The Behavioral Consequences of Service Quality

Abstract

If service quality relates to retention of customers at the aggregate level, as other research has indicated, then evidence of its impact on customers’ behavioral responses should be detectable. The authors offer a conceptual model of the impact of service quality on particular behaviors that signal whether customers remain with or defect from a company. Results from a multicompany empirical study examining relationships from the model concerning customers’ behavioral intentions show strong evidence of their being influenced by service quality. The findings also reveal differences in the nature of the quality-intentions link across different dimensions of behavioral intentions. The authors’ discussion centers on ways the results and research approach of their study can be helpful to researchers and managers.

Full text

Valarie
A.
Zeithaml,
Leonard
L.
Berry,
&A.
Parasuraman
The
Behavioral
Consequences
of
Service
Quality
If service quality relates to retention of customers at the aggregate level, as other research has indicated, then ev-
idence of its impact on customers' behavioral responses should be detectable. The authors offer a conceptual
model of the impact of service quality on particular behaviors that signal whether customers remain with or defect
from a company. Results from a multicompany empirical study examining relationships from the model concerning
customers' behavioral intentions show strong evidence of their being influenced by service quality. The findings
also reveal differences in the nature of the quality-intentions link across different dimensions of behavioral inten-
tions. The authors' discussion centers on ways the results and research approach of their study can be helpful to
researchers and managers.
Delivering quality service is considered an essential
strategy for success and survival in today's competi-
tive environment (Dawkins and Reichheld 1990;
Parasurarnan, Zeithaml, and Berry 1985; Reichheld and
Sasser 1990; Zeithaml, Parasuraman, and Berry 1990). Dur-
ing the 1980s, the primary emphasis
of
both academic and
managerial effort focused on determining what service qual-
ity meant to customers and developing strategies to meet cus-
tomer expectations (e.g., Parasuraman, Zeithaml, and Berry
1985, 1988). Since then, many
organizations-including
those whose primary offerings involve physical goods such
as automobiles or
computers-have
instituted measurement
and management approaches to improve their service. The
service-quality agenda has now shifted and reconfigured to
include other issues. The issue of highest priority today in-
volves understanding the impact of service quality on profit
and other financial outcomes
of
the organization (Greising
1994; Rust, Zahorik, and Keiningham 1995).
Executives
of
many companies in the 1980s were will-
ing to trust their intuitive sense that better service would
lead to improved financial success and thus committed re-
sources to improving service prior to having documentation
of
the financial payoff.
Some
of
these companies, such as
Federal Express and Xerox, have been richly rewarded for
their efforts (Germano 1992; Kearns and Nadler 1992). But
executives in other companies have been reluctant to invest
in service improvements without solid evidence
of
their fi-
nancial soundness. And in the current era
of
downsizing and
streamlining, interest in tools to ascertain and monitor the
payoff from service investments is high.
Valarie
A.
Zeithaml
is
Principal,
Partners
for
Service
Excellence,
a
consult-
ing
firm
specializing
in
strategy,
measurement,
and
implementation
ofser-
vice
quality.
Leonard
L.
Berry
is
JCPenney
Chair
of
Retailing
Studies
and
Professor
of
Marketing,
Texas
A&M
University.
A.
Parasuraman
is
Professor
and
Holder
of the James
W.
McLamore
Chair
in
Marketing,
University
of
Miami.
The
authors
thank
the
editor
and
five
anonymous
JM
reviewers
for
their
constructive
comments
and
suggestions
on
earlier
drafts
of
this
arti-
cle.
They
also
thank
the
Marketing
Science
Institute
and
four
ofits
corpo-
rate
sponsors
for
supporting
the
research
on
which
this
article
is based.
Journal
of
Marketing
Vol. 60 (April
1996),31-46
Research on the relationship between service quality and
profits has begun to accumulate, and
one
thing is clear:
The
link between service quality and profits is neither straight-
forward nor simple (Greising 1994; Zahorik and Rust 1992).
The
intermediate links between service quality and profits
have not been well understood. To delineate the complex re-
lationship between these two variables, researchers and
managers must investigate and understand many
other
rela-
tionships, each
of
which is an integral part of the composite.
One
such
relationship--between
service quality and behav-
ioral
intentions-is
the primary focus
of
our present re-
search. In the remainder
of
this introductory section, we pro-
vide a general overview
of
the extant knowledge about the
link between service quality and profits. We then outline
our
specific objectives and how our study attempts to extend
current knowledge.
Seminal studies using the PIMS (Profit Impact
of
Mar-
ket Strategy) data set have uncovered significant associa-
tions among service quality, marketing variables, and prof-
itability. Findings from these studies show that companies
offering superior service achieve higher-than-normal market
share growth (Buzzell and Gale 1987), that the mechanisms
by which service quality influences profits include increased
market share and premium prices (Phillips, Chang, and
Buzzell 1983), and that businesses in the top quintile of rel-
ative service quality on average realize an 8% higher price
than their competitors (Gale 1992). Evidence from compa-
nies large enough to have multiple outlets also suggest a
positive quality-profitability relationship: The Hospital Cor-
poration
of
America found a strong link between perceived
quality
of
patient care
and
profitability across its many hos-
pitals (Koska 1990); and the Ford Motor
Company
has
demonstrated that dealers with high service-quality scores
have higher-than-normal profit, return on investment, and
profit per new vehicle sold (Ford
Motor
Company 1990).
Although the previous findings document the financial
and strategic impact of service quality across firms or out-
lets, the evidence is often too general to answer the ques-
tions foremost in executives' minds:
If
Iinvest in service
Service
Quality
/31

quality, will it
payoff
for my company? How will service
quality payoff? How much should we invest in service qual-
ity to receive the best return? In addressing such questions,
researchers (Fornell and Wernerfelt 1987, 1988; Rust and
Zahorik 1993; Zahorik and Rust 1992) distinguish between
offensive effects (capturing new customers) and defensive
effects (retaining customers). Determining the offensive im-
pact of service quality parallels the age-old search for the
advertising-sales connection. Service quality's
effects-
similar to advertising's
effects-are
cumulative, and there-
fore evidence of the link may develop slowly. And, similar
to advertising, service quality is one of many
variables-in-
cluding pricing, advertising, efficiency, and
image-that
si-
multaneously influence profits. Furthermore, spending on
service per se does not guarantee results, because strategy
and execution must both be considered.
On the other hand, evaluating the defensive impact of
service quality through customer retention promises to help
companies gauge the financial impact of service quality. The
relationship between retention and profits recently has been
estimated by a variety of researchers (e.g., Anderson and
Sullivan 1990; Fornell and Wernerfelt 1987, 1988; Reich-
held and Sasser 1990) and companies (e.g., IBM). If the re-
lationship between service quality and retention can be sim-
ilarly documented, the financial implications for a given
company or even a given service initiative can be calibrat-
ed. Zahorik and Rust (1992) distinguish among five tasks
that must be completed to model the impact of service on
profits:
(I)
identifying the key service attributes to include
in the model, (2) selecting the most important attributes, (3)
modeling the link between programs and attitudes, (4) mod-
eling behavioral response to service programs, and (5) mod-
eling the impact of service programs on profits.
The research we describe involves the first four tasks
that Zahorik and Rust (1992) propose and concentrates on
the fourth, namely, modeling behavioral response to quality
service. All four of these tasks are firmly in the domain of
marketing and the first three have been studied extensively
in the last decade (for a review, see Zahorik and Rust 1992).
In contrast, the fourth attribute, the impact of service quali-
ty on behavioral response, has been the subject of only a few
marketing studies to date (Boulding et al. 1993; Cronin and
Taylor 1992).
The underlying premise of our article is that if service
quality relates to retention
of
customers at the aggregate
(i.e., firm) level, as other research has suggested, then evi-
dence of its impact on customers' behavioral responses
should be detectable. The consequences of service-quality
perceptions on individual-level behavioral intentions can be
viewed as signals of retention or defection and are desirable
to monitor. With that in mind, our objectives are four-fold:
I. To summarize existing evidence about the behavioral con-
sequences of service quality at the individual customer
level.
2. To offer a conceptual model of the impact of service quality
on particular behaviors that signal whether customers re-
main with or defect from the company.
3. To report the results of an empirical study examining rela-
tionships between service quality and customers' behavioral
intentions.
32/
Journal
of
Marketing,
April
1996
4. To suggest a research agenda whereby information about in-
dividual-level behavioral consequences of service quality
can be monitored and linked to sales and customer-retention
data to provide ongoing evidence of the financial impact of
service quality.
In addressing these objectives, we provide a concise
synthesis of the extant literature on the subject and extend
the literature in three significant ways. First, our study in-
volves a comprehensive (multicompany/multi-industry) ex-
amination of service quality's impact at the individual-con-
sumer level rather than at the company/industry level, as is
the case in most previous studies. Second, in addition to ex-
amining the general relationship between service quality
and behavioral intentions, we explore changes in the
strength of this relationship that are due to potential moder-
ating effects of different levels of service relative to cus-
tomers' expectation levels. Third, we incorporate a more ex-
tensive multiple-item behavioral-intentions measure than
has been used in previous research and examine service
quality's impact on specific types of behavioral intentions.
Conceptual Framework and
Hypotheses
Background
Lowering customer defection rates can be profitable to com-
panies. In fact, research has shown that it is a more prof-
itable strategy than gaining market share or reducing costs.'
For example, in an empirical study linking customer satis-
faction to profits, Fornell and Wernerfelt (1987, 1988) ex-
amine the impact of complaint-handling programs on cus-
tomer retention and conclude that marketing resources are
better spent keeping existing customers than attracting new
ones. In support of this position, Reichheld and Sasser
(1990, p. 105) assert that customer defections have a
stronger impact on a company's profits than "scale, market
share, unit costs, and many other factors usually associated
with competitive advantage." For this reason, they extol the
benefits of zero customer defections as an overall company
performance standard:
Ultimately, defections should be a key performance mea-
sure for senior management and a fundamental component
of incentive systems. Managers should know the compa-
ny's defection rate, what happens to profits when the rate
moves up or down, and why defections occur (p.
III).
Research and company efforts to quantify the financial impact
of defection and retention have intensified in recent years.
Financial impact
of
defection. When customers are lost,
new ones must be attracted to replace them, and replacement
comes at a high cost. Capturing new customers is expensive
"This is not to say that companies should focus on customer re-
tention to the exclusion of strategies to attract new customers. For
instance, share-building strategies should be a high priority for
companies that are new entrants or operate in emerging markets.
However, for companies with an established customer base (espe-
cially in mature markets with entrenched competitors) the net re-
turn on investments could be much higher for retention strategies
than for strategies to attract new customers.

FIGURE 1
The Behavioral and Financial Consequences of Service Quality
r---------------------------------------------------------------------------,
I
I
I
LFocus
of
present study
- - - -
-~
Empirical links demonstrated in macro
studies
for it involves advertising, promotion, and sales costs, as
well as start-up operating expenses. New customers are
often unprofitable for a period of time after acquisition: In
the insurance industry, for example, the insurer typically
does not recover selling costs until the third or fourth year of
the relationship. Capturing customers from other companies
is also an expensive proposition: Anderson and Sullivan
(1990) find that a greater degree of service improvement is
necessary to make a customer switch from a competitor than
to retain a current customer.
Financial impact
of
retention. The longevity of a cus-
tomer's relationship favorably influences profitability. Cus-
tomers who remain with a firm for a period of years because
they are pleased with the service are more likely than short-
term customers to buy additional services and spread favor-
able word-of-mouth communication. The firm also may be
able to charge a higher price than other companies charge,
because these customers value maintaining the relationship.
The initial costs of attracting and establishing these cus-
tomers have already been absorbed and, due to experience-
curve effects, they often can be served more efficiently (Re-
ichheld and Sasser 1990). Rose (1990) supports this view,
contending that profit on credit card services purchased by a
ten-year customer is on average three times greater than for
a five-year customer.
Although the financial impacts of defection and reten-
tion have been studied at a macro level (i.e., company or in-
dustry level), the micro-level (i.e., individual-level) process-
es through which these impacts occur have not been well
understood. To attempt to fill this void, we develop and test
±1
Ongoing Revenue
Increased Spending
Price Premium
Referred Customers
::l
Decreased Spending
Lost Customers
Costs to Attract New
Customers
a conceptual model focusing on individual-level behavioral
consequences of service quality.
A Model
of
the Behavioral Consequences
of
Service Quality
Figure 1 is a conceptual model that depicts the behavioral
consequences of service quality as intervening variables be-
tween service quality and the financial gains or losses from
retention or defection. The left portion of the model is at the
level of the individual customer and proposes that service
quality and behavioral intentions are related and, thus, that
service quality is a determinant of whether a customer ulti-
mately remains with or defects from a company.
Starting on the left, the model begins with a customer's
assessment of service quality and posits that when service
quality assessments are high, the customer's behavioral inten-
tions are favorable, which strengthens his or her relationship
with the company. When service quality assessments are low,
the customer's behavioral intentions are unfavorable and the
relationship is more likely to be weakened. Behavioral inten-
tions can be viewed as indicators that signal whether cus-
tomers will remain with or defect from the company.
Some of the links in Figure I (shown by dotted arrows)
have been demonstrated empirically in several aggregate-
level studies using overall multicompany analysis (e.g.,
Buzzell and Gale 1987; Gale 1992; Reichheld and Sasser
1990). However, the mediating roles of behavioral inten-
tions and actual behavior on the relationship between ser-
vice quality and financial performance are not well under-
stood, especially at the individual-customer level. We at-
tempt to add to our knowledge in this regard by undertaking
Service
Quality
I33

an in-depth conceptual and empirical examination of the
first link in the sequence of effects posited in Figure I. As
we discuss in subsequent sections, multiple measures of ser-
vice quality and behavioral intentions were operationalized
and used in surveys of customers from four different com-
panies. For ease of exposition in this section, the dependent
construct is split broadly into favorable and unfavorable be-
havioral intentions.
Favorable behavioral intentions. Certain behaviors sig-
nal that customers are forging bonds with a company. When
customers praise the firm, express preference for the compa-
ny over others, increase the volume of their purchases, or
agreeably pay a price premium, they are indicating behav-
iorally that they are bonding with the company. Recent re-
search offers some evidence that customer satisfaction and/or
service-quality perceptions positively affect intentions to be-
have in these ways. However, most of the research opera-
tionalizes behavioral intentions in a unidimensional way
rather than delineate specific types of behavior. For example,
Cronin and Taylor (1992), using a single-item purchase-in-
tention scale, find a positive correlation with service quality
and customer satisfaction. Anderson and Sullivan (1990), in
analyzing data from a study of customer satisfaction among
Swedish consumers, find that stated repurchase intention is
strongly related to stated satisfaction across product cate-
gories. A study conducted by Woodside, Frey, and Daly
(1989) uncovers a significant association between overall pa-
tient satisfaction and intent to choose the hospital again.
Several studies have examined the association between
service quality and more specific behavioral intentions. In
previous studies (see Parasuraman, Berry, and Zeithaml
1991a; Parasuraman, Zeithaml, and Berry 1988), we find a
positive and significant relationship between customers'
perceptions of service quality and their willingness to rec-
ommend the company. Boulding and colleagues (1993), in
one of two studies they conducted, find a positive correla-
tion between service quality and a 2-item measure of repur-
chase intentions and willingness to recommend. In a second
study involving university students, they find strong links
between service quality and behavioral intentions that are of
strategic importance to the school, including saying positive
things about the school, planning to contribute money to the
class pledge on graduation, and planning to recommend the
school to employers as a place from which to recruit.
Individual companies are also monitoring the impact of
service quality on selected behavioral intentions. For exam-
ple, Northwest Airlines found that the preference index (i.e.,
the preference for Northwest Airlines as the airline passengers
like to fly) increased substantially in 1992, compared to 1991,
following a major company effort to improve service. As
measured in random surveys, preference rose in Minneapolis
(from 70% to 75%), Detroit (from 49% to 59%), and Mem-
phis (from 48% to 63%) (Executive Report on Customer Sat-
isfaction 1992). Toyota found that intent to repurchase a Toy-
ota automobile increased from a base of 37% to 45% with a
positive sales experience, from 37% to 79% with a positive
service experience, and from 37% to 91% with both positive
sales and service experiences (McLaughlin 1993).
34/
Journal
of
Marketing,
April
1996
By integrating research findings and anecdotal evidence,
a list of specific indicators of favorable behavioral inten-
tions can be compiled. These include saying positive things
about the company to others (Boulding et al. 1993), recom-
mending the company or service to others (Parasuraman,
Berry, and Zeithaml 1991a; Parasuraman, Zeithaml, and
Berry 1988; Reichheld and Sasser 1990), paying a price pre-
mium to the company, and remaining loyal to the company
(LaBarbera and Mazursky 1983; Newman and Werbel 1973;
Rust and Zahorik 1993). Loyalty may be manifested in mul-
tiple ways; for example, by expressing apreference for a
company over others, by continuing to purchase from it, or
by increasing business with it in the future.
Unfavorable behavioral intentions. Customers perceiv-
ing service performance to be inferior are likely to exhibit
behaviors signaling they are poised to leave the company or
spend less with the company. These behaviors include com-
plaining, which is viewed by many researchers as a combi-
nation of negative responses that stem from dissatisfaction
and predict or accompany defection (Richins 1983;
Scaglione 1988).
Complaining behavior itself is conceptualized as multi-
faceted. According to Singh (1988), dissatisfaction leads to
consumer-complaining behavior (CCB) that is manifested in
voice responses (such as seeking redress from the seller),
private responses (negative word-of-mouth communica-
tion), or third-party responses (taking legal action). His
three-dimensional typology of complaining behavior,
founded on the object of the complaints (seller, friend, third
party), is statistically superior to previous models of CCB.
Maute and Forrester (1993) find strong support for a three-
way classification of dissatisfaction responses based on Hir-
shman's (1970) exit, voice, and loyalty responses (loyalty
being the decision to remain with the company despite dis-
satisfaction). Solnick and Hemenway (1992) observe that
though voice and exit (in their view the two main behavioral
manifestations of dissatisfaction) can be substitutes for each
other, they often occur together. In the context of a health
maintenance organization, they find that complaining cus-
tomers were four and one-half times more likely to leave the
plan voluntarily than noncomplaining customers.
Specific indicators of unfavorable behavioral intentions
suggested by the preceding discussion include different
types of complaining (e.g., complaining to friends or exter-
nal agencies) and contemplation of switching to competi-
tors. Another indicator of eventual defection is a decrease in
the amount of business a customer does with a company.
Differential impact
of
service-quality levels. Although
superior service is likely to foster favorable behaviors and
reduce the likelihood of unfavorable behaviors, an impor-
tant unresolved issue is the service-quality level that com-
panies must target to have the desired impact on behaviors.
How much service quality is enough to retain customers? Is
there a level of service beyond which there are diminishing
returns in terms of strengthening behavioral intentions?
Does the degree of association between service quality and
behavioral intentions change at different quality levels?
Little published evidence directly addresses these ques-
tions. However, a study by Gale (1992), which quantitative-

ly assesses the relationship between level of service quality
and willingness to purchase at AT&T, offers some indirect
insight. Of AT&T's customers who rated the company's
overall quality as excellent, over 90% expressed willingness
to purchase from AT&T again. For customers rating the ser-
vice as good, fair, or poor, the percentages decreased to
60%, 17%, and 0%, respectively. According to these data,
willingness to repurchase increased at a steeper rate (i.e., by
43%) as the service-quality rating improved from fair to
good than when it went from poor to fair (17%) or from
good to excellent (30%). These results suggest that the im-
pact of service quality on willingness to repurchase is most
pronounced in some intermediate level of service quality.
Coyne (1989, p. 73), however, makes the opposite predic-
tion on the basis of research relating to the impact of customer
satisfaction with service in a consumer-durable context:
There appear to be thresholds of service for affecting cus-
tomer behavior.... When satisfaction rose above a certain
threshold, repurchase loyalty climbed rapidly. In contrast,
when satisfaction fell below a different threshold, cus-
tomer loyalty declined equally rapidly. However, between
these thresholds, loyalty was relatively flat. I believe this
twin threshold framework applies to a wide variety
of
ser-
vice situations.
A similar categorization of service levels follows one de-
finition of service quality in the literature-the extent to
which a service meets or exceeds customer expectations
(Parasuraman, Zeithaml, and Berry 1985,
1988)-and
from
recent research explicating the expectations construct as two
levels of expectations (Zeithaml, Berry, and Parasuraman
1993). The first level is desired service, which is the level of
service the customer hopes to receive, consisting of a blend of
what the customer believes can and should be delivered. The
second, lower level of expectations is adequate service,
which is the level of service the customer will accept. Ade-
quate service is the minimum service a company can provide
and still hope to meet customers' basic needs. A zone of tol-
erance, bounded on the lower end by adequate service and on
the upper end by desired service, captures the range of service
within which a company is meeting customer expectations.
Although the zone-of-tolerance framework seems struc-
turally similar to Coyne's (1989) twin-threshold framework,
the managerial implications of the two frameworks are dif-
ferent. Coyne, invoking the flat satisfaction-loyalty relation-
ship he hypothesizes between the two thresholds, suggests
that unless a company already has a strong reputation for
service, it may not benefit by improving service much be-
yond the lower threshold:
"If
a company is already above
the minimum acceptable threshold, but nearer the lower end
of the service satisfaction band, investments to incremental-
ly change position may not be warranted" (p. 75). In con-
trast, we have argued previously (see Parasuraman, Berry,
and Zeithaml 1991b, p. 47) that firms operating within the
zone of tolerance, while possibly enjoying competitive ad-
vantage, should continue to improve service, even to the
point of exceeding the desired service level:
'To
develop a
true customer franchise-unwavering customer
loyalty-
firms must exceed not only the adequate service level but
also the desired service level." Although we do not refer to
the slope of the service performance-loyalty relationship,
our prior recommendation implies an upward-sloping
(rather than flat) relationship within the zone of tolerance.
Available evidence suggests that the sensitivity of behav-
ioral intentions to changes in service quality is likely to vary
from below to within to above the zone of tolerance, though
there is no consensus about the nature of this variation across
the three regions of quality. A key empirical question is
whether the relationship between behavioral intentions and
service quality is flat or upward sloping within the zone of
tolerance and, if it is upward sloping, whether or not it is
steeper than the relationship below and above the zone.
The discussion in the preceding sections implies that,
though service quality is positively associated with favor-
able behavioral intentions and negatively related to unfavor-
able behavioral intentions, customers' perceptions of the
service relative to their adequate and desired service levels
moderate these associations. More formally, we posit,
H(
The service quality-behavioral intentions relationship (a)
is positive (negative) for favorable (unfavorable) behav-
ioral intentions and (b) has a different slope below and
above the zone of tolerance relative to within it.
Impact
of
problem experience
and
resolution. Another
aspect of service provision that can influence behavioral in-
tentions involves the problem experience of customers.
When customers encounter service problems, these experi-
ences are likely to affect behavioral intentions adversely.
However, the impact of problem resolution on customers'
intentions is less clear. One view, based primarily on anec-
dotal evidence, is that superior problem resolution forges
stronger bonds between customers and the company than
would exist had no service problem occurred. For example,
J. W. Marriott, chief executive officer of the Marriott hotel
chain, states: "Sometimes those [disgruntled] customers
whom you make that extra effort to gain back become the
most loyal customers that you have" (Lovelock 1994, p.
214). The reasoning underlying this view seems to be that a
service problem gives a company the opportunity to demon-
strate its commitment to customer service through excellent
recovery efforts. On the other hand, empirical evidence sug-
gests that service failures may weaken the customer-compa-
ny bond even when the problem is resolved satisfactorily
(Bolton and Drew 1992). We report (see Zeithaml, Parasur-
aman, and Berry 1990) that customers who experienced no
recent service problem with a company have significantly
better service-quality perceptions than customers who expe-
rienced a recent service problem that was satisfactorily re-
solved. A plausible explanation for this finding is that satis-
factory problem-resolution service, though perhaps pleasing
to customers, does not cause them to forget the service fail-
ure. And the memory of the failed service negatively affects
customers' overall perception of the company's service. The
existing empirical evidence on this question leads to our
second hypothesis:
Hz: Favorable (unfavorable) behavioral intentions are (a) high-
est (lowest) for customers experiencing no service prob-
lem; (b) next highest (lowest) for customers experiencing
service problems that are resolved, and (c) lowest (highest)
Service
Quality
/35

for customers experiencing service problems that are not
resolved.
In summary, the first hypothesis suggests a positive
(negative) relationship between service quality and favor-
able (unfavorable) behavioral intentions, the strength of
which is different below and above the zone of tolerance rel-
ative to that within it.
HI,
along with H2o is depicted in Fig-
ure 2, which details the portion of the behavioral conse-
quences model on which we focus in the present study.
Methodology
Sample Design
and
Mail Survey
Four companies that provide services to end or business cus-
tomers were sponsors of the research study. Questionnaires
were mailed to business customers of a computer manufac-
turer, as well as to end customers of a retail chain, automo-
bile insurer, and life insurer. The sponsoring companies gen-
erated mailing lists from their current customer bases. The
retail chain, automobile insurer, and life insurer each pro-
vided random samples of 2400 customers. The computer
manufacturer provided a larger random sample of 5270 cus-
tomers, because it wanted to conduct its own detailed, seg-
ment-by-segment analysis following the completion of the
main study. A total of 12,470 questionnaires were mailed.
Surveys were mailed with a cover letter and postage-
paid return envelope to all customers in the sample. The
cover letter appeared on company letterhead and was signed
by a senior company official. Respondents were requested
to return completed questionnaires to a marketing research
company hired to assist with data collection and coding. A
reminder postcard was sent two weeks after mailing the
questionnaires.
Overall response rate was 25% (3069 questionnaires).
Company-specific response rates were 30% (1566 question-
naires) for the computer manufacturer; 22% (522 question-
naires) for the retail chain; 24% (568 questionnaires) for the
automobile insurer; and 17% (413 questionnaires) for the
life insurer. Demographic profiles of the respondent samples
were reviewed by managers in the respective companies and
considered to be representative of their customer bases.
Survey Instrument
Operationalization
of
service quality. Several measures
of service quality were included in the questionnaire: (1) an
overall, single-item rating scale with anchors at 1 (extreme-
ly poor) and 9 (extremely good); (2) a multiple-item scale of
perceived service from an expanded version of the
SERVQUAL scale we originally developed (see Parasura-
man, Zeithaml, and Berry 1988) and later refined (see Para-
suraman, Berry, and ZeithamI199Ia); and (3) two categori-
FIGURE 2
Hypothesized
Effects
of
Service
Quality
on Behavioral
Intentions
Performance
Relative to Adequate
and
Desired Service
Problem
Resolved?
Perceived Service
Performance
361
Journal
of
Marketing,
April
1996
Favorable
• Say positive things
•Recommend company
•Remain loyal to company
•Spend more with company
· Pay price premium
Unfavorable
· Say negative things
·Switch to another company
•Complain to external agencies
· Do less business with company

cal questions to measure whether respondents had experi-
enced a recent service problem with the company and, if so,
whether the problem was resolved to their satisfaction.
The second measure (i.e., the revised SERVQUAL bat-
tery) represented the service dimensions of reliability (five
items), responsiveness (three items), assurance (four items),
empathy (four items), and tangibles (five items). Consistent
with the expanded conceptualization of customers' service
expectations (Zeithaml, Berry, and Parasuraman 1993), re-
spondents were asked to indicate their adequate- and de-
sired-service levels in addition to their perceptions of each
SERVQUAL item. Thus, separate ratings of adequate, de-
sired, and perceived service were obtained on three 9-point
scales (I =low, 9 =high) arranged as three adjacent
columns next to the SERVQUAL battery on the
questionnaire.I
The questionnaire containing the SERVQUAL battery
with the three columns of ratings used in this study was one
of three different questionnaire formats evaluated in a larg-
er methodological study (Parasuraman, Zeithaml, and Berry
1994a). As such, the adequate-, desired-, and perceived-ser-
vice scores used in the present study were based on a partial
sample from each company (the other two questionnaire for-
mats did not produce separate scores for these variables). A
total of 1009 questionnaires contained scores for the ade-
quate-, desired-, and perceived-service variables: 498 from
the computer manufacturer, 188 from the retail chain, 191
from the automobile insurer, and 132 from the life insurer.
All three questionnaire formats contained measures for the
remaining study variables (overall service quality, behav-
ioral intentions, and incidence of service-problem experi-
ence and satisfactory problem resolution). Therefore, scores
for these variables were based on the full sample.
Operationalization
of
behavioral intentions. Previous re-
search has not captured the full range of potential behaviors
likely to be triggered by service quality. Cronin and Taylor
(1992) focus solely on purchase intentions and measure the
construct with a single-item scale. In the first of two studies
by Boulding and colleagues (1993), repurchase intentions
and willingness to recommend were the only two behavioral
intentions measured. In the second study, involving service
quality of an educational institution, they used a 6-item scale
comprised largely of education-specific items, such as intent
to contribute money to the class pledge and intent to recom-
mend the school to employers as a place to recruit.
A 13-item battery was developed to gauge a wider range
of behavioral intentions than have been suggested in the lit-
erature or by anecdotal evidence from companies.3 This bat-
tery included items to capture several facets of behavioral
intentions not incorporated in previous service-quality stud-
ies: likelihood of paying a price premium and remaining
loyal to a company even when its prices go up, intent to do
more business with the firm in the future, and complaint in-
2The adequate- and desired-level service ratings were used in
defining the lower and upper boundaries of the zone of tolerance to
verify the differential slopes predicted by H1b for the quality-in-
tentions relationship.
3A copy of the instrument containing the behavioral-intentions
and service-quality questions can be obtained from the third author.
tentions when service problems occur. The 13 items were
grouped into four a priori categories: word-of-mouth com-
munications, purchase intentions, price sensitivity, and com-
plaining behavior. (These groupings were not made known
to respondents.) The last two categories contained items not
included in prior service-quality research. Each item was ac-
companied by a 7-point likelihood scale
(I
=not at all like-
ly, and 7 =extremely likely).
Analyses, Results, and Discussion
Dimensions
of
Behavioral Intentions
Factor analysis of the behavioral-intentions battery was con-
ducted to examine the dimensionality of the items. Because
the battery was designed to represent four categories of be-
havioral intentions, a four-factor solution was obtained sep-
arately for each company and subjected to oblique rotation
to allow for potential correlation among the categories. The
item clusters implied by the factor loadings differed from
the a priori clusters and varied somewhat across the four
companies. The general patterns of loadings suggested that
a five-factor solution may help reconcile these differences.
A five-factor solution produced an unambiguous factor pat-
tern that was consistent across all companies. This consis-
tent pattern suggested a reconfiguration of the 13 items into
five dimensions: loyalty to company (loyalty), propensity to
switch (switch), willingness to pay more (pay more), exter-
nal response to problem (external response), and internal re-
sponse to problem (internal response). In Table I, we present
the reconfigured behavioral-intentions battery, and in Table
2, we present the factor-loading matrices supporting it,
along with reliability coefficients for its multiple-item
components.
Of the five factors, loyalty (with five items) and pay
more (with two items) exhibit consistent patterns of load-
ings across the four companies. Switch (with two items) and
external response (with three items) also display a moderate
to high degree of uniformity in factor loadings. The final di-
mension, internal response, contains just one item that loads
on the fifth factor.
The 5-item loyalty scale has excellent internal consis-
tency, which is evidenced by alphas ranging from .93 to .94
across the four companies. The 3-item external-response
scale has alphas of at least .6, with the values in two of the
four companies exceeding the threshold of .7 that Nunnally
(1978) suggested. The 2-item scales measuring switch and
pay more have somewhat weaker alphas, with several values
falling below .6, perhaps because of too few items in the re-
configured factors. In general, the alpha score for loyalty is
high, particularly for an early study. The alpha scores for the
other three factors with multiple items range from adequate
to weak, indicating the need to add items to the scale in fur-
ther research.
Although the factor structure of the behavioral-inten-
tions battery differs somewhat from the a priori specifica-
tion, the loadings support the dichotomy in behavioral in-
tentions of favorable and unfavorable categories. The largest
factor, loyalty, contains five favorable behavioral-intentions
items: saying positive things about the company, recom-
Service
Quality
/37

Behavioral-
Intentions
Dimension
Loyalty
Switch
Pay More
External Response
Internal Response
Item
Label
11
12
13
14
15
16
17
18
19
110
111
112
113
TABLE 1
Behavioral-Intentions Batterya
Item Wording
Say positive things about XYZ to other people.
Recommend XYZ to someone who seeks your advice.
Encourage friends and relatives to do business with XYZ.
Consider XYZ your first choice to buy
---
services.
Do more business with XYZ in the next few years.
Do less business with XYZ in the next few years
(-).
Take some of your business to a competitor that offers better prices
(-).
Continue to do business with XYZ if its prices increase somewhat.
Pay a higher price than competitors charge for the benefits you currently
receive from XYZ.
Switch to a competitor if you experience a problem with XYZ's service.
Complain to other customers if you experience a problem with XYZ's service.
Complain to external agencies, such as the Better Business Bureau, if you
experience a problem with XYZ's service.
Complain to XYZ's employees if you experience a problem with XYZ's service.
aThe items were grouped as follows in the a priori categorization of the battery: Word-of-Mouth
Communications-I
1,
12,
13;
Purchase Inten-
tions -
14,
IS,
16;
Price Sensitivity -
17,
18, 19;
Complaining Behavior -
110,
111,
112,
113.
Each item was accompanied by a 7-point likelihood
scale (1 =not at all likely and 7 =extremely likely). Items identified with a -:» were reverse scored.
mending the company to someone who seeks advice, en-
couraging friends and relatives to do business with the com-
pany, considering the company the first choice from which
to buy services, and doing more business with the company
in the next few years. Pay more contains two favorable
items: continuing to do business with the company even if
its prices increase somewhat and paying a higher price than
competitors charge for the benefits currently received from
the company.
The second and fourth factors comprise all unfavorable
behavioral-intentions items. Switch contains two of these:
doing less business with the company in the next few years
and taking some business to a competitor that offers better
prices. External response includes items that relate to expe-
riencing a service problem: switching to a competitor, com-
plaining to other customers, and complaining to external
agencies such as the Better Business Bureau.
The interpretation of internal response, the fifth factor
with one item (complaining to the company's employees if
a service problem is experienced), is unclear. Customers
more favorably disposed toward a company may be more
likely to complain internally to give the company a "second
chance." Conversely, disgruntled customers with an unfa-
vorable image of the company may be more likely to com-
plain internally to vent their frustrations. The equivocal in-
terpretation of this factor and its being represented by just
one item undermine its meaningfulness on conceptual and
psychometric grounds. As such, we deleted this single-item
measure from all subsequent analyses.
Relationship Between Service Quality
and
Behavioral Intentions
The first hypothesis predicted a positive (negative) quality-
intentions relationship for favorable (unfavorable) behav-
ioral intentions, with different slopes below and above the
381
Journal
of
Marketing,
April
1996
zone of tolerance relative to within it. This hypothesis was
tested by using multiple regression analysis to examine si-
multaneously
(I)
whether the slope of the relationship with-
in the zone of tolerance was significantly different from zero
and (2) whether this slope differed significantly from the
slopes below and above the zone of tolerance. In accordance
with procedures discussed by Cohen and Cohen (1983,
Chapter 8) for conducting this type of analysis, the follow-
ing regression equation was estimated:
(I)
Y=Bo+Bd,d j+
Bdzd
z+BIX +Bzd,X +B3d zX +E,
where
Y=behavioral-intentions score;
X=service-quality score;
dl=dummy variable with a value
of
Iif the perceived service is
below the zone of tolerance, 0 otherwise;
dz=dummy variable with a value of Iif the perceived service is
above the zone of tolerance, 0 otherwise;
Bs =unstandardired regression coefficients; and
E=error term.
The coefficients in Equation I that are relevant for examin-
ing the first hypothesis are B I,B2,and B3•Specifically, B,
represents the slope of the quality-intentions relationship
within the zone of tolerance, whereas B2and B3represent
changes in B, below and above the zone of tolerance, re-
spectively. Thus, B I+B2represents the slope below the
zone, and B, +B3represents the slope above the zone.
Service quality (the key independent variable X) was
operationalized in two ways: as the rating on the 9-point
overall-quality (OQ) scale and as a weighted-average per-
ceived performance (WP) score across the SERVQUAL di-
mensions. Of late there has been debate in the literature
about the most appropriate way to operationalize service
quality (cf. Brown, Churchill, and Peter 1993; Cronin and

TABLE 2
Factor Loading Matrices and Reliability Coefficients (Alphas) for Behavioral-Intentions Dimensionsa
Computer Manufacturer Retail Chain Automobile Insurer Life Insurer Combined Sample
B-Iltemsb
F1
F2
F3 F4 F5
F1
F2 F3
F4
F5
F1
F2
F3 F4
F5
F1
F2
F3 F4 F5
F1
F2 F3 F4 F5
Loyalty [a] [.93]
[.94] [.94]
[.93]
[.94]
11
93 --- -94 ----91 --- - 97 -- - -96
12
97 - - - - 94 - - - - 94 ----95 ----95
13
93 - - - - 93 --- - 95 - - - -94 ----94
14
65 -- - -79 - - - - 87 - - - - 87 - - - - 79
15
63 -- - -83 - - - - 78 - - - -62 --32 -73
Switch [a] [.67] [.53] [.63]
[049]
[.61]
16
-71 - - - - 73 - - - - 90 - - - - 95 - - - - 72
17
-83 - - - - 85 - - - - 74 - - - - - - 70 - - 83
Pay More [a] [.73]
[.60]
[.68] [.52] [.69]
18
- - 79 - - - -75 - - - - 77 - - - -70 - - --75
19
- - 88 --- - 89 - - - - 94 - - --93 - - --92
Extemal Response
[a]
[.60]
[.67] [.76] [.77]
[.70]
110
- - -66 - - - - 88 - - - - 83 ----82 --- - 74
111
- - -71 - - - - 82 - - - - 81 ----76 - - - - 78
112
---84 - - - - 33 60 - - - 76 -- - -61 45 - - - 79
Intemal Response [a] HH H H H
113
- - - - 99 -- - -95 - - - - 99 - - - - 85 - - - - 99
aNumbers within brackets are reliability coefficients. The other numbers are factor loadings obtained after oblique rotation of the initial solutions (all loadings have been multiplied by 100). Load-
ings of less than .3 have been omitted. The total variance extracted by the five factors is 77%, 79%, 80%, and 78%, and the average interfactor correlation is .23, .22, .21, and .14 for the com-
en puter manufacturer, retail chain, automobile insurer, and life insurer, respectively.
CD
bBehavioral-intentions labels
11
through 113correspond to those of the items listed in Table 1.
...
<
c:;"
CD
0
c:
e!.
~
-
w
CD

Taylor 1992; Parasuraman, Berry, and Zeithaml 1993; Para-
suraman, Zeithaml, and Berry 1994b; Teas 1993). The cen-
tral issue in this debate is whether service quality should be
measured as the difference between customers' perceptions
and expectations ratings or simply as the perceptions rat-
ings. Although this issue continues to be debated, there is
some agreement that a study's purpose may influence the
choice of which measure to use: The perceptions-only oper-
ationalization is appropriate if the primary purpose of mea-
suring service quality is to attempt to explain the variance in
some dependent construct; the perceptions-minus-expecta-
tions difference-score measure is appropriate if the primary
purpose is to diagnose accurately service shortfalls (Para-
suraman, Zeithaml, and Berry 1994a). The purpose of our
present study is the former. Moreover, as we discuss subse-
quently, the two expectations measures (i.e., the adequate-
and desired-service levels) were independently incorporated
into the analysis to operationalize the two dummy variables
d( and dz.Therefore, the ratings from the SERVQUAL por-
tion of the survey were used to operationalize service quali-
ty as weighted-average performance scores, rather than dif-
ference scores.
To determine WP, a perceived performance rating was
first computed for each SERVQUAL dimension by averag-
ing the ratings on the items forming the dimension. (The co-
efficient alpha values for reliability [five items], responsive-
ness [three items], assurance [four items], empathy [four
items] and tangibles [five items] ranged from .80 to .96
across the four samples.) To obtain the WP score, the aver-
age performance ratings for the dimensions were then
weighted by the relative importance of the dimensions. To
measure the relative importance of the five dimensions, re-
spondents were asked to allocate 100 points among the di-
mensions according to how important each dimension was
to them in evaluating a company's service. The relative
points allocated to the dimensions were used as weights in
computing the WP score.
The dummy variables d( and dzwere operationalized by
comparing each respondent's WP score with his or her
weighted-average adequate- and desired-service scores
(computed using a procedure similar to that used in deter-
mining WP). The d( value was I if WP was less than the
weighted-average adequate-service score, 0 otherwise. The
dzvalue was 1 if WP was greater than the weighted-average
desired-service score, 0 otherwise.
The regression analysis was performed separately for
the four companies, as well as for the combined sample. In
each instance, two equations were estimated for each be-
havioral-intentions dimension: one using WP scores and the
second using OQ scores as values for the independent vari-
able X. The average score across items comprising the be-
havioral-intentions dimension represented the dependent
variable Y. In Table 3, we summarize the regression-analy-
sis results pertaining to the first hypothesis.
The regression coefficients in the first two columns of
Table 3 (B Ivalues) offer strong support for the hypothesized
quality-intentions links within the zone of tolerance. The co-
efficients for WP and OQ are all in the hypothesized direc-
tions-positive
for loyalty and pay more and negative for
40I
Journal
of
Marketing,
April
1996
switch and external
response-and,
with few exceptions, are
statistically significant at p<.01. The pattern of BIvalues
across the four companies suggests that the effects are gen-
erally stronger for loyalty and switch than for pay more and
external response. The results for the combined sample pro-
vide additional insight into the relative influences of service
quality on the four behavioral-intentions dimensions: the
strongest effects of both WP and OQ are on loyalty (.70 and
.55), followed by switch (-.67 and -.47), pay more (.43 and
.37), and external response (-.28 and -.21) in that order.
The regression coefficients in Table 3 in the columns for
Bzand B3values pertain to the differential effects predicted
by H ( (the statistically significant Bzand B3coefficients are
boldfaced). For each behavioral-intentions dimension with-
in a given company, the Bzcoefficients for WP and OQ have
the same sign except in a few instances. Similarly, the signs
of the B3coefficients are identical for WP and OQ. The sta-
bility in the signs of the slope-change coefficients across
two different service-quality measures is encouraging in
terms of drawing inferences about the direction of changes
in the quality-intentions link below and above the zone of
tolerance. However, support for the strength of these
changes is mixed, as is evidenced by the pattern of statisti-
cal significance of these coefficients. Therefore, based on
the presence of significant coefficients for at least one of the
two service-quality measures (WP and OQ), only the fol-
lowing inferences seem warranted.
In the computer-manufacturer sample, the quality-inten-
tions relationship for loyalty and switch is flatter above the
zone of tolerance (implying diminished sensitivity to quali-
ty improvements beyond the desired-service level), but is
unchanged below the zone of tolerance. The relationship for
pay more is flatter both below and above the zone. In the re-
tail-chain sample, the relationship for loyalty, switch, and
external response is flatter below the zone of tolerance but
remains unchanged above the zone of tolerance (implying
undiminished returns for quality beyond the desired-service
level). In the automobile-insurer sample, the relationship for
loyalty is steeper below the zone of tolerance but remains
unchanged above the zone. However, the relationship for
switch is flatter below the zone and considerably steeper
above the zone, which implies that there are increasing pay-
offs as service improves from below to within to above the
zone. The relationship for external response in the automo-
bile-insurer sample is similar to that in the retail-chain sam-
ple (i.e., flatter below the zone but unchanged above it). All
of the slope-change coefficients in the life-insurer sample
are nonsignificant. This lack of significance may be due to
insufficient data points below and above the
zone-only
15
respondents in the life-insurer sample had WP scores below,
and only 8 had WP scores above the zone. A similar defi-
ciency may account for the lack of significance of any of the
B3coefficients in the retail-chain sample; only 16 respon-
dents had WP scores above the zone.
In the combined sample, the quality-intentions relation-
ship for loyalty and switch is flatter below but remains un-
changed above the zone. Thus, exceeding the adequate-ser-
vice threshold can sharply increase the payoffs (in terms of
fostering customer loyalty and curtailing propensity to

TABLE 3
Regression Analysis Results
Change in
Slope:
aThe 61 values are significant at p<.01 unless otherwise stated.
bThe adjusted A-squared values are for the regression model specified in Equation 1; the values are significant at p<.01 unless otherwise stated.
Slope
Within
Zone
of
Tolerance (B1)8
Adjusted
R-squared
Valuesb
WP OQ
.37 .41
.14 .15
.16 .19
.08 .07
.39 .46
.20 .15
.19 .17
.20 .19
.50 .47
.28 .28
.11 .15
.20 .13
.68 .62
.28 .30
.08(p<.1)
.12 (p <.05)
.03 (ns) .00 (ns)
.45 .48
.21 .20
.16 .18
.11 .10
OQ
.12 (ns)
.03 (ns)
.15 (ns)
.45 (ns)
.20 (ns)
-.23
(p <.1)
.36 (p <.1)
-.52
(p <.01)
-.23
(ns)
.29 (ns)
-.28
(ns)
1.27 (ns)
1.12 (ns)
.62 (ns)
-.01 (ns)
-.69
(p <.05)
-.13
(ns)
-.15
(ns)
.05 (ns)
-.34
(p <.05)
Above
Zone
of
Tolerance (B3)
.52 (ns)
.92 (ns)
.25 (ns)
.16 (ns)
WP
.21 (ns)
.15 (ns)
.65 (ns)
.22 (ns)
-.29
(ns)
.74 (ns)
-.85
(ns)
-.15
(ns)
.33 (ns)
-.29
(ns)
-.15
(ns)
-.73(p<.1)
-.14
(ns)
-.10
(ns)
.15 (ns)
-.25
(ns)
OQ
.29 ip «.05)
.06 (ns)
.52 (p <.05)
.25 (p <.1)
.24 (ns)
.05 (ns)
.03 (ns)
-.09
(ns)
-.02
(ns)
.08 (ns)
-.25
(p <.05)
-.11 (ns)
.12 (ns)
.02 (ns)
.01 (ns)
.16 (p <.05)
-.07
(ns)
.14 (ns)
.13 (ns)
-.10
(ns)
Below
Zone
of
Tolerance (B2)
-.09
(ns)
.23 (ns)
-.23
(p <.1)
.43 ip «.05)
.55 (p <.1)
WP
-.08
(ns)
.06 (ns)
.15 (ns)
-.33
(p <.1)
.46 w«.05)
-.22
(ns)
-.08
(ns)
.45 (p <.1)
-.15
(ns)
.10 (ns)
.14 (ns)
-.05
(ns)
-.12
(p <.1)
.35 (p <.01)
-.10
(ns)
OQ
-.12
(ns)
.56
-.33
.18 (ns)
.58
-.44
.54
-.25
(p <.05)
.49
-.63
.32
-.14
(ns)
.52
-.37
.38
-.57
.55
-.47
.37
WP
.70
-.67
.43
-.2
-.57
.67
-.55
.56
-.32
(ns)
.78
-.87
.39
-.47
.78
-.69
.43
.72
-.45
.40
Independent
Variable (X)
Computer Manufacturer
Loyalty
Switch
Pay More
External
Response
Retail Chain
Loyalty
Switch
Pay More
External
Response
Automobile
Insurer
Loyalty
Switch
Pay More
External
Response
Life Insurer
Loyalty
Switch
Pay More
External
Response
All Companies
Loyalty
Switch
Pay More
External
Response
-.28
-.21
f
~.
o
c
~
~
-
"'"
...

TABLE 4
Mean Scores for Service Quality and Behavioral Intentions
Service Qualitya Behavioral Intentionsb
Company
Computer Manufacturer
Retail Chain
Automobile Insurer
Life Insurer
aMean scores on a 9-point scale.
bMean scores on a 7-point scale.
WP
7.3
6.5
7.8
7.6
OQ
6.8
6.2
7.3
7.0
Loyalty
5.3
4.9
5.6
5.0
Switch
3.8
3.9
3.1
3.1
Pay More
3.9
3.2
3.5
3.4
External
Response
3.6
4.3
4.2
3.8
switch). However, the combined-sample results for the pay
more dimension reveal considerable flattening of the quali-
ty-intentions relationship above the zone of tolerance. In
fact, the slope for the OQ-pay more relationship changes
from .37 below the desired-service level to just .03 (.37 -
.34) above. Thus, companies wishing to improve service be-
yond the desired-service level should do so cautiously and
cost-effectively, because recouping the added expense by
charging price premiums may not be a viable option. The
quality-intentions relationship for external
response-
which, as indicated by its BIcoefficients, is flatter within
the zone than for the other three
dimensions-remains
un-
changed below and above the zone as well. Thus, relative to
the other dimensions, external response appears much less
affected by changes in quality over a wide range.
The pattern of adjusted R-squared values in the last two
columns of Table 3 offer two noteworthy insights based on
the overall ability of service-quality-related variables (db
dz, and X) to explain the variation in scores on each behav-
ioral-intentions dimension. First, the relationship of quality
(both WP and OQ) with loyalty and switch is consistently
stronger in the two pure-service companies (automobile and
life insurers) than in the two product companies (computer
manufacturer and retail chain); however, the reverse is true
for the quality-pay more relationship: The relationship is
consistently stronger in the two product companies than in
the two pure-service companies (for additional analyses, see
the Appendix). Second, the quality-pay more relationship is
consistently weaker than the quality-loyalty relationship in
all four companies and the combined sample. We examine
the implications of these insights subsequently.
In Table 4, we summarize the mean scores for service
quality and behavioral intentions by company. An across-
company comparison of the mean-score patterns provides
additional support for inferring that service quality is asso-
ciated positively with favorable behavioral intentions and
negatively with unfavorable behavioral intentions. With few
exceptions, the better a company's service-quality scores,
the higher are its loyalty and pay more means and the lower
are its switch and external response means. To illustrate, the
retail chain's WP and OQ scores are considerably lower than
the corresponding scores for the automobile insurer. Match-
ing behavioral-intentions data show that the retail chain's
customers are less loyal and less willing to pay
more-and
42/
Journal
of
Marketing,
April
1996
more prone to switch and complain
externally-than
the au-
tomobile insurer's customers.
Impact
of
Service-Problem Experience
and
Resolution on Behavioral Intentions
H2predicts that customers experiencing no service problems
have the best behavioral-intentions scores (highest for fa-
vorable intentions and lowest for unfavorable
intentions-
H2a) , customers experiencing problems that were resolved
would have intermediate scores (H2b) , and customers with
unresolved service problems would have the worst scores
(H2e) . To test this hypothesis, the combined sample was clas-
sified into three groups of respondents: those experiencing
no recent service problems; those experiencing problems
that were resolved; and those experiencing problems that
were not resolved. Analysis of variance was conducted to
determine whether scores on each behavioral-intentions di-
mension differed across the groups. The F-values for all four
ANOVAs were significant at p<.001. Eight prespecifed
contrasts (first-group mean versus second-group mean and
second-group mean versus third-group mean for each of the
four behavioral-intentions dimensions) were also evaluated.
In Table 5, we present the group means and the significance
levels for the planned contrasts.
The alpha level for testing the significance of individual
contrasts was reduced by applying the Bonferroni correction
to ensure that the overall probability of Type I error across
all eight contrasts did not exceed .05 (for details, see foot-
note b in Table 5). The evidence in Table 5 fully supports the
second hypothesis for the loyalty, switch, and external re-
sponse dimensions, and partially supports it for the pay
more dimension. The findings clearly show that customers
experiencing no service problems have the strongest levels
of loyalty intentions and the weakest switch and external re-
sponse intentions. However, their pay more intentions are
not significantly higher than those of customers experienc-
ing service problems that were resolved satisfactorily.
Among customers experiencing recent service problems,
those receiving satisfactory resolution have significantly
higher loyalty and pay more intentions, and significantly
lower switch and external response intentions, than those
with unresolved problems. Thus, effective service recovery
significantly improves all facets of behavioral intentions.
However, with the possible exception of the pay more di-
mension, the improvements do not restore intentions to the

levels expressed by customers not experiencing service
problems. These results are consistent with those from a
study in which Bolton and Drew (\ 992) examine the impact
of problem experience and resolution on telephone cus-
tomers' evaluation of billing service: Customers rated the
service substantially lower if they had experienced a billing
problem, and the effect of satisfactorily resolving the prob-
lem did not completely offset its negative impact.
Discussion and Implications
We developed a conceptual model of the behavioral and fi-
nancial consequences of service quality (Figure I. A portion
of the
model-the
quality-intentions
link-was
empirically
examined at the individual-customer level in a multicompa-
ny context. Two distinctive features of the study's empirical
component were the development of a more extensive be-
havioral-intentions battery than has been used in previous
research and the investigation of changes in the quality-in-
tentions link at different service levels relative to customers'
expectations. The study's findings have important implica-
tions for researchers and managers.
Directions for Further Research
The distinctive aspects of the empirical study contribute
several new insights whose implications we subsequently
explore. However, our findings also reveal certain weak-
nesses with methodological implications. First, the behav-
ioral-intentions battery developed here, though more com-
prehensive than intentions scales used in previous studies,
needs further development. In particular, more items are
needed to strengthen the reliability of three of its compo-
nents, namely, switch, pay more, and external response.
With additional items, the scales should be reevaluated for
their psychometric properties. Consideration should also be
given to augmenting and including in the battery internal re-
sponse-the
component that was eliminated because it had
only one item subject to equivocal interpretation. As we pre-
viously mentioned, customers favorably disposed toward a
company may complain to give it a second chance, while
customers unfavorably disposed may also complain merely
to vent their frustrations. Therefore, in expanding this com-
ponent, it would be useful (from a diagnostic standpoint) to
add items that capture why customers are likely or unlikely
to complain. For example, respondents could be asked to
rate the likelihood of the following (on the same 7-point
scale used in the behavioral-intentions battery):
•
Complain
to XYZ's
employees
about a
service
problem
be-
cause I am
confident
they
will
resolve
the
problem.
•
Complain
to XYZ's
employees
about a
service
problem
to
help
relieve
myfrustration (reverse
scored).
Second, additional research is needed to examine further
our tentative insights pertaining to intercompany differences
in the quality-intentions relationship and the changes in its
slope below and above the zone of tolerance relative to
within it. Although the total sample size was large for each
company, the subsamples of respondents below and above
the zone were relatively small, and this possibly contributed
to the lack of significance of some of the slope-change co-
efficients. Obtaining larger samples of respondents below
and above the zone in further studies would facilitate a more
robust examination of changes in the quality-intentions rela-
tionship. One option for doing so is to select samples from
companies that are well known for their excellent (or poor)
service. Another option is to devise a suitable quota-sam-
pling procedure to ensure large enough subsamples below,
within, and above the zone. Multicompany research using
such sampling procedures is needed for more definitive con-
clusions about the intriguing differences uncovered in this
study, which concern changes in the quality-intentions link
within and across dimensions and companies.
In addition to addressing the previous issues, further re-
search should also focus on aspects of the conceptual model
not examined here. For example, the association between
behavioral intentions and remaining with or defecting from
TABLE 5
Mean Behavioral-Intentions Scores for
Respondents
Classified According to Service Problem
Experience
Significance Levels for
Mean
Scores for
Customers
Experiencinga
Planned
contrastse
Service Problems Service Problems
Behavioral- No Service That Were That Were Group 1Group 2
Intentions Problems Resolved Not Resolved Mean versus Mean versus
Dimension (Group 1; n=2153) (Group 2; n=455) (Group 3; n=346) Group 2Mean Group 3Mean
Loyalty 5.47 5.01 4.11 .000 .000
Switch 3.35 4.00 4.49 .000 .000
Pay More 3.76 3.63 3.11 .036 .000
External
Response 3.70 3.95 4.43 .000 .000
aThe behavioral-intentions scores are on a 7-point likelihood scale.
bThe reported significance levels are for one-tailed tests, because H2implies directional comparisons of group means. Because multiple con-
trasts were evaluated to test this hypothesis, the Bonferroni correction was applied to the customary alpha level of .05 to control the Type I
error rate. Specifically, the alpha level was lowered by a factor of eight (the total number of planned contrasts) to yield a critical alpha level of
.006 for testing the significance of each contrast (Myers 1979, pp. 298-300). At this reduced alpha level, seven of the eight planned contrasts
are significant; the sole exception is the Group 1 versus Group 2 contrast for the pay more dimension.
Service
Quality
I 43

the company merits study. Rust and Zahorik (1993)
sug~est
ways to investigate this link, including panel data, longitu-
dinal analysis with customers, and cross-sectional surveys
asking customers about previous and current providers.
~d
ditional cross-sectional research might ask customers to
10-
dicate not only their behavioral intentions but also their ac-
tual behaviors. For example, customers could be asked
whether they have said positive things about the company
(actual behaviors) instead of how likely they would be to
say positive things (behavioral intentions). Such research
also needs to be supplemented with longitudinal research to
verify the causal direction
of
the quality-intentions link.
Data from studies tracking service quality and behavioral in-
tentions over time can be analyzed to determine the impact
of service quality in a given period on behavioral intentions
in subsequent periods.
If
the longitudinal data set also contains information for
individual customers on variables such as purchase frequen-
cy and volume and new-customer referrals, the impact .of
service quality on actual behavior can be traced. Companies
that have information systems linking customer data and
purchase data could also examine increases or decreases in
spending that result from different levels
of
service quality.
This type of research would provide direct evidence
of
the
financial impact of service quality at the individual level.
An intriguing finding worthy
of
further research is the
pattern
of
across-company differences implied by the differ-
ences in the adjusted R-squared values for the various qual-
ity-intentions regression equations (Table 3). As previously
highlighted, the quality-intentions link for the loyalty and
switch dimensions is consistently stronger for the two pure-
service companies than for the two product companies,
whereas the reverse is true for pay more. Is it possible that
the role
of
service within a firm's total offering (i.e., core
versus supplemental component) is a plausible explanation
for this pattern of differences? Because service is all that a
pure-service provider, such as a life insurance company, de-
livers in exchange for customers' money, customers' com-
mitment to the company might be extremely responsive to
service-quality improvements; however, these customers'
willingness to pay more may not be as responsive, because
they may feel they have, in effect, already paid for high-
quality service. Alternatively, the pay more findings may
simply reflect customers' general reluctance to pay for in-
surance services and may not apply to pure services overall.
In contrast, because service is not the core
of
what a
product company sells to customers, their commitment to
the company may be less sensitive to changes in service
quality (especially if product quality is mediocre); however,
customers may be somewhat more willing to pay more for
better service, because they may consider service to be an
extra feature. To what extent and under what circumstances
are these speculations likely to be true? Furthermore, what
is the nature and extent
of
the impact of factors other than
the service component (e.g., price, product characteristics)
on customers' behavioral intentions? Additional conceptual
and empirical research addressing these issues can improve
our understanding of the behavioral consequences of service
quality.
44/
Journal
of
Marketing,
April
1996
Managerial Implications
The overall findings offer strong empirical support for the
intuitive notion that improving service quality can increase
favorable behavioral intentions and decrease unfavorable in-
tentions. The findings demonstrate the importance
of
strate-
gies that can steer behavioral intentions in the right direc-
tions, including striving to meet customers' desired-service
levels (rather than merely performing at their adequate-ser-
vice levels), emphasizing the prevention
of
service prob-
lems, and effectively resolving problems that do occur.
However, multiple findings suggest that companies wanting
to improve service, especially beyond the desired-service
level, should do so in a cost-effective manner: the quali-
ty-pay more relationship in the combined sample, while
u~