22
Evaluating stability of the performance-satisfaction relationship across selected lodging market segments Haemoon Oh and Miyoung Jeong Department of Hospitality and Tourism Management, Isenberg School of Management, University of Massachusetts Amherst, Amherst, Massachusetts, USA Abstract Purpose – The purpose of this paper is to illustrate new methods of examining structural differences among segmented markets beyond comparing merely univariate variable mean scores, so as to help marketers and researchers gain better insights into segment differences for meaningful strategy development. Design/methodology/approach – A comprehensive dataset covering various lodging market segments was constructed from Tripadvisor.com. The data then were sorted into lodging customer segments by star rating, type of operation, and level of price charged. Structural equation modeling with the 22 log-likelihood difference test was conducted to illustrate how effectively the differences, if any, of market segments could be assessed in contrast to the traditional mean-score comparison approach. Findings – Guest satisfaction was influenced by the same performance variable to the same magnitude and direction across different lodging segments examined. Such stability in the amount of influence of performance on guest satisfaction was true even in the fact that the variable mean scores were significantly different across the market segments. Research limitations/implications – The traditional approach to examining segment differences via univariate mean scores could be one set of results, while the effect-based difference assessments in this paper resulted in another. Developing marketing strategies based on the effect-based segment differences, as illustrated in this paper, is considered more effective than the traditional mean-based approach. One limitation of this paper could be use of a secondary dataset with limited scope of the model employed for an illustrative purpose. Another limitation is that the sample characteristics are unknown due to the nature of a secondary dataset. The examination of the market segments was also limited to those based on only three popular variables. Originality/value – The paper is a fresh attempt to examine market segment differences through the effect of one variable on another. The paper advances the methods of hospitality and tourism research for examining segment differences beyond the traditional univariate mean-based examination approach. The methodological illustration is applicable to a vast majority of different theoretical frameworks known in the hospitality and tourism field. Use of the assessment method illustrated in this paper also requires future market segmentation studies to rely more on theories than data. Keywords Business performance, Customer satisfaction, Market segmentation, Hotels Paper type Research paper Introduction The lodging industry has been competing on guest satisfaction and such competition has become more intense in recent decades. While companies with limited resources have struggled to find ways to satisfy guests’ ever-increasing demands, those with more The current issue and full text archive of this journal is available at www.emeraldinsight.com/0959-6119.htm Performance- satisfaction relationship 953 Received 27 June 2009 Revised 19 October 2009, 5 February 2010 Accepted 13 March 2010 International Journal of Contemporary Hospitality Management Vol. 22 No. 7, 2010 pp. 953-974 q Emerald Group Publishing Limited 0959-6119 DOI 10.1108/09596111011066626

Evaluating stability of the performance‐satisfaction relationship across selected lodging market segments

  • Upload
    miyoung

  • View
    212

  • Download
    0

Embed Size (px)

Citation preview

Evaluating stability of theperformance-satisfaction

relationship across selectedlodging market segments

Haemoon Oh and Miyoung JeongDepartment of Hospitality and Tourism Management,

Isenberg School of Management, University of Massachusetts Amherst,Amherst, Massachusetts, USA

Abstract

Purpose – The purpose of this paper is to illustrate new methods of examining structural differencesamong segmented markets beyond comparing merely univariate variable mean scores, so as to helpmarketers and researchers gain better insights into segment differences for meaningful strategydevelopment.

Design/methodology/approach – A comprehensive dataset covering various lodging marketsegments was constructed from Tripadvisor.com. The data then were sorted into lodging customersegments by star rating, type of operation, and level of price charged. Structural equation modeling withthe22 log-likelihood difference test was conducted to illustrate how effectively the differences, if any, ofmarket segments could be assessed in contrast to the traditional mean-score comparison approach.

Findings – Guest satisfaction was influenced by the same performance variable to the samemagnitude and direction across different lodging segments examined. Such stability in the amount ofinfluence of performance on guest satisfaction was true even in the fact that the variable mean scoreswere significantly different across the market segments.

Research limitations/implications – The traditional approach to examining segment differencesvia univariate mean scores could be one set of results, while the effect-based difference assessments inthis paper resulted in another. Developing marketing strategies based on the effect-based segmentdifferences, as illustrated in this paper, is considered more effective than the traditional mean-basedapproach. One limitation of this paper could be use of a secondary dataset with limited scope of themodel employed for an illustrative purpose. Another limitation is that the sample characteristics areunknown due to the nature of a secondary dataset. The examination of the market segments was alsolimited to those based on only three popular variables.

Originality/value – The paper is a fresh attempt to examine market segment differences through theeffect of one variable on another. The paper advances the methods of hospitality and tourism researchfor examining segment differences beyond the traditional univariate mean-based examinationapproach. The methodological illustration is applicable to a vast majority of different theoreticalframeworks known in the hospitality and tourism field. Use of the assessment method illustrated in thispaper also requires future market segmentation studies to rely more on theories than data.

Keywords Business performance, Customer satisfaction, Market segmentation, Hotels

Paper type Research paper

IntroductionThe lodging industry has been competing on guest satisfaction and such competitionhas become more intense in recent decades. While companies with limited resourceshave struggled to find ways to satisfy guests’ ever-increasing demands, those with more

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/0959-6119.htm

Performance-satisfactionrelationship

953

Received 27 June 2009Revised 19 October 2009,

5 February 2010Accepted 13 March 2010

International Journal ofContemporary Hospitality

ManagementVol. 22 No. 7, 2010

pp. 953-974q Emerald Group Publishing Limited

0959-6119DOI 10.1108/09596111011066626

resources have made systematic investments, such as periodic guest surveys andsatisfaction guarantee programs, to manage guest satisfaction more competitively(Shoemaker and Lewis, 1999). It is not uncommon today that lodging companies, smallor large, implement certain forms of satisfaction research as a part of ongoingoperational efforts to measure guest satisfaction. Such efforts may range from a simplecomment card placed in guest rooms to a formal satisfaction research programoutsourced to firms like Gallup and J.D. Power and Associates. Apparently, high guestsatisfaction still remains one of the most glorified goals for most lodging operationstoday (Strauss, 2004).

A close look at the industry’s practice as well as a broad review of research tomeasure guest satisfaction indicates that guest satisfaction derives largely from thecompany’s positive performance. The company’s performance has taken many formsin both concept and operationalization, be it performance as compared to expectations(Oh and Parks, 1997; Oliver, 1997), emotional experience (Barsky and Nash, 2003),value perceptions (Mattila, 1999; Oh, 1999), service quality (Parasuraman et al., 1994;Saleh and Ryan, 1991), or just perceived performance itself (Cronin and Taylor, 1992;Lewis, 1985). Although divergent in approaches, these performance measurementefforts are commonly grounded in a multivariate model in that the company’sperformance is measured on multiple attributes or dimensions to predict anotherphenomenon such as overall guest satisfaction (Lewis, 1985; Saleh and Ryan, 1992).

While guest satisfaction programs continue to proliferate in the lodging industry,whether operational performance predicts guest satisfaction consistently acrossdifferent market segments has received limited academic and practical inquiries.The lodging market is often segmented by level of service, level of price or room ratescharged, and/or type of operation. Market segmentation by level of service, forexample, is reflected most popularly in the Mobil Travel Guide’s five-star ratingsystem or other similar hotel rating systems. Segmentation by price tends to becorrelated with that by star rating systems, although price perceptions may varyacross geographic market situations. For type of operation, the common dichotomy ofchain-affiliated vs independent operations is widespread, with the former associatedmore strongly with branded hotels than the latter (Holverson and Revaz, 2006). Suchcommonly practiced market segmentations are based on the assumption that eachsegment is distinct from another, resulting in various marketing implications. Thisassumption, however, has eluded systematic inquiries, although the practical relevanceof such inquiries is promising.

Of substantive theoretical and practical interests, too, is whether the predictiverelationship between lodging performance and guest satisfaction is consistent in itsdirection and strength across market segments. In particular, when the company’sperformance is measured on multiple attributes, would the extent to which eachattribute’s performance predicts guest satisfaction remain constant across marketsegments as well as between the attributes within the same market segment? Studiesare needed to determine whether operators can rely on the same attribute to the sameor different extent for predicting guest satisfaction when dealing with differentsegments. Such studies will also allow reassessing generalizability of theperformance-satisfaction (P-S) relationship across various lodging markets. Althoughthe differential effects of performance attributes on satisfaction have been examinedin many lodging studies (Fang et al., 2008; Matzler et al., 2008), their between-market

IJCHM22,7

954

segment stability has not received systematic evaluations in the relevant literature,especially in the context of lodging operations and guest satisfaction.

Using a large dataset developed from Tripadvisor.com, a web site recentlypopularized for providing travelers’ feedback on various lodging and travel-relatedexperiences, this study examined the related issues. First, this study shows thatmarketers need to examine differences among segmented markets beyond comparingmerely mean scores of performance attributes and guest satisfaction to betterunderstand how each attribute predicts guest satisfaction in different market segments.Second, this study reveals that even if travelers’ perceptions of a lodging company’sperformance and their satisfaction level may differ significantly across marketsegments, the strength and direction of the P-S relationship may not necessarily be so.Tested additionally in this study is whether the P-S relationship remains constant notonly across different lodging segments but also among key performance attributeswithin each segment. In essence, the central inquiry of this study is about the predictivestability of the same performance attribute in different market situations, beyond meresignificant tests of either mean equality or differences of performance and satisfaction inisolation.

1. Lodging market segmentsMarket segmentation is typically driven by industry’s needs to better suit its offerings tothe heterogeneous demands of customers. Hospitality research on market segmentationhas been largely method- or data-driven without establishing a conspicuous stream ortradition of related research (Oh et al., 2004). Hence, studies of market segmentation aresomewhat fragmentary and our review in this section focuses on the three popularsegmentation variables (i.e. star rating, price, and type of operations) for the purpose ofour study.

As most marketing management texts illustrate (Kotler, 1991; Shoemaker et al., 2006),the consumer market can be segmented in a variety of ways using differentsegmentation criteria. In certain markets, it is observed that segmentation tends to entaila highly “vertical” structure, while a more “horizontal” structure of segmentation is thecase in other markets (Bolton and Myers, 2003; Mentzer et al., 2004). Marketsegmentation can be said vertical when the segments are ordered along the increasing ordecreasing magnitude of a particular segmentation criterion. When consumers aregrouped by their income, age, or education level, for example, the segmentation tends tobe vertical. In contrast, horizontal segmentation tends to produce market segments as aresult of differentiating consumers by the diversity of a target criterion attribute.Consider, for example, market segments based on marital status, gender, lifestyle, andpreference.

The lodging market is frequently segmented along the vertical layers of qualityand quantity of amenity, service, and experience for customers (Kutzner et al., 2009;Weaver et al., 2009). Such segmentation is perhaps most popularly reflected in the starrating system and its variations (http://goeurope.about.com/cs/hotels/a/hotel_stars.htm). While star ratings of hotels are often inconsistently used (www.usatoday.com/travel/columnist/burbank/2005-08-23-column_x.htm), the Mobil Travel Guide(http://mobiltravelguide.howstuffworks.com) provides the star rating system as abasic platform for segmenting the lodging market in its 50 years of tradition.For example, well-known internet intermediaries such as Hotels.com, Expedia.com,

Performance-satisfactionrelationship

955

and Orbitz.com use star ratings to segment lodging markets or properties: one star foreconomy hotels and motels, two stars for value properties, three stars for mid-scalehotels, four stars for deluxe/upscale hotels, and five-stars for luxury hotels (www.orbitz.com/pagedef/content/hotel/popupStarRatingGuide.jsp?popupsDisabled¼false).Apparently, the star rating system corresponds closely to another method of verticallysegmenting the lodging market into the economy, budget, mid-scale, upscale, and luxurymarkets, as used by companies like Smith Travel Research (www.hospitalitynet.org/organization/17001110.search?query¼smithþtravelþresearch).

Price or room rate is another variable that is conveniently used to segment thelodging market vertically. Price influences consumer choice in various stages of apurchase decision. Marketers often heuristically view segments of homogeneousdemands by different price levels (Bojanic, 1996; Varini et al., 2003). Although thelodging market may be segmented roughly into high-, mid-, and low-priced markets(Knutson et al., 1992), it is not clear exactly where the actual price breaks should occur.The reason is that price, especially in the sense of a nominal price, is an elusiveconcept for the purpose of understanding consumer behavior (Monroe, 1990). That is,consumers interpret a price in many different ways due to numerous variables affectingthe interpretation. Discretionary income, perceived gain and loss (or sacrifice) in thetransaction, perceived demands of the product, and market economy are just a fewexamples of such influential variables. To illustrate, the same hotel room rate can beinterpreted quite differently by the same consumer depending upon which geographicmarket the hotel is located. While caution is needed in the price-based approachto market segmentation, it is intuitive that price-based segmentation results willcorrelate highly with star rating or similar segmentation results (Henley et al., 2004;Osman et al., 2009).

A horizontal segmentation of the lodging market may follow grouping customersby the property or brand they prefer to patronize. Some customers are more consciousabout using branded products than others; they may prefer staying at a brandedhotel to venturing to a relatively unknown property. One heuristic way to observebrand-conscious lodging customers is to study customer groups patronizingchain-affiliated in contrast to independently branded (“independent” hereafter)properties. In the lodging market, it is common that many branded properties tend tobe affiliated with a chain, while most independent hotels tend to be perceived lessbranded (Holverson and Revaz, 2006; Imrie and Fyall, 2000). That is, brand-consciouscustomers tend to prefer staying at chain-affiliated hotels to settling with non-chainindependent hotels. This is particularly likely when the customer is unfamiliar with thedestination, is uncertain about quality of the hotel, and is under time pressure tomake a purchase decision, among other reasons (Slevitch and Sharma, 2008; Wong andYeh, 2009).

Despite the widely practiced lodging market segmentation by star rating, pricelevel, and somewhat less frequently type of operations, studies assessing segment-wisedifferences along these variables are scarce. Moreover, the majority of hospitalitymarket segmentation studies compared segmented markets based limitedly onattribute-specific ratings (Griffin et al., 1996; Leahy et al., 2009; Varini et al., 2003;Weaver et al., 2009). Few studies investigated segment differences beyond mereattribute-based ratings. New in this study was an attempt to assess how the strengthof the proven theoretical relationship between performance (i.e. an attribute)

IJCHM22,7

956

and satisfaction (i.e. another attribute) changes across the lodging markets segmentedby the three variables (for theoretical expositions on the relationship (Oh and Parks,1997; Szymanski and Henard, 2001). By doing so, we also evaluate the stability orgeneralizability of the P-S relationship across different market situations.

2. Effects of star rating, price, and chain affiliation on the P-S relationshipThe marketing and hospitality literature is replete with reports on how performanceperceptions affect customer satisfaction. Szymanski and Henard (2001) providea meta-analytic review of the relationship, while Oh and Parks (1997) provided ahospitality-specific review of the topic. In general, the P-S relationship is rooted inexpectancy disconfirmation theory in which satisfaction is posited to derive from apositive confirmation of company performance against expectations held prior tothe purchase (Oliver, 1981). When negatively confirmed against expectations, companyperformance is known to result in customer dissatisfaction. This relationship has beenshown to be linear in most cases, although boundary conditions for such a linearrelationship are relatively unknown. It is important for advancing theories thattheoretical relationships are understood in tandem with their boundary conditions(Oakley et al., 2004).

Market segmentation variables are logical candidates as boundary conditions ormoderators of a theoretical relationship. In practice, a market is segmented in a way tomaximize known structural differences among the segmented markets. Thus, whendealing with segmented markets, a generalized belief or assumption is that customersbelonging to different segments will exhibit significant differences on the chosensegmentation criteria relevant to the marketer’s interests. Tests of such differences areoften demonstrated in post hoc analysis of variance (ANOVA) in market segmentationstudies (Chen, 2001; Gustin and Weaver, 1993; Oh and Jeong, 1996). Nevertheless, suchdifference tests have been limited to univariate or multivariate mean scores, withoutbeing extended to theoretical relationships such as the one between performance andsatisfaction. Thus, if applied to theoretical relationships between variables in addition toonly the mean scores of the variables, explicit tests of between-segment differences willadvance our understanding of the theoretical relationships. For the purpose of thisstudy, therefore, we asked: would the P-S relationship remain consistent, or vary, in bothdirection and magnitude across market segments in addition to the differences invariable mean scores?

The direction and magnitude of the P-S relationship may remain constant acrossstar-rated lodging segments, even if such segmentation suggests the opposite. Whileevidence is abundant showing differences in the mean scores of such variables asperformance and guest satisfaction across star-rated lodging markets (Henley et al.,2004; Ingram and Daskalakis, 1999; Law and Hsu, 2006; Ryan and Huimin, 2007), studiesexamining stability of the P-S relationship (again, in terms of its direction andmagnitude) across the star segments are rare. As postulated in expectancydisconfirmation theory, satisfaction is a joint function of expectations andperformance perceptions (Churchill and Surprenant, 1982; Oliver, 1981; Szymanskiand Henard, 2001). Thus, as star rating goes up, so do expectations, followedcorrespondingly by performance perceptions (Carman, 1990; Fernandez and Bedia,2004; Ingram and Daskalakis, 1999). Satisfaction is consequential. What this impliesis that the theoretical P-S relationship may be preserved through the systematic

Performance-satisfactionrelationship

957

operations of expectations, perceptions, and satisfaction in the same direction acrossdifferent star ratings. Therefore, we expect that the P-S relationship will remain constantacross the star-rated market segments; it would not vary just as the mean scores ofperformance (“P”) and guest satisfaction (“S”) might do.

Price is not likely to undermine the stability of the P-S relationship, either. Priceis known to affect many behavioral judgments such as quality, value, and perhapssatisfaction (Monroe and Lee, 1999; Rao and Monroe, 1989). While price effects may besalient on individual variable mean scores (Bojanic, 1996; Knutson et al., 1992), the sameeffects on theoretical relationships such as the P-S relationship are not well documented.In the economy of imperfect information (Monroe, 1990), price often sets up expectationsof performance, which accompanies a corresponding level of satisfaction (O’Neill andMattila, 2006). This is particularly likely in reality because in most cases, customersfreely choose an accommodation option as a result of a successful trade-off between priceand expected performance (see Rao and Monroe, 1989 for meta-analysis results). Hence,the P-S relationship across price-based lodging market segments is expected to be stablein both direction and magnitude as long as the price range is within the latitude of anormal market economy. In this study, we investigate the stability of the P-Srelationship across lodging markets segmented by actual prices, with the expectationthat it will stay constant as a result of the customer’s synchronous adjustments of priceand performance expectations.

Finally, the segment-specific differences in the P-S relationship may be tenable whenthe market is segmented by type of operations, i.e. chain-affiliated vs independent, asmuch as such market segmentation reflects the degree of branding or quality offeringsas discussed earlier (O’Neill and Xiao, 2006; Weber, 2001; Yeung and Law, 2004). Thecustomer who is loyal to branded chain hotels may process the hotel’s performance moreelaborately than the customer who patronizes unbranded independent hotels. Such anelaborate processing may cause deeper involvement in information processing leadingto increased reliance on the processed information for subsequent judgments andfeelings such as satisfaction (Bettman, 1979). This is particularly true if the customerintends to remain loyal to the hotel in the future (Barsky and Nash, 2005; Imrie and Fyall,2000). Therefore, it is generally expected that the P-S relationship will be stronger in thechain-affiliated lodging segment than in the independent lodging segment.

3. The P-S model and effect assessmentsFigure 1 shows the P-S model used in this study, along with plans for testing stability ofthe relationships across both the market segments and variables. In the model, which is agraphical representation of a regression model, four performance variables – room,service, price-value, and cleanliness – predict overall guest satisfaction. These fourperformance variables were chosen for illustrative purposes and they are known to beimportant predictors of guest satisfaction in the lodging industry (Lewis, 1985; Weber,2001). The direction and magnitude of each relationship is captured in beta. Identical testprocedures were applied to all lodging segments examined in this study. First, we testedwhether the customers of different segments came from the same population, which wasdone by testing the equality of the total data matrix (

PG, i.e. equality of variances andcovariances of all variables, where

Pdenotes the data matrix and superscript G

indicates group or lodging segment).

IJCHM22,7

958

Second, we tested whether the direction and magnitude of the P-S relationship wasconstant across different lodging segments (BG, where B denotes the vector of effectbi and i ¼ 1, 2, 3, and 4). This test was followed by two logically sequential tests: onewas whether the error of the model (i.e. prediction error) was equal across the lodgingsegments (CG, where C means the vector of model errors – each segment-specificmodel had one value for this in our study) and the other whether both each relationshipand prediction error were equivalent across the lodging segments (BG þ CG).

Tested next was the equality of the P-S relationship across the four performancevariables within each segment (bG

i ), followed by a series of pairwise equality tests ofthe relationship for more detailed results. These results could tell whether the differentperformance variables determined guest satisfaction to the same degree within eachsegment. Finally, equality of all P-S relationships both across and within lodgingsegments was tested simultaneously (BG þ bG

i ). This test could show whether

Figure 1.The P-S prediction model

for testing invariance

β1

β2

β3

β4

Room performance

Overall guestsatisfaction

Service performance

Price-value performance

Cleanliness performance

Invariance to be tested (“G” indicates group or market segment):

1. Invariance of variance and covariance matrices: ΣG1 = ΣG2 = ΣG3 =…

2. Invariance of effects or relationships across segments:

a. β1G1 = β1

G2 = β1G3 = …

b. β2G1 = β2

G2 = β2G3 = …

c. β3G1 = β3

G2 = β3G3 = …

d. β4G1 = β4

G2 = β4G3 = …

3. Invariance of model errors: ΨG1 = ΨG2 = ΨG3 = …

4. Invariance of both relationships and model errors across segments:

a. β1G1 and ΨG1 = β1

G2 and ΨG2 = β1G3 and ΨG3 = …

b. β2G1 and ΨG1 = β2

G2 and ΨG2 = β2G3 and ΨG3 = …

c. β3G1 and ΨG1 = β3

G2 and ΨG2 = β3G3 and ΨG3 = …

d. β4G1 and ΨG1 = β4

G2 and ΨG2 = β4G3 and ΨG3 = …

5. Invariance of relationships within segment (or equality of regression weights):

a. β1G1 = β2

G1 = β3G1 = β4

G1

b. β1G2 = β2

G2 = β3G2 = β4

G2, etc.

6. Invariance of relationships both across and within segment simultaneously:

β1Gj = β2

Gj = β3Gj = β4

Gj (where j = 1, 2, …, k industry segments)

Performance-satisfactionrelationship

959

the direction and magnitude of all P-S relationships were identical not only across thelodging segments but also within each segment.

One of the study’s main objectives was to test the equality of between-segment P-Srelationships. Such tests would reveal whether the same P-S relationship wasgeneralizable across different lodging segments. Tests like these follow a multi-groupanalysis approach (Joreskog and Sorbom, 2006), providing straightforward proceduresand results as compared to the traditional ordinary least square procedure. Weemployed a multi-group analysis coupled with the logic of the x 2 difference test( Joreskog and Sorbom, 2006).

In a general sense, thex 2 difference test is conducted to determine whether additionalconstraints on a model are acceptable, which is determined based on the model fitindices increasing with the constraints imposed. In this study, the P-S relationships wereconstrained to be equal across different lodging segments. This (alternative) model withconstraints is called a nested model and, theoretically, its model fit is going to get onlyworse due to the constraint(s) forced in the data. Note that a nested model will gain indegrees of freedom corresponding to the number of constraints imposed. The test logichere is that the nested model is acceptable – i.e. the imposed constraint(s) is statisticallyvalid – if the fit of the nested model does not deteriorate more than the theoreticallyexpected threshold value following the x 2 distribution.

4. DataTo test whether the P-S relationship would be constant across lodging marketsegments, we needed a comprehensive dataset covering the lodging segments by starratings, type of operation, and level of price. Collecting broadly structured data like thiswould be prohibitive in terms of time and cost; suitable secondary data were rare.Hence, we extracted and organized customer reviews posted on TripAdvisor.comduring a recent three-week period. TripAdvisor LLC is a company based in Needham,Massachusetts, collecting and publishing customers’ feedback for various lodgingfacilities, travel and vacation packages, travel guides, and others. The company makesavailable on its web site both qualitative and quantitative review summaries providedvoluntarily by lodging customers around the world. Customers provide theirevaluations of lodging companies they experienced by using seven variables on afive-point rating scale (5 – best or most satisfied). The seven variables are room,service, price-value, cleanliness, pool, dining, and overall satisfaction. We also visitedthe web site of every reviewed hotel to collect supplementary information such as thetype of hotel operation and the average room rate charged by the hotel as posted on thehotel web site during the three-week period.

For the purpose of this study, we used only five of the seven performance variables;pool and dining were excluded because they were not standard items for many lowerend lodging segments. The reduced set of five measurement variables was in line withthose variables included in other lodging performance studies such as Lewis (1985) andWeber (2001) and industry surveys such as that of J.D. Power and Associates andLockyer (2005).

Our data covered TripAdvisor.com-rated top 20 US tourist destinations with anexclusion of three destinations. The chosen destinations were Atlanta (31 hotels), Austin(40), Branson (28), Chicago (45), Fort Lauderdale (41), Houston (41), Indianapolis (41),Kissimmee (32), Los Angeles (44), Miami Beach (42), Myrtle Beach (28), New Orleans (36),

IJCHM22,7

960

New York (50), Orlando (50), San Antonio (38), San Diego (45), and San Francisco (50).These top destinations were chosen as they were believed to accommodate variouslodging needs of travelers and, hence, the results could be more generalized. Weexcluded Island of Hawaii, Maui, and Las Vegas, because the hotels in the first twodestinations were outliers in prices while Las Vegas was thought to represent somewhatatypical lodging needs. Another reason that we focused on these popular destinationswas that many of the less popular destinations did not either feature a sufficient varietyof hotels covering diverse lodging segments or result in a sufficient number of customerreviews for many individual hotels.

The unit of our analyses was lodging property, i.e. individual hotel, aggregatingcustomer reviews at the hotel level. The data were organized based on the threesegmentation variables (i.e. star rating, type of operation, and price) after we extractedcustomer reviews for ten hotels per each of the five-star ratings within each selected city.This data method was expected to result in 50 hotels per city for a total expected samplesize of 1,000 hotels (20 destinations £ five-star ratings £ ten hotels per city). Thisprocedure was to assure eventually a sufficiently large sample size for each lodgingsegment for the planned analyses. However, the three destinations were alreadyexcluded; in addition, many destinations did not host either a sufficiently large numberof hotels in some categories (e.g. five-star hotels) or a sufficiently large number ofcustomer reviews for some chosen hotels. Consequently, the sample size did not realizeour goal of 50 hotels in most selected destinations, resulting in 682 hotels from17 destinations. These hotels were also organized by type of operation and level of price,based on the additional information we collected directly from each hotel’s web site.

The resulting dataset was considered large enough to produce generalizable resultsas well as cover the various lodging segments focused in our investigation. The682 hotels represented 1,66,261 rooms (mean ¼ 255; median ¼ 151; range ¼ 15-2,880,per hotel) and 61,997 customer reviews (mean ¼ 91; median ¼ 53; range ¼ 1-1,035). Theaverage room rate was $211.91with the median of $174.50, and a range of $37-$1,326.At least, 28 hotels (Branson and Myrtle Beach; up to 50 hotels for New York, Orlando,and San Francisco) and 516 customer reviews (Branson; up to 12,471 reviews forNew York) represented each chosen city in the data. The univariate distribution of thefive variables (i.e. room, service, price-value, cleanliness, and satisfaction) wascomparable and close to normal, with the skewness values falling within þ 1.2 andkurtosis within þ 2.0 (Neter et al., 1990). These distributional characteristics, coupledwith the mean scores shown in Table I, indicated that the data could represent a broadspectrum of lodging evaluations, not just extremely angry or happy customers who weremore likely to voice their opinions. Additional preliminary analyses provided evidencefor face validity of the data. For example, the correlation between star ratings (i.e. fivecategories) and price levels (four categories) was 0.68, statistically significant asexpected ( p , 0.001). Higher star-rated hotels tended to charge higher prices(x2 ¼ 442.2, df ¼ 12, p , 0.001). Independent hotels tended to charge higher roomrates than chain-affiliated hotels ($244.27 vs $201.09; t ¼ 3.32, p , 0.001). Perhaps, thescale of economy with chain-affiliated hotels could give this pricing advantage.

5. Mean score-based assessments of segment differencesTo understand differences among the lodging market segments by the variable meanscores, we compared the variable mean scores by lodging segment via one-way

Performance-satisfactionrelationship

961

Nu

mb

erof

rev

iew

sS

atis

fact

ion

Roo

mR

atin

g1

Ser

vic

eP

rice

-val

ue

Cle

anli

nes

sS

egm

ent

Nu

mb

erof

hot

els

Mea

nM

edia

nM

ean

SD

Mea

nS

DM

ean

SD

Mea

nS

DM

ean

SD

Overall

682

9053

3.80

0.56

3.82

0.58

3.91

0.50

3.76

0.50

4.00

0.56

On

est

ar13

452

163.

21a

0.62

3.20

a0.

633.

39a

0.56

3.36

a0.

623.

40a

0.64

Tw

ost

ars

165

5931

3.76

b0.

453.

74b

0.49

3.83

b0.

443.

80b

0.42

3.95

b0.

48T

hre

est

ars

170

9959

4.10

c0.

364.

07c,

d0.

374.

15c

0.34

4.04

c0.

364.

20c

0.36

Fou

rst

ars

144

130

853.

92d

0.42

4.02

c0.

424.

07c

0.35

3.77

b0.

434.

19c

0.37

Fiv

est

ars

6914

112

44.

10c

0.37

4.17

d0.

374.

15c

0.35

3.76

b0.

334.

32d

0.38

F-v

alu

e281

.71

87.9

878

.14

43.8

480

.31

Operation

type

Ch

ain

-affi

liat

ed51

172

423.

760.

573.

810.

593.

890.

513.

730.

513.

970.

57In

dep

end

ent

171

146

953.

920.

503.

870.

543.

990.

493.

860.

464.

080.

51t-

val

ue2

3.31

1.15

ns

2.22

2.91

2.30

Price

level3

Les

sth

an$1

0013

634

123.

42a

0.58

3.42

a0.

533.

55a

0.51

3.53

a0.

553.

59a

0.55

$100

-$19

9.99

267

6435

3.79

b0.

563.

77b

0.59

3.90

b0.

513.

82b

0.55

3.97

b0.

58$2

00-$

299.

9914

513

410

73.

89b

0.43

3.96

0.46

4.00

b0.

423.

80b

0.43

4.10

0.41

$300

orm

ore

134

156

115

4.11

0.37

4.18

d0.

414.

190.

323.

85b

0.38

4.34

d0.

36F-v

alu

e243

.29

51.0

744

.69

13.2

753

.96

Notes:1

Th

ere

sult

sof

one-

way

AN

OV

A’sposthoc

wit

hin

-seg

men

tra

ng

ete

sts

are

den

oted

inal

ph

abet

sup

ersc

rip

ts(p

,q,r

and

s);2

allF

andt-

val

ues

are

stat

isti

call

ysi

gn

ifica

nt

(p,

0.05

),ex

cep

tw

her

ein

dic

ated

as“n

s”(¼

not

sig

nifi

can

t);3

dai

lyro

omra

tep

oste

don

the

hot

el’s

web

site

;th

em

ean

sw

ith

the

sam

eal

ph

abet

sup

ersc

rip

tsar

en

otsi

gn

ifica

ntl

yd

iffe

ren

t(p

.0.

05),

wh

ile

thos

ew

ith

dif

fere

nt

alp

hab

etsu

per

scri

pts

are

sig

nifi

can

tly

dif

fere

nt

(p,

0.05

);fo

rex

amp

le,o

ver

all

gu

est

sati

sfac

tion

wit

hth

ree-

and

fiv

e-st

arh

otel

sis

not

sig

nifi

can

tly

dif

fere

nt,

wh

ile

itis

sig

nifi

can

tly

dif

fere

nt

from

that

wit

hh

otel

sin

dif

fere

nt

star

rati

ng

s;ov

eral

lsa

tisf

acti

onw

ith

thes

eot

her

star

-rat

edh

otel

sis

all

sig

nifi

can

tly

dif

fere

nt;

see

the

tex

tfo

rm

ore

det

aile

din

terp

reta

tion

Table I.Descriptive statistics andtests of mean differences

IJCHM22,7

962

ANOVA and t-tests. This is a routine way of analyzing data in most marketsegmentation studies. Although this basic level of analysis was not the main purposeof this study, it was necessary to show how the relationship-focused, instead of meanscore-focused, analysis of this study could make a different case and contribute to theexisting body of literature. The results, appear, in Table I. Overall, the mean scoresranged from 3.76 for price-value to 4.00 for cleanliness. Note that the sample size of thefive-star hotel category was relatively small, due perhaps to the comparatively smallnumber of high-end hotels operating in most destinations. The one-way ANOVAresults indicated that the five-star-rated segments received significantly differentevaluations on both satisfaction and the four performance variables.

The pattern of mean-score differences for the five lodging segments was notconsistent across the five measurement variables. For example, the level of satisfactionwas lowest with guests who patronized one-star hotels, but it was highest with thosewho experienced both three- and five-star hotels. Guest satisfaction with four-starhotels was significantly lower than that with three-star hotels, although it was higherthan that with the hotels in the two lowest star-rated segments.

For room performance, the upper three-star segments received higher evaluationsthan their two lower counterparts. The three- and five-star segments appeared to lead theindustry in room performance, followed by the four-star segment. Note that althoughthe room performance of the three- and four-star lodging segments was not different, thefour-star segment’s room performance was lower than the five-star segment’s. Also,room performance was higher in the two-star segment than in the one-star segment.

Similarly, guests perceived three-, four-, and five-star segment hotels as providingbetter services than one- and two-star hotels. The former three hotel segments receivedthe same high mean score for service performance. However, the two-star hotelsegment received a more positive evaluation for service performance than the one-starhotel segment did.

The three-star hotels appeared to lead the lodging industry in price-value perceptionsas they scored highest on the price-value variable. Guests perceived no significantdifferences in the value they received from two-, four-, and five-star hotels whose value,though, was rated higher than that of one-star hotels. Although it is generally assumedthat higher star-rated hotels tend to provide more to guests than their lowercounterparts, guests may not only consider what they get but also what they have togive up (e.g. price paid) when judging the value of hotel offerings (Rao and Monroe, 1989;Zeithaml, 1988). Our data show that guests get the most value from three-star hotels.

As the star rating went up, so did guests’ perceptions of cleanliness. Cleanliness wasbest rated for the five-star hotel segment, followed by the three- and four- starsegments which were rated performing equally. Guests’ perception of cleanliness waslowest for the one-star segment. These results seem to be consistent with generalexpectations of lodging customers.

Comparisons by type of operation showed that independent hotels were ratedgenerally higher than chain-affiliated hotels on all performance variables used in thisstudy. Guests staying at independent hotels were more satisfied than those at chainhotels. These could be caused by the significantly better performance of independenthotels on service, price-value, and cleanliness than that of chain-affiliated hotels. Bothgroups of hotels performed equally well on room, however. Interestingly, even ifindependent hotels charged more (as discussed earlier) than chain hotels, guests still

Performance-satisfactionrelationship

963

perceived higher value from independent hotels than from chain hotels. Thus,differentiation between these two lodging segments seemed to occur not by roomperformance but rather by superior service, price-value, and cleanliness, which in turnseemed to result in higher satisfaction.

The bottom block of Table I shows that the four price segments were hierarchicallydifferent in mean scores of the five variables. That is, as the average room rates wentup, so did guest satisfaction and hotel performance ( p , 0.001); all ANOVA F-valuestesting for segment-wise differences were statistically significant. Guest satisfactionfor the $100-199.99 segment was virtually the same as that for the $200-299.99segment; these two segments satisfied guests more than the less-than-$100 pricesegment did, but less than the $300-or-more price segment did. The same patternemerged with service performance; the two middle price segments performed onservice equally, but they performed better than the lowest price segment and worsethan the highest price segment. For room, all four price-based lodging segmentsperformed differently; the higher the price, the better room performance perceived. Thesegment-specific performance pattern was also apparent in the case of cleanliness. Forprice-value performance, the upper three price segments performed equally well andbetter than the lowest price segment. This particular result contests the conventionalbelief that the cheaper the price, the better the value. Combined with the results ofsegments by star ratings, the results of price-value perceptions indicate that guestsseem to form price-value perceptions differently from the other performance variablesexamined in this study.

6. The P-S relationship-based assessment of segment differencesTable II summarizes segment-specific regression results for which guests’ overallsatisfaction was regressed onto the four performance variables in each lodging segment.The general conclusion is that the four performance variables were indeed strongdeterminants of guest satisfaction in all segments, although there were a few minorvariations with no systematic patterns. The error term of the model was statisticallysignificant in all lodging segments, as theoretically expected. Multicollinearity was not acritical concern for the model, as the variance inflation factor was lower than 6 in allsegments. All parameter estimates were in the expected direction (Neter et al., 1990).

For the five-star-rated segments, the regression model performed well with R 2

values ranging from 0.63 for the three-star segment to 0.86 for the one-star segment. Allthe four performance variables were significant predictors of guest satisfaction in boththe three- and four-star lodging segments. However, cleanliness appeared to be not asignificant predictor of guest satisfaction in both the one-star and five-star segments.Room performance also did not play a role in predicting guest satisfaction in thetwo-star hotel segment. Within each segment, the magnitude of the P-S relationshipvaried across the four performance variables; likewise, the magnitude of the same P-Srelationship across the five-star segments appeared different. These seeminglydifferent relationship strengths are to be subjected to formal statistical tests, in thefollowing section.

The same regression model performed relatively strongly for the two lodgingsegments by type of operation, as shown in R 2- values of 0.82 for the independent and0.84 for the chain-affiliated segment. In the latter segment, all four performancevariables were significant predictors of guest satisfaction. Cleanliness, however,

IJCHM22,7

964

SegmentUnstandardized

estimateaStandardized

estimatea t-valueb

Star ratingOne star (R 2 ¼ 0.86) (n ¼ 129) 0.17 0.18 2.00

0.29 0.26 3.590.44 0.43 5.750.11 0.11 1.25ns

0.06 0.25 7.98Two stars (R 2 ¼ 0.74) (n ¼ 162) 0.12 0.12 1.64ns

0.34 0.30 4.840.31 0.31 4.630.20 0.20 2.540.05 0.25 8.95

Three stars (R 2 ¼ 0.63) (n ¼ 170) 0.24 0.25 3.510.15 0.13 2.040.33 0.32 4.890.20 0.20 2.480.05 0.22 9.16

Four stars (R 2 ¼ 0.75) (n ¼ 144) 0.23 0.24 3.080.37 0.34 5.200.29 0.28 4.490.18 0.18 2.170.05 0.22 8.43

Five stars (R 2 ¼ 0.72) (n ¼ 69) 0.36 0.36 3.560.39 0.35 3.900.24 0.23 2.410.04 0.04 0.36ns

0.04 0.18 5.81Operation typeChain-affiliated (R 2 ¼ 0.84)(n ¼ 505) 0.27 0.28 6.32

0.28 0.25 7.130.32 0.29 8.930.18 0.18 3.910.05 0.17 15.83

Independent (R 2 ¼ 0.82) (n ¼ 169) 0.17 0.18 2.660.48 0.44 7.110.26 0.23 4.840.10 0.10 1.50ns

0.05 0.15 9.14Price levelLess than $99.99 (R 2 ¼ 0.85) 0.28 0.28 3.44(n ¼ 132) 0.32 0.29 4.57

0.36 0.35 4.760.13 0.13 1.55ns

0.05 0.20 8.07$100-199.99 (R 2 ¼ 0.83) (n ¼ 263) 0.17 0.17 2.72

0.29 0.26 4.920.36 0.34 6.680.19 0.19 3.200.06 0.21 11.41

(continued )

Table II.Results of regressionanalyses by industry

segment

Performance-satisfactionrelationship

965

did not determine guest satisfaction in the independent hotel segment. The P-Srelationship appeared to have different magnitudes across the four performancevariables within each segment as well as across the two segments, again thedifferences which are to be tested shortly.

For the lodging segments by price level, the model resulted in R 2- values rangingfrom 0.70 for the highest to 0.85 for the lowest price segment. Interestingly, the amountof variance in guest satisfaction that was explained by the four performance variablesdecreased as the price segment went up. This could mean that higher room rates mighthave made customers become more elaborate evaluators to consider additionalperformance variables than the four included in the model. This may also be thathigher price hotels offer more evaluation cues affecting the guest’s performanceevaluations and satisfaction. Room, service, and price/value performance significantlydetermined guest satisfaction in all price segments. Again, however, cleanliness wasnot a significant predictor of guest satisfaction in three of the four price segments.Results are suggestive of differing magnitudes in the P-S relationship across the fourperformance variables as well as the four price level segments.

7. Assessments of stability of the P-S relationshipThe results of mean difference tests in Table I and segment-specific regressionanalyses in Table II seemed to collectively suggest that the lodging segments by starratings, type of operation, and price level result in different guest perceptions oflodging performance and satisfaction. Some of these findings are not new and areconsistent with the marketer’s expectation. However, the question of whether aparticular performance variable determined guest satisfaction to the same extent as theother performance variables across different segments necessitated a differentapproach to analyze the data. Testing stability of the P-S relationship using nestedmodels via the x 2 difference test was effective in this case, because such a test methodwould provide straightforward answers even in the presence of a few insignificantparameter estimates as reported in Table II.

SegmentUnstandardized

estimateaStandardized

estimatea t-valueb

$200-299.99 (R 2 ¼ 0.76) (n ¼ 145) 0.26 0.27 3.730.29 0.26 4.090.39 0.37 6.390.04 0.04 0.47ns

0.05 0.17 8.46$300 or more (R 2 ¼ 0.70) (n ¼ 134) 0.28 0.28 3.77

0.41 0.37 5.230.24 0.23 3.800.09 0.09 1.10ns

0.04 0.16 8.13

Notes: aEntries are the parameter estimates in order of room (b1), service (b2), price-value (b3),cleanliness (b4), and model error (C) that were obtained from regressing overall guest satisfaction ontoroom, service, price-value, and cleanliness; bexcept where indicated as “ns” (– not significant),the t-values are statistically significant ( p , 0.05)Table II.

IJCHM22,7

966

Table III summarizes test results. For the lodging segments by star rating, a generalconclusion is that the direction and magnitude of the P-S relationship were constantacross the five-star segments as well as among the four performance variables withineach segment. Customers patronizing different star-rated hotels seemed to come fromdifferent populations ( p . 0.05; test result A.1 in Table III), but their reliance on thefour performance attributes in judging their satisfaction was consistent as shown inthe insignificant results of testing the equality of the P-S relationship and associatedmodel error (A.2-A.4). Moreover, the four performance variables determined guest

Equality tested x2 df p-value

A. Five-star rating segments1. Total data matrix (SG1-5) 158.29 60 0.002. Effects across segments (BG1-5) 15.93 16 0.463. Error of models (CG1-5) 4.34 4 0.364. Both effects and errors (BG1-5 þ CG1-5) 19.01 20 0.525. Effects within each segment (bG

1-4) 24.07 15 0.065.1. Effects within one-star segment only 8.90 3 0.03

5.1.1. Room vs price-value 4.11 1 0.045.1.2. Price-value vs cleanliness 6.71 1 0.01

5.2. Effects within two-star segment only 4.60 3 0.205.3. Effects within three-star segment only 3.03 3 0.395.4. Effects within four-star segment only 3.36 3 0.345.5. Effects within five-star segment only 4.19 3 0.24

6. Effects both across and within segments (BG1-5 þ bG1-4) 28.78 19 0.06

B. Two operation type segments1. Total data matrix (SG1-2) 34.29 15 0.002. Effects across segments (BG1-2) 8.49 4 0.083. Error of models (CG1-2) 1.21 1 0.274. Both effects and errors (BG1-2 þ CG1-2) 8.65 5 0.125. Effects within each segment (bG

1-4) 19.17 6 0.015.1. Effects within chain hotels only 5.08 3 0.17

5.1.1. Price-value vs cleanliness 4.91 1 0.035.2. Effects within independent hotels only 14.09 3 0.00

5.2.1. Service vs cleanliness 11.64 1 0.005.2.2. Room vs service 7.91 1 0.005.2.3. Service vs price-value 4.83 1 0.03

6. Effects both across and within segments (BG1-2 þ bG1-4) 19.92 7 0.01

C. Four price level segments1. Total data matrix (SG1-4) 89.34 45 0.002. Effects across segments (BG1-4) 10.82 12 0.543. Error of models (CG1-4) 4.15 3 0.254. Both effects and errors (BG1-4 þ CG1-4) 14.06 15 0.525. Effects within each segment (bG

1-4) 23.80 12 0.025.1. Effects within less than $99.99 hotels 3.45 3 0.335.2. Effects within $100-$199.99 hotels 5.97 3 0.11

5.2.1. Room vs price-value 4.20 1 0.045.3. Effects within $200-$299.99 hotels 8.72 3 0.03

5.3.1. Price-value vs cleanliness 8.43 1 0.005.4. Effects within $300 or more hotels 5.67 3 0.13

5.4.1. Service vs cleanliness 5.59 1 0.026. Effects both across and within segments (BG1-4 þ bG

1-4) 28.17 15 0.02

Table III.Results of invariance

tests for effects acrossindustry segments

Performance-satisfactionrelationship

967

satisfaction in an equal magnitude in each star segment (A.5). Segment-specific testsindicated that the relationship stability was evident in all, but one-star, segments(A.5.1-5.5). The same relationship stability across the four performance variables couldnot be permitted for the one-star segment (A.5.1), primarily because in this segmentprice-value performance determined guest satisfaction more strongly than either roomor cleanliness performance did (A.5.1.1 and A.5.1.2; see Table II for the effect sizes).Overall, it could not be rejected, as we expected, that the P-S relationship was constantacross both the five-star segments and the four performance variables (A.6).

The second block of Table III shows test results for the lodging segments by type ofoperation. Customers staying at chain-affiliated hotels came from a different customerpopulation, compared to those patronizing independently operated hotels (B.1). As wasin the star-rated segments, each performance variable determined guest satisfaction tothe same degree in the two lodging segments (B.2-B.4, p < 0.08, B.2). This result issomewhat against what we and most marketers would have expected. However, withineach segment the P-S relationship across the four performance variables wassignificantly different (B.5). Although the direction and magnitude of the P-Srelationship across the four performance variables was generally invariant in the chainhotel segment (B.5.1), additional pairwise comparisons indicated that price-valueperformance was a stronger predictor of guest satisfaction than cleanliness performance(B.5.1.1). In the independent hotel segment, however, the same relationship invariance(i.e. stability) did not hold (B.5.2.); specifically, service performance was a strongerpredictor of guest satisfaction than cleanliness (B.5.2.1), room (B.5.2.2), and price-valueperformance (B.5.2.3.). Overall, the magnitude of the P-S relationship was different bothbetween the two lodging segments and across the four performance variables withineach segment, when they were considered simultaneously (B.6).

Our final set of invariance tests was conducted for the lodging segments by pricelevel and the results appear in the bottom block of Table III. Again, customers whopatronized hotels in different price levels belonged to different lodging customerpopulations (C.1). Nonetheless, the same performance variable determined guestsatisfaction in an equal magnitude (and direction) across the four price-based lodgingsegments (C.2-C.4), as we predicted earlier. The extent to which the four performancevariables determined guest satisfaction within each price segment was, however,significantly different (C.5), due largely to the $200-299.99 price segment (C.5.3) in whichprice-value performance determined guest satisfaction more strongly than cleanlinessperformance did (C.5.3.1). In the other price-based segments, the four performancevariables determined guest satisfaction in an equal magnitude (C.5.1, C.5.2, and C.5.4);in pairwise equality tests, however, price-value performance appeared to be a strongerpredictor of guest satisfaction than room performance in the $100-$199.99 segment(C.5.2.1) and service performance stronger than cleanliness performance in the highestprice segment (C.5.4.1). Finally, the direction and magnitude of the P-S relationship wasnot equal across the four price-based lodging segments as well as across the fourperformance variables within each segment, when they were tested simultaneously.

8. Discussions and implicationsInfluencing the customer’s behavior remains a fundamental premise in marketing(Kotler et al., 2006), even if most hospitality market segmentation studies haveoverlooked such a premise in their study design and analysis methods. In this study,

IJCHM22,7

968

we attempted to show how to assess differences (i.e. heterogeneous demands) ofsegmented markets based on how the market was influenced, beyond merely examiningvariable mean score-based differences. Viewing or assuming lodging segments to bedistinct based only on differences in univariate mean scores could not tell about thecustomer’s decision process underlying the mean scores. Many researchers, though,often concluded that markets were different when univariate mean scores werestatistically different across market segments (Barsky and Nash, 2005; Carman, 1990;Ryan and Huimin, 2007). Such mere mean score-based conclusions do not permitmarketers to understand, for example, how differently the customer’s satisfaction ofdifferent segments were influenced by the same set of performance variables examined.Few studies have addressed these issues, especially based on a comprehensive datasetcovering lodging market segments broadly.

Our findings indicate that the extent to which the same performance variabledetermines guests’ overall satisfaction is equally strong (i.e. constant or stable) acrosslodging segments by star rating, type of operation, and price level. Such stability wasevident despite the fact that the mean values of performance variables and guestsatisfaction were statistically different across the segments. In other words, guestsatisfaction was equally influenced by the same performance variable across thesemarket segments. Lodging marketers, therefore, should not assume heterogeneity ofdifferent market segments simply because of the mean score-based differences, whichwas done in many previous hospitality market segmentation studies. Differences in thevariable mean scores could come from the lodging industry’s mere productdifferentiation in accordance with prices charged. Implications are twofold. First,lodging marketers now know that guest satisfaction may be influenced by the samevariable consistently (in both direction and magnitude) across different marketsegments in spite of the significant differences in mean performance scores. Second,marketers need to examine how some critical marketing outcomes, such as guestsatisfaction or repurchase intention, are influenced in different market segments whenattempting to develop segment-specific market strategies. Mean score-basedexaminations of segment differences and subsequent strategy development may notproduce effective marketing outcomes.

It is interesting that each of the four performance variables determines guestsatisfaction to the same degree across all the lodging market segments examined inthis study. One straightforward explanation for this is that, despite the differences inperformance and satisfaction mean scores, lodging customers’ reliance on the sameperformance attributes in forming satisfaction judgments may be constant regardlessof the type of property they patronize. Another implication of the results may be thatthe four performance variables used in our study no longer serve as sources ofdifferentiation among the lodging segments. Since guest satisfaction seems to derivefrom the four variables equally across the examined segments, lodging marketers needto look for the different variables or attributes that can positively influence theircustomer in their unique ways. Of course, this is not to say that positive performanceon the four variables is not important; it is indeed important to guest satisfaction, asshown in the statistical significance of each variable and R 2 in the segment-specificregression models (Table II). We mean that good performance on these variables maybe a norm, but not sufficient enough to assure a competitive advantage throughdifferentiation in performance leading to guest satisfaction.

Performance-satisfactionrelationship

969

The stability of the P-S relationship may not generalize across different performancevariables within the same market segment. Only in the five-star-rated hotel segmentsdid the four performance variables determine guest satisfaction in a uniform directionand magnitude, with a slight variation in the one-star segment. This between-variablestability of the P-S relationship did not hold for the two sets of lodging marketssegmented by type of operation and price. This finding is not new, though, and themessage for lodging marketers is that different performance variables have differentlevels of impact on guest satisfaction. It is necessary that marketers take asegment-specific approach to understand the weight each performance variable has onguest satisfaction and make resource allocation decisions based on the characteristicsof each segment.

9. Conclusions and suggestionsThe findings of this study extend previous lodging performance and satisfactionresearch in several important ways:

. Mean score-based differences do not necessarily suggest that desired marketingoutcomes (e.g. guest satisfaction) are influenced by the same input (e.g. the fourvariables) to the same extent (i.e. stability) across the market segments.Marketers need to examine segment differences not only in the variable meansbut also in light of a cause-and-effect relationship as demonstrated in this study.Only after understanding such process-based differences can marketers developeffective segment-specific strategies.

. The four performance variables used by Tripadvisor.com (i.e. room, service,price-value, and cleanliness) seem to possess strong predictive validity as shownin the generally high R 2-values. Nonetheless, these variables may not necessarilybe powerful enough for understanding segment-specific satisfaction drivers.Rather, they seem to function as basic dimensions of performance in which hotelsshould perform well to stay competitive. Further research is necessary toexamine industry-wide normative predictors vs competitive differentiators ofguest satisfaction.

. Despite the statistically significant differences in the mean scores of the fourperformance variables, each variable’s predictive power of guest satisfaction wasequivalent across the various lodging segments. Mean scores only should not bethe basis of segmenting the lodging market, as widely practiced in previoushospitality market segmentation studies. Instead, understanding the changes inthe marketing input and output relationship across segments may produce moreinfluential marketing strategies.

. This study confirmed that lodging operators need to allocate resourcesdifferently across the four and perhaps other variables to improve theirperformance. This was particularly true for the one-star, independent hotelswhose average daily rates were below $99 (Table III). Segment-specific, carefulselection of predictors of guest satisfaction is necessary in some lodging marketsegments like these to effectively achieve and manage guest satisfaction.

. The three criteria that are widely used to segment the lodging industry (i.e. starratings, type of operation, and price level) may not be as useful as reflected in theirpopularity in practice if the segmentation focus is on the P-S relationship rather

IJCHM22,7

970

than the mean scores of individual variables. Researchers need to seek additionalanswers to these issues and develop practical guidelines for differentiatingmarket segments based on various marketing input-output relationships. It ishoped that this study provided a useful illustrative case future research.

Cautions are due to several limitations of this study. First, our investigation wasfocused on the lodging segments by only three perhaps most popular variables,namely, star rating, type of operation, and price level. Different companies may employdifferent variables to segment their market and similar analyses may be necessary toobtain applicable results. In addition, our segmentation approach, especially by pricelevel, might have not been effective, as price perceptions fluctuate across differentdestinations. Nonetheless, it should be noted that the primary objective of our studywas to illustrate an effective way to analyze data for segment-wise differences and,thus, the analysis approach we demonstrated should be applicable to different pricemarkets. Along such extensions, future research may also explore other predictors andconsequences of guest satisfaction. Food, convention facilities, or property ambience,for example, may qualify strong predictors if the hotel features them as core marketingstrategies. Although Tripadvisor.com did not provide data on repurchase andword-of-mouth intention, inclusion of these marketing outcome variables may generateadditional useful insights.

Second, generalizability of our findings may be limited due to our conveniencesampling methods. Although we attempted to cover the lodging industry across theentire USA as broadly as possible, the source data limited our coverage as we could notsecure a sufficient number of customer reviews for some selected destinations andhotels. The same line of reasoning points to a plausible selection bias in the samplebase used in this study. It is apparent that the customers who had access to the internetvoluntarily posted their review comments which became the input data for this study.Readers must consider this limitation when interpreting the findings.

Finally, the nature of the secondary data used in this study did not allow examiningthe characteristics of the sample and their potential implications for the results. Hence,it is not clear whether the customers who voluntarily registered their review commentson the Tripadvisor.com site had any systematically common characteristics that couldalter our interpretation of the findings. Although we collected customer reviewcomments from a variety of destinations and hotels, representation of the US lodgingcustomer population could not be guaranteed with our sampling method. Interestingly,the mean score distributions shown in Table I suggest that a typical issue insatisfaction research – only extremely happy or angry customers tend to raise theirvoice – was not an apparent concern for our data as the mean scores fell mostly in themid to mid-high ranges rather than extremely high or low ranges of the scale. Sizablestandard deviations for most measurement variables also support such a conclusion.We hope future research will improve on our limitations in this study.

References

Barsky, J. and Nash, L. (2003), “Customer satisfaction: applying concepts to industry-widemeasures”,CornellHotel&RestaurantAdministrationQuarterly, Vol. 44 Nos 5/6, pp. 173-83.

Barsky, J. and Nash, L. (2005), “Survey finds guest satisfaction higher at overseas hotels”, Hoteland Motel Management, 12 December, available at: www.HotelMotel.com

Performance-satisfactionrelationship

971

Bettman, J.R. (1979), Information Processing Theory of Consumer Choice, Addison-WesleyEducational, Glenview, IL.

Bojanic, D. (1996), “Consumer perceptions of price, value and satisfaction in the hotel industry:an exploratory study”, Journal of Hospitality & Leisure Marketing, Vol. 4 No. 1, pp. 5-22.

Bolton, R.N. and Myers, M.B. (2003), “Price-based global market segmentation for services”,Journal of Marketing, Vol. 67 No. 3, pp. 108-20.

Carman, J. (1990), “Consumer perceptions of service quality and an assessment of theSERVQAUL dimensions”, Journal of Retailing, Vol. 66 No. 1, pp. 33-55.

Chen, J.S. (2001), “Norwegians’ preferences for U.S. lodging facilities: market segmentationapproach”, Journal of Travel & Tourism Marketing, Vol. 9 No. 4, pp. 69-82.

Churchill, G.A. Jr and Surprenant, C. (1982), “An investigation into the determinants of customersatisfaction”, Journal of Marketing Research, Vol. 19, pp. 491-504.

Cronin, J.J. Jr and Taylor, S.A. (1992), “Measuring service quality: a reexamination andextension”, Journal of Marketing, Vol. 56, pp. 55-68.

Fang, M., Tepanon, Y. and Uysal, M. (2008), “Measuring tourist satisfaction by attribute andmotivation: the case of a nature-based resort”, Journal of Vacation Marketing, Vol. 14No. 1, pp. 41-56.

Fernandez, M. and Bedia, A. (2004), “Is the hotel classification system a good indicator of hotelquality? An application in Spain”, Tourism Management, Vol. 25, pp. 771-5.

Griffin, R., Shea, L. and Weaver, P. (1996), “How business travelers discriminate betweenmid-priced and luxury hotels: an analysis using a longitudinal sample”, Journal ofHospitality & Leisure Marketing, Vol. 4 No. 2, pp. 63-75.

Gustin, M.E. and Weaver, P.A. (1993), “The mature market: underlying dimensions and groupdifferences of a potential market for the hotel industry”, FIU Hospitality Review, Vol. 11No. 2, pp. 49-60.

Henley, J., Cotter, M. and Herrington, J. (2004), “Quality and pricing in the hotel industry:the Mobil ‘star’ and hotel pricing behavior”, International Journal of Hospitality & TourismAdministration, Vol. 5 No. 4, pp. 53-65.

Holverson, S. and Revaz, F. (2006), “Perceptions of European independent hoteliers: hard and softbranding choices”, International Journal of Contemporary Hospitality Management, Vol. 18No. 5, pp. 398-413.

Imrie, R. and Fyall, A. (2000), “Customer retention and loyalty in the independent mid-markethotel sector: a United Kingdom perspective”, Journal of Hospitality & Leisure Marketing,Vol. 7 No. 3, pp. 39-54.

Ingram, H. and Daskalakis, G. (1999), “Measuring quality gaps in hotels: the case of Crete”,International Journal of Contemporary Hospitality Management, Vol. 11 No. 1, pp. 24-30.

Joreskog, K.G. and Sorbom, D. (2006), LISREL 8.8: Structural Equation Modeling with theSIMPLIS Command Language, Scientific Software International, Chicago, IL.

Knutson, B., Stevens, P., Patton, M. and Thompson, C. (1992), “Consumers’ expectations forservice quality in economy, mid-price and luxury hotels”, Journal of Hospitality & LeisureMarketing, Vol. 1 No. 2, pp. 27-43.

Kotler, P. (1991), Marketing Management: Analysis, Planning, Implementation, and Control,7th ed., Prentice-Hall, Englewood Cliffs, NJ.

Kotler, P., Bowen, J. and Makens, J. (2006), Marketing for Hospitality and Tourism, 5th ed.,Prentice-Hall, Englewood Cliffs, NJ.

IJCHM22,7

972

Kutzner, D., Wright, P.A. and Stark, A. (2009), “Identifying tourists’ preferences for aboriginaltourism product features: implications for a northern first nation in British Columbia”,Journal of Ecotourism, Vol. 8 No. 2, pp. 99-114.

Law, R. and Hsu, C. (2006), “Importance of hotel website dimensions and attributes: perceptionsof online browsers and online purchasers”, Journal of Hospitality & Tourism Research,Vol. 30 No. 3, pp. 295-312.

Leahy, J., Shugrue, M., Daigle, J. and Daniel, H. (2009), “Local and visitor physical activitythrough media messages: a specialized benefits-based management application at AcadiaNational Park”, Journal of Park and Recreation Administration, Vol. 27 No. 3, pp. 59-77.

Lewis, R.C. (1985), “The market positioning: mapping guests’ perceptions of hotel operations”,Cornell Hotel & Restaurant Administration Quarterly, Vol. 26 No. 2, pp. 86-99.

Lockyer, T. (2005), “The perceived importance of price as one hotel selection dimension”,Tourism Management, Vol. 46 No. 4, pp. 529-37.

Mattila, A. (1999), “An examination of factors affecting service recovery in a restaurant setting”,Journal of Hospitality & Tourism Research, Vol. 23 No. 3, pp. 284-98.

Matzler, K., Fuller, J., Renzl, B., Herting, S. and Spath, S. (2008), “Customer satisfaction withAlpine ski areas: the moderating effects of personal, situational, and product factors”,Journal of Travel Research, Vol. 46 No. 4, pp. 403-13.

Mentzer, J., Myers, M.B. and Cheung, M. (2004), “Global market segmentation for logisticsservices”, Industrial Marketing Management, Vol. 33 No. 2, pp. 15-30.

Monroe, K.B. (1990), Pricing: Making Profitable Decisions, 2nd ed., McGraw-Hill, New York, NY.

Monroe, K.B. and Lee, A.Y. (1999), “Remembering versus knowing: issues in buyers’ processing ofprice information”, Journal of the Academy of Marketing Science, Vol. 27 No. 2, pp. 207-25.

Neter, J., Wasserman, W. and Kutner, M.H. (1990), Applied Linear Statistical Models, 3rd ed.,Irwin, Homewood, IL.

Oakley, J.L., Calder, B.J. and Iacobucci, D. (2004), “Customer satisfaction across organizationalunits”, Journal of Service Research, Vol. 6 No. 3, pp. 231-42.

Oh, H. (1999), “Service quality, customer satisfaction, and customer value: a holistic perspective”,International Journal of Hospitality Management, Vol. 18 No. 1, pp. 67-82.

Oh, H. and Jeong, M. (1996), “Improving marketers’ predictive power of customer satisfaction onexpectation-based target market levels”,HospitalityResearch Journal, Vol. 19 No. 4, pp. 65-85.

Oh, H. and Parks, S.C. (1997), “Customer satisfaction and service quality: a critical review of theliterature and research implications for the hospitality industry”, Hospitality ResearchJournal, Vol. 20 No. 3, pp. 35-64.

Oh, H., Kim, B. and Shin, J. (2004), “Hospitality and tourism marketing: recent developments inresearch and future directions”, International Journal of Hospitality Management, Vol. 23,pp. 425-7.

Oliver, R.L. (1981), “Measurement and evaluation of satisfaction process in retail setting”, Journalof Retailing, Vol. 57, pp. 25-48.

Oliver, R.L. (1997), Satisfaction: A Behavioral Perspective on the Consumer, Irwin/McGraw-Hill,New York, NY.

O’Neill, J.W. and Mattila, A.S. (2006), “Strategic hotel development and positioning”, CornellHotel & Restaurant Administration Quarterly, Vol. 47 No. 2, pp. 146-54.

O’Neill, J.W. and Xiao, Q. (2006), “The role of brand affiliation in hotel market value”, CornellHotel & Restaurant Administration Quarterly, Vol. 47 No. 3, pp. 210-23.

Performance-satisfactionrelationship

973

Osman, H., Hemmington, N. and Bowie, D. (2009), “A transactional approach to customer loyaltyin the hotel industry”, International Journal of Contemporary Hospitality Management,Vol. 21 No. 3, pp. 239-50.

Parasuraman, A., Zeithaml, V.A. and Berry, L.L. (1994), “Alternative scales for measuring servicequality: a comparative assessment based on psychometric and diagnostic criteria”, Journalof Retailing, Vol. 70 No. 3, pp. 201-30.

Rao, A.R. and Monroe, K.B. (1989), “The effect of price, brand name, and store name on buyers’perceptions of product quality: an integrative review”, Journal of Marketing Research,Vol. 26, pp. 351-7.

Ryan, C. and Huimin, G. (2007), “Perceptions of Chinese hotels”, Cornell Hotel & RestaurantAdministration Quarterly, Vol. 48 No. 4, pp. 380-91.

Saleh, F. and Ryan, C. (1991), “Analyzing service quality in the hospitality industry using theSERVQUAL model”, The Service Industries Journal, Vol. 11, pp. 324-43.

Saleh, F. and Ryan, C. (1992), “Client perceptions of hotels: a multi-attribute approach”, TourismManagement, Vol. 13, pp. 163-6.

Shoemaker, S. and Lewis, R.C. (1999), “Customer loyalty: the future of hospitality marketing”,International Journal of Hospitality Management, Vol. 18 No. 4, pp. 345-70.

Shoemaker, S., Lewis, R.C. and Yesawich, P.C. (2006), Marketing Leadership in Hospitality andTourism: Foundations and Practice, 4th ed., Prentice-Hall, Upper Saddle River, NJ.

Slevitch, L. and Sharma, A. (2008), “Management of perceived risk in the context of destinationchoice”, International Journal of Hospitality & Tourism Administration, Vol. 9 No. 1,pp. 85-103.

Strauss, K. (2004), “Are your satisfied customers loyal, too?”, Hotels, August, pp. 14-15.

Szymanski, D.M. and Henard, D.H. (2001), “Customer satisfaction: a meta-analysis of theempirical evidence”, Journal of the Academy of Marketing Science, Vol. 29 No. 1, pp. 16-35.

Varini, K., Engelmann, R., Claessen, B. and Schleusener, M. (2003), “Evaluation of the price-valueperception of customers in Swiss hotels”, Journal of Revenue and Pricing Management,Vol. 2 No. 1, pp. 47-60.

Weaver, P.A., McCleary, K.W., Han, J. and Blosser, P. (2009), “Identifying leisure travel marketsegments based on preference for novelty”, Journal of Travel & TourismMarketing, Vol. 26Nos 5/6, pp. 568-84.

Weber, K. (2001), “Association meeting planners’ loyalty to hotel chains”, International Journal ofHospitality Management, Vol. 20, pp. 259-75.

Wong, J. and Yeh, C. (2009), “Tourist hesitation in destination decision making”, Annals ofTourism Research, Vol. 36 No. 1, pp. 6-23.

Yeung, T. and Law, R. (2004), “Extending the modified heuristic usability evaluation techniqueto chain and independent hotel websites”, International Journal of HospitalityManagement, Vol. 23, pp. 307-13.

Zeithaml, V.A. (1988), “Consumer perceptions of price, quality, and value: a means-end modeland synthesis of evidence”, Journal of Marketing, Vol. 52, pp. 2-22.

Corresponding authorMiyoung Jeong can be contacted at: [email protected]

IJCHM22,7

974

To purchase reprints of this article please e-mail: [email protected] visit our web site for further details: www.emeraldinsight.com/reprints