Sie sind auf Seite 1von 12

Tourism Management 30 (2009) 51–62

Contents lists available at ScienceDirect

Tourism Management
journal homepage: www.elsevier.com/locate/tourman

The effects of perceived justice on recovery satisfaction, trust,


word-of-mouth, and revisit intention in upscale hotels
Taegoo (Terry) Kim a,1, Woo Gon Kim b, *, Hong-Bumm Kim c, 2
a
BK21 Institute for HRD in Tourism Research for Jeju International Free City, Cheju National University, 66 Jejudaehakno, Jeju Special Self-Governing Province,
Jeju City 690-756, Republic of Korea
b
International Center for Hospitality Research, Dedman School of Hospitality, College of Business, Florida State University, 1 Champions Way, UCB 4100,
P.O. Box 3062541, Tallahassee, FL 32306-2541, USA
c
College of Hospitality and Tourism, Sejong University, Kwang-jin Gu, Gun-ja Dong 98, Seou 143-747, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: The purpose of this study is to assess the relative influences of distributive (DJ), procedural (PJ), and
Received 13 October 2007 interactional (IJ) justices on customer satisfaction with service recovery and to examine the relationship
Accepted 8 April 2008 between recovery satisfaction and subsequent customer relationships: trust, word-of-mouth (WOM),
and revisit intention. On-site surveys were administered to collect data from hotel guests who stayed,
Keywords: and experienced a service failure, at five-star hotels. The effect of DJ on satisfaction with service recovery
Service failure was stronger than those of PJ and IJ. Since DJ, PJ, and IJ have significant effects on trust, WOM, and revisit
Service recovery
intention through recovery satisfaction, recovery satisfaction was found to be an important mediating
Complaint handling
Perceived justice
variable. In addition, the mediational role of trust between recovery satisfaction and WOM/revisit
Trust intention is substantial. Thus, in a case where strong trust is formed between the service provider and
Nested model the customer, a long-term relationship can be expected.
Model comparison Ó 2008 Elsevier Ltd. All rights reserved.
Model parsimony

1. Introduction Grönroos (1990, p. 5) defined service recovery processes as ‘‘those


actions in which a firm engages to address a customer complaint
No matter how excellent the service a hotel delivers, every hotel regarding a perceived service failure.’’ Some researchers suggest
still often makes mistakes in meeting the expectations of today’s that a company’s service recovery effort can restore customer sat-
hotel guests, who tend to be more demanding and less loyal than isfaction (Smith & Bolton, 1998; Smith, Bolton, & Wagner, 1999),
ever before. Service failures are inevitable due to the high ‘‘people reinforce customer relationships, and eventually attain customer
factor’’ of services in the hospitality business (Susskind, 2002). patronage (Blodgett, Granbois, & Walters, 1993; Blodgett, Hill, &
Service failure can prompt customer dissatisfaction with the service Tax, 1997; Chebat & Slusarczyk, 2005; Smith & Bolton, 1998; Wirtz
provider, which could possibly lead to customer complaints. Hotel & Mattila, 2004). Superb post-failure recovery efforts can induce
guests who are dissatisfied with the service encounter could pos- ‘‘service recovery paradox,’’ a term coined by McCollough and
sibly show any number of the following four reactions: exit silently, Bharadwaj (1992), whereby customers experiencing a service fail-
spread a negative word-of-mouth (WOM), voice their complaints to ure perceive a higher level of post-recovery satisfaction than those
the operator, or continue to patronage the same property despite who did not experience a service failure encounter at all.
their dissatisfaction (Karande, Magnini, & Tam, 2007; Susskind, Past research reveals that effective complaint handling is posi-
2002; Tax, Brown, & Chandrashekaran, 1998). tively related to customer loyalty and the subsequent customer
retention, which eventually leads to long-term profitability (Brown,
Cowles, & Tuten, 1996; Reichheld & Sasser, 1990; Tax et al., 1998).
Since hotel companies cannot prevent all customer complaints,
* Corresponding author. Tel.: þ1850 644 8242; fax: þ1850 644 5565. they have recently devoted a great deal of money and effort to learn
E-mail addresses: tgkim62@cheju.ac.kr (T. (Terry) Kim), wkim@cob.fsu.edu
(W.G. Kim), kimhb@sejong.ac.kr (H.-B. Kim).
how to better manage and respond to customer complaints about
1
Tel.: þ8264 754 3199; fax: þ8264 754 3198. service failures. Hotel companies have implemented service re-
2
Tel.: þ822 3408 3172; fax: þ822 3408 3312. covery programs that empower employees to immediately solve

0261-5177/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tourman.2008.04.003
52 T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62

problems by offering dissatisfied customers appropriate compen- and substitution (Lewis & McCann, 2004). DJ in service recovery has
sation (Stoller, 2005). For instance, the Hampton Inn has imple- been measured by the ‘‘justice,’’ ‘‘fairness,’’ ‘‘need,’’ ‘‘value,’’ and
mented a service guarantee program which resulted in one of the ‘‘reward’’ of outcomes (Blodgett et al., 1997; Chebat & Slusarczyk,
highest customer retention rates in the hotel industry and, sub- 2005; Smith et al., 1999; Wirtz & Mattila, 2004).
sequently, a high economic return (Ettorre, 1994). Each year
Hampton Inn spends approximately 0.5% of its total room revenue 2.1.2. Procedural justice (PJ)
on dissatisfied guests. For every $1 invested, the Hampton Inn re- PJ refers to ‘‘the perceived fairness of policies, procedures, and
alizes $7 in additional revenue from a new customer or a disgrun- criteria used by decision makers to arrive at the outcome of a dis-
tled one who would not have returned without the atonement pute or negotiation’’ (Blodgett et al., 1997, p. 189). That is, the per-
(Stoller, 2005). Ritz-Carlton is considered a leader in thoroughly ceived PJ for service recovery means the customer’s perception of
managing guest complaints by empowering all its employees to justice for the several stages of procedures and processes needed to
resolve all customer disputes up to $2000 per day (Stoller, 2005). recover the failed service (Mattila, 2001). PJ is the customer per-
A justice theory framework appears to gain popularity in ception of the company’s systematic and appropriate process in
explaining how guests evaluate service providers’ reactions to solving the service problems or managing the customer com-
service failure/recovery. Perceived justice is a multi-dimensional plaints, and it includes the company’s policies and disciplines
concept comprising three dimensions: distributive (DJ), procedural (Smith et al., 1999), including ‘‘timeliness,’’ ‘‘promptness,’’ ‘‘ap-
(PJ), and interactional justice (IJ). The application of the justice proach,’’ ‘‘flexibility,’’ ‘‘procedure control,’’ ‘‘outcome control,’’
framework facilitates a deepening theoretical understanding of the ‘‘right policy and execution,’’ and ‘‘appropriate method’’ for the
service provider and the customer relationship dynamics (Collie, specific contents of PJ in service recovery (Blodgett et al., 1997;
Sparks, & Bradley, 2000). Notwithstanding the recent advances Chebat & Slusarczyk, 2005; Maxham III & Netemeyer, 2002; Smith
concerning the effects of perceived justice on post-recovery be- et al., 1999; Wirtz & Mattila, 2004).
havior, there is still room to learn how a service provider’s recovery
efforts affect subsequent customer relationships with the company. 2.1.3. Interactional justice (IJ)
Although some studies have been performed on service recovery in IJ in service recovery is related to the way customers involved in
the service industry field, such as the hospitality industry (e.g., a failed service are handled, and it means the evaluation of the de-
DeWitt, Nguyen, & Marshall, 2008; Karande et al., 2007; Karatepe, gree to which the customers have experienced justice in human
2006; Mattila & Patterson, 2004; Ok, Back, & Shanklin, 2005; Smith interactions from the employees of service firms during the recovery
et al., 1999; Sparks & Fredline, 2007; Wirtz & Mattila, 2004; Yuksel, process (Sparks & McColl-Kennedy, 2001). Tax et al. (1998, p. 62)
Kilinc, & Yuksel, 2006), only a few studies have attempted to adopt conceptualized IJ as ‘‘the perceived fairness of interpersonal treat-
the relationship marketing approach to explain the relationship ment that people receive during the enactment of procedures.’’ The
between perceived justice with service recovery, recovery satis- specific methods suggested for IJ in service recovery are ‘‘courtesy,’’
faction, and three relationship outcome variables: trust, WOM, and ‘‘respect,’’ ‘‘interest,’’ ‘‘careful listening,’’ ‘‘effort,’’ ‘‘trust,’’ ‘‘explana-
revisit intention. Considering the importance of the outcomes of tion,’’ ‘‘empathy,’’ ‘‘apology,’ and ‘‘communication’’ (Blodgett et al.,
the relationship management after service recovery efforts, this 1997; Mattila, 2001; Smith et al., 1999; Wirtz & Mattila, 2004).
study attempts to fill the gap by linking the perceived justice of the
recovery efforts following service failure with important customer 2.2. Perceived justice and satisfaction with the recovery
relationships.
The purpose of this study, then, is to assess the relative effects of Extant research indicated that perceived justice of the recovery
DJ, PJ, and IJ on recovery satisfaction and to examine the relation- efforts could affect satisfaction with the recovery (Blodgett et al.,
ships between recovery satisfaction and customer relationship 1997; Chebat & Slusarczyk, 2005; Mattila & Patterson, 2004;
variables: trust, WOM, and revisit intention. Satisfaction with Karatepe, 2006). As reported in Table 1, many of them insisted that
recovery serves as the mediator between the three dimensions of the degree of the influence would vary according to the types of
perceived justice and post-recovery relationship variables. justice itself. For instance, Smith et al. (1999) concluded that DJ has
a stronger impact on satisfaction than the other two types of jus-
2. Literature review and hypotheses tice. On the other hand, Blodgett et al. (1997) stated that IJ was the
most important determinant of satisfaction followed by DJ. Con-
2.1. Perceived justice with service recovery sistent with the findings of Blodgett et al. (1997), Karatepe (2006)
disclosed that IJ has a greater influence on hotel guests’ satisfaction
The justice theory of Adams (1963) has been adopted as with complaint handling than do DJ and PJ. Ok et al. (2005), on the
a powerful vehicle in predicting customer satisfaction or dissatis- other hand, showed that PJ had the largest influence on recovery
faction after they have experienced service failure. Table 1 sum- satisfaction in a restaurant followed by DJ and then IJ. Thus, the
marizes the existing empirical literature that examines the effect of above discussion on perceived justice with service recovery leads to
perceived justice on recovery satisfaction and relationship outcome the following hypotheses:
variables in hospitality as well as in a variety of other industries.
Hypothesis 1a. DJ has a positive effect on satisfaction with the
Generally, the perceived justice dimensions with service recovery
recovery.
include DJ, PJ, and IJ (Blodgett et al., 1997; Chebat & Slusarczyk,
2005; Clemmer & Schneider, 1996; Karatepe, 2006; Ok et al., 2005; Hypothesis 1b. PJ has a positive effect on satisfaction with the
Smith et al., 1999; Tax et al., 1998). recovery.
Hypothesis 1c. IJ has a positive effect on satisfaction with the
2.1.1. Distributive justice (DJ)
recovery.
Smith et al. (1999, pp. 358–359) defined DJ as ‘‘the allocation of
costs and benefits in achieving equitable exchange relationships.’’
DJ has generally dealt with the outcomes given to the customers 2.3. Recovery satisfaction and trust
during service recovery, which include such monetary rewards as
refunds for failed service, discounts, coupons, and free-of-charge Many researchers have studied customer satisfaction with
(Mattila, 2001; Sparks & McColl-Kennedy, 2001), and adjustment service recovery (e.g., Blodgett, Wakefield, & Barnes, 1995; Brown
T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62 53

et al., 1996; Goodwin & Ross, 1992; Maxham III, 2001; McCollough, on motivation to improve a mutually beneficial buyer–seller
Berry, & Yadav, 2000; Ok et al., 2005; Smith & Bolton, 1998; Smith relationship. Oh (2002) demonstrated that trust in restaurant
et al., 1999; Tax et al., 1998). In this research, customer satisfaction customers has a significantly positive effect on relationship in-
with service recovery means a positive status of emotion perceived tention and repurchase intention. In light of the aforementioned
by customers in the process and result of recovering the failed evidence, the following hypotheses are proposed:
service. Davidow (2000, p. 478) defines satisfaction with complaint
Hypothesis 5. Trust has a positive effect on WOM.
handling as ‘‘the customer’s overall affective feeling about the firm
as a result of the firm’s complaint handling.’’ Hypothesis 6. Trust has a positive effect on revisit intention.
Trust builds when the customer has confidence in a service
provider’s reliability and integrity (Morgan & Hunt, 1994; Wong &
Sohal, 2002). Trust is a requisite in service marketing for main- 3. Method
taining the relationship between customers and service providers,
because customers often have to make a purchase decision before 3.1. Measures
they actually experience the service (Berry & Parasuraman, 1991).
As satisfaction arises from meeting or exceeding the expectations of Multiple item scales were used to measure each construct in this
the customer, satisfaction over time strengthens the reliability of study. If possible, validated scales from previous literature were
the service provider and cultivates trust (Ganesan, 1994; Tax et al., employed after a slight modification. In this study, the items in all
1998). Previous research confirms that satisfaction with service scales, except for WOM, revisit intention, and one item for DJ, were
recovery serves as a predictor of trust (Ok et al., 2005; de Ruyter & measured on a seven-point Likert-type scale ranging from ‘‘strongly
Wetzels, 2000; Tax et al., 1998). Thus, the following hypothesis is disagree’’ (1) to ‘‘strongly agree’’ (7). WOM and revisit intention were
proposed: measured on a seven-point Likert-type scale anchored with ‘‘least
likely’’ (1) and ‘‘extremely likely’’ (7). One item to measure DJ,
Hypothesis 2. Recovery satisfaction has a positive effect on trust. ‘‘compared to what you expected, the offer received (e.g., discount)
was,’’ was anchored with ‘‘less than expected’’ (1) and ‘‘more than
expected’’ (7). A total of 25 items were used to capture the seven
2.4. Recovery satisfaction, word-of-mouth, and revisit intention
latent constructs: three exogenous (perceived justice dimensions)
and four endogenous constructs (recovery satisfaction and
Previous empirical research suggests that recovery satisfaction
relationship outcome variables: trust, WOM, and revisit intention).
is strongly associated with customer loyalty (de Ruyter & Wetzels,
As shown in Table 2, the DJ was measured by a four-item scale
2000; Karatepe, 2006). Extant research also reveals that effective
adopted from Blodgett et al. (1997) and Smith et al. (1999). The PJ
complaint handling is positively related to subsequent customer
was measured by a four-item scale adapted from Blodgett et al.
retention (Reichheld & Sasser, 1990; Smith & Bolton, 1998; Tax
(1997) and Karatepe (2006). To measure the IJ construct, we used
et al., 1998). Maxham III and Netemeyer (2002) showed that re-
a five-item scale adapted from Karatepe (2006), Smith et al. (1999),
covery satisfaction has a significantly positive effect on spreading
and Tax et al. (1998). The recovery satisfaction scale was measured
positive WOM. Furthermore, Tax et al. (1998) claimed that
by a four-item scale adapted from Brown et al. (1996), as well as
recovery satisfaction has a positive effect on a customer’s revisit
from Maxham III and Netemeyer (2002). The four-item scale for
intention.
trust was adapted from Morgan and Hunt (1994) and Wong and
If recovery from service failure is effective, the customer
Sohal (2002). The WOM scale was measured by a two-item scale
satisfaction increases and revisit intention will rise. However, if
adapted from Mattila (2001) and Wong and Sohal (2002). Finally,
recovery from service failure is not executed fairly, recovery
the two-item scale, revisit intention, was adapted from Mattila
satisfaction will decrease, and ultimately WOM referral and
(2001) and Maxham III and Netemeyer (2002).
revisit intention will also decrease. Thus, the following hypoth-
eses are proposed:
3.2. Sampling frame and data collection
Hypothesis 3. Recovery satisfaction has a positive effect on WOM.
Hypothesis 4. Recovery satisfaction has a positive effect on revisit Data were collected from hotel guests in Seoul, Korea. We chose
intention. five-star hotels as our sample, because successful service recovery
from failure is a high priority for luxury hotel managers to maintain
long-term guest relationships. Researchers contacted the directors
2.5. Trust, word-of-mouth, and revisit intention of marketing for all 11 luxury hotels, but only 4 of the hotels
allowed a research team to survey their guests. The four hotels are
Morgan and Hunt (1994) developed a model of relationship affiliated with an international chain brand and primarily cater to
marketing that conceptualized trust as the key mediating variable to business travelers. A screening question was asked to see if re-
the development of long-term customer relationships. Gwinner, spondents encountered any service failure at a five-star hotel dur-
Gremler, and Bitner (1998) indicated that the psychological benefit ing the past six months. The on-site self-administered survey was
of belief and trust is more important than special treatment or social distributed to only those qualified Korean hotel guests. The ques-
benefit in the relationship between service firms and customers. tionnaire was adapted from original English questionnaire items
Therefore, customers developing trust with service providers have and then translated into Korean. Two research faculties, who were
proper reasons to maintain the relationship with them. Because of educated in the U.S., and a native Korean were involved in the
these inherent characteristics, a service firm (such as a hotel) is able translation process. A group of six hotel managers and four research
to employ trust as the most powerful marketing tool to build cus- faculties participated in a pilot test to refine the instrument by
tomer loyalty. In particular, when service providers recover in a way clarifying any ambiguous expressions, awkward wordings, or dis-
that builds customer trust, the perceived risk in complaining to the tortions of the original meanings. The survey provides respondents
providers in the future is likely to be reduced (DeWitt et al., 2008). a monetary incentive for completing and returning the question-
Garbarino and Johnson (1999) suggest that trust and commit- naire, and out of 1000 surveys, the response rate was 47.8%, rep-
ment both affect the future purchase intentions of an exchange resenting a total of 478 returned questionnaires. Out of the 478
partner. Selnes (1998) confirmed that trust has a positive influence collected, 27 questionnaires were either incomplete or the answers
54 T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62

Table 1
Previous studies on the perceived justice, recovery satisfaction, and related-relationship outcome variables

Author/s Sample Research design Notable findings


DeWitt et al. (2008) Restaurants and hotel Between-subjects Perceived justice following a service recovery had a
customers experiment with positive effect on customer trust. In addition, trust
hypothetical scenarios following service recovery had a positive effect on
involving service customer loyalty (customer loyalty was employed
failures two perspectives: attitudinal and behavioral loyalty.
Justice perception construct was used as single global
construct instead of three individual dimensions
of justice).

Karande et al. (2007) Online consumer panel Between-subjects Simple-effect tests showed that procedural justice had
(of a large reputed experiment with a significant positive effect on overall post-failure
marketing research hypothetical scenarios satisfaction for the airline and hotel scenairos and
company) as hotel involving service combined sample. Customers had higher overall
and customers failures post-failure satisfaction when they had higher
procedural justice perceptions (for the airline, hotel,
and combined samples) than lower procedural justice
perceptions (for the airline, hotel, and combined
samples) across all samples.

Sparks and Fredline (2007) Community members Between-subjects A referential explanation of a service failure had a
who had previously experiment with different impact on customer evaluations of customer
experienced staying hypothetical satisfaction and intended loyalty than a justification.
in luxury hotels scenarios involving More thorough accounts of a service failure resulted in
service failures higher levels of customer satisfaction and intended
loyalty. Account adequacy for a service failure is greater
in contexts of low versus high outcome severity. Under
the high severity condition an explanation of a service
failure reduced the level of counterfactual thinking.

Karatepe (2006) Hotel customers Retrospective Distributive, procedural, and interactional justice affect
self-report survey complaint satisfaction and complaint loyalty, where
of past dissatisfactory interactional justice has a stronger impact on complaint
experiences satisfaction and complaint loyalty than other types of justice.

Yuksel et al. (2006) Hotel customers Between-subjects There are more differences than similarities in complaining
experiment with behaviors of British, Israeli, Dutch and Turkish hotel customers.
hypothetical scenarios These countries had a favorable attitude toward complaining.
involving service A moderate relationship between attitudes and choice of
failures complaining behaviors was found. Hotel customers with
favorable attitudes to complaining were more likely to engage
in voice behavior, whereas hotel customers with negative
attitudes were more likely to display switch or loyalty behavior.

Ok et al. (2005) Members of community Mixed-design Distributive, procedural, and interactional justice affect service
service and religious experiment with recovery satisfaction, which in turn affects trust and overall
group as casual hypothetical scenarios satisfaction. Here, procedural justice has a stronger impact
restaurant customers on service recovery satisfaction than other types of justice.

Chebat and Slusarczyk (2005) Bank customers Telephone interview Interactional justice (e.g., courtesy) demonstrated direct effect
survey of past on loyalty-exit behavior. Distributive and procedural justice
dissatisfactory (i.e., timeliness) have only indirect effect on loyalty through
experiences the mediating effects of negative and positive emotions.

Mattila and Patterson (2004) Undergraduate students Quasi-experimental Compensation seemed to drive customers’
(American, Thai, and and between-subjects fairness perceptions, in particular with American consumers.
Malay) as casual dining with hypothetical Offerings an explanation
restaurant customers scenarios for the failure had a positive impact on customer perceptions
regardless of the customers’ cultural orientation. Perceived
fairness was directly linked to post-recovery satisfaction.

Maxham III (2001) Barber shop customers Mixed-design Moderate to high service recovery efforts notably enhanced
experiment with post-failure satisfaction, purchase intent, and positive WOM.
hypothetical On the other hand, poor service recoveries appeared
scenarios to aggravate the dissatisfaction resulted from a service failure.

McCollough et al. (2000) Airline passengers Between-subjects Distributive and interactional justice affected satisfaction with
experiment with a particular experience.
hypothetical scenarios
involving service failures

Smith et al. (1999) Undergraduate students Mixed-design Distributive, procedural, and interactional justice affected service
and hotel customers experiment with encounter satisfaction. In particular, the impact of distributive
hypothetical justice on recovery satisfaction appeared to be stronger than
scenarios that of procedural and interactional justice.

Tax et al. (1998) Employees as customers Cross-sectional Interactional, distributive, and procedural justice affected
survey capturing satisfaction with complaint handling. In particular, interactional
retrospective justice has a stronger impact on satisfaction with complaint
evaluations of handling than other types of justice.
past complaints
T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62 55

Table 1 (continued)

Author/s Sample Research design Notable findings

Smith and Bolton (1998) Undergraduate students Between-subjects Cumulative satisfaction and repatronage intent increased as
and hotel customers experiment with transactional satisfaction increases.
hypothetical scenarios

Blodgett et al. (1997) Consumers recruited Quasi-experiment Interactional justice can compensate for lower levels of
from church groups with hypothetical distributive justice. In addition, interactional justice was
scenarios the most important determinant of satisfaction followed
by distributed justice.

Clemmer and Schneider (1996) Fast food restaurant, Retrospective Distributive justice was the most important determinant
bank, and hospital self-report survey of customer satisfaction, followed by procedural justice
customers of past dissatisfactory and interactional justice. Thus, procedural and interactional
experiences justice may be secondary to distributive justice.

Blodgett et al. (1995) Retail customers Retrospective When customers perceived a lack of distributive or
self-report survey interactional justice, they became unsatisfied, and this led
of past dissatisfactory to a high post-redress negative WOM and to a low
experiences intention to re-patronize.

Blodgett et al. (1993) University staff Retrospective Overall perceived justice affected negative WOM and
self-report survey repatronage intentions. In addition, overall perceived
of past dissatisfactory justice was found to mediate the effects of likelihood of
experiences success, attitude toward complaining, product importance,
and stability and controllability on complaining behavior.

Goodwin and Ross (1992) Undergraduate Between-subjects Apology (i.e., interactional justice) and voice appeared to
students experiment with enhance fairness and satisfaction perceptions in the ‘‘favorable
hypothetical scenarios outcome’’ condition, when consumers were offered a discount
or gift after service failure. When no tangible offering was made,
apology and voice had lesser effect and in some instances were
associated with lower perceptions of fairness and satisfaction.

Table 2
Construct measurements

Constructs and measurement items References


Distributive justice (x1) Blodgett et al. (1997); Smith et al. (1999)
 Compared to what you expected, the offer received (e.g., discount) was
 Taking everything into consideration, the manager’s offer was quite fair
 Given the circumstances, I feel that the hotel has offered adequate compensation
 The customers did not get what they deserved (i.e., regarding a refund, coupon,
and room upgrade, etc.) (R)
Procedural justice (x2) Blodgett et al. (1997); Karatepe (2006)
 My complaint was handled in a very timely manner
 My complaint was not resolved as quickly as it should have been (R)
 The procedure for handling my complaint was complicated (R)
 Employees made an effort to adjust the procedure of handling my complaint
according to my needs
Interactional justice (x3) Karatepe (2006); Smith et al. (1999); Tax et al. (1998)
 Employees were courteous to me
 Employees’ communication with me was appropriate
 Employees put the proper effort into resolving my problem
 Employees showed a real interest in trying to be fair
 Employees showed concern
Recovery satisfaction (h1) Brown et al. (1996); Maxham III and Netemeyer (2002)
 Overall, I am satisfied with the service I received
 I am satisfied with the manner in which the service failure was resolved
 This hotel’s response to the service failure was better than expected
 I now have a more positive attitude toward this hotel
Trust (h2) Morgan and Hunt (1994); Wong and Sohal (2002)
 This hotel’s employees can be trusted at all times
 This hotel’s employees have a high level of integrity
 This hotel’s employees made every effort to fulfill the promises made
to its customers
 Overall, this hotel is reliable
Word-of-mouth (h3) Mattila (2001); Wong and Sohal (2002)
 I would recommend this hotel to other people
 I would tell other people positive things about this hotel
Revisit intention (h4) Mattila (2001); Maxham III and Netemeyer (2002)
 I consider this hotel as my first choice compared to other hotels
 I have a strong intention to visit this hotel again

Note: R items were reverse scored.


56 T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62

were found to be unreliable, leaving a remaining 451 question- and less than $65,000, 14.2% between $25,000 and less than
naires that were retained for further data analysis. $45,000, and 9.5% less than $25,000.
The aforementioned sample characteristics exhibit that the re-
3.3. Data analysis spondents of this study are rather equally distributed in terms of
gender. However, age distribution shows that the majority of
Data were analyzed using the two-step approach (Anderson & respondents are in their 30s and 40s. The general picture of our
Gerbing, 1988). In the first step, a confirmatory factor analysis (CFA) sample can be depicted as adult, highly educated, affluent pro-
was performed to determine whether the measured variables re- fessionals. These sample characteristics show that our respondents
liably reflected the hypothesized latent variables. In the second represent typical Korean baby-boomers.
step, path analyses were tested to determine overall model fit, path
coefficients significance, and explanatory power (R2). 4.2. Measurement model
More specifically, because a competing model (Model 2) is
nested within a proposed model (Model 1), a c2 difference test was The adequacy of the measurement model was examined by
used. Also, to compare the proposed model (Model 1) with the confirmatory factor analysis (CFA). Table 4 demonstrates the results
competing model (Model 2), path analysis was used to compare the of the measurement model. Most of the model fit indices from CFA
two models in terms of overall model fit (including absolute fit demonstrated a good fit with c2 (df) ¼ 648.10 (254), normed c2 (c2/
measures, incremental fit measures, as well as parsimony fit mea- df) ¼ 2.55, goodness-of-fit index (GFI) ¼ 0.90, root mean square
sures), path coefficients significance, and explanatory power. error of approximation (RMSEA) ¼ 0.07, adjusted goodness-of-fit
index (AGFI) ¼ 0.86, normed fit index (NFI) ¼ 0.94, Tucker–Lewis
4. Results index (TLI) ¼ 0.95, and comparative fit index (CFI) ¼ 0.96. Although
the c2 statistic shows significant p-value, the relatively large sample
4.1. Respondents’ demographic profile size of our study (N ¼ 451) offsets the seriousness of the effect of
statistic to the validity of measurement model (Anderson &
The demographic profile of the respondents is presented in Gerbing, 1982). Collectively, the results of the CFA satisfy the rec-
Table 3. The sample consisted of 51.7% male and 48.3% female re- ommended level of goodness of fit, which implies that the mea-
spondents. Approximately 28.8% were single or divorced and 71.2% surement model generally fits the sample data well.
were married. In terms of age, 40.6% of the respondents were 40–49 In this study, an effort to purify the measurement scales was per-
years old, 23.9% were 30–39, 22.4% were 50 or older, and 13.1% were formed in the stage of employing the CFA for the purpose of delineating
younger than 30. The majority of the respondents (77.8%) were more valid constructs. Scale items displaying insignificant and low
highly educated, holding at least a college degree; 13.7% had some values of factor loadings were deleted from the measurement variables
graduate education. In terms of occupation, the largest group in the of each scale. It can also be noted that scale reliability is an important
distribution (31.9%) was administrative or managerial workers, measure of scale adequacy. When scale reliability is high, variables
followed by 23.3% professionals (such as doctors, lawyers, pro- measuring a single factor share a high degree of common variance
fessors, and researchers), 17.3% sales or service industry-related, (Sanders & Premus, 2005). The Cronbach’s alphas of the six constructs
15.5% self-employed workers, and 5.8% technical workers. With range from 0.83 (WOM) to 0.97 (revisit intention), which shows that
regard to annual household income, 35.3% of the respondents the measures are internally consistent because they are all above the
earned more than US$65,000, followed by 41.0% between $45,000 recommending criterion of 0.70 (Nunnally, 1978). Fornell and Larcker
(1981) suggest that composite construct reliability may be an alter-
native criterion to Cronbach’s alpha in examining the internal consis-
Table 3 tency of the variables in the measurement model. The composite
Profile of the respondents
construct reliability of each construct ranges from 0.70 (WOM) to 0.93
Variables Frequency (N ¼ 451) (revisit intention), exceeding a suggested minimum of 0.70 (Bagozzi,
N % 1980). The results show that all scales of the measurement model
Gender demonstrate adequate internal consistency for further analysis of the
Male 233 51.7 construct model.
Female 218 48.3
Martial status 4.3. Convergent and discriminant validity
Single or divorced 130 28.8
Married 321 71.2
Age (year) Convergent validity exists when a researcher’s proposed scale
Less than 30 59 13.1 items for the same construct are correlated strongly with significance.
30-39 108 23.9 The standardized loadings and the squared multiple correlation (SMC)
40-49 183 40.6
for the measurement items and the constructs were examined as
50 or older 101 22.4
Education level evidences of convergent validity (Bollen, 1989). Significant factor
Associate degree or below 100 22.2 loadings for a specific construct suggest a confirmation of convergent
College 289 64.1 validity such that indicators for a given construct should be at least
Graduate school 62 13.7 moderately correlated among themselves. All items loaded signifi-
Occupation
Self-employed/owner 70 15.5
cantly on their constructs with the lowest critical ratio (t-value) of
Administrative/managerial 144 31.9 15.61 (p < 0.001), indicating that the specific measurement variables
Professional 105 23.3 are sufficient in their representation of the constructs (Hair, Anderson,
Sales/service 78 17.3 Tatham, & Black, 1998). In addition, the convergent validity can be
Technical 26 5.8
achieved when the SMC between the items and the construct is
Others 28 6.2
Annual household income (US$) greater than the cut-off value of 0.40 (Bollen, 1989; Kline, 1986). Since
Less than $25,000 43 9.5 all the SMC of each item surpassed the recommended level, evidence
$25,000 – Less than $45,000 64 14.2 of convergent validity for the tested items is provided.
$45,000 – Less than $65,000 185 41.0 Evidence of discriminant validity is demonstrated when
$65,000 or more 159 35.3
measures of conceptually different constructs are not strongly
T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62 57

Table 4
Measurement model results

Constructs and measurement items Standardized loadings t-Values SMCs


Distributive justice (x1; a ¼ 0.91; AVE ¼ 0.56; CCR ¼ 0.84)
Compared to what you expected, the offer received (e.g., discount) was 0.80 16.00 0.64
Taking everything into consideration, the manager’s offer was quite fair 0.84 16.87 0.70
Given the circumstances, I feel that the hotel has offered an adequate compensation 0.94 – 0.89
I did not get what I deserved (i.e., regarding a refund, coupon, room upgrade) (R) 0.95 17.83 0.90

Procedural justice (x2; a ¼ 0.93; AVE ¼ 0.55; CCR ¼ 0.86)


My complaint was handled in a very timely manner 0.73 21.69 0.54
My complaint was not resolved as quickly as it should have been (R) 0.96 42.32 0.92
The procedure for handling my complaint was complicated (R) 0.97 – 0.93
Employees made an effort to adjust the procedure of handling my complaint 0.74 21.15 0.55
according to my needs

Interactional justice (x3; a ¼ 0.92; AVE ¼ 0.49; CCR ¼ 0.83)


Employees were courteous to me 0.97 – 0.93
Employees’ communication with me was appropriate 0.85 30.97 0.72
Employees put the proper effort into resolving my problem 0.68 19.52 0.46
Employees showed a real interest in trying to be fair 0.96 47.48 0.92
Employees showed a concern 0.65 18.07 0.42

Recovery satisfaction (h1; a ¼ 0.93; AVE ¼ 0.52; CCR ¼ 0.82)


Overall, I am satisfied with the service I received 0.80 22.59 0.64
I am satisfied with the manner in which the service failure was resolved 0.85 25.13 0.71
This hotel’s response to the service failure was better than expected 0.89 – 0.79
I now have a more positive attitude toward this hotel 0.88 26.38 0.78

Trust (h2; a ¼ 0.86; AVE ¼ 0.51; CCR ¼ 0.75)


This hotel’s employees can be trusted at all times 0.80 15.61 0.63
This hotel’s employees have a high level of integrity 0.72 16.11 0.52
This hotel’s employees make every effort to fulfill the promises made 0.77 17.62 0.59
to its customers
Overall, this hotel is reliable 0.87 – 0.76

Word-of-mouth (h3; a ¼ 0.83; AVE ¼ 0.54; CCR ¼ 0.70)


I would recommended this hotel to other people 0.85 21.72 0.72
I would tell other people positive things about this hotel 0.85 – 0.71

Revisit intention (h4; a ¼ 0.97; AVE ¼ 0.88; CCR ¼ 0.93)


I consider this hotel as my first choice compared to other hotels 0.97 44.00 0.94
I have a strong intention to visit this hotel again 0.97 – 0.93

Model fit measures
c2 (df) 648.10 (254)
Normed c2 2.55
GFI 0.90
RMSEA 0.07
AGFI 0.86
NFI 0.94
TLI 0.95
CFI 0.96

Note: All standardized loadings are significant at the 0.01 level or better. GFI ¼ goodness-of-fit index, RMSEA ¼ root mean square error of approximation, AGFI ¼ adjusted
goodness-of-fit index, NFI ¼ normed fit index, TLI ¼ Tucker–Lewis index, CFI ¼ comparative fit index, SMC ¼ squared multiple correlation, AVE ¼ average variance extracted,
CCR ¼ composite construct reliability. Normed c2 ¼ c2/df. R items were reverse scored. CCR ¼ (S standardized loadings)2/(S standardized loadings)2 þ (S indicator mea-
surement error), AVE ¼ (S squared standardized loadings)/(S squared standardized loadings) þ (S indicator measurement error).

correlated among themselves. To assess the discriminant validity, parameters of the model. In comparison with values suggested by
the square root of the average variance extracted (AVE) in each Hu and Bentler (1999), the overall fit of the model to the data
construct is compared to the correlation coefficients between two resulted acceptable statistics as follows: c2 (df) ¼ 27.15 (10),
constructs (Fornell & Larcker, 1981). One pair of scales between normed c2 ¼ 2.72, GFI ¼ 0.98, RMSEA ¼ 0.07, AGFI ¼ 0.95,
trust and WOM appeared to have relatively high correlation NFI ¼ 0.99, TLI ¼ 0.97, and CFI ¼ 0.99.
(F ¼ 0.71) (see Table 5). The variances extracted for both constructs All structural path estimates were significant at the 0.01 level,
were 0.51 and 0.54, respectively (see Table 4). Since the two vari- where the signs of all structural paths were consistent with the
ances are greater than the square of the correlation coefficients hypothesized relationships among the latent constructs. The model
(F2 ¼ 0.50), a discriminant validity is confirmed. These results accounted for a substantial proportion of the variance (R2) in four
exhibit that the variables in our measurement model appear to endogenous variables: 47.2% of variance in recovery satisfaction,
have acceptable levels of discriminant validity. 34.1% in trust, 63.7% in WOM, and 52.5% in revisit intention (see
Fig. 1).
4.4. Path model testing Hypothesis testing generally led to support for all of the hy-
potheses proposed. Regarding the relationships between perceived
4.4.1. Proposed model (Model 1) justice and recovery satisfaction, distributive (Hypothesis 1a:
The results of the structural model tested are illustrated in Fig. 1. g11 ¼ 0.49, p < 0.01), procedural (Hypothesis 1b: g12 ¼ 0.17,
Using the path analysis, the predicted relationships among exoge- p < 0.01), and interactional perception (Hypothesis 1c: g13 ¼ 0.24,
nous and endogenous constructs were tested. The maximum like- p < 0.01), all appeared to have significant influences on recovery
lihood (ML) estimation was employed to estimate the structural satisfaction with complaint handling. It is noteworthy here that the
58 T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62

Table 5
Correlation estimates, construct means, and standardized deviations

Constructs Number of items M SD 1 2 3 4 5 6 7


1. Distributive justice (x1) 4 3.94 1.46 1.00
2. Procedural justice (x2) 4 3.84 1.44 0.32 1.00
3. Interactional justice (x3) 5 4.14 1.38 0.21 0.28 1.00
4. Recovery satisfaction (h1) 4 3.97 1.47 0.61 0.40 0.40 1.00
5. Trust (h2) 4 4.29 1.27 0.33 0.33 0.41 0.59 1.00
6. Word-of-mouth (h3) 2 4.31 1.35 0.41 0.31 0.39 0.68 0.71 1.00
7. Revisit intention (h4) 2 4.60 1.41 0.33 0.29 0.42 0.63 0.67 0.70 1.00

Note: Composite scores were calculated by averaging scores items representing that construct. The scores range from 1 to 7. A higher score indicates a more favorable response.
All correlations are significant at the 0.01 level or better.

effect of perceived DJ on recovery satisfaction is stronger than those endogenous variables: 43.3% of variance in recovery satisfaction,
of perceived PJ and IJ. 34% in trust, 42.7% in WOM, and 42.2% in revisit intention (see
As hypothesized, recovery satisfaction was positively associated Fig. 2).
with trust (Hypothesis 2: b21 ¼ 0.54, p < 0.01), WOM (Hypothesis 3:
b31 ¼ 0.37, p < 0.01), and revisit intention (Hypothesis 4: b41 ¼ 0.36, 4.4.3. Model comparison
p < 0.01). This shows that a customer’s recovery satisfaction would be The model comparison should focus on assessing model fit and
affected more significantly by DJ than by either PJ or IJ. Furthermore, compare the fit of competing and theoretically plausible models
trust had a significant effect on WOM (Hypothesis 5: b32 ¼ 0.52, (Kelloway, 1998). As both the proposed model (Model 1) and the
p < 0.01) and revisit intention (Hypothesis 6: b42 ¼ 0.46, p < 0.01), competing model (Model 2) fit the original data appropriately, a c2
which would imply the significant mediating role of trust in the causal difference test was employed to determine if one of these struc-
link from recovery satisfaction to WOM and revisit intention. tures performed better than the other (Anderson & Gerbing, 1988;
Bentler & Bonett, 1980) (see Figs. 1 and 2).
4.4.2. Competing model (Model 2) As shown in Table 6, the c2 statistics between the two models
In addition to the proposed model (Model 1) illustrated in Fig. 1, were compared using the proposed model (Model 1) as a reference
a competing model (Model 2) was tested (see Fig. 2). In the com- point. The c2 statistics’ difference between the two models was
peting model (Model 2), recovery satisfaction was presumed to significant (Dc2 ¼ 13.94, Ddf ¼ 2, p < 0.001), which indicated that
have only an indirect effect on WOM and revisit intention; that is, adding the direct paths from recovery satisfaction to WOM and
the competing (full-mediation) model (Model 2) was nested within revisit intention does improve its fit, though both models had met
the proposed (partial-mediation) model (Model 1), emphasizing the fit criteria. In this type of model comparison, parsimony fit
the mediating role of trust from recovery satisfaction to WOM and measures such as CFI (Bentler, 1990), the Akaike information cri-
revisit intention. The competing model (Model 2) achieved a good terion (AIC) (Akaike, 1987; Arbuckle, 2006; Jöreskog & Sörbom,
level of fit (c2 (df) ¼ 41.09 (12), normed c2 ¼ 3.42, GFI ¼ 0.97, 1993), and the Browne–Cudeck criterion (BCC) (Arbuckle, 2006;
RMSEA ¼ 0.08, AGFI ¼ 0.90, NFI ¼ 0.97, TLI ¼ 0.93, CFI ¼ 0.98), Browne & Cudeck, 1989) are appropriate for model comparison
which reinforces the intervening role of trust for final performance because these assess parsimony as well as fit (Rust, Lee, & Valente,
of service recovery efforts. 1995). As shown in Table 6, in the case of the proposed model
All structural path estimates were significant at the 0.01 level, (Model 1), CFI is 0.99, and in the case of the competing model
where the signs of all structural paths were consistent with the (Model 2), CFI is 0.98. As bigger value of this criteria indicates
hypothesized relationships among the latent constructs. The model a better fit of the model, these results indicated a preference for the
accounted for a substantial proportion of the variance (R2) in four proposed model (Model 1) over the competing model (Model 2). In

D3
Distributive Word-of-mouth
0.66
justice (η3)
(ξ1) (R2 = 0.64)
H1a (γ11): 0.49** **
D1
0.66** **
0.75
D2
Recovery
Procedural H1b (γ12): 0.17** satisfaction
** Trust 0.86
0.43** justice (η2)
(η1) 2
(ξ2) (R = 0.34)
(R2 = 0.47)

**
0.50**
D4
H1c (γ13): 0.24** **
Revisit
Interactional 0.54
intention
justice
(η4)
(ξ3)
(R2 = 0.53)

Fig. 1. Results of the proposed model (Model 1). Model fit measures: c2 (df) ¼ 27.15 (10), normed c2 ¼ 2.72, GFI ¼ 0.98, RMSEA ¼ 0.07, AGFI ¼ 0.95, NFI ¼ 0.99, TLI ¼ 0.97, CFI ¼ 0.99.
Note: Normed c2 ¼ c2/df. GFI ¼ goodness-of-fit index, RMSEA ¼ root mean square error of approximation, AGFI ¼ adjusted goodness-of-fit index, NFI ¼ normed fit index,
TLI ¼ Tucker–Lewis index, CFI ¼ comparative fit index. **p < 0.01.
T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62 59

D3
Distributive Word-of-mouth 0.79
justice (η3)
(ξ1) (R2= 0.43)
γ11: 0.49**
D1
0.66** β32: 0.74**
0.84
D2
Recovery
Procedural γ12: 0.17** β21: 0.54** Trust 0.90
0.33** justice
satisfaction
(η2)
(η1)
(ξ2) (R2= 0.34)
(R2= 0.43)

0.43** β42: 0.67**


γ13: 0.25** D4
Revisit
Interactional intention 0.58
justice
(η4)
(ξ3) (R2 = 0.42)

Fig. 2. Results of the competing model (Model 2). Model fit measures: c2 (df) ¼ 41.09 (12), normed c2 ¼ 3.42, GFI ¼ 0.97, RMSEA ¼ 0.08, AGFI ¼ 8.90, NFI ¼ 0.97, TLI ¼ 0.93, CFI ¼ 0.98.
Note: Normed c2 ¼ c2/df. GFI ¼ goodness-of-fit index, RMSEA ¼ root mean square error of approximation, AGFI ¼ adjusted goodness-of-fit index, NFI ¼ normed fit index,
TLI ¼ Tucker–Lewis index, CFI ¼ comparative fit index. **p < 0.01.

the case of the proposed model (Model 1), AIC is 64.15 and BCC is 4.4.4. Direct and indirect effects
64.78, while the values for the competing model (Model 2) are The direct and indirect effects in the proposed model (Model 1)
AIC ¼ 81.09 and BCC ¼ 81.90, respectively. As smaller values of were further examined in an effort to gain in-depth insights into
these criteria indicate a better fit of the model (Kelloway, 1998), the customer’s WOM and revisit intention process with service
these results indicate a preference for the proposed model (Model recovery (see Table 6). Since the proposed model (Model 1) dem-
1) over the competing model (Model 2). In addition, parsimony onstrated a better fit than the competing model (Model 2), the
goodness-of-fit index (PGFI) [the proposed model (Model 1): 0.78; interpretation of direct and indirect effects focuses on the proposed
the competing model (Model 2): 0.74] and parsimony normed fit model (Model 1) herewith.
index (PNFI) [the proposed model (Model 1): 0.88; the competing Perceived DJ (0.27), PJ (0.06), and IJ (0.14) affected trust positively
model (Model 2): 0.83], which assesses the parsimony fit of com- via recovery satisfaction, where the DJ perception had the largest ef-
peting models (Kelloway, 1998), favor the proposed model (Model fect on trust via recovery satisfaction. The results show that the effect
1). Also, for this type of model comparison, the most common of PJ on trust via recovery satisfaction was relatively small, whereas
statistical test is normed c2 (Rust et al., 1995). As shown in Table 6, the effect of DJ on trust via recovery satisfaction was substantial.
in the case of the proposed model (Model 1), normed c2 is 2.72, Table 6 presented the mediational roles of recovery satisfaction
in the case of the competing model (Model 2), normed c2 is 3.42. and trust between DJ (0.32), PJ (0.09), and IJ (0.16) and WOM. In
As smaller value of this criteria indicates a better parsimony as particular, the effect of perceived DJ on WOM via recovery satis-
well as a better fit of the model, this result indicated a preference faction and trust was found to be the greatest. The data further
for the proposed model (Model 1) over the competing model suggested DJ (0.31), PJ (0.09), and IJ (0.14) exerted positive effects on
(Model 2). revisit intention via recovery satisfaction and trust. In particular, DJ
More specifically, the indirect effects of recovery satisfaction on perception had the largest effect on revisit intention via recovery
WOM and revisit intention via trust appeared to be relatively satisfaction and trust. The results also show that the effect of PJ on
smaller than the direct ones. In sum, compared to the competing revisit intention via recovery satisfaction and trust was not im-
model (Model 2), the exploratory power of the proposed model portant, but the effects of DJ and IJ on revisit intention via recovery
(Model 1) had increased by adding the two direct effects of re- satisfaction and trust were substantial.
covery satisfaction on WOM and revisit intention: the largest in- The direct effect of recovery satisfaction on WOM appeared to be
crement in R2 was 0.21 for WOM [0.64 for the proposed model significant (0.37). Also, the indirect effect via trust was found to be
(Model 1) – 0.43 for the competing model (Model 2)] and 0.11 [0.53 apparent (0.29). Therefore, trust served as an important mediating
for the proposed model (Model 1) – 0.42 for the competing model variable between recovery satisfaction and WOM. The results show
(Model 2)] for revisit intention. Overall, the results suggest that in that the direct effect of recovery satisfaction on revisit intention via
a comparison between the proposed model (Model 1) interpreting trust was noticeable with 0.36, and the indirect effect of recovery
trust as a mediating role between the recovery satisfaction and satisfaction on revisit intention via trust was noticeable with 0.25.
both WOM and revisit intention and the competing model (Model Thus, these results indicate that the mediational role of trust
2), the proposed model (Model 2) is slightly superior to the com- between recovery satisfaction and revisit intention is substantial.
peting model (Model 1).
In summary, the models shown in Figs. 1 and 2 can be ordered 5. Discussion and conclusions
in a nested sequence (nested model) to compare model fit,
exploratory power, and parameter estimates as shown in Table 6. Repeat guests are an essential asset to any successful hotel
The difference in c2 and degrees of freedom (df), the fit statistics business. The most effective way to retain repeat guests is to attain
including absolute fit measures, incremental fit measures, and high customer expectations or to provide a service that exceeds the
parsimony fit measures, as well as exploratory power, indicate that customers’ expectations. Unfortunately, however, flawless cus-
the proposed model (Model 1) is slightly superior to the competing tomer service may be a practically unattainable goal, because the
model (Model 2). hotel business has such unique characteristics as simultaneous
60 T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62

Table 6 higher customer loyalty, whereas the poor responses may prompt
Comparison of the proposed model (Model 1) and the competing model (Model 2) customers to switch hotels. Thus, an effective effort for service re-
Paths Proposed model Competing model covery after experiencing faulty service must be carefully planned
(Model 1) (Model 2) and carried out in order to establish a long-term relationship with
Path coefficients (t-values) the customers.
Direct effects The empirical results testing the relationships between per-
H1a (g11): distributive justice / 0.49 (13.74)** 0.49 (13.74)** ceived justice with service recovery, recovery satisfaction, trust,
recovery satisfaction WOM, and revisit intention demonstrated that all hypothesized
H1b (g12): procedural justice / 0.17 (4.58)** 0.17 (4.58)**
relationships were supported. The impact of DJ on recovery satis-
recovery satisfaction
H1c (g13): interactional justice / 0.24 (6.92)** 0.25 (6.92)**
faction appears to be stronger than that of PJ or IJ, which is consis-
recovery satisfaction tent with the previous findings of Smith et al. (1999) and Maxham III
H2 (b21): recovery satisfaction / trust 0.54 (13.97)** 0.54 (13.97)** and Netemeyer (2002). The significant role of DJ in affecting cus-
H3 (b31): recovery satisfaction / word- 0.37 (10.56)** tomer satisfaction would be supported by fair distributive treatment
of-mouth
such as ‘‘refunds,’’ ‘‘discounts,’’ ‘‘coupons,’’ ‘‘upgrading a room,’’ and
H4 (b41): recovery satisfaction / revisit 0.36 (8.96)**
intention ‘‘offering free room and/or food,’’ which are surely important in
H5 (b32): trust / word-of-mouth 0.52 (15.03)** 0.74 (23.45)** winning back satisfaction from disgruntled hotel guests. The impact
H6 (b42). trust / revisit intention 0.46 (11.46)** 0.67 (19.03)** of IJ on recovery satisfaction is found to be stronger than that of PJ
Indirect effects (Davidow, 2003; Karatepe, 2006). To proceed appropriate IJ, hotel
Distributive justice / trust 0.27 (7.10)** 0.29 (7.11)** staffs may be alerted to respond by offering apologies (Smith et al.,
Procedural justice / trust 0.06 (1.82)n.s. 0.07 (1.85)n.s. 1999), showing empathy and attentiveness (Karatepe, 2006), or
Interactional justice / trust 0.14 (2.65)** 0.14 (2.66)**
appearing courteous and respectful (Blodgett et al., 1997). In all, our
Distributive justice / word-of-mouth 0.32 (8.72)** 0.22 (5.14)**
Procedural justice / word-of-mouth 0.09 (1.91)n.s. 0.08 (1.84)n.s. results imply that when hotel management should exert the service
Interactional justice / word-of-mouth 0.16 (4.26)** 0.11 (2.35)* recovery efforts, they should monitor all three justice dimensions of
Recovery satisfaction / word-of-mouth 0.29 (7.29)** 0.43 (10.01)** recovery outcomes: fair monetary compensation for discontented
Distributive justice / revisit intention 0.31 (7.51)** 0.20 (4.13)** guests (i.e., DJ), existing recovery policy and process to resolve the
Procedural justice / revisit intention 0.09 (1.90)n.s. 0.04 (1.74)n.s.
Interactional justice / revisit intention 0.14 (4.14)** 0.08 (1.79)n.s.
service failure (i.e., PJ), and employee–guest communications (i.e.,
Recovery satisfaction / revisit intention 0.25 (6.93)** 0.39 (10.14)** IJ) (Tax et al., 1998). Regarding the PJ, hotel employees must keep in
mind that timeliness in resolving conflicts (Blodgett et al., 1997) and
Model fit measures
Absolute fit measures high customer voice and the neutrality of the recovery process
c2 27.15 41.09 (Karande et al., 2007; Sparks & McColl-Kennedy, 2001) must be
df 10 12 handled carefully.
Normed c2 2.72 3.42 Our findings also show that the effect of recovery satisfaction on
GFI 0.98 0.97
trust is significant and positive (Ganesan, 1994; Garbarino & John-
RMSEA 0.07 0.08
son, 1999; Tax et al., 1998), and in turn, the trust is positively as-
Incremental fit measures
sociated with WOM and revisit intention (Garbarino & Johnson,
AGFI 0.95 0.90
NFI 0.99 0.97
1999; Kim, Han, & Lee, 2001; Oh, 2002). Particularly, the important
TLI 0.97 0.93 role of trust as an antecedent of WOM and revisit intention co-
incides with Kim et al.’s (2001) findings that trust and satisfaction
Parsimony fit measures
CFI 0.99 0.98 play important roles in stimulating customers’ revisit intentions. In
PGFI 0.78 0.74 the hotel industry, appropriate relationship marketing through
PNFI 0.88 0.83 establishing trust is important to form, maintain, and improve
AIC 64.15 81.09 a sound business relationship with customers. Once satisfaction
BCC 64.78 81.90
Dc2 13.94**
with service recovery is met, trust gives confidence to customers
Ddf 2 about the service provider, so future behavior of the customer can
be voluntarily and positively geared toward the company through
Exploratory power
R2 Recovery satisfaction 0.47 0.43 revisit intention, WOM references, and recommendations.
R2 Truth 0.34 0.34 The significant and positive effect of recovery satisfaction on
R2 Word-of-mouth 0.64 0.43 WOM and revisit intention confirms that service recovery efforts
R2 Revisit intention 0.53 0.42 after failure could ultimately enhance the customer relationship
Note: c2 difference test was computed based on comparison to the proposed model (McCollough, 2000; McDougall & Levesque, 2000). This corrobo-
(Model 1). GFI ¼ goodness-of-fit index, RMSEA ¼ root mean square error of ap- rates that a customer can build on a positive feeling by the hotel’s
proximation, AGFI ¼ adjusted goodness-of-fit index, NFI ¼ normed fit index,
efforts of effective service recovery, and the relationship between
TLI ¼ Tucker–Lewis index, CFI ¼ comparative fit index, PGFI ¼ parsimony goodness-
of-fit index, PNFI ¼ parsimony normed fit index, AIC ¼ Akaike information criterion, the service provider and customer can be fortified (Smith et al.,
BCC ¼ Browne-Cudeck criterion. Normed c2 ¼ c2/df. n.s. indicates not significant. All 1999; Tax et al., 1998). The effective service recovery would be one
texts in bold are used to provide a list of criteria that were used to distinguish model of the most important factors in attaining customer repatronage
1 and model 2. (Blodgett et al., 1997; Wirtz & Mattila, 2004).
*Significant at p < 0.05.
**Significant at p < 0.01.
The results of this study provide useful insights into the be-
havior of hotel guests who have experienced faulty service with the
subsequent follow-up action of service recovery. The study results
production and consumption and intangibleness (Collie et al., should help to enlighten the efforts of any hotel manager who
2000). An effective recovery strategy after service breakdown, pursues to ensure that the guests receiving service recovery efforts
therefore, can play an important role in maintaining or increasing perceive a high level of satisfaction. Some more practical implica-
long-term customer loyalty (Brown et al., 1996; McCollough et al., tions are suggested herewith. First, considering the utmost im-
2000; Maxham III, 2001; Ok et al., 2005; Smith & Bolton, 1998). portance of DJ, hotel management should develop specific
When a service breakdown occurs, the effective reactions of the monetary compensation guidelines while training both full- and
service providers can help create even stronger bonds and lead to part-time employees to quickly and properly react to various
T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62 61

service failure situations. A comprehensive training program can Future research needs to adopt a longitudinal design to trace the
facilitate in selecting the appropriate level of compensation (room relationship between customers’ perceived justice associated with
discount versus upgrading a room). Depending on different levels service recovery over time, because most relationship variables
of complaints and different situations of guests, hotel management may be measured more accurately in the time-series design.
needs to train front desk staff to offer proper arrays of compensa- Second, future research may focus on the empirical validation of
tions with similar values and let their guests select the best com- the recovery paradox across different segments (e.g., accommo-
pensation. Since compensation with less economic value could be dation, restaurant, casino, theme park, and other recreation busi-
more appealing to a displeased guest, offering optimal compensa- nesses) of the tourism industry. Also, the recovery paradox may not
tion with less cost will be a win–win remedy for both an unhappy be realized, because the impact of the first critical failure is too
guest and hotel. For example, a room upgrade may be a better re- devastating to be overcome by recovery efforts in some sectors of
dress for a corporate business traveler who is making a business the tourism industry. However, the recovery paradox could be
trip using business account. However, the same guest would most found in other segments of tourism industry, so that further
likely prefer a room discount to a room upgrade when he/she is on research may confirm it across different types of businesses (e.g.,
vacation. resort hotels vs. downtown hotels or luxury restaurants vs. quick-
In order to enhance the PJ practice, a training program should service restaurants) within a segment of the tourism industry.
focus on instilling the proper procedures and the correct policies by Third, the results of this study need to be cautiously generalized
reacting to customer problems quickly and handling guest com- because the sample is limited to hotel guests from a single country
plaints in a timely manner. In addition, hotel managers should in Asia. It will be interesting to see the findings gleaned from
implement a training program, which clearly illustrates the ex- a similar study conducted on hotel guests from Western cultures.
emplary reactions improving the IJ practice via teaching how to Also, future researchers should conduct cross-cultural studies to
properly treat angry guests, demonstrating empathy and atten- explore the role of culture in justice theory.
tiveness, and offering a genuine apology. Situational questions and Fourth, although this study, like the majority of similar studies,
role playing can be an effective way for guest-contact employees to presumes satisfaction as an antecedent of trust (Garbarino &
learn what to do and what not to do in non-routine situations. Hotel Johnson, 1999; Morgan & Hunt, 1994), another study presumes that
managers should consider suitable rewards and recognition for trust was established as an antecedent of satisfaction (Anderson &
their staff’s smart choice of exemplary recovery efforts to stimulate Narus, 1990). Further research would be able to validate a more
their voluntary participation (Karatepe, 2006). solid basis of causal relationships among recovery satisfaction and
Second, guest-contact employees should be empowered in such trust.
a way as to provide a quick recovery resolution for any service Finally, future research may consider some significant moder-
breakdown. It is important for hotels to incorporate empowerment ating variables between the three dimensions of perceived justice
and mentoring procedures into their operating manuals by clearly and satisfaction, such as the length of relationship with the hotel,
delineating the critical issues such as the maximum monetary in- frequency of hotel stay, and the purpose of customers’ travels.
centive of contact employees without supervisor approval; the
promptness of the responses; and the employees’ behavioral re- Acknowledgement
sponses of showing empathy and compassion, of giving an apology,
and of allowing flexibility of complaining procedures (Tax et a1., The authors would like to thank the three anonymous reviewers
1998). When hotel managers empower their guest-contact em- of Tourism Management for their helpful insights in improving this
ployees, it is important for them to clearly identify authority and paper.
decision-making boundaries. These boundaries need to be revisited
by employees and managers when staff empowerment efforts were
not successful. A caveat for managers is to meet basic needs of References
employees and to make sure that employees are not under- Adams, J. S. (1963). Toward an understanding of inequity. Journal of Abnormal and
compensated, undertrained, overworked, and underappreciated Social Psychology, 67(5), 422–436.
(Karatepe, 2006). Under those negative circumstances, empowered Akaike, H. (1987). Factor analysis and AIC. Psychometrika, 52(3), 317–332.
Anderson, J. C., & Gerbing, D. W. (1982). Some methods of respecifying measure-
employees are not likely to demonstrate any discretionary energy
ment models to obtain unidimensional construct measurement. Journal of
nor to make any extra efforts to remedy service failures. Managers Marketing Research, 19(4), 453–460.
need to coach their employees and to provide proper training and Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A
learning opportunities to assure their employees make better de- review and recommended two-step approach. Psychological Bulletin, 103(3),
411–423.
cisions in the future. It is critical for managers to have a strong trust Anderson, J. C., & Narus, J. A. (1990). A model of distributor firm and manufacturer
in their staff’s capability to respond to complaints or service fail- firm working partnerships. Journal of Marketing, 54(1), 42–58.
ures. It is also crucial for managers to avoid micro-management; Arbuckle, J. L. (2006). AMOS 7.0 user’s guide. Chicago: SPSS.
Bagozzi, R. P. (1980). Causal models in marketing. New York: John Wiley & Sons.
this avoidance will enhance the employees’ morale and team sprit Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychology
leading to customer satisfaction and long-term bottom line. Bulletin, 107(2), 238–246.
Finally, the ultimate goal of service recovery is not limited to Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the
analysis of covariance structures. Psychological Bulletin, 88(3), 588–606.
preventing the loss of customers, but rather to maintain a long- Berry, L. L., & Parasuraman, A. (1991). Marketing services: Competing through quality.
term cooperative relationship with customers. Our findings, then, New York: Free Press.
will be useful to hotel managers for developing procedures that Blodgett, J. G., Granbois, D. H., & Walters, R. G. (1993). The effects of perceived
justice on complaints’ negative word-of-mouth behavior and repatronage
maximize the customer satisfaction with service recovery and intentions. Journal of Retailing, 69(4), 399–427.
subsequently augment long-term customer relationships. Blodgett, J. G., Hill, D. J., & Tax, S. S. (1997). The effects of distributive, procedural,
and interactional justice on postcomplaint behavior. Journal of Retailing, 73(2),
185–210.
6. Limitations and future research directions
Blodgett, J. G., Wakefield, K. L., & Barnes, J. H. (1995). The effects of customer service
on consumer complaining behavior. Journal of Services Marketing, 9(4), 31–42.
Of course, there are some limitations that this study could not Bollen, K. A. (1989). Structural equations with latent variables. New York: John Wiley
escape. First, the results of this study might be affected by re- & Sons.
Brown, J. T., Cowles, D. L., & Tuten, T. L. (1996). Service recovery: Its value and
spondents’ memory bias, because the study was based on the cross- limitations as a retail strategy. International Journal of Service Industry Man-
sectional survey to respondents who had stayed in five-star hotels. agement, 7(5), 32–46.
62 T. (Terry) Kim et al. / Tourism Management 30 (2009) 51–62

Browne, M. W., & Cudeck, R. (1989). Single sample cross-validation indices for attribution based theories. In C. T. Allen, T. J. Madden, T. A. Shimp, R. D. Howell,
covariance structures. Multivariate Behavioural Research, 24(4), 445–455. G. M. Zinkham, & D. D. Heisley (Eds.), Marketing theory and applications (pp.
Chebat, J. C., & Slusarczyk, W. (2005). How emotions mediate the effects of per- 119). Chicago: American Marketing Association.
ceived justice on loyalty in service recovery situation: an empirical study. McDougall, G. H. G., & Levesque, T. (2000). Customer satisfaction with services: putting
Journal of Business Research, 58(5), 664–673. perceived value into the equation. Journal of Service Marketing, 14(5), 392–410.
Clemmer, E. C., & Schneider, B. (1996). Fair service. In T. A. Swartz, D. E. Bowen, & S. Mattila, A. S. (2001). Emotional bonding and restaurant loyalty. The Cornell Hotel
W. Brown (Eds.), Advances in services marketing and management (pp. 109–126). and Restaurant Administration Quarterly, 42(6), 73–79.
Greenwich, CT: JAI Press. Mattila, A. S., & Patterson, P. G. (2004). Service recovery and fairness percep-
Collie, T. A., Sparks, B., & Bradley, G. (2000). Investing in interactional justice: tions in collectivist and individuals contexts. Journal of Service Research,
a study of the fair process effect within a hospitality failure context. Journal of 6(4), 336–346.
Hospitality and Tourism Research, 24(4), 448–472. Maxham, J. G., III (2001). Service recovery’s influence on customer satisfaction,
Davidow, M. (2000). The bottom line impact of organizational responses to cus- positive word-of-mouth, and purchase intentions. Journal of Business Research,
tomer complaints. Journal of Hospitality & Tourism Research, 24(4), 473–490. 54(1), 11–24.
Davidow, M. (2003). Have you heard the word? The effect of word of mouth on Maxham, J. G., III, & Netemeyer, R. G. (2002). Modeling customer perceptions of
perceived justice, satisfaction and repurchase intentions following complaint complaint handling over time: the effects of perceived justice on satisfaction
handling. Journal of Consumer Satisfaction, Dissatisfaction and Complaining and intent. Journal of Retailing, 78(4), 239–252.
Behavior, 16(3), 67–80. Morgan, R. M., & Hunt, S. D. (1994). The commitment-trust theory of relationship
DeWitt, T., Nguyen, D. T., & Marshall, R. (2008). Exploring customer loyalty marketing. Journal of Marketing, 58(3), 20–38.
following service recovery. Journal of Service Research, 10(3), 269–281. Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York: McGraw-Hill.
Ettorre, B. (1994). Phenomenal promises that mean business. Management Review, Oh, H. (2002). Transaction evaluations and relationship intentions. Journal of Hos-
83(3), 18–23. pitality and Tourism Research, 26(3), 278–305.
Fornell, C., & Larcker, D. (1981). Evaluating structural equation models with un- Ok, C., Back, K., & Shanklin, C. W. (2005). Modeling roles of service recovery
observable and measurement error. Journal of Marketing Research, 18(1), 39–50. strategy: a relationship-focused view. Journal of Hospitality and Tourism Re-
Ganesan, S. (1994). Determinants of long-term orientation in buyer–seller search, 29(4), 484–507.
relationships. Journal of Marketing, 58(2), 1–19. Reichheld, F. F., & Sasser, W. E. (1990). Zero defections: quality comes to services.
Garbarino, E., & Johnson, M. S. (1999). The different role of satisfaction, trust, and Harvard Business Review, 68(5), 105–111.
commitment in customer relationships. Journal of Marketing, 63(2), 70–87. Rust, R. T., Lee, C., & Valente, E. (1995). Comparing covariance structure models:
Goodwin, C., & Ross, I. (1992). Consumer responses to service failures: Influences of a general methodology. International Journal of Research in Marketing, 12(4),
procedural and interactional fairness perceptions. Journal of Business Research, 279–291.
25(2), 149–163. de Ruyter, K., & Wetzels, M. (2000). Customer equity considerations in service re-
Grönroos, C. (1990). Relationship marketing approach to the marketing function in covery: a cross-industry perspective. International Journal of Service Industry
service contexts: the marketing and organizational behavior influence. Journal Management, 11(1), 91–108.
of Business Research, 20(1), 3–12. Sanders, N. R., & Premus, R. (2005). Modeling the relationship between firm IT
Gwinner, K. P., Gremler, D. D., & Bitner, M. J. (1998). Relational benefits in services capability, collaboration, and performance. Journal of Business Logistics, 26(1),
industries: the customer’s perspective. Journal of the Academy of Marketing 1–23.
Science, 26(2), 101–114. Selnes, F. (1998). Antecedents and consequences of trust and satisfaction in buyer–
Hair, J. F., Anderson, R. E., Taltam, R. L., & Black, W. C. (1998). Multivariate data seller relationships. European Journal of Marketing, 32(3/4), 305–322.
analysis with readings (5th ed.). New York: Prentice-Hall. Smith, A. K., & Bolton, R. N. (1998). An experimental investigation of customer re-
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure actions to service failure and recovery encounters: paradox or peril. Journal of
analysis: conventional criteria versus new alternatives. Structural Equation Service Research, 1(1), 65–81.
Modeling, 6(1), 1–55. Smith, A. K., Bolton, R. N., & Wagner, J. (1999). A model of customer satisfaction with
Jöreskog, K. G., & Sörbom, D. (1993). LISREL 8.0: User’s reference guide. Chicago: SSI. service encounters involving failure and recovery. Journal of Marketing Research,
Karande, K., Magnini, V. P., & Tam, L. (2007). Recovery voice and satisfaction after 36(3), 356–373.
service failure: an experimental investigation of mediating and moderating Sparks, B. A., & McColl-Kennedy, J. R. (2001). Justice strategy options for increased
factors. Journal of Service Research, 10(2), 187–203. customer satisfaction in a service recovery setting. Journal of Business Research,
Karatepe, O. M. (2006). Customer complaints and organizational responses: The 54(3), 209–218.
effects of complaints’ perceptions of justice on satisfaction and loyalty. In- Sparks, B., & Fredline, L. (2007). Providing an explanation for service failure:
ternational Journal of Hospitality Management, 25(1), 69–90. context, content, and customer responses. Journal of Hospitality and Tourism
Kelloway, E. K. (1998). Using LISREL for structural equation modeling: A researcher’s Research, 31(2), 241–260.
guide. Thousand Oaks, CA: Sage. Stoller, G. (2005). Companies give front-line employees more power: travel industry
Kim, W. G., Han, J. S., & Lee, E. (2001). Effects of relationship marketing on repeat workers fix customer problems with cash, free rooms. <http://www.usatoday.
purchase and word of mouth. Journal of Hospitality and Tourism Research, 25(3), com/money/companies/management/2005-06-26-service-usat_x.htm>.
272–288. Susskind, A. (2002). I told you so! Restaurant customers’ word-of-mouth commu-
Kline, P. (1986). A handbook of test construction: Introduction to psychometric design. nication patterns. The Cornell Hotel and Restaurant Administration Quarterly,
London: Methuen. 43(2), 75–85.
Lewis, B. R., & McCann, P. (2004). Service failure and recovery: evidence from the Tax, S. S., Brown, S. W., & Chandrashekaran, M. (1998). Customer evaluations of
hotel industry. International Journal of Contemporary Hospitality Management, service complaint experiences: implication for relationship marketing. Journal
16(1), 6–17. of Marketing, 62(2), 60–76.
McCollough, M. A. (2000). The effect of perceived justice and attributions re- Wirtz, J., & Mattila, A. S. (2004). Consumer responses to compensation, speed of
garding service failure and recovery on post-recovery customer satisfaction recovery and apology after a service failure. International Journal of Service In-
and service quality attributes. Journal of Hospitality and Tourism Research, dustry Management, 15(2), 150–166.
24(4), 423–447. Wong, A., & Sohal, A. (2002). An examination of the relationship between trust,
McCollough, M. A., Berry, L. L., & Yadav, M. S. (2000). An empirical investigation of commitment, and relationship quality. International Journal of Retail & Distri-
customer satisfaction after service failure and recovery. Journal of Service bution Management, 30(1), 34–50.
Research, 3(2), 121–137. Yuksel, A., Kilinc, U. K., & Yuksel, F. (2006). Cross-national analysis of hotel cus-
McCollough, M. A., & Bharadwaj, S. G. (1992). The recovery paradox: an examination tomers’ attitudes toward complaining and their complaining behaviours.
of consumer satisfaction in relation to disconfirmation, service quality, and Tourism Management, 27(1), 11–24.

Das könnte Ihnen auch gefallen