The Effect of Compensation on Repurchase Intentions inService Recovery�
Dhruv Grewal a,∗, Anne L. Roggeveen a, Michael Tsiros b,c
a Babson College, Babson Park, MA 02457, United Statesb University of Miami, Miami, FL 33133, United States
c ALBA Graduate Business School, Athens, Greece
Received 22 January 2008; received in revised form 29 April 2008; accepted 10 June 2008
bstract
To explore when the presence of compensation enhances repurchase intentions after a service failure, the authors use an experimental procedurend evaluate the impact of compensation in different stability and locus of responsibility conditions. Findings from three studies using scenariosrom different service industries indicate that compensation is necessary only when the company is responsible for the failure and the failure
ccurs frequently. If the failure occurs infrequently or the company is not responsible, compensation does not affect repurchase intentions. Theesults further demonstrate that stability and locus of responsibility attributions influence the perceived equity of the exchange, which mediates theffectiveness of compensation as a recovery effort. The authors discuss the theoretical and managerial implications.
2008 New York University. Published by Elsevier Inc. All rights reserved.
Repur
1Raapewb(tdeec
eywords: Service recovery; Compensation; Attribution; Consumer behavior;
A recent review by Grewal and Levy (2007) calls for furtheresearch to investigate the components of service recovery andheir interactive effects on repurchase intentions. In this research,e respond to their call by investigating how the content of
n explanation for a failure may influence the effectiveness ofompensation as a recovery strategy. Compensating customers,common service recovery strategy, can help dissipate con-
umer anger and dissatisfaction after a service failure (Bitner,ooms, and Tetreault 1990). However, offering compensationithout an explanation often indicates an admission of guilt
nd results in more negative evaluations (Bitner 1990). Thus,firm’s explanation appears to represent a necessary part of a
ompensation recovery strategy; we argue that understandingow the content of the explanation influences the effective-ess of compensation as a recovery strategy can be a key
onsideration.
An explanation for a failure affects attributions (Bitner 1990),hich in turn influence overall customer evaluations (Folkes
� All authors contributed equally to this research and are listed in alphabeticalrder.∗ Corresponding author.
984; Folkes, Koletsky, and Graham 1987; Tsiros, Mittal, andoss 2004) and may result in compensation enhancing evalu-tions in some conditions but not in others. As such, retailersnd service providers must understand the joint effects of arovided explanation and offered compensation on consumervaluations (Bolton, Grewal, and Levy 2007). In this research,e focus on company-provided explanations pertaining to sta-ility (is the cause likely to reoccur?) and locus of responsibilitywho is responsible?) (Tsiros et al. 2004; Weiner 1985). Explana-ions regarding stability and locus of responsibility should helpetermine whether compensation provides an effective recov-ry strategy, because these conditions create differing levels ofquity. And equity is expected to mediate the effectiveness ofompensation.
Understanding the effectiveness of compensation representsn important issue because despite prior research showing effec-iveness can vary in different conditions (e.g., Bitner 1990;mith, Bolton, and Wagner 1999), we lack a clear understand-
ng of exactly when compensation works, and more importantlyhen it does not work. In addition, managers need to know if
here are conditions in which compensation enhances repurchasententions, as well as conditions in which it has no impact. Withuch an understanding, a company can make strategic decisionsbout when to compensate customers.
Critical incident studies of service failures and recoveryncounters identify compensation as an effective recovery strat-gy (Bitner et al. 1990; Hoffman, Kelley, and Chung 2003;elley, Hoffman, and Davis 1993). In general, these studies sug-est that compensating customers after a service failure leadso more favorable consumer responses, either by dissipatingheir anger and dissatisfaction or by enhancing their overallxperience (Bitner et al. 1990). However, in some conditions,ompensation has no impact on evaluations. Without knowl-dge of when compensation relative to no compensation has anmpact, the company cannot make effective strategic decisionsbout when to compensate customers.
As we summarize in Table 1,1 various articles explore thempact of compensation, though only two (Bitner 1990; Smithnd Bolton 1998) focus on attributions. Bitner (1990) experi-entally investigates how employee-provided explanations and
ompensation affect control and stability attributions by pre-enting travelers at an airport with scenarios related to a travel.lthough she focuses on the main effects of explanation and
ompensation, she finds an unexpected interaction effect on attri-utions of control. That is, when compensation is offered, anyxplanation (indicating internal or external blame for the fail-re) reduces attributions of control. But without an explanation,ttributions of control increase. Bitner suggests this result mayccur because when compensation is offered with no explana-ion, it appears as an admission of guilt. No interactive effectsetween the explanation provided and compensation emerge fortability attributions, but the results indicate that stability attri-utions decrease when compensation is offered, regardless ofhe explanation (internal, external, none).
In addition, Bitner (1990) examines the impact of control andtability attributions on satisfaction and finds that when cus-omers perceive that the firm has control over the cause, theyre more dissatisfied than when they believe the firm has noontrol; when customers perceive the cause of the failure is sta-le, they also are more dissatisfied than when they believe theailure is rare. However, Bitner does not investigate how the pres-nce of compensation may influence these findings. Nor doesitner’s study specifically consider the joint effects of compen-
ation and explanation on satisfaction or behavioral intentions.itner (1990, p. 80) therefore concludes her article with a call for
urther research to “determine the robustness and boundaries ofhe model and the results.” We address this call by manipulatinghe locus of responsibility and stability attributions directly in therovided explanation and then measuring the interactive effectsf these factors, with compensation, on repurchase intentions.
The other article that considers both attributions and com-ensation in terms of service recoveries investigates how failure,ecovery, and attributions about failure stability influence cumu-
1 To provide a fuller context of studies of compensation in service recoveryiterature, we highlight the methodology, independent and dependent variables,nd key findings as they pertain to the main (Conlon and Murray 1996; Mattila001) and interactive (Harris et al. 2006; McColl-Kennedy, Daus, and Sparks003; Wirtz and Mattila 2004) effects of compensation.
rtc1sbttt
iling 84 (4, 2008) 424–434 425
ative satisfaction with the service provider and repatronagententions. Smith and Bolton (1998) had customers who hadsed a particular type of service provider (restaurant or hotel)valuate a scenario related to that type of service provider. Theynd that a single recovery (a proxy for compensation) can havesubstantial impact on customers’ overall level of satisfactionith the firm and ultimately their intentions to repatronize. In
erms of stability, they find mixed results. In a restaurant setting,valuations are lower when consumers believe the service failures likely to occur again, but in a hotel setting, stability attributionso not affect evaluations. However, similar to Bitner (1990) andmith and Bolton (1998) fail to investigate the interactive effectf compensation and stability.
As evidenced by such previous research, responsibilityBitner 1990) and stability (Bitner 1990; Smith and Bolton998) of failures represent important factors that explicate howompensation may influence repurchase intentions. We there-ore test these two factors in a series of three experiments. Werst develop and test our key prediction regarding the interac-
ive effect of stability, compensation, and locus of responsibilityn Study 1. Next, we replicate findings regarding the effectf the critical compensation by stability interaction (company-esponsible condition) using a different compensation conditionStudy 2) and a different context (Study 3).
quity and the stability by compensation interaction
Equity may explain how consumers respond to serviceecoveries (e.g., Alexander 2002; DeRuyter and Wetzels 2000;oodwin and Ross 1992; Smith et al. 1999; Susskind 2002),
uch that the effectiveness of recovery efforts may be a func-ion of equity in the exchange (Oliver and Swan 1989). Serviceailure and recovery create an exchange in which the consumerxperiences a loss due to the service failure and the firm attemptso make up for it in the form of a recovery (Smith et al. 1999).n general, to retain customers, companies must ensure that theecovery effort provides a benefit that the consumer believesquitably makes up for his or her loss (Adams 1965; Deutsch985). In the case of core service failures (e.g., cancellation offlight), the firm must fix the problem quickly (Parasuraman,erry, and Zeithaml 1991), but simply fixing the problem (e.g.,ooking the customer on the next flight) may not be enough.onsumers also may expect to be compensated for the harmone (e.g., 3 hr spent waiting) to preserve the equity of their rela-ionship with the company. Compensation is the most common
ethod used to restore equity (Walster, Berscheid, and Walster973).
Whether compensation is necessary to restore equity to theelationship is expected to vary as a function of the stability ofhe failure. If the failure is perceived to occur frequently (stable),onsumers anticipate the same outcome in the future (Weiner985, 1986). In general, consumers believe that stable problemshould be corrected by the company, and if the company has not
een able to correct these stable problems the company shouldry and make up for the customers’ loss in some way. Thus, ifhe failure repeatedly occurs, but customers are compensatedhey are likely to view the company as somewhat responsive
426D
.Grew
aletal./JournalofRetailing
84(4,2008)
424–434Table 1Summary of compensation in service recovery studies
Study Methodology IV DV Finding*
Bitner (1990) Experiment Explanation Control attribution Interactive effect of compensation and explanation issignificant on control attributions but insignificant forstability attributions. The interactive effects are notreported for satisfaction, quality, or behavioral intentions
Recovery and stability are significantpredictors of satisfaction and repurchaseintentions in restaurants; recovery issignificant in hotel context, but stability is not
Stability Repurchase intention
Smith, Bolton, andWagner (1999)
Experiment Failure (type andmagnitude)
Distributive justice The interaction of compensation and severity is onlypredicted and modeled for distributive justice. The resultsof this interaction are mixed. In a restaurant setting,compensation enhances justice when the failure is severe,but the interaction for moderate compensation is notsignificant. In the hotel context, the effect of moderatecompensation is more effective when the failure is lesssevere, but the interactive effect of high compensation andseverity is not significant
Recovery strategyincludingcompensation
Procedural justice
Interactional justiceSatisfaction
Mattila (2001) Experiment Service type Distributive justice Main effect of compensation. Interactionof compensation with service type(satisfaction, loyalty, procedural justice,and interactional justice)
Compensation Procedural justiceMagnitude of failure Interactional justice
Satisfaction withrecoveryLoyalty
Smith and Bolton(2002)
Experiment Service failure (typeand magnitude)
Satisfaction Medium compensation is more beneficial ina no-emotion group in a restaurant setting.High compensation is equally effective inboth emotion groups in the restaurant setting.In the hotel setting, compensation is not asignificant influence on satisfaction
Service recoverystrategies (includingcompensation)Emotion (measured,not manipulated)
D. Grewal et al. / Journal of RetaTa
ble
1(C
onti
nued
)
Stud
yM
etho
dolo
gyIV
DV
Find
ing*
Hof
fman
,Kel
ley,
and
Chu
ng(2
003)
CIT
55pe
rcen
tof
reco
very
resp
onse
stra
tegi
esin
volv
eso
me
form
ofco
mpe
nsat
ion
McC
oll-
Ken
nedy
,D
aus,
and
Spar
ks(2
003)
Exp
erim
ent
Con
cern
,voi
ce,a
ndco
mpe
nsat
ion
amou
nt(h
igh
vs.l
ow)
Satis
fact
ion,
futu
rein
tent
ions
Com
pens
atio
nef
fect
sar
em
oder
ated
byge
nder
ofpr
ovid
eran
dge
nder
ofcu
stom
er
Gen
der
ofse
rvic
epr
ovid
erE
ffor
tper
cept
ions
Gen
der
ofcu
stom
er
Wir
tzan
dM
attil
a(2
004)
Exp
erim
ent
Com
pens
atio
n(2
0pe
rcen
t,no
com
p)Sa
tisfa
ctio
nSa
tisfa
ctio
n:T
hree
-way
inte
ract
ion.
Whe
nco
mpe
nsat
ion
ispr
ovid
ed,t
hetw
om
ain
effe
cts
are
sign
ifica
nt.F
orth
epo
st-p
urch
ase
mea
sure
s,no
com
pens
atio
nef
fect
soc
cur
Res
pons
esp
eed
Post
-pur
chas
ere
spon
seA
polo
gy(p
rovi
ded,
notp
rovi
ded)
Har
ris,
Gre
wal
,Moh
r,an
dB
ernh
ardt
(200
6)E
xper
imen
tE
nvir
onm
ent(
onlin
evs
.offl
ine)
Satis
fact
ion
with
rem
edy
The
effe
ctof
reco
very
rem
edy
isgr
eate
rin
the
offli
nese
tting
(com
pare
dto
onlin
ese
tting
)R
ecov
ery
(fee
vs.
com
p)Po
st-p
urch
ase
inte
ntio
nsSe
tting
(air
line,
bank
)
*T
here
sults
ofth
ese
stud
ies
are
rest
rict
edto
the
effe
cts
ofco
mpe
nsat
ion.
itatualsatwpe
boit1cttAtCamsTs
Hcff
Li
suttcsfe
fl1rrqt(
iling 84 (4, 2008) 424–434 427
n trying to restore equity to the relationship. Additionally, dueo post-failure compensation consumers may view repurchases less risky, since they expect the company will compensatehem in an attempt to restore equity to the relationship if a fail-re were to occur again. In contrast, if a stable failure occurs,nd the company does not try to make up for the customersoss, customers are likely to view the company has as unrespon-iveness. They likely view this condition as inequitable (Seidersnd Berry 1998) and, as a result, have lower repurchase inten-ions (Folkes et al. 1987). Thus, when a stable failure occurs,e expect that equity will be damaged and anticipate that com-ensation should help restore equity to the relationship, whichnhances repurchase intentions.
If the company explanation indicates that the failure is unsta-le, consumers are likely to recognize that it is an infrequentccurrence and that the company probably could not have antic-pated it. The very nature of an unstable failure dictates thathe future may not be the same as the immediate past (Weiner985, 1986). Thus, when a failure is ascribed to an unstableause, consumers factor the infrequency of the occurrence intoheir evaluation of the situation and are less likely to ques-ion the equity of the transaction (Seiders and Berry 1998).s a result they are more likely to give the service provider
he benefit of the doubt and not expect to be compensated.onsequently, simply fixing the core service failure providesn equitable solution to customers and may be sufficient toaintain their repurchase intentions. In other words, compen-
ation is unlikely to have an effect on repurchase intentions.hus, we propose a moderating impact of stability on compen-ation.
1. A stability by compensation interaction exists, such thatompensation results in higher repurchase intentions when theailure is ascribed to a stable cause but has no effect when theailure is ascribed to an unstable cause.
ocus of responsibility by stability by compensationnteraction
The interaction between stability and compensation certainlyhould exist when the firm is responsible for the service fail-re. However, we consider it equally important to understandhe effects when the firm is perceived as not responsible forhe failure (e.g., the cause of the failure is external to theompany, “The flight has been delayed because of a snow-torm”) to determine whether the perceived stability of theailure still alters the effectiveness of compensation as a recoveryffort.
Regardless of whether the company is responsible for theailure, when the failure is unstable, consumers should be lessikely to question the equity of the transaction (Seiders and Berry998), and compensation is not required to enhance equity andepurchase intentions. Furthermore, when the company is not
esponsible for the failure, consumers should be less likely touestion the failure, which reduces the need to restore equityo the customer–company relationship through compensationBitner 1990; Widmier and Jackson 2002). Specifically, we
4 f Reta
essth
Hs
Hsaa
Hrn
Nc
M
D
tTflsct
P
voiimff
thlouitpbr
atflp
ts
lbotl(pcp
M
dZrauiiws
R
M
p(ct(
R
ciAdspwspMcpaFi
28 D. Grewal et al. / Journal o
xpect that when a firm is not responsible for a failure, compen-ation is not required to restore equity, regardless of the failuretability (e.g., even if snowstorms are a frequent occurrence,he company could not have prevented the storm). Formally, weypothesize:
2. A three-way interaction among stability, locus of respon-ibility, and compensation exists, such that:
2a. When the company is responsible for the failure, compen-ation results in higher repurchase intentions when the failure isscribed to a stable cause but has no effect when the failure isscribed to an unstable cause.
2b. When the company is not responsible for the failure,egardless of the stability of the occurrence, compensation haso effect on repurchase intentions.
ote that H2a reflects the basic interaction between stability andompensation predicted in H1.
Study 1
ethod
esignParticipants were 251 students who received class credit for
heir participation in this 2 × 2 × 2 between-subjects design.he locus of responsibility (company is responsible [shortage ofight crew] versus company is not responsible [weather]), thetability of the cause of the problem (stable versus unstable), andompensation (compensation versus no compensation) providehe between-subjects factors.
rocedureWe use a scenario-based experimental approach, which alle-
iates difficulties associated with the observation or enactmentf service failure and recovery incidents in the field, such as eth-cal considerations, as well as the managerial undesirability ofntentionally imposing service failures on consumers. Further-
Participants read a scenario that described them arriving athe airport to leave for vacation, only to find that their flightad been cancelled. It also explained the cause of the cancel-ation (weather versus shortage of flight crew) and the stabilityf the problem (stable: common occurrence versus unstable:ncommon occurrence). An airline agent acknowledged thenconvenience but assured participants that they would be ableo take the next flight, which would depart in 3 hr. In one com-ensation condition, participants were offered a $10 voucher toe spent at any of the restaurants in the airport; in the other, theyeceived no compensation.
The scenario context was picked to be a flight delay since
ll participants had traveled on airlines (average number ofrips taken on an airline per year was 6.36). Further, sinceight delays are a common occurrence with airlines and ourarticipants frequently travel, it is realistic to envision that
aoMM
iling 84 (4, 2008) 424–434
hey have experienced flight delays and can relate to thecenario.
A pretest had confirmed that the two causes of the cancel-ation appropriately manipulated locus of responsibility. In theetween-subjects design with high stability, participants readne of the two scenarios for the cause of the cancellation,hen rated their level of agreement with the statement “the air-ine is responsible for the inconvenience” on a five-point scale1 = strongly disagree, 5 = strongly agree). As the results show,articipants view the airline as more responsible for the flight-rew shortage (3.85) than for the weather (2.19; F(1,32) = 13.31,< .01).
easuresParticipants also rate their level of agreement (1 = strongly
isagree, 5 = strongly agree) with two statements adapted fromeithaml, Berry, and Parasuraman (1996), designed to measure
epurchase intentions (“I will recommend this airline to a friend”nd “I will fly this airline again in the future”). Finally, the manip-lation checks are aided recall measures the participants use tondicate whether the airline caused the inconvenience, whethert was a common problem for the airline, and whether a voucheras offered. The subjects responded to each measure on a yes/no
cale.
esults
anipulation checksThe manipulations work as intended. Significantly more peo-
le indicate that the problem is common in the stable conditionχ2 = 20.73, p < .001), that the airline caused the problem in theompany-responsible locus condition (χ2 = 33.78, p < .001), andhat a meal voucher was offered in the compensation conditionχ2 = 122.82, p < .001).
epurchase intentionsThe correlations between the two items measuring repur-
hase intention was r = .73. We ran an ANOVA with repurchasententions as the dependent variable and provide the overallNOVA results in Table 2 and means in Table 3. As we pre-icted in H1, there is a significant three-way interaction amongtability, responsibility, and compensation (F(1,243) = 5.69,< .05), as plotted in Fig. 1a. Follow-up contrasts reveal thathen the company is responsible for the failure and the failure is
table, offering compensation (versus offering none) enhancesarticipants’ repurchase intentions (Mcompensation = 2.42,no compensation = 1.85; F(1,243) = 7.8, p < .01). When the
ompany is responsible for the failure and failure is unstable,articipants have similar evaluations, regardless of whether theyre compensated (Mcompensation = 2.44, Mno compensation = 2.79;(1,243) = 3.09, ns), in support of H2a. As we predicted, and
n support of H2b, when the company is not responsible, there
re no differences among compensation conditions, regardlessf the stability of the failure (stable: Mcompensation = 3.12,no compensation = 3.05; F < 1; unstable: Mcompensation = 2.83,no compensation = 2.73; F < 1).
D. Grewal et al. / Journal of Reta
Table 2Analysis of variance results (Studies 1–3)
Effect F(1,243) p η
Study 1Locus of responsibility (L) 30.64 .00 .33Stability (S) .77 .38 .06Compensation (C) .94 .33 .06L × S 15.07 .00 .24L × C .01 .92 .01S × C 4.83 .03 .14L × S × C 5.69 .02 .15
S × C (company resp.) 10.38 .00 .20S × C (company not resp.) .02 .89 .009
Effect F(1,107) p η
Study 2: company responsible (airline context)Stability (S) 22.14 .00 .41Compensation (C) 6.35 .01 .24S × C 4.36 .04 .20
Effect F(1,214) p η
Study 3: company responsible (restaurant context)Stability (S) 4.03 .01 .13
D
w
TM
S
S
S
foerntordfsns
M
D
ufup
Compensation (C) 12.20 .01 .23S × C 4.38 .04 .20
iscussion
Consistent with our hypotheses, we find a significant three-ay interaction, such that when the company is responsible
able 3eans (Studies 1–3)
Company responsible Company not responsible
Stable Unstable Stable Unstable
tudy 1Compensation
Repurchase Intentions 2.422.44 3.122.83
No compensationRepurchase Intentions 1.852.79 3.052.73
Company responsible
Stable Unstable
tudy 2Compensation
Repurchase intentions 2.87 3.30Equity 3.67 3.93
No compensationRepurchase intentions 2.11 3.23Equity 2.80 3.77
tudy 3Compensation
Repurchase intentions 2.75 2.74Equity 3.78 4.07
No compensationRepurchase intentions 2.14 2.58Equity 1.87 3.03
P
sf(coawtoadmt
R
M
l(ivt
R
c(fe
iling 84 (4, 2008) 424–434 429
or the failure, stability attributions moderate the effectivenessf compensation. A stable failure requires compensation tonhance repurchase intentions, but compensation does not affectepurchase intentions for an unstable failure. When a company isot responsible, compensation does not affect repurchase inten-ions, regardless of the stability of the failure. To generalizeur stability by compensation results (within the company-esponsible condition), we next replicate our results using aifferent compensation condition (Study 2), as well as a dif-erent service failure context (Study 3). We felt that the airlinecenario could be too specific, and in order to enhance the robust-ess of the findings we replicate the interaction in a restaurantervice recovery scenario.
Study 2
ethod
esignParticipants in this 2 × 2 between-subjects design were 116
ndergraduate and graduate students who received class creditor their participation. The stability of the problem (stable versusnstable) and compensation (none versus 50 percent off coupon)rovide the between-subjects factors.
rocedureIn a procedure similar to that of Study 1, participants read a
hort scenario and answered the dependent measures. However,or this study, the cause of the cancellation remains constantshortage of flight crew). In one compensation condition, noompensation is offered, whereas in the other, participants areffered a $175 discount off their next flight. Because an aver-ge ticket price was $350, according to the scenario, the couponould result in savings of approximately 50 percent off their next
icket purchase. The manipulation checks include “The shortagef flight-crew members is a common problem for this specificirline” and “The value of the coupon is very high” (1 = stronglyisagree, 5 = strongly agree). We do not use a compensationanipulation check question in the “none” (no voucher) condi-
ion.
esults
anipulation checksThe manipulations work as intended. Subjects view the prob-
em as more common in the stable than in the unstable conditionMstable = 3.51, Munstable = 2.33, F(1,106) = 42.93, p < .001), andn the compensation condition, participants view the coupon asaluable (3.33, significantly higher than the neutral point of 3,(54) = 2.26, p < .05).
epurchase intentionsThe correlations between the two items measuring repur-
hase intention (r = .81) and the two items measuring equityr = .45) are both significant. We used two statements adaptedrom Tax (1993), designed to measure their perceptions of thequity (“The airline was concerned about my inconvenience”
430 D. Grewal et al. / Journal of Retailing 84 (4, 2008) 424–434
sults.
arcwapfiworbatpvlbd
itc
eawtwr(eFd
Fig. 1. (a) Study 1 results (airline context). (b) Study 2 re
nd “The airline handled the problem appropriately”. The cor-elation between the factors was also significant (r = .56). Aonfirmatory factor analysis shows that a two-factor solution, inhich the two repurchase intentions measures load on one factor
nd the two equity items load on a separate factor, is more appro-riate (χ2
(1) = 13.79, p < .01) and provides a significantly bettert than the one-factor solution (see Table 4). We run an ANOVAith repurchase intentions as the dependent variable, composedf the two previously described items. The overall ANOVAesults appear in Table 3. We find a significant interactionetween stability and compensation (F(1,107) = 4.36, p < .05),s plotted in Fig. 1b. Follow-up contrasts reveal that whenhe problem is stable, offering compensation (versus no com-ensation) enhances repurchase intentions (Mcompensation = 2.87
ersus Mnone = 2.11; F(1,107) = 10.53, p < .01). When the prob-em is unstable, there is no difference in repurchase intentionsetween those who receive compensation and those whoo not (Mcompensation = 3.30 versus Mnone = 3.23; F < 1). Thus,
asti
(c) Study 3 results. Contrast is statistically significant.
n support of H1, compensation enhances repurchase inten-ions, but only when the failure is ascribed to a stableause.
As suggested previously, the process that generates differ-nt repurchase intentions appears to entail the perceived equityssociated with the recovery attempt. To test this relationship,e ran an ANOVA with equity as the dependent variable. Similar
o the repurchase intentions results, we find a significant two-ay interaction (F(1,107) = 5.44, p < .05). Follow-up contrasts
eveal that when the problem is stable, offering compensationversus no compensation) causes participants to express higherquity perceptions (Mcompensation = 3.67 versus Mnone = 2.80;(1,107) = 16.29, p < .001), but when the problem is unstable, noifference appears in the equity perceptions between those who
re compensated and those who are not (Mcompensation = 3.93 ver-us Mnone = 3.77; F < 1). Therefore, equity may be a mediator ofhe effects of both explanations and compensation on repurchasententions (Baron and Kenny, 1986).
I will fly this airline again in the future .88 .98I will recommend this airline to a friend .92 .81The airline handled the problem
appropriately.53 .67
The airline was concerned about myinconvenience
.55 .74
Goodness of fitχ2 14.75 .96Degrees of freedom 2 1Normed fit index .93 .99Root mean squared error of .24 .01
msIcnpadaToirmto
D
trcfrciFeIs
M
D
c
ecWp
P
aET(tia2a
tetwata5hh
r“or(e5chtorct(
R
M
i(it
approximationAkaike information criterion 38.75 18.37Bayesian information criterion 39.84 19.45
To test this mediation relationship further, we ran a pathodel, which shows a mediation effect of stability, compen-
ation, and the interaction between stability and compensation.n the full model (with mediator), the direct effects of stability,ompensation, and their interaction on repurchase intentions areot statistically significant (.15, .06, and −.10, respectively; all’s > .10). When the direct paths from stability, compensation,nd the interaction to the mediator (equity) are set to 0, theirirect effects to the dependent variable (repurchase intentions)re significant (.57, .45, and −.48, respectively; all p’s < .01).his relationship indicates that equity acts as a full mediatorf stability, compensation, and their interaction on repurchasententions. The Sobel test and its Aroinan version confirm theseelationships (all zs > 4.0, p < .01). Consistent with our pathodel analysis, these results provide evidence of the full media-
ion of the effect of stability, compensation, and their interactionn repurchase intentions by equity.
iscussion
In the Study 2 scenarios, the provided explanation indicateshat the company is responsible for the failure, and the resultseplicate our prediction that stability moderates the effect ofompensation on repurchase intentions, such that when theailure is ascribed to a stable cause, compensation enhancesepurchase intentions. When the failure is ascribed to an unstableause, compensation has no effect, so compensating consumersn such situations is an ineffective use of company resources.urthermore, equity represents the process through which bothxplanations and compensation influence repurchase intentions.n the next study, we again replicate our findings using a differentervice setting.
Study 3
ethod
esignParticipants, 218 undergraduate students, received class
redit for their participation in this 2 × 2 between-subjects
t(wc
iling 84 (4, 2008) 424–434 431
xperiment. Stability of the cause of the problem and whetherompensation is offered represent the between-subjects factors.e hold the locus of responsibility constant, such that the com-
any is always responsible for the failure.
rocedureParticipants read a scenario that described them arriving
t a restaurant with a friend to celebrate a special occasion.ven though they had a reservation, their table was not ready.he scenarios vary the stability of the problem; in the stable
unstable) condition, the hostess informed the participant thathis happens frequently (rarely). The hostess acknowledged thenconvenience and informed them that the table will be ready inpproximately 35 min. Finally, half the participants received a0 percent off their bill, whereas the other half were not offeredny compensation.
The scenario context was picked to be in a restaurant set-ing since participants dine out for dinner regularly (one averageight times per month). Additionally, we found that 9.7 min washe average for how long respondents considered a reasonableait time when they have a reservation. Thus, the scenario withwaiting time of 35 min represented something unreasonable to
hem. Finally, we found that participants viewed the scenarios something which was realistic (3.70, 1 = very unrealistic,= very realistic), possible to happen (4.07, 1 = impossible toappen, 5 = possible to happen), and which they could envisionappening to them (3.70, 1 = not easily, 5 = very easily).
Participants rate their agreement with a four-item scale ofepurchase intentions adapted from Zeithaml et al. (1996): (1)I will consider this restaurant as my first choice when goingut to celebrate a special occasion in the future,” (2) “I willecommend this restaurant to someone who seeks my advice,”3) “I will recommend this restaurant to a friend,” and (4) “I willat at this restaurant again in the future” (1 = strongly disagree,= strongly agree). Finally, participants complete manipulationhecks, namely, aided recall measures (yes/no option) regardingow frequently it happens that a table has not to been ready at theime of the reservation for this restaurant and whether they wereffered anything for the inconvenience of not having the tableeady at the time of the reservation. In addition, participants alsoompleted manipulation checks questions indicating the degreeo which the stability of the cause and compensation were present1 = strongly disagree, 5 = strongly agree).
esults
anipulation checksThe manipulations work as intended. More participants
ndicate that the problem is common in the stable conditionχ2
(1) = 115.54, p < .001) and that it occurs more frequentlyn the stable than in the unstable condition (4.00 vs. 2.20,(214) = 11.19, p < .001). Also, more participants indicate that
hey were offered compensation in the compensation conditionχ2
(1) = 115.49, p < .001) and that they agreed that compensationas offered in the compensation versus in the no compensation
ondition (3.95 vs. 1.65, t(216) = 16.47, p < .001).
4 f Reta
R
biiletFaF
moiriacafoippce
D
taicfc
psη
Fenocatsvc
tiMu
sa
gm(phrebwd
espeWsSl
eemiaBeeafedwitjnr
voe
32 D. Grewal et al. / Journal o
epurchase intentionsThe four-repurchase intention items indicate high relia-
ility (α = .89). As we predict in H1, a significant two-waynteraction marks stability and compensation for repurchasententions (F(1,214) = 4.38, p < .05), as plotted in Fig. 1c. Inine with H1 and Studies 1 and 2, offering compensationnhances repurchase intentions when the failure is ascribedo a stable cause (Mcompensation = 2.75, Mno compensation = 2.14;(1,214) = 15.74, p < .001) but not when it stems fromn unstable cause (Mcompensation = 2.74, Mno compensation = 2.58;(1,214) = .94, p > .30).
Similar to Study 2, we perform a mediation test via a pathodel. In the full model (with mediator), the direct effects
f stability, compensation, and their interaction on repurchasententions are not statistically significant (.05, .07, and −.12,espectively; all p’s > .10). When the direct paths from stabil-ty, compensation, and the interaction to the mediator (equity)re set to 0, their direct effects to the dependent variable (repur-hase intentions) are significant (.17, .19, and −.30, respectively;ll p’s < .01). This relationship indicates that equity acts as aull mediator of stability, compensation, and their interactionn repurchase intentions. The Sobel test (Sobel, 1982) andts Aroinan version confirm these relationships (all zs > 3.0,< .01). Consistent with our path model analysis, these resultsrovide evidence of the full mediation of the effect of stability,ompensation, and their interaction on repurchase intentions byquity.
iscussion
We again find a significant two-way interaction and replicatehe findings of Studies 1 and 2, such that when the failure isscribed to a stable cause, compensation enhances repurchasententions, whereas when the failure is ascribed to an unstableause, compensation does not affect repurchase intentions. Tourther generalize the results we ran another study in a hotelontext and the results support this interaction.2
We calculate the effect sizes associated with the four contrastsertaining to the enhancing effect of compensation in the high-tability condition in Studies 1–3 and footnoted Study (Study 1= .30, Study 2 η = .41, Study 3 η = .34, footnoted Study η = .42).ollowing procedures by Rosenthal and Rosnow (1984), we firststablish that the four effect sizes are homogeneous (χ2
(3) = .72,s); the average weighted η is .36. Next, we test the significancef the overall relationship using Rosenthal and Rosnow’s (1984)ombining p-value technique. The results indicate that the over-ll relationship is significant at .0001. Finally, we calculatedhe file drawer n and find that it would take more than 45 null
tudies to reduce the significance level to .05. These results pro-ide considerable confidence that the overall enhancing effect ofompensation when stability is high and the company is respon-
2 We find the same stability and compensation interaction for repurchase inten-ions (F(1, 83) = 3.87, p = .05). Offering compensation enhances repurchasententions when the failure is ascribed to a stable cause (Mcompensation = 2.34,
no compensation = 1.28; F(1,83) = 9.55, p < .01) but not when it stems from annstable cause (Mcompensation = 2.03, Mno compensation = 1.99; F < 1).
bmautFeeu
iling 84 (4, 2008) 424–434
ible for both large and significant, as well as unlikely to be annomaly.
General discussion
During the past couple of decades, the role of services hasained prominence, and recent articles suggest marketing hasoved from a product-dominant to a service-dominant logic
Lusch, Vargo, and O’Brien 2007). In the domain of productroblems, standardized procedures tend to dictate how firmsandle defects (e.g., warranty policies), and the area of productecovery appears fairly well developed. However, service recov-ry has not been defined as well and may be handled differentlyy service providers in the same field (e.g., two hotel chainsith different recovery policies) or even in the same firm (e.g.,ifferent managers of a fast-food restaurant).
Compensating customers is a common service recovery strat-gy that can help dissipate anger and dissatisfaction after aervice failure (Bitner et al. 1990). However, we present a com-lex three-way interaction that indicates when compensationnhances repurchase intentions, as well as when it has no impact.e develop these predictions on the basis of the locus of respon-
ibility and the stability ascribed to the failure; in addition, withtudies 1 and 2, we demonstrate that equity serves as an under-
ying process.Despite the prevalence of compensation as a recovery strat-
gy (Bitner et al. 1990; Hoffman et al. 2003), relatively fewxperimental investigations consider the moderating factors thatay influence its effectiveness. Researchers who undertake such
nvestigations tend to consider a mix of dependent variablesnd arrive at conflicting results (e.g., Bitner 1990; Smith andolton 1998; Smith et al. 1999). The key insight derived fromxisting literature is that the effect of compensation may be mod-rated by certain variables. With our series of studies, we predictnd demonstrate that when a company is responsible for theailure, the effectiveness of compensation as a recovery strat-gy varies depending on the stability of the failure. That is, weemonstrate that compensation enhances repurchase intentionshen the company is responsible for the failure and the failure
s stable. But when the failure is an infrequent occurrence orhe company is not responsible for it, customers are satisfiedust with an explanation. In these cases, compensation is notecessary, and because it has no impact, it becomes a wastedesource.
The influence of compensation on repurchase intentions thusaries as a function of the locus of responsibility and the stabilityf the failure. Companies must maintain well-developed recov-ry strategies to manage consumers’ post-failure evaluations,ut they also need to know exactly when to use them. Managersust learn the conditions in which compensation is (or is not)
n effective recovery tool, especially as firms continue to eval-ate the effectiveness of their market activities on the basis ofheir bottom lines or returns on investment (Ambler et al. 2001).
urthermore, managers must weigh the costs of service recov-ry strategies relative to their benefits; the costs of such plansasily can increase to astronomical levels. Helping consumersnderstand the cause of the service failure by providing expla-
f Reta
nte
irapccna
stomsdaimsohoe
tDtbestcsBoe
cdevmtwcAwtcaa
a
rs(
tsrveearf
qmpht
iOtDPta
A
A
A
B
B
B
B
C
D. Grewal et al. / Journal o
ations thus offers a powerful and inexpensive tool with whicho manage the effectiveness and efficiency of service recoveryfforts.
Although our research provides some interesting insightsnto the role of explanation and compensation, additionalesearch might explore the ability of the firm to influence thesettributions (e.g., convert them from stable to unstable). Werovide the failure attributions to our respondents, but whenustomers face actual failures, they may generate their ownausal attributions. The effectiveness of the company expla-ation may be a function of how well it can influence suchttributions.
In addition, the effectiveness of compensation as a recoverytrategy may depend on other factors, such as globality attribu-ions (Hess, Ganesan, and Klein 2007), whether code-switchingccurs (i.e., going off script) when the compensation offer isade (Schau, Dellande, and Gilly 2007), or the type of con-
umer. For example, frequent users of a service may react veryifferently than infrequent users after a failure. The type andmount of compensation offered to, say, business travelers toncrease their repurchase intentions may differ from that which
ust be offered to pleasure travelers. It is therefore important toegment consumers into different types; however, our scenariosnly include one segment. Further research should investigateow different segments (e.g., frequent flyers, business travelers,ccasional pleasure travelers) react to different service recoveryfforts.
It also would be useful to examine the impact of the emo-ions people experience as a result of a failure (e.g., Menon andube 2004; Smith and Bolton 2002). Our research is limited
o scenario-based studies which may evoke more cognitivelyased responses than the emotional reactions a person experi-nces when in an actual service experience. Thus, future researchhould examine how the type of emotions, negative versus neu-ral versus positive affect, could be influential. The effect ofompensation and explanation in our studies may decline if con-umers experience very strong negative emotions. As Smith andolton (2002) indicate, the effect also may vary as a functionf the industry, importance of the situation, or length of thexperience.
Furthermore, we rely only on experimental scenarios thatomplement field study results provided by previous critical inci-ent studies (Bitner et al. 1990; Kelley et al. 1993). Althoughxperimental scenarios have several advantages over field sur-eys, their external validity may be limited. For example, weanipulate attributions as stable or internal and have evidence
hat the manipulations work, but it is hard to know whether theyould have worked in a similar fashion in a real setting or if
onsumers would have been more skeptical of the explanations.dditional research therefore might conduct field experimentalork in which actual service providers manipulate the alterna-
ive explanations in person or through call centers; such researchould monitor the effects of future behaviors, such as the length
nd profitability of the relationship, as well as the share of walletnd word-of-mouth behaviors.
We also focus on the presence or absence of compensationnd operationalize its presence differently in each study. Further
D
D
iling 84 (4, 2008) 424–434 433
esearch should explicitly examine the role of levels of compen-ation (e.g., 5, 20, 50, 100 percent) and types of compensatione.g., monetary, non-monetary).
Finally, we examine only the effects of equity, or expec-ations pertaining to fairness. Our measure was a two-itemcale which tapped into empathy and distributive justice. Futureesearch may need to use broader, multi-item measures of thearious components of justice. Additional research might alsoxamine the role of other types of expectations, such as thexpectation that the company will compensate customers afterservice failure. The mediating effect of an expectation of
eceiving compensation might be quite pronounced for stableailures—perhaps even regardless of the locus of responsibility.
Thus, we have only just begun to explore the very importantuestion of how to address service failures in a cost-effectiveanner, clearly a crucial issue for managers. As suggested by the
receding discussion, much more work remains; we thereforeope this article serves as a springboard for further research inhis area.
Acknowledgements
The authors acknowledge the helpful suggestions of partic-pants of research forums at Babson College, Baruch College,hio State University, Temple University, and Virginia Tech and
he detailed feedback provided by Ruth Bolton, Mike Brady,avid Hardesty, Gita Johar, Noreen Klein, Michael Levy, A.arasuraman, and Arun Sharma. The authors also appreciate
he suggestions of the editors Rajiv Dant and Jim Brown and thenonymous reviewers.
References
dams, J. Stacy (1965), “Inequity in Social Exchange,” New York: AcademicPress, Inc, 267–99.
lexander, Elizabeth C. (2002), “Consumer Reactions to Unethical ServiceRecovery,” Journal of Business Ethics, 36 (March), 223–37.
mbler, Tim, Flora Kokkinaki, Stefano Puntoni, Debra Riley (2001), “AssessingMarket Performance: The Current State of Metrics,” Center for MarketingWorking Paper No. 01-903, London Business School.
aron, Reuben M. and A.D. Kenny (1986), “The Moderator–Mediator VariableDistinction in Social Psychological Research: Conceptual, Strategic, andStatistical Considerations,” Journal of Personality and Social Psychology,51 1173–82.
itner, Mary Jo (1990), “Evaluating Service Encounters: The Effects of PhysicalSurroundings and Employee Responses,” Journal of Marketing, 54 (April),69–82.
itner, Mary Jo, Bernard H. Booms and Mary Stanfield Tetreault (1990), “TheService Encounter: Diagnosing Favorable and Unfavorable Incidences,”Journal of Marketing, 54 (January), 71–84.
olton, Ruth N., Dhruv Grewal and Michael Levy (2007), “Six Strategies forCompeting through Service: An Agenda for Future Research,” Journal ofRetailing, 83 (1), 1–4.
onlon, Donald E. and Noel M. Murray (1996), “Customer Perceptions ofCorporate Responses to Product Complaints: The Role of Explanations,”Academy of Management Journal, 39 (4), 1040–56.
eRuyter, Ko and Martin Wetzels (2000), “Customer Equity Considerations inService Recovery: A Cross-Industry Perspective,” International Journal ofService Industry Management, 11 (1), 91–108.
eutsch, Morton (1985), “Distributive Justice,” New Haven, CT: Yale UniversityPress.
4 f Reta
F
F
G
G
H
H
H
K
L
M
M
M
O
P
R
S
S
S
S
S
S
S
T
T
W
W
W
W
34 D. Grewal et al. / Journal o
olkes, Valerie S. (1984), “Consumer Reactions to Product Failure: An Attri-butional Approach,” Journal of Consumer Research, 10 (March), 398–409.
olkes, Valerie S., Susan Koletsky and John Graham (1987), “A Field Studyof Causal Inferences and Consumer Reaction: The View from the Airport,”Journal of Consumer Research, 13 (March), 534–9.
oodwin, Cathy and Ivan Ross (1992), “Consumer Responses to ServiceFailures: Influence of Procedural and Interactional Fairness Perceptions,”Journal of Business Research, 25 149–63.
rewal, Dhruv and Michael Levy (2007), “Retailing Research: Past, Present,and Future,” Journal of Retailing, 83 (4), 447–64.
arris, Katherine E., Dhruv Grewal, Lois A. Mohr and Kenneth L. Bernhardt(2006), “Consumer Responses to Service Recovery Strategies: The Mod-erating Role of Online versus Offline Environment,” Journal of BusinessResearch, 59 (April), 425.
ess, Ronald L. Jr., Shankar Ganesan and Noreen M. Klein (2007), “Interac-tional Service Failures in a Pseudorelationship: The Role of OrganizationalAttributions,” Journal of Retailing, 83 (1), 79–95.
offman, K. Douglas, Scott W. Kelley and Beth C. Chung (2003), “A CITInvestigation of Servicescape Failures and Associated Recovery Strategies,”Journal of Services Marketing, 17 (4), 322–40.
elley, Scott W., K. Douglas Hoffman and Mark A. Davis (1993), “A Typology ofRetail Failures and Recoveries,” Journal of Retailing, 69 (Winter), 429–52.
usch, Robert F., Stephen L. Vargo and Matthew O’Brien (2007), “Compet-ing Through Service: Insights from Service-Dominant Logic,” Journal ofRetailing, 83 (1), 5–18.
attila, Anna S. (2001), “The Effectiveness of Service Recovery ina Multi-Industry Setting,” Journal of Services Marketing, 15 (6/7),583–97.
cColl-Kennedy, Janet R., Catherine S. Daus and Beverley A. Sparks (2003),“The Role of Gender in Reactions to Service Failure and Recovery,” Journalof Service Research, 6 (August), 66.
enon, Kalyani and Laurette Dube (2004), “Service Provider Responses toAnxious and Angry Customers: Different Challenges, Different Payoffs,”Journal of Retailing, 80 (3), 229–38.
liver, Richard L. and John E. Swan (1989), “Consumer Perceptions of Interper-sonal Equity and Satisfaction in Transactions: A Field Survey Approach,”Journal of Marketing, 53 21–35.
arasuraman, A., Leonard L. Berry and Valarie A. Zeithaml (1991), “Under-
standing Customer Expectations of Service,” Sloan Management Review, 39(Spring), 39–48.
osenthal, Robert and Ralph L. Rosnow (1984), “Essentials of BehavioralResearch: Methods and Data Analysis,” New York: Mc-Graw-Hill BookCompany.
Z
iling 84 (4, 2008) 424–434
chau, Hope Jensen, Stephanie Dellande and Mary C. Gilly (2007), “The Impactof Code Switching on Service Encounters,” Journal of Retailing, 83 (1),65–78.
eiders, Kathleen and Leonard L. Berry (1998), “Service Fairness: What it Isand Why it Matters,” Academy of Management Executive, 12 (May), 8–20.
mith, Amy K. and Ruth N. Bolton (1998), “An Experimental Investigation ofCustomer Reactions to Service Failure and Recovery Encounters: Paradoxor Peril?,” Journal of Service Research, 1 (1), 65–81.
mith, Amy K. and Ruth N. Bolton (2002), “The Effect of Customers’ EmotionalResponses to Service Failures on Their Recovery Effort Evaluations andSatisfaction Judgments,” Journal of the Academy of Marketing Science, 30(1), 5–23.
mith, Amy K., Ruth N. Bolton and Janet Wagner (1999), “A Model of Cus-tomer Satisfaction with Service Encounters Involving Failure and Recovery,”Journal of Marketing Research, 36 (August), 356–72.
obel, Michael E. (1982), “Asymptotic Intervals for Indirect Effects in Struc-tural Equations Models,” in Sociological Methodology, Leinhart S. ed. SanFrancisco, CA: Jossey-Bass, 290–313.
usskind, Alex M. (2002), “I Told You So! Restaurant Customers’ Word-of-Mouth Communication Patterns,” Cornell Hotel and RestaurantAdministration Quarterly, 43 (2), 75–8.
ax, Stephen S. (1993), “The Role of Perceived Justice in Complaint Reso-lutions: Implications for Services and Relationship Marketing,” DoctoralDissertation, Arizona State University.
siros, Michael, Vikas Mittal and William T. Ross Jr. (2004), “The Role of Attri-butions in Customer Satisfaction: A Re-Examination,” Journal of ConsumerResearch, 31 (September), 476–83.
alster, Elaine, Ellen Berscheid and G. William Walster (1973), “New Direc-tions in Equity Research,” Journal of Personality and Social Psychology, 25(February), 151–76.
einer, Bernard (1985), “An Attributional Theory of Achievement Motivationand Emotion,” Psychological Review, 92 (October), 548–73.
(1986), “An Attributional Theory of Motivation andEmotion,” New York: Springer-Verlag.
idmier, Scott and Donald W. Jackson Jr. (2002), “Examining the Effects of Ser-vice Failure, Customer Compensation, and Fault on Customer Satisfactionwith Salespeople,” Journal of Marketing Theory and Practice, 10 (Winter),63–74.
irtz, Jochen and Anna S. Mattila (2004), “Consumer Responses to Compensa-
tion, Speed of Recovery and Apology After a Service Failure,” InternationalJournal of Service Industry Management, 15 (2), 150–66.
eithaml, Valarie A., Leonard L. Berry and A. Parasuraman (1996), “The Behav-ioral Consequences of Service Quality,” Journal of Marketing, 60 (April),31–47.
