reexamining the dimensionality of brand loyalty: a …

REEXAMINING THE DIMENSIONALITY OFBRAND LOYALTY: A CASE OF THE CRUISE

INDUSTRY

Xiang (Robert) LiJames F. Petrick

ABSTRACT. This study revisits the dimensional structure of the brand loyalty construct.Following recent developments in loyalty studies, this research conceptualizes brand loyalty as afour-dimensional construct comprising of cognitive, affective, conative, and behavioral loyalty. It isproposed that the first three dimensions collectively form a higher order factor—namely attitudinalloyalty—which then leads to behavioral loyalty. However, this conceptualization is not supportedby the data. Alternatively, a modified model—based on the traditional conceptualization thatattitudinal loyalty is a first-order, one-dimensional construct—was found to better fit the data.Thus, this study revalidates the traditional two-dimensional conceptualization of loyalty. It alsocontributes to the literature by introducing and validating a five-item attitudinal loyalty measure.

KEYWORDS. Attitudinal loyalty, cognitive loyalty, affective loyalty, conative loyalty,behavioral loyalty, brand loyalty

The recent increase in brand loyaltyresearch seems to have echoed the emer-gence of the relationship marketing para-digm (Morais, Dorsch, & Backman, 2005),which emphasizes the importance of estab-lishing relationships between customersand businesses (Gronroos, 1994; Sheth &Parvatlyar, 1995). Nevertheless, brand loy-alty research has been consistently criti-cized for lacking theoretical grounding andconceptual depth (Dimanche & Havitz,1994; Iwasaki & Havitz, 2004; Jacoby &Chestnut, 1978; Oliver, 1999; Pritchard,Havitz, & Howard, 1999). It is particularlydisquieting that no consensus has been

reached on what loyalty is. That is, whatcomponents should be included when con-ceptualizing or measuring customers’brand loyalty, and where to draw the linebetween loyalty and its antecedents oroutcomes. Moreover, the vast majority ofprevious loyalty studies have focused onconsumer goods, while the advent of the‘‘service economy’’ (Gummersson, 2002) or‘‘experience economy’’ (Pine & Gilmore,1999) has called for more research onservices. Therefore, this study seeks tosystematically examine the conceptualdomain and structure of brand loyalty ina tourism service context.

Journal of Travel & Tourism Marketing, Vol. 25(1) 2008Available online at http://jttm.haworthpress.com

# 2008 by The Haworth Press. All rights reserved.68 doi: 10.1080/10548400802164913

Xiang (Robert) Li is Assistant Professor, School of Hotel, Restaurant, and Tourism Management,University of South Carolina, Columbia, SC 29208, USA (E-mail: [email protected]). James F. Petrickis Associate Professor, Department of Recreation, Park and Tourism Science, Texas A&MUniversity, College Station, TX 77843, USA. (E-mail: [email protected]).

An earlier draft of this paper was presented at the 2007 Travel and Tourism Research Association(TTRA) Annual Conference in Las Vegas, NV.

One sector in need of retaining loyalcustomers is the cruise industry, which istraditionally characterized by a high level ofrepurchase (i.e., behavioral loyalty) (Petrick,2004). To continue the current marketbalance and to block potential competitorsfrom entry, major cruise companies havebeen investing heavily on cruise capacityexpansion (Lois, Wang, Wall, & Ruxton,2004). This growth in berths has made itimperative for the industry—among otherthings—to retain its current clientele, andimprove repurchase rate, to maintain presentoccupancy rates (Miller & Grazer, 2003).Thus, it seems that research focusing oncustomer loyalty may provide operationalsignificance to the cruise industry.

This paper seeks a better understanding ofthe structure of cruisers’ brand loyalty.Specifically, the study will examine thedimensionality of loyalty, and identify mea-sures of loyalty from a multidimensionalperspective. Theoretical significance aside,exploring the structure of loyalty mayprovide guidance to the measurement andmanagement of loyalty.

LITERATURE REVIEW

Traditional View

The loyalty construct has been a centralresearch topic among marketing scholars(Rundle-Thiele, 2005). Until recently, theconceptualization of loyalty has beenadopted from three major approaches(Jacoby & Chestnut, 1978; Morais, 2000;Rundle-Thiele, 2005). It has been suggestedthat loyalty may refer to customers’ beha-vioral consistency, attitudinal predispositiontoward purchasing a brand, or both.

Behavioral Loyalty

The majority of early loyalty studies tooka behavioral approach, and interpretedloyalty as synonymous with repeat purchase.This was grounded on a stochastic view ofconsumer behavior (Rundle-Thiele, 2005),

which proposes that consumer behavior, aswell as market structure, are characterizedby randomness rather than rationality (Bass,1974). Tucker (1964, p. 32) went so far as toassert that ‘‘no consideration should be givenwhat the subject thinks or what goes on inhis central nervous system; his behavior isthe full statement of what brand loyalty is.’’More recently, Ehrenberg (1988) contendedthat researchers should understand howpeople make brand purchases, before under-standing why people buy. Finally, from ameasurement perspective, O’Mally (1998,p. 49) suggests that behavioral measures ofloyalty provide ‘‘a more realistic picture ofhow well the brand is doing vis-a-viscompetitors...’’

A major criticism of the behavioral loyaltyapproach is that it fails to distinguishcustomers making purchase decisionsbecause of genuine brand preference, fromthose who purchase solely for convenience orcost reasons (Back, 2001). In other words,underlying customers’ repeat brand purchasemay be inertia (i.e., repeat brand purchasesfor the sake of saving time and energy;Assael, 2004), rather than the customer-brand bond (Fournier, 1998). Furthermore,due to inconsistency between behavioralmeasures, one customer classified as a loyalclient based on Method A, may be classifiedas disloyal by Method B (Morais, 2000).Thus, several researchers have argued thatthe loyalty phenomenon cannot be ade-quately understood without measuring indi-viduals’ attitude toward a brand (Backman& Crompton, 1991; Day, 1969; Dick & Basu,1994).

Attitudinal Loyalty

The stochastic philosophy essentiallymaintains that marketers are unable toinfluence buyer behavior in a systematicmanner. In contrast, the deterministic philo-sophy suggests that behaviors do not justhappen, they can be ‘‘a direct consequence ofmarketers’ programs and their resultingimpact on the attitudes and perceptions heldby the customer’’ (Rundle-Thiele, 2005,

Xiang (Robert) Li and James F. Petrick 69

p. 38). Researchers holding a deterministicview hence advocate the need to understandthe loyalty phenomenon from an attitudinalperspective.

Guest (1944) was arguably the firstresearcher to propose the idea of measuringloyalty as an attitude. He used a singlepreference question asking participants toselect the brand they liked the best, among agroup of brand names. A number ofresearchers followed his approach, and con-ceptualized loyalty as attitudes, preferences,or arguably purchase intentions—all ofwhich can be considered as a function ofpsychological processes (Jacoby & Chestnut,1978). Terms such as cognitive loyalty (Jarvis& Wilcox, 1976) and intentional loyalty(Jain, Pinson, & Malhotra, 1987) subse-quently emerged to capture different compo-nents of the psychological processes. Morerecently, Reichheld (2003) argued that loy-alty may be assessed using only one vari-able—‘‘willingness to recommend’’ (which isotherwise considered as an attitudinal loy-alty outcome).

A major criticism of the attitudinal loyaltyapproach is that it lacks power in predictingactual purchase behavior, even though arecent meta-analysis on attitude-behaviorstudies (Kraus, 1995) reported that attitudessignificantly predict future behavior(Rundle-Thiele, 2005). It has been foundthat using attitudinal loyalty alone may notcapture the entirety of the loyalty phenom-enon (Morais, 2000). Meanwhile, someauthors have suggested that the limitedexplanatory power of attitudinal loyaltycould be the result of intervening influencesfrom other factors constraining purchasebehaviors (Backman & Crompton, 1991).

Composite Loyalty

The foregoing review implies that neitherthe behavioral nor attitudinal loyaltyapproach alone provides a satisfactoryanswer to the question ‘‘what is loyalty?.’’Day (1969) argued that genuine loyalty isconsistent purchase behavior rooted inpositive attitudes toward the brand. His

two-dimensional conceptualization of loy-alty suggested a simultaneous considerationof attitudinal loyalty and behavioral loyalty,which profoundly influenced the direction ofloyalty research (Jacoby & Chestnut, 1978;Knox & Walker, 2001).

A number of researchers have operatio-nalized loyalty using a composite approach(Backman & Crompton, 1991; Dick & Basu,1994; Morais, Dorsch, & Backman, 2004;Petrick, 2004; Pritchard et al., 1999; Selin,Howard, Udd, & Cable, 1988; Shoemaker &Lewis, 1999). For instance, Dick and Basuconceptualized loyalty as the relationshipbetween relative attitude (attitudinal dimen-sion) and repeat patronage (behavioraldimension). They maintained that true brandloyalty only exists when consumer beliefs,affect, and intention all point to a focalpreference toward the brand or serviceprovider. In leisure literature, Backman andCrompton (1991) conceptualized psycholo-gical attachment and behavioral consistencyas two dimensions of loyalty. Their findingsrevealed that ‘‘attitudinal, behavioral, andcomposite loyalty capture the loyalty phe-nomenon differently’’ (p. 217). To date,although some researchers still conceptualizeloyalty as a unidimensional construct, thevast majority of researchers have adoptedthe composite loyalty approach.

Recent Conceptual Development

As loyalty research has evolved, thedominant two-dimensional conceptualiza-tion has been challenged (see Jones andTaylor, 2007; and Rundle-Thiele, 2005 forcomprehensive reviews). It has been sug-gested that the two-dimensional conceptua-lization provides inadequate guidance forpractitioners designing loyalty programs(Rundle-Thiele, 2005). Further, the dimen-sionality issue warrants attention as market-ers who misunderstood the conceptualdomain and structure of loyalty may ‘‘bemeasuring the wrong things in their attemptsto identify loyal customers; be unable to linkcustomer loyalty to firm performance mea-sures; and be rewarding the wrong customer

70 JOURNAL OF TRAVEL & TOURISM MARKETING

behaviors or attitudes when designing loy-alty programs’’ (Jones & Taylor, 2007, p. 36).

Many new conceptualizations of loyaltyare somewhat influenced by Oliver’s work(Oliver, 1997, 1999). Oliver followed thesame cognition-affect-conation structure asDick and Basu (1994), but suggested thatloyalty formation is more likely to be anattitudinal development process, and thatcustomers may demonstrate different levelsof loyalty in different stages of this process.Thus, Oliver implied that loyalty is neither adichotomy (loyalty vs. no loyalty), normulticategory typology (e.g., low, spurious,latent, and high loyalty), but a continuum.Specifically, Oliver (1997, 1999) posited thatthe loyalty-building process starts from somecognitive beliefs (cognitive loyalty), followedby affective loyalty (i.e., ‘‘I buy it because Ilike it’’), to conative loyalty (i.e., ‘‘I’mcommitted to buying it’’), and finally actionloyalty (i.e., actual ‘‘action inertia’’).Although the temporal sequence of loyaltyformation remains controversial (Rundle-Thiele, 2005), a number of researchers haveadopted Oliver’s four-dimensional loyaltyconceptualization (Back, 2001; Harris &Goode, 2004; Jones & Taylor, 2007; Lee,2003; McMullan & Gilmore, 2003).

For instance, Harris and Goode (2004)operationalized and tested Oliver’s 4-facetmeasure in two online service scenarios(purchasing books and flight tickets). Theauthors concluded that the hypothesizedcognitive-affective-conative-action loyaltysequence provided a better fit of the datathan other possible variations. In a similarvein, McMullan and Gilmore (2003) devel-oped a 28-item scale to measure the fourphases of loyalty, following Oliver’s con-ceptualization. Their empirical test in arestaurant-dining context supported thefour-dimensional conceptualization.

Back (2001) agreed with most of Oliver’s(1997, 1999) development on the traditionaltwo-dimensional view. However, based onthe tripartite model of attitude structure(Breckler, 1984), he argued that cognitive,affective, and conative loyalty are essen-tially three components of the traditional

attitudinal loyalty construct; and all threeshould lead to action/behavioral loyalty.Furthermore, Back argued that the cogni-tive, affective, and conative phases of loyaltymight not be a sequential formation process,as suggested by Oliver (1997, 1999). ToBack, the three aspects are more likely tobe independent factors of attitudinal loyaltyattributable to unique variance. Empiricaltesting revealed that both affective andconative loyalty were positively associatedwith behavioral loyalty, while cognitiveloyalty was not (Back, 2001; Back & Parks,2003). Notably, although he maintained thatcognitive, affective, and conative loyaltywere three elements of attitudinal loyalty;Back did not measure the overarchingconstruct of attitudinal loyalty, or includeit in his model.

Lee (2003) also adopted part of Oliver’sconceptualization. However, she argued that‘‘the cognitive stage is more likely to be anantecedent to loyalty rather than loyaltyitself’’ (p. 22). Thus, Lee’s loyalty measurecontained three dimensions: attitudinal,conative, and behavioral loyalty. Her studylent partial support to the three-dimensionalconceptualization. Although conative loyaltywas significantly and positively influenced byattitudinal loyalty, the direct effect of cona-tive loyalty on behavioral loyalty was foundto be negative, which was opposite of thehypothesized direction. Lee postulated thatthis negative relationship might be the resultof perceived constraints.

More recently, Jones and Taylor (2007)explored the dimensionality of customerloyalty. The authors suggested that withcognitive components of loyalty gettingmore attention, recent marketing literatureseems to support a three-dimensional con-ceptualization of loyalty (cognitive, attitudi-nal, and behavioral). Parallel to this, theinterpersonal psychology literature has tra-ditionally adopted a two-dimensional (beha-vioral and cognitive) conceptualization ofinterpersonal commitment—a construct clo-sely akin to loyalty. Jones and Taylor’s studysupported a two-dimensional loyalty con-struct, in which behavioral loyalty remains


as one dimension, while attitudinal andcognitive loyalty are combined into onedimension. A closer look at Jones andTaylor’s measures indicates that what theycalled ‘‘attitudinal loyalty’’ might be termed‘‘affective loyalty’’ in Oliver’s terminology,while their behavioral loyalty was essentiallyconative loyalty. Thus, Jones and Taylorrevealed a conative versus cognitive/affectiveloyalty structure.

Overall, it seems consensus has not beenreached on the specific structure of, ordimensions contained in the loyalty con-struct (Table 1). Nevertheless, recent discus-sion on loyalty dimensionality broadens,rather than invalidates the traditional two-dimension view.

The Proposed Model

Based on the foregoing review, the presentpaper attempts to integrate previous findingsand propose a conceptual model of loyaltydimensionality (Figure 1). Following recentconceptual development (Harris & Goode,2004; McMullan & Gilmore, 2003; Oliver,1999), the present research conceptualizesloyalty as a four-dimensional constructcomprising of cognitive, affective, conative,and behavioral components. The first threecomponents collectively represent the attitu-dinal aspect of loyalty. Together they form ahigher order factor termed attitudinal loy-alty, which then leads to behavioral loyalty.Since the behavioral aspect of loyalty hasbeen well supported and documented(Backman & Crompton, 1991; Cunningham,1956; Iwasaki & Havitz, 2004; Morais et al.,2004; Pritchard et al., 1999), the focus of thepresent paper is on the breakdown of theattitudinal aspect of loyalty.

The operational definition of brand loy-alty, and its four components are listedbelow:

N Cognitive Loyalty: The existence of be-liefs that (typically) a brand is prefer-able to others (Harris & Goode, 2004).

N Affective Loyalty: The customer’sfavorable attitude or liking toward the

service brand/provider based on satis-fied usage (Harris & Goode, 2004).

N Conative Loyalty: Behavioral intentionto repurchase the service brand char-acterized by a deep brand-specificcommitment (Harris & Goode, 2004).

N Behavioral Loyalty: The frequency ofrepeat or relative volume of same-brandpurchase (Tellis, 1988).

N (Brand) Loyalty: ‘‘A deeply held psy-chological commitment to rebuy orrepatronize a preferred product/serviceconsistently in the future, thereby caus-ing repetitive same-brand or samebrand-set purchasing, despite situa-tional influences and marketing effortshaving the potential to cause switchingbehavior’’ (Oliver, 1999, p. 34).

The model is developed from marketing,social psychology, and leisure literature. Thefour-dimensional structure originated fromOliver’s (1997, 1999) conceptualization.However, following Back (2001), the presentpaper argues that the first three dimensionsare three independent components of attitu-dinal loyalty—an overarching construct.This argument is theoretically grounded onthe widely accepted tripartite model ofattitude structure (Breckler, 1984; Eagly &Chaiken, 1993; Reid & Crompton, 1993).The tripartite model suggests that there arethree components of people’s attitudes:cognition, affect, and behavioral intention.The three components of attitude are inde-pendent of each other, and each exhibitsunique variance that is not shared by theother two (Bagozzi, 1978). Further, somehave argued that attitudes do not have toembrace all three components at the sametime (Tian, 1998). Thus, the three compo-nents may not be sequential as suggested byOliver (1997, 1999).

As a development of Back’s model (2001),which only contains first-order factors, thepresent model included attitudinal loyalty asa higher-order factor. This is also theoreti-cally grounded on the tripartite model ofattitude structure (Breckler, 1984). Finally,the attitude-behavior linkage (i.e., attitudinal


TABLE 1. Competing New Conceptualizations on Loyalty Dimensionality

Relationship Selected Studies

Loyalty building is a continuum, starting fromcognitive loyalty, followed by affective loyalty, toconative loyalty, and finally action (behavioralloyalty).

(Harris & Goode, 2004; McMullan & Gilmore,2003; Oliver, 1999; Oliver, 1997)

Loyalty, a higher order factor, is comprised of twodimensions: a behavioral element, and acombined attitudinal/cognitive element.

(Jones & Taylor, 2007)

Cognitive loyalty, affective loyalty, and conativeloyalty are 3 components of the traditionalattitudinal loyalty construct, and all 3 should leadto action/behavioral loyalty.

(Back, 2001; Back & Parks, 2003)

Loyalty building starts from affective loyalty,which leads to conative loyalty and thenbehavioral loyalty.

(Lee, 2003)

Xiang

(Robert)

Li

andJam

esF

.P

etrick73

loyalty leading to behavioral loyalty) hasbeen both theoretically and empiricallyestablished in the past (Ajzen, 1991;Albarracin, Johnson, Fishbein, &Muellerleile, 2001; Dick & Basu, 1994).

It is believed that the proposed concep-tualization is congruous with the traditionaltwo-dimension view of loyalty, which hasbeen widely accepted across disciplines, andhas generated meaningful results. A majordevelopment is that the present conceptuali-zation suggests that attitudinal loyalty is ahigher-order factor, comprising of cognitive,affective, and conative dimensions. Inessence, the proposed model incorporates,rather than invalidates the traditional two-dimensional view of loyalty.

RESEARCH METHODS

Instrument Development

The survey questionnaire was developedbased on a literature review, as well asextensive personal communications withleading loyalty researchers in the fields ofmarketing and leisure studies. To enhancethe quality of this review, the authors alsoposted a request for updated loyalty (orcommitment) literature on the Ameri-can Marketing Association Listserv, whichgenerated valuable inputs from scholars all

over the world. After the initial version ofthe questionnaire was developed, 14 expertswere invited to review and pretest theinstrument. Further, a shortened question-naire was pilot tested among three under-graduate classes (N 5 114). The finalinstrument was developed based on theexpert panel’s suggestions and pilot testresults.

In this study, three 7-point Likert-type scales proposed by Back (2001) andBack and Parks (2003) were used to mea-sure cognitive loyalty, affective loyalty,and conative loyalty, respectively (seeTable 2). Action or behavioral loyalty, fol-lowing the most frequently-used approach,was measured by proportion of brandpurchase (Cunningham, 1956; Iwasaki &Havitz, 1998). Specifically, this was opera-tionalized as the number of cruises therespondent had taken with the focal cruiseline in the past 3 years, divided by the totalnumber of cruises s/he had taken during thattime.

Online Panel Survey

This study utilized an online panel survey,which is a fairly commonplace method inmarketing research (Dennis, 2001;Deutskens, de Jong, de Ruyter, & Wetzels,2006; Duffy, Smith, Terhanian, & Bremer,2005; Hansen, 2005; Sparrow & Curtice,

FIGURE 1. The Proposed Structure of Brand Loyalty


2004; Van Ryzin, 2004). Online surveypanels ‘‘are made up of individuals who areprerecruited to participate on a more or lesspredictable basis in surveys over a period oftime’’ (Dennis, 2001, p. 34). Despite itsobvious advantage in cost efficiency andspeed, some researchers have expressedconcern regarding the validity of informa-tion collected from online panel studies—particularly due to the potential for samplingbias (Duffy et al.; McWilliams & Nadkarni,2005). Some researchers have even arguedthat repeat and paid participation in surveysmight bias online survey panelists’ attitudesand behaviors, and make them closer to‘‘professional respondents’’ (Dennis, 2001).However, a series of recent studies (Dennis,2001; Deutskens et al., 2006; Duffy et al.,2005) have revealed that, despite minordifferences, online panel and traditionalmethodologies generate equivalent resultsin most cases. Since the representativenessof public opinion is not the primary concernof the study, the authors deemed onlinepanel surveys appropriate for this study.

The Survey Process

The survey was conducted from March 15to 22, 2006. Participants of this study were

currently active cruisers, who took a cruisevacation in the past 12 months. FollowingCruise Lines International Association(CLIA) (2005), the authors specified fourdemographic and behavioral characteristicsof the sample when acquiring the onlinepanel. Participants of this study were cruisetravelers who cruised at least once in the past12 months, were over 25 years old, and had ahousehold income of $25,000 or more.Moreover, a 50–50 gender distribution wasdesired. For survey design purposes, onlyresponses about CLIA member cruise lines(CLIA, 2006b) were collected. These linesmake up 95% of the overall North Americacruise market (CLIA, 2006a). Further, cruiselines, rather than specific ships were chosento ensure that participants’ responses were atthe brand level.

The survey started from a screeningquestion, asking whether the respondenttook a cruise vacation in the past 12 monthsor not. Respondents who said ‘‘Yes’’ werepresented a list of CLIA’s member lines(CLIA, 2006b), and asked which line theycruised with on their most recent cruisevacation. Clicking any of the cruise com-pany names would lead the respondent tothe actual survey, which was customized

TABLE 2. Scale Wording and Measurement Property

Scale Itemsa Coeff. a (Back& Parks, 2003)

Coeff. a(Current)

Mean SD

Cognitive Loyalty (COG) 0.85 0.92cog1 ,name. provides me superior service quality as

compared to other cruise lines5.18 1.60

cog2 I believe ,name. provides more benefits than othercruise lines in its category

4.90 1.64

cog3 No other cruise line performs better services than,name.

4.27 1.88

Affective Loyalty (AFF) 0.87 0.94aff1 I love cruising with ,name. 5.49 1.61aff2 I feel better when I cruise with ,name. 4.64 1.77aff3 I like ,name. more than other cruise lines 4.60 1.90Conative Loyalty (CON) 0.86 0.90con1 I intend to continue cruising with ,name. 5.56 1.67con2 I consider ,name. my first cruising choice 4.91 1.95con3 Even if another cruise line is offering a lower rate, I still

cruise with ,name.4.00 1.98

aAll items were measured on 7-point scales.


to the brand being chosen. Those who hadnot cruised with any of CLIA cruise lines inthe past 12 months were thanked and askedto disregard the survey. A technical mechan-ism was used to ensure that all questions hadto be answered before submission. Thesurvey took approximately 12 minutes tocomplete.

The sample size needed for this study wasmainly determined by Cohen’s (1988) poweranalysis. Following MacCallum, Browne,and Sugawara (1996); the minimum samplesize for the proposed model (df 5 32) isapproximately 350, in order to achievepower of 0.80.

RESULTS

The aforementioned procedure yielded atotal of 727 responses, or, a response rate of31.8% out of 2,283 email invitations thatwere sent. The response rate of the presentstudy compares favorably to other onlinepanel studies (Zoomerang, 2005). Theauthors took a conservative approach anddeleted 61 invalid responses. Further,responses from 112 first-time cruisers wereexcluded. Thus, the effective sample size forthe present study was 554.

Sample Characteristics

Respondents were mostly male (55.8%),had an average age of 53.9, were dominantlywhite (91.7%), and married (80.5%). Abouttwo thirds (63.9%) had a college degree ormore and the median income was $75,000 to$100,000. On average, respondents hadtaken 8.3 cruises with 3.4 different lines intheir lifetime. For their brand purchasehistory (i.e., experiences with the specificcruise line they chose), respondents hadtaken an average of 3.1 cruises with thecruise line, and had a history of 6.2 yearscruising with that line.

Nonresponse bias was checked by com-paring three demographic characteristics(age, gender, and household income) of therespondents to those of the 2,283 people

invited to the survey. Overall, no significantbias was detected. Further, sampling biaswas checked by comparing respondents’demographic statistics to those of averagecruise passengers, as reported in CLIA’s2004 Cruise Market Profile (CLIA, 2005). Itseemed that respondents of this study weredemographically similar to typical cruisers,but slightly more active in behavior.

Modeling and Hypotheses Testing

A structural equation modeling (SEM)procedure was employed to analyze thedata. The analysis followed guidelines sug-gested by Byrne (2001) and Ullman (2001).Before testing the model, a variety ofpractical issues were checked—includingsample size, missing values, univariate andmultivariate outliers, continuous scales, lin-earity, univariate and multivariate normal-ity, and so on. The only detected issue wasthat Mardia’s (1970) normalized estimate ofmultivariate kurtosis was fairly large, whichsuggested the data might have a multi-variate nonnormal distribution. Oneapproach to dealing with multivariate non-normal data is nonparametric bootstrapping(Byrne, 2001; Kline, 2005). Thus, bootstrapresults based on 500 bootstrap samples arereported in the following section. Further,inter-correlations between major constructswere obtained, as recommended by Hatcher(1994). It was found that cognitive, affec-tive, and conative loyalty had exceedinglyhigh correlations (all . 0.97). This will beaddressed later.

The SEM procedure was conducted infour stages: (a) testing the proposed model,(b) model comparison, (c) model modifica-tion, and (d) assessing validity and reliabil-ity.

Stage 1: Testing the Proposed Model

To examine the proposed model, a second-order confirmatory factor analysis (CFA) wasemployed. A second-order factor model positsthat the first-order factors estimated (i.e.,cognitive, affective, and conative loyalty) areactually caused by a broader and more


encompassing construct (in this case, attitudi-nal loyalty). Hair, Anderson, Tatham, andBlack (1998) suggested that second-order CFAmodels allow for a stronger statement aboutthe dimensionality of a construct than tradi-tional approaches.

The second-order CFA model was testedfollowing a procedure recommended byByrne (2001). First, the identification of thehigher order portion of the model wasaddressed, since this part of the model wasinitially just-identified with 3 first-orderfactors. As suggested by Byrne, this problemcan be solved by placing equality constraintson certain parameters known to yield esti-mates that are approximately equal, throughthe application of the critical ratio difference(CRDIFF) method. It was found that theestimated values of the higher order residualsrelated to affective (20.0031) and conativeloyalty (20.021) were almost identical, andthe computed critical ratios for differencesbetween the two residuals were 20.703(absolute value , 1.96). Thus, it was decidedto constrain the variance of the residualsrelated to affective and conative loyalty to beequal. The hypothesized model, with theequality constraints specified, is presented inFigure 2.

The next step involved obtaining thegoodness-of-fit statistics and modificationindices (MI) (Sorbom, 1986) related to thehypothesized model. Since most researchershave argued that chi-square is highly sensi-tive to sample size, it has been suggested thatthe use of multiple indices may collectivelypresent a more realistic picture of model fit(McDonald & Ringo Ho, 2002). FollowingByrne’s (2001) recommendation, GFI(acceptable when . 0.9; Hu & Bentler,1995), CFI (acceptable when . 0.9;Bentler, 1990),, and RMSEA (acceptablewhen , 0.1; Browne & Cudeck, 1993) werechosen to assess model fitness. Also includedwere the normed chi-square (NC) (x2/df,acceptable when , 5; Bollen, 1989), and theBollen-Stine bootstrap x2 (BSboot)—the chi-square test based on Bollen and Stine’s(1992) bootstrap procedure.

Considering the model was neither toolarge nor complex, the goodness-of-fit sta-tistics indicated a poor fit (see Table 3). Themultiple large MI values further evidencedthat there could be substantial misfit in thehypothesized second-order model structure.Further, the MI results were fairly complex,and did not present a meaningful solution toimprove the model fit.

FIGURE 2. Hypothesized Second-Order Model


Stage 2: Model Comparison

For years, statisticians have called for theuse of alternative models (i.e., comparing theperformances of rival a priori models) inmodel specification and evaluation (Bagozzi& Yi, 1988; Joreskog & Sorbom, 1996;MacCallum & Austin, 2000). Thus, theauthors examined alternative loyalty con-ceptualizations by testing a series of compet-ing models (Table 1). These included:

N Rival Model 1: Oliver’s four-dimen-sional sequential model (Harris &Goode, 2004; McMullan & Gilmore,2003; Oliver, 1997, 1999);

N Rival Model 2: Back’s four-dimen-sional first-order model (Back, 2001;Back & Parks, 2003);

N Rival Model 3: Lee’s three-dimensionalsequential model (Lee, 2003); and

N Rival Model 4: The traditional two-dimensional model (Backman &Crompton, 1991; Day, 1969; Dick &Basu, 1994; Jacoby & Chestnut, 1978;Pritchard et al., 1999).

Table 3 displays the fitness statistics ofthese models. It seems that the fitness levelsof all these models were no different from, oreven worse than the hypothesized one. Inother words, none of the models provided agood fit of the data. In light of these results,it was decided that exploratory analysisshould be used to purify measures(Churchill, 1979).

Stage 3. Model Modification

Following Churchill’s (1979) recommen-dation, an exploratory factor analysis (EFA)was employed to identify the potential

pattern of the nine items, which weresupposed to measure cognitive, affective,and conative loyalty. Note that the EFAresults should and would only serve as areference for the present discussion onloyalty dimensionality. It was found thatthe nine items in discussion all loaded on asingle dimension, instead of the three dimen-sions hypothesized. Next, Cronbach’s alpha,and alpha-if-item-deleted analysis was alsoperformed. The Cronbach’s alpha for thenine items was quite high, and deleting anyone of the items would have little effect onalpha.

The EFA results seemed to support theone-dimension conceptualization of attitu-dinal loyalty. Further, recall that the inter-correlations among cognitive, affective, andconative loyalty were exceptionally high (allexceeding 0.97). Kline (2005) suggested thatwhen two factors have a correlation over0.85, they may not be accommodated in onestructural equation model, as the twofactors demonstrate poor discriminantvalidity (Rundle-Thiele, 2005), and couldcause SEM to be statistically unstable. Putsimply, they may be measuring the sameconstruct. These results implied that thetraditional one-dimensional conceptualiza-tion of attitudinal loyalty was theoreticallyand statistically more solid than the pro-posed model.

Moreover, the alpha-if-item-deleted ana-lysis showed that when all nine items wereused to measure one single first-order factor,they might be redundant with each other.Byrne (2001, p. 134), in her discussion onmodel modification, suggested ‘‘error corre-lations between item pairs are often anindication of perceived redundancy in itemcontent.’’ To solve such problems, some

TABLE 3. Goodness-of-Fit Statistics of the Models

x2(df) NC BSboot CFI RMSEA GFI

The Proposed Model 479.193 (32) 14.975 0.002 0.934 0.159 0.83Rival Model 1 480.497(33) 14.561 0.002 0.934 0.157 0.829Rival Model 2 2731.295 (33) 82.761 0.002 0.605 0.385 0.610Rival Model 3 356.977 (13) 27.460 0.002 0.920 0.219 0.838Rival Model 4 495.104 (35) 14.146 0.002 0.933 0.154 0.829


researchers have suggested that deletingquestionable items could be an effectiveway to improve a measurement model with-out sacrificing its theoretical meaningfulness(Bentler & Chou, 1987; Byrne, 2001; Morais,Backman, & Dorsch, 2003). Further,Hatcher (1994) recommended that to avoidexcessive complexity in measurement mod-els, researchers may limit the number ofindicators used to measure one latent vari-able to around four. Netemeyer, Bearden,and Sharma (2003) also maintained thatshorter scales are typically preferred.

In light of these recommendations, it wasconcluded that the initial misfit of RivalModel 4 might be due to redundant items,and deleting these items may generate abetter measure of one-dimensional attitudi-nal loyalty. This modification process,though post hoc in nature, strictly followedrecommended procedures (Bentler & Chou,1987; Byrne, 2001; Hatcher, 1994). Itemsassociated with questionable MIs, insignif-icant paths (if at all), large standardizederrors—and most importantly, conceptual orsemantic fuzziness—were considered as can-didates for deletion.

Specifically, this deletion process startedwith CON3, which had the largest standarderror, and a comparatively weaker path.Two other items—AFF1 and CON1—weresubsequently deleted, as both items wereassociated with multiple significant MIs. Infact, several expert panelists mentioned inthe pilot test phase that AFF1 was somewhatconfusing. Finally, COG1 was deleted basedon its comparatively large residuals, andweak loadings, as well as its semanticredundancy with the other two cognitiveitems. This process resulted in a one-dimen-sional loyalty measure containing five items:COG2 (‘‘I believe ,name. provides morebenefits than other cruise lines in its cate-gory’’), COG3 (‘‘No other cruise line per-forms better services than ,name.’’), AFF2(‘‘I feel better when I cruise with ,name.’’),AFF3 (‘‘I like ,name. more than othercruise lines’’), and CON2 (‘‘I consider,name. my first cruising choice’’). Thefive-item model, with x2(5, N 5 554) 5

26.131, p , 0.001, CFI 5 0.994, GFI 50.982, RMSEA 5 0.087, demonstrated goodfit.

Finally, the modified loyalty model wastested in a structural equation model, withattitudinal loyalty as an exogenous variable,and behavioral loyalty as an endogenousvariable (see Figure 3). The model, with x2(9,N 5 554) 5 52.399, p , 0.001, CFI 5 0.988,GFI 5 0.969, RMSEA 5 0.093, demon-strated a good fit of the data. However, itwas noted that the RSMC

2 (0.115) ofBEHLOY was fairly low, which indicatedthat attitudinal loyalty accounted for only asmall portion of the variance associated withbehavioral loyalty.

Stage 4. Assessing Validity and Reliability

The preceding procedure, though post hocin nature, essentially generated a five-itemscale measuring attitudinal loyalty. Beforedrawing final conclusions, the authorsdeemed it necessary to examine the psycho-metric properties of this measure. First,convergent validity of indicators is evidencedby the ability of the scale items to load on itsunderlying construct (Bagozzi, 1994).Convergent validity may be further evi-denced if each indicator’s standardizedloading on its posited latent construct isgreater than twice its standard error(Anderson & Gerbing, 1988). All itemsunder investigation met these two require-ments.

Second, discriminant validity may beassessed by comparing the average variance

FIGURE 3. Exploring the Relationship BetweenAttitudinal Loyalty and Behavioral Loyalty


extracted (AVE) for the focal measure with asimilar, but conceptually different, construct;and the square of the correlation between thetwo factors (Hatcher, 1994; Netemeyer et al.,2003). Discriminant validity is demonstratedif both AVEs are greater than the squaredcorrelation. This requirement was satisfiedafter checking the AVEs and the squaredcorrelation value for the attitudinal loyaltymeasure and three similar, but conceptuallydifferent constructs (satisfaction, quality,and value) (see Table 4). Thus, discriminantvalidity of the scale was established.

Third, scale reliability was checked inmultiple ways. These included Cronbach’scoefficient alpha (a values need to exceed 0.7;Nunnally & Bernstein, 1994), indicatorreliability (RSMC

2 needs to exceed 0.5;Fornell & Larcker, 1981), composite relia-bility (the recommended cutoff point is 0.6;Bagozzi & Yi, 1988), and AVE (AVE needsto exceed 0.5; Fornell & Larcker, 1981). It

was found that the five-item measure met allthese requirements.

Finally, nomological validity is consideredto be established when the proposed measuresuccessfully predicts other constructs thatprevious literature suggests it should predict(Netemeyer et al., 2003). To test it, theauthors ran three regression models, whereattitudinal loyalty (operationalized as themean of the five items) was modeled aspredictors of three behavioral outcomes. Thethree variables—all of which have beensuggested as loyalty outcomes—includedrepurchase intention (Morais et al., 2004),willingness to recommend (Dick & Basu,1994), and complaining behavior (Davidow,2003). As shown in Table 5, in all three models,attitudinal loyalty’s effect on the dependentvariables was statistically significant, and itseffects were consistent with what has beenpreviously observed (Davidow, 2003; Dick &Basu, 1994; Morais et al., 2004; Petrick, 2004;

TABLE 4. Correlations Between Major Constructs

VAL QUA ATTLOY SAT

Value (VAL)d 0.849a 0.630c 0.551 0.623Quality (QUA)e 0.794b 0.929 0.567 0.663Attitudinal Loyalty(ATTLOY)

0.742 0.753 0.873 0.555

Satisfaction (SAT)f 0.789 0.814 0.745 0.841

aThe diagonal entries (in italics) represent the average variance extracted by the construct.b The correlations between constructs are shown in the lower triangle.c The upper triangle entries represent the variance shared (squared correlation) between constructs.d Measured by Sirdeshmukh, Singh, and Sabol’s (2002) four-item, 7-point scale.e Measured by Petrick’s (2002) four-item, 7-point subscale of his SERV-PERVAL scale.f Measured by Spreng, MacKenzie, and Olshavsky’s (1996) four-item, 7-point scale.

TABLE 5. Summary of Regression Analyses

Dependent Variable B SE B F df R2 Radj2

Repurchase Intention a 0.552 0.016 .827*** 1195.218 553 0.684 0.683Willingness toRecommend b

1.288 0.043 0.785*** 883.765 553 0.616 0.615

Complaining Behavior c 20.0766 0.029 20.112** 6.962 553 0.012 0.011

Note. ** p , .01, *** p , .001.a Measured by Grewal, Monroe, and Krishnan’s (1998) two-item, 5-point scale.b Measured by Reichheld’s (2003) one-item, 11-point scale.c Measured by Rundle-Thiele’s (2005) seven-item, 7-point scale.


Rundle-Thiele, 2005). These provide furthersupport for the validity of the scale.

Combined, tests on the convergent, dis-criminant and nomological validity, and thereliability of the five-item measure showedthat it served as a good measure of thesingle-dimensioned attitudinal loyalty con-struct. It was thus concluded that the five-item measure, measuring attitudinal loyaltyas a single-dimension, first-order construct,demonstrated better fit of data than thehypothesized second-order model.

CONCLUSIONS ANDIMPLICATIONS

This study attempted to explore thedimensional structure of the loyalty con-struct. Following recent developments inloyalty studies (Back, 2001; Jones &Taylor, 2007; Oliver, 1997, 1999), loyalty inthis paper was conceptualized as a four-dimensional construct, comprising of cogni-tive, affective, conative, and behavioralloyalty. Further, this paper postulated thatthree components of loyalty (cognitive,affective, and conative loyalty) collectivelyformed a higher order factor—namely atti-tudinal loyalty. However, this conceptualiza-tion was not supported by the data. Acompeting model based on the traditionalconceptualization that attitudinal loyalty is aone-dimensional, first-order factor wasfound to provide a better fit of the datathan other possible variations. Further, thepaper supported the attitudinal loyalty-behavioral loyalty link (Ajzen, 1991;Albarracin et al., 2001; Dick & Basu,1994). Nevertheless, the relatively low var-iance of behavioral loyalty explained byattitudinal loyalty suggests that the atti-tude-behavior link may be moderated byother factors, which is also consistent withprevious studies (Back, 2001; Dick & Basu,1994).

In sum, this study supported the tradi-tional two-dimensional conceptualization ofloyalty, which maintains that loyalty has anattitudinal and a behavioral component

(Backman & Crompton, 1991; Cunningham,1956; Iwasaki & Havitz, 2004; Morais et al.,2004; Pritchard et al., 1999). Moreover, thisfinding seems to be congruent with psychol-ogy literature on interpersonal commitment,which has consistently suggested that pro-relationship acts (i.e., commitment) have twocomponents—behavioral and cognitive(Jones & Taylor, 2007). Findings are alsosimilar to Jones and Taylor, who concludedthat ‘‘…regardless of the target (friend,spouse, service provider), loyalty captures,in essence, what Oliver (1999) referred to as‘what the person does’ (behavioral loyalty)and the psychological meaning of the rela-tionship (attitudinal/cognitive loyalty)’’(p. 45).

While the two-dimensional conceptualiza-tion of brand loyalty is not new to marketingor psychology researchers, what the presentresults reveal is that the two dimensionsmight be more complex than previouslysuggested. Remaining in the final five-itemattitudinal loyalty measure are cognitive,affective, and conative components; whichis consistent with the tripartite model ofattitude structure in the psychology litera-ture (Breckler, 1984; Eagly & Chaiken, 1993;Reid & Crompton, 1993). One might spec-ulate that although these three aspects ofloyalty loaded in the same dimension, theycould account for unique aspects of theconstruct. Admittedly, the present resultsmay also imply that the respondents couldn’ttell the differences between cognitive, affec-tive, and conative loyalty; even though thesecomponents make conceptual sense.

In addition to clarifying the conceptualstructure of customers’ brand loyalty, thisresearch also contributes to the literature byintroducing and validating a five-item atti-tudinal loyalty measure. The scale wasdeemed to be theoretically and psychome-trically sound, and might be used in futureloyalty research.

Although this study is primarily theore-tical, it is believed that the revealed con-ceptual structure of customer brand loyaltymay provide insights for cruise management.Although the data did not support the


MacBook

proposed multidimensional structure of atti-tudinal loyalty, the final five-item scale doescontain cognitive, affective, and conativecomponents. For many service providerswho focus primarily on the technical aspectsof their services (i.e., helping customers buildcognitive belief), this suggests that theyshould include affective and conative infor-mation in their marketing messages. Further,the relatively low variance of behavioralloyalty explained by attitudinal loyaltysuggests that simply winning customers’positive attitude does not necessarily leadto positive outcomes. Consumer behavior isextremely complicated, and marketers needto better understand other moderators to theattitude-behavior link.

Facing more sophisticated customers andchallenged by more aggressive competitors,cruise line management, as well as manyother tourism sectors, have invested tremen-dous resources to retain and reward loyalcustomers. The present scale provides afeasible tool for identifying, and potentiallysegmenting loyal and disloyal customers.Information generated via this tool may helpmanagers design loyalty programs, andreward the right type of customer attitudesand behaviors (Jones & Taylor, 2007). Itmay also facilitate the benchmarking ofcustomers’ loyalty within, and across differ-ent tourism services.

LIMITATIONS AND FUTURERESEARCH

The present results may be limited torespondents who participated in this study,and who cruised at least once with one ofCLIA’s member lines in the past 12 months.Further research is necessary in order todetermine whether the conceptual structurecan be generalized to cruise passengers inother cultures and geographic regions, otherrecreationists, and ultimately consumers ofdifferent services.

Another limitation of this study is it didnot consider differences in cruise lines.Employing different marketing strategies

and loyalty programs and targeting differentmarket segments, the cruise lines used in thisstudy might exhibit considerable differencesaffecting customer loyalty building. It isuncertain whether and how these ‘‘noises’’will influence the theoretical relationshipssuggested. It is quite possible that the currentresults are very different at the individualcruise line level; and that by combiningcruise lines, the present results cannot beapplied at the individual cruise line level.

The five-item attitudinal loyalty scale usedin this study, though demonstrating goodvalidity and reliability, was generated frompost hoc analyses. Admittedly, the originalpurpose of this paper is to examine thedimensionality of the loyalty construct, notscale development. Thus, the study is furtherlimited by not going through a completescale development process (Churchill, 1979;Netemeyer et al., 2003).

Yet, in conclusion, it is believed that thisstudy contributes to the literature by system-atically reviewing and empirically examiningrecent conceptual developments on loyaltydimensionality. As a result, the traditionaltwo-dimensional loyalty conceptualizationwas revalidated, and a five-item attitudinalloyalty scale was generated. It is hoped thatthese findings will provide new insights forcustomer loyalty research, measurement, andmanagement.

ENDNOTE

1. The negative residuals here, considering theirmagnitude, may be treated as 0 (Kline, 2005).

REFERENCES

Ajzen, I. (1991). The theory of planned behavior.Organizational Behavior and Human DecisionProcesses, 50(2), 179-211.

Albarracin, D., Johnson, B. T., Fishbein, M., &Muellerleile, P. A. (2001). Theories of reasoned actionand planned behavior as models of condom use: Ameta-analysis. Psychological Bulletin, 127(1), 142-161.

Anderson, J. C., & Gerbing, D. W. (1988). Structuralequation modeling in practice: A review and


recommended two-step approach. PsychologicalBulletin, 103(3), 411-423.

Assael, H. (2004). Consumer behavior: A strategicapproach. Boston: Houghton Mifflin.

Back, K. (2001). The effects of image congruence oncustomer satisfaction and brand loyalty in the lodgingindustry. Unpublished doctoral dissertation, ThePennsylvania State University.

Back, K., & Parks, S. C. (2003). A brand loyalty modelinvolving cognitive, affective, and conative brandloyalty and customer satisfaction. Journal ofHospitality & Tourism Research, 27(4), 419-435.

Backman, S. J., & Crompton, J. L. (1991). Theusefulness of selected variables for predictingactivity loyalty. Leisure Sciences, 13(3), 205-220.

Bagozzi, R. P. (1978). The construct validity of theaffective, behavioral, and cognitive components ofattitude by analysis of covariance structures.Multivariate Behavioral Research, 13(1), 9-31.

Bagozzi, R. P. (1994). Principles of marketing research.Oxford, UK: Basil Blackwell.

Bagozzi, R. P., & Yi, Y. (1988). On the evaluation ofstructural equation models. Journal of the Academyof Marketing Science, 16(1), 74-94.

Bass, F. M. (1974). The theory of stochastic preferenceand brand switching. Journal of MarketingResearch, 11(1), 1-20.

Bentler, P. M. (1990). Comparative fit indexes instructural models. Psychological Bulletin, 107(2),238-246.

Bentler, P. M., & Chou, C.-P. (1987). Practical issuesin structural modeling. Sociological Methods &Research, 16(1), 78-117.

Bollen, K. A. (1989). Structural equations with latentvariables. New York: Wiley.

Bollen, K. A., & Stine., R. A. (1992). Bootstrappinggoodness-of-fit measures in structural equationmodels. Sociological Methods and Research, 21(2),205-229.

Breckler, S. J. (1984). Empirical validation of affect,behavior, and cognition as distinct components ofattitude. Journal of Personality and SocialPsychology, 47(6), 1191-1205.

Browne, M. W., & Cudeck, R. (1993). Alternativeways of assessing fit. In K. A. Bollen & J. S. Long(Eds.), Testing structural equations models (pp. 445-455). Newbury Park, CA: Sage.

Byrne, B. M. (2001). Strucutral equation modeling withAMOS: Basic concepts, applications, and program-ming. Mahwah, NJ: Lawrence Erlbaum Associates.

Churchill, G. A. J. (1979). A paradigm for developingbetter measures of marketing constructs. Journal ofMarketing Research, 16(1), 64-73.

Cohen, J. (1988). Statistical power analysis for thebehavioral sciences (2nd ed.). Hilldale, NJ:Lawrence Erlbaum Associates.

Cruise Lines International Association. (2005). CLIA’s2004 cruise market profile: Report of findings.Retrieved April 8, 2006, from http://www.cruising.org/press/research/2004%20market%20Profile%20Presentation_files/frame.htm

Cruise Lines International Association. (2006a). AboutCLIA. Retrieved April 8, 2006, from http://www.cruising.org/about.cfm

Cruise Lines International Association. (2006b). Cruiselines & ship profiles. Retrieved March 11, 2006, fromhttp://www.cruising.org/CruiseLines/index.cfm

Cunningham, R. M. (1956). Brand loyalty—What,where, how much? Harvard Business Review, 34(1),116-128.

Davidow, M. (2003). Organizational responses tocustomer complaints: What works and whatdoesn’t. Journal of Service Research, 5(3), 225-250.

Day, G. S. (1969). A two-dimensional concept ofbrand loyalty. Journal of Advertising Research, 9(3),29-35.

Dennis, J. M. (2001). Are Internet panels creatingprofessional respondents? Marketing Research,13(2), 34-38.

Deutskens, E. C., de Jong, A., de Ruyter, K., &Wetzels, M. (2006). Comparing the generalizabilityof online and mail surveys in cross-national servicequality research. Marketing Letters, 17(2), 119-136.

Dick, A. S., & Basu, K. (1994). Customer loyalty:Toward an integrated framework. Journal of theAcademy of Marketing Science, 22(2), 99-113.

Dimanche, F., & Havitz, M. E. (1994). Consumerbehavior and tourism: Review and extension offour study areas. Journal of Travel & TourismMarketing, 3(3), 37-57.

Duffy, B., Smith, K., Terhanian, G., & Bremer, J.(2005). Comparing data from online and face-to-face surveys. International Journal of MarketResearch, 47(6), 615-639.

Eagly, A. H., & Chaiken, S. (1993). The psychology ofattitudes. Orlando, FL: Harcourt BraceJovanovich.

Ehrenberg, A. S. C. (1988). Repeat buying: Theory andapplication London: Charles Griffin and Co.

Fornell, C., & Larcker, D. F. (1981). Evaluatingstructural equation models with unobservable vari-ables and measurement error. Journal of MarketingResearch, 18(1), 39-50.

Fournier, S. (1998). Consumers and their brands:Developing relationship theory in consumerresearch. Journal of Consumer Research, 24(4),343-373.

Grewal, D., Monroe, K. B., & Krishnan, R. (1998).The effects of price-comparison advertising onbuyers’ perceptions of acquisition value, transac-tion value, and behavioral intentions. Journal ofMarketing, 62(2), 46-59.


Gronroos, C. (1994). From marketing mix to relation-ship marketing: Towards a paradigm shift inmarketing. Asia-Australia Marketing Journal, 2(1),9-29.

Guest, L. (1944). A study of brand loyalty. Journal ofApplied Psychology, 28(1), 16-27.

Gummersson, E. (2002). Relationship marketing and anew economy: It’s time for deprogramming.Journal of Services Marketing, 16(7), 585-589.

Hair, J. F., Jr., Anderson, R. E., Tatham, R. L., &Black, W. C. (1998). Multivariate data analysis (5thed.). Upper Saddle River, NJ: Prentice-HallInternational, Inc.

Hansen, T. (2005). Understanding consumer onlinegrocery behavior: Results from a Swedish study.Journal of Euromarketing, 14(3), 31-58.

Harris, L. C., & Goode, M. M. H. (2004). The fourlevels of loyalty and the pivotal role of trust: Astudy of online service dynamics. Journal ofRetailing, 80(2), 139-158.

Hatcher, L. (1994). A step-by-step approach to using theSAS system for factor analysis and structuralequation modeling. Cary, NC: SAS Institute Inc.

Hu, L. T., & Bentler, P. M. (1995). Evaluating modelfit. In R. H. Hoyle (Ed.), Structural equationmodeling: Concepts, issues, and applications (pp.76-99). Thousand Oaks, CA: Sage.

Iwasaki, Y., & Havitz, M. (1998). A path analyticmodel of the relationships between involvement,psychological commitment, and loyalty. Journal ofLeisure Research, 30(2), 256-280.

Iwasaki, Y., & Havitz, M. (2004). Examining relation-ships between leisure involvement, psychologicalcommitment and loyalty to a recreation agency.Journal of Leisure Research, 36(1), 45-72.

Jacoby, J., & Chestnut, R. (1978). Brand loyaltymeasurement and management. New York: Wiley.

Jain, A. K., Pinson, C., & Malhotra, N. K. (1987).Customer loyalty as a construct in the marketing ofbanking services. International Journal of BankMarketing, 5(3), 49-72.

Jarvis, L. P., & Wilcox, J. B. (1976). Repeat purchasingbehavior and attitudinal brand loyalty: Additionalevidence. In K. L. Bernardt (Ed.), Marketing:1776–1976 and beyond (pp. 151-152). Chicago:American Marketing Association.

Jones, T., & Taylor, S. F. (2007). The conceptualdomain of service loyalty: How many dimensions?Journal of Services Marketing, 21(1), 36-51.

Joreskog, K. G., & Sorbom, D. (1996). LISREL 8user’s reference guide. Chicago: Sci. Software Int.

Kline, R. B. (2005). Principles and practice of structuralequation modeling (2nd ed.). New York: Guilford.

Knox, S., & Walker, D. (2001). Measuring andmanaging brand loyalty. Journal of StrategicMarketing, 9(2), 111-129.

Kraus, S. J. (1995). Attitudes and the prediction ofbehavior: A meta-analysis of the empirical litera-ture. Personality and Social Psychology Bulletin,21(1), 58-75.

Lee, J. (2003). Examining the antecedents of loyalty in aforest setting: Relationships among service quality,satisfaction, activity involvement, place attachment,and destination loyalty. Unpublished doctoral dis-sertation, The Pennsylvania State University.

Lois, P., Wang, J., Wall, A., & Ruxton, T. (2004).Formal safety assessment of cruise ships. TourismManagement, 25(1), 93-109.

MacCallum, R. C., & Austin, J. T. (2000).Applications of structural equation modeling inpsychological research. Annual Review ofPsychology, 51(1), 201-226.

MacCallum, R. C., Browne, M. W., & Sugawara, H.M. (1996). Power analysis and determination ofsample size for covariance structure modelingPsychological Methods, 1(2), 130-149.

Mardia, K. V. (1970). Measures of multivariateskewness and kurtosis with applications.Biometrika, 57(3), 519-530.

McDonald, R. P., & Ringo Ho, M.-H. (2002).Principles and practice in reporting structuralequation analyses. Psychological Methods, 7(1),64-82.

McMullan, R., & Gilmore, A. (2003). The conceptualdevelopment of customer loyalty measurement: Aproposed scale. Journal of Targeting, Measurementand Analysis of Marketing, 11(3), 230-243.

McWilliams, E. G., & Nadkarni, N. (2005).Differences in reported travel behavior for anonline panel versus mail panel [Electronic version].e-Review of Tourism Research (eRTR), 3, 16-17.

Miller, A. R., & Grazer, W. F. (2003). Complaintbehavior as a factor in cruise line losses: An analysisof brand loyalty. Journal of Travel & TourismMarketing, 15(1), 77-91.

Morais, D. B. (2000). Reconceptualization of loyaltyunder a resource investment perspective: A study ofgroup leaders in the leisure service industry.Unpublished doctoral dissertation, ClemsonUniversity, SC.

Morais, D. B., Backman, S. J., & Dorsch, M. J. (2003).Toward the operationalization of resource invest-ments made between customers and providers of atourism service. Journal of Travel Research, 41(4),362-374.

Morais, D. B., Dorsch, M. J., & Backman, S. J. (2004).Can tourism providers buy their customers’ loyalty?Examining the influence of customer-providerinvestments on loyalty. Journal of TravelResearch, 42(3), 235-243.

Morais, D. B., Dorsch, M. J., & Backman, S. J. (2005).Building loyal relationships between customers and


providers: A focus on resource investments. Journalof Travel & Tourism Marketing, 18(1), 49-57.

Netemeyer, R. G., Bearden, W. O., & Sharma, S.(2003). Scaling procedures: Issues and applications.Thousand Oaks, CA: Sage Publications.

Nunnally, J., & Bernstein, I. (1994). Psychometrictheory (3rd ed.). New York: McGraw-Hill.

Oliver, R. L. (1997). Satisfaction: A behavioralperspective on the consumer. New York: Irwin/McGraw-Hill.

Oliver, R. L. (1999). Whence consumer loyalty.Journal of Marketing, 63(4), 33-44.

O’Mally, L. (1998). Can loyalty schemes really buildloyalty? Marketing Intelligence and Planning, 16(1),47-65.

Petrick, J. F. (2002). Development of a multi-dimensional scale for measuring the perceived valueof a service. Journal of Leisure Research, 34(2), 119-134.

Petrick, J. F. (2004). Are loyal visitors desired visitors?Tourism Management, 25(4), 463-470.

Pine, B. J., & Gilmore, J. H. (1999). The experienceeconomy. Boston: Harvard Business School Press.

Pritchard, M. P., Havitz, M. E., & Howard, D. (1999).Analyzing the commitment-loyalty link in servicecontexts. Journal of the Academy of MarketingScience, 27(3), 333-348.

Reichheld, F. F. (2003). The one number you need togrow. Harvard Business Review, 81(12), 2-10.

Reid, L. J., & Crompton, J. L. (1993). A taxonomy ofleisure purchase decision paradigms based on levelof involvement. Journal of Leisure Research, 25(2),182-202.

Rundle-Thiele, S. (2005). Loyalty: An empiricalexploration of theoretical structure in two servicemarkets. Unpublished doctoral dissertation,University of South Australia, Adelaide, Australia.

Selin, S. W., Howard, D. R., Udd, E., & Cable,T. T. (1988). An analysis of consumer loyalty tomunicipal recreation programs. Leisure Sciences,10(3), 217-223.

Sheth, J. N., & Parvatlyar, A. (1995). Relationshipmarketing in consumer markets: Antecedents andconsequences. Journal of the Academy of MarketingScience, 23(4), 255-271.

Shoemaker, S., & Lewis, R. C. (1999). Customerloyalty: The future of hospitality marketing.International Journal of Hospitality Management,18(4), 345-370.

Sirdeshmukh, D., Singh, J., & Sabol, B. (2002).Consumer trust, value, and loyalty in relationalexchanges. Journal of Marketing, 66(1), 15-37.

Sorbom, D. (1986). Model modification (ResearchReport No. 86-3). Uppsala, Sweden: University ofUppsala, Department of Statistics.

Sparrow, N., & Curtice, J. (2004). Measuring theattitudes of the general public via internet polls: Anevaluation. International Journal of MarketResearch, 46(1), 23-44.

Spreng, R. A., MacKenzie, S. B., & Olshavsky, R. W.(1996). A reexamination of the determinants ofconsumer satisfaction. Journal of Marketing, 60(3),15-32.

Tellis, G. J. (1988). Advertising exposure, loyalty, andbrand purchase: A two-stage model of choice.Journal of Marketing Research, 25(2), 134-144.

Tian, S. (1998). Testing the efficacy of an attitudinalprocess model of the relationship between servicequality and visitor satisfaction in a tourism context.Unpublished doctoral dissertation, Texas A&MUniversity, College Station, TX.

Tucker, W. T. (1964). The development of brandloyalty. Journal of Marketing Research, 1(3), 32-35.

Ullman, J. B. (2001). Structural equation modeling. InB. G. Tabachnick & L. S. Fidell (Eds.), Usingmultivariate statistics (4th ed., pp. 653-771).Needham Heights, MA: Allyn & Bacon.

Van Ryzin, G. G. (2004). The measurement of overallcitizen satisfaction. Public Performance &Management Review, 27(3), 9-28.

Zoomerang (2005). ZoomPanel and Sample SalesTraining [Powerpoint file]. San Francisco: Author.

SUBMITTED: June 28, 2007FINAL REVISION SUBMITTED:

September 18, 2007ACCEPTED: September 27, 2007REFEREED ANONYMOUSLY


reexamining the dimensionality of brand loyalty: a …

Documents