quality to
long-term attitudes and that of consumer satisfaction to trans-
action-specific judgments. We suggest that by limiting the
domains of service quality and consumer satisfaction, such
a restriction could enhance our understanding of how these
constructs interact in consumer decision-making processes.
REFERENCES
Babakus, Emin and Gregory
W.
Boiler (1992), "An Empirical As-
sessment of the SERVQUAI^ Scale," Joumal of Business Re-
search,
24, 253-68.
and W. Glenn Mangold (1992), "Adapting the
SERVQUAL Scale to Hospital Services: An Empirical Inves-
tigation," Health Service Research, 26 (6), 767-80.
Bitner, Mary Jo (1990), "Evaluating Service Encounters: The Ef-
fects of Physical Surroundings and Employee Responses," Jour-
nal of Marketing, 54 (2), 69-82.
Bolton, Ruth N. and James H. Drew (1991), "A Multistage Model
of Customers' Assessments of Service Quality and Value,"
Joumal of
Consumer
Research, 17 (4), 375-84.
Boulding, William, Ajay Kalra, Richard Staelin, and Valarie A.
Zeithaml (1993),' 'A Dynamic Process Model of Service Qual-
ity: From Expectations to Behavioral Intentions," Joumal of
Marketing Research, 30 (February 1993), 7-27.
Brown, Tom J., Gilbert A. Churchill, Jr., and J. Paul Peter (1992),
"Improving the Measurement of Service Quality," A.C.
Nielsen working paper #91-4, University of Wisconsin—
Madison, Madison, WI.
Carman, James M. (1990), "Consumer Perceptions of Service
Quality: an Assessment of the SERVQUAL Dimension," Jour-
nal of Retailing, 66 (1), 33-55.
Cronin, J. Joseph, Jr. and Steven A. Taylor (1992), "Measuring Ser-
vice Quality: A Reexamination and Extension," Joumal of
Marketing, 56 (3), 55-68.
Gardial, Sarah Fisher, D. Scott Clemons, Robert B.
Woodruff,
David W. Schumann and Mjiry Jane Bums (1994), "Compar-
ing Consumers' Recall of Prepurchase and Postpurchase Prod-
uct Evaluation Experiences," Joumal of
Consumer
Research,
forthcoming.
, Robert B.
Woodruff,
Mary Jane Bums, David
W.
Schu-
mann, and Scot Clemons (1994), "Comparison Standards: Ex-
ploring Their Variety and the Circumstances Surrounding
Their Use," Joumal of
Consumer
Satisfaction,
Dissatisfaction,
and
Complaining
Behavior, forthcoming.
Hawes, J. M. and C.P. Rao (1985), "Using Importance-Perfor-
mance Analysis to Develop Health Care Marketing Strate-
gies,"
Joumal of Health Care Marketing, 5 (4), 19-25.
Mazis, Michael B., Olli T. Ahtola, and R. Eugene Klippel (1975),
"A Comp^son of Four Multi-Attribute Models in the Predic-
tion of Consumer Attitudes," Joumal of
Consumer
Research,
2 (June), 38-52.
Oliver, Richard L. (1980), "A Cognitive Model of the Antece-
dents and Consequences of Satisfaction Decisions," Joumal of
Marketing Research, 17 (November), 460-9.
(1993),
"A Conceptual Model of Service Quality and
Service Satisfaction: Compatible Goals, Different Concepts,"
in Advances in Services Marketing and Management: Re-
seareh and Practice, Vol. 2, Teresa A. Swartz, David E.
Bowen, and Stephen W. Brown, eds. Greenwich, CT: JAI
Press.
Parasuraman, A., Valarie Zeithaml, and Leonard Berry (1985), "A
Conceptual Model of Service Quality and Its Implications for
Future Research," Joumal of Marketing, 49 (Fall) 41-50.
Parasuraman, A., Valarie Zeithaml, and Leonard Berry (1988),
"SERVQUAL: A Multi Item Scale for Measuring Consumer
Perception of Service Quality," Journal of Retailing, 64
(Spring) 12-W.
, , and (1991), "Refinement and Re-
assessment of the SERVQUAL Scale," Joumal of Retailing,
67 (4), 420^50.
-, and (1994), "Reassessment of Ex-
pectations as a Comparison Standard in Measuring Service
Quality: Implications for Further Research," Joumal of Market-
ing, 5S
{January),
111-24.
Patterson, Paul G. and Lester W. Johnson (1993). "Disconfirma-
tion of Expectations and the Gap Model of Service Quality: An
Integrated Paradigm,'' Joumal of
Consumer
Satisfaction,
Dis-
satisfaction, and
Complaining
Behavior, forthcoming.
Peter, J. Paul, Gilbert A. Churchill, Jr., and Tom J. Brown (1992),
"Caution in the Use of Difference Scores as Measures of Mar-
keting Constructs," A. C. Nielsen working paper #91-2, Uni-
versity of Wisconsin—Madison, Madison, WI.
Zeithaml, Valarie, Leonard Berry, and A. Parasuraman (1993),
"The Nature and Determinant of Customer Expectation of Ser-
vice,
Journal of the Academy of Marketing Science, 21 (1),
1-12.
SERVPERF Vmus SERVQUAL /131