DOI: Conjoint Analysis of Tourist Choice of Hotel ...Conjoint Analysis of Tourist Choice of Hotel...

INTERNATIONAL JOURNAL OF TOURISM RESEARCHInt. J. Tourism Res. (2012)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/jtr.1899

Conjoint Analysis of Tourist Choice ofHotel Attributes Presented in TravelAgent BrochuresRubén Huertas-Garcia1,*, Marta Laguna García2 and Carolina Consolación31Department of Economics and Business Administration, Barcelona University, Barcelona, Spain2Department of Administration and Marketing Research, Valladolid University, Segovia, Spain3Department of Business Administration, Technical University of Catalonia, Barcelona, Spain

ABSTRACT

The purposes of this study are to demonstratehow it is possible to determine which attributesare the most important in the final choices oftourists who use a travel agent brochure as asource of information and how these attributesinfluence perceived value in a pre-purchasestage.We conduct the study in three phases: (i) aqualitative study, (ii) an experiment usingchoice-based conjoint analysis by means of afractional factorial experimental design and(iii) another experiment using a full factorialderived from the same design. Results suggestadvertisement size, a hotel’s starred rating andprice influence perceived value at this stage. Thepresence of a positive combined effect of priceand advertising was found. Implications anddirections for future applications are offered,focusing particularly on marketing services.Copyright © 2012 John Wiley & Sons, Ltd.

Received 05 November 2010; Revised 29 May 2012; Accepted11 June 2012

Keywords: perceived value; choice-based conjointanalysis; tourist brochure; experiments in blocks.

INTRODUCTION

Despite the increasing importance of theInternet as a communication and distri-bution instrument, brochures continue

*Correspondence to: Rubén Huertas-García, Departmentof Economics and Business Administration, BarcelonaUniversity, Avda. Diagonal, 696, Barcelona, Spain.E-mail: [email protected]

to be a common source of information for tourists(Bieger et al., 2000; Andereck, 2005; Molina andEsteban, 2006). The influence information pub-lished in brochures has on the decision-makingprocess means brochures play an important rolein promoting tourist businesses and destinations.Several studies analyse the influence varioussources of information have on destination choice.Andereck (2005) argued that evaluations of bro-chures by tourists influence interest in visiting adestination, and these evaluations influence thechoice of destination as well. Hsieh and O’Leary(1993) reported that brochures are the third mostcommon source of information used by travellers,after interpersonal communication from relativesand friends. In one of the most technologicallyadvanced countries, Yamamoto and Gill (1999)found that brochures are one of the two mostimportant sources of information used by Japanesetravellers who buy travel packages.More recently, Chiou et al. (2008) suggested

that traditional brochures greatly influencebehaviour in verbalizer consumers, those whoprefer to process written or verbal informationfrom sources such as relatives, friends or travelagents over pictorials. Ortega and Rodriguez(2007) found that brochures are one of the mostimportant audio-visual formats for Spanishtourists seeking destination information. Molinaand Esteban (2006) proposed a model ofbrochure utility and studied its significance asimage generators. Clearly, the brochure is notan extinct product.Despite the importance of tourist brochures as

communication tools, little is known about theireffectiveness. Some researchers call attention tolarge amounts of money spent by tourism distri-buters in brochure production but rarely

Copyright © 2012 John Wiley & Sons, Ltd.

R. Huertas-Garcia, M. Laguna García and C. Consolación

demonstrate their usefulness or effectiveness inmeeting objectives (Hodgson, 1993; Andereck,2005). Little is known about the effect of touristbrochures as tools to promote destination andtourism organizations. There is also little empir-ical evidence concerning effects generated by lay-outs presented to tourists on brochures, especiallyin the pre-purchase stage. The literature containssufficient evidence to suggest that the way infor-mation is presented, the layout, influences con-sumer evaluations and purchase intentions(Munger and Grewal, 2001; Rewtrakunphaiboonand Oppewal, 2008). The purpose of this study isto demonstrate that it is possible to determine themost important attributes people consider whenchoosing a hotel and using a travel agentbrochure as a source of information. This paperreports results from a two-step experiment wherethe design of the second was derived from theresults of the first. This process helps determinewhat kind of information displayed by a hotelin the tourist brochures influences a consumer’sperceived value and choice.

THE TOURIST CHOICE PROCESS ANDINFORMATION SEARCHES

Choosing a holiday destination is not a simpledecision because selection is derived from a setof independent choices. It is a complex, multi-faceted process in which choices are interrelatedand evolve (Dellaert, et al., 1998). It is well estab-lished in consumer behaviour literature that thepurchase decision process consists of five stages:problem recognition, information search, productoptions evaluation, purchase decision and post-purchase support (Kotler, 2000). The purchaseprocess for a traveller begins when a consumerrecognizes a need or unsolved problem, havingreceived external stimulus usually from touristadvertisements that trigger thoughts on makinga purchase.In the context of tourism services, the purchase

process is much more complex because travelincludes various services (e.g. destination, trans-portation and hotel). Nicolau and Mas (2006)suggested that the process of choosing a touristdestination involves two phases of sequentialdecisions. The first phase is formed by twochronological decisions, the decision of whetherto go on holiday and length of stay. The second


phase involves selection of destination, a condi-tional decision because choice depends on lengthof holiday. Various socio-demographic factorsincluding income, number of family members,age, environment and attitudes toward holidaysinfluence both decision phases.Havingmade a se-lection, the decision process continues with aseries of derivative decisions generallymade priorto a trip: (i) accommodations, (ii) travel company,(iii) transportation and (iv) changes to adjustlength of trip. The choice of accommodation isoften made at a later stage in the decision-makingprocess for a tourist trip.

The next stage involves searching for informa-tion about characteristics offered by servicecompanies and their prices. Consumers use bothinternal and external information sources whenmaking purchase decisions. Consumers typic-ally use internal information from memory asthe first source and then move to externalsearches by using a variety of sources, a processrequiring considerable effort. Travellers oftenvisit several travel agencies or search for infor-mation on the Internet. The two primarycategories of destination factors are related to(i) activities tourists can experience at a destin-ation such as sports, dining, historic sites, night-life, shopping and outdoor recreation, and(ii) other destination characteristics such asprice, accessibility, climate, health and safety,residential attitudes toward tourists, languagebarriers, availability and quality of accommoda-tion, and air quality (Dellaert et al., 1998).

All sources of travel information includingbrochures help travellers make informed deci-sions. Planning and searching for informationoffer tourists the opportunity to reduce uncer-tainty risk and disappointing experiences (Roehland Fesenmaier, 1992). Fodness and Murray(1999) suggested that for companies operating ina competitivemarket, it is essential to understandthe search information process tourists follow.The more unfamiliar a destination is, the moretime and effort must be spent on pre-purchaseinformation searching (Andereck, 2005). Wicksand Schuett (1991) found that travellers who usebrochure information as an aid when planningtravel continue to use brochures as a guide orreference throughout a trip.

Once the tourist gathers enough information,the next step in the purchase decision process isevaluation. This is the third stage, involving

Int. J. Tourism Res. (2012)DOI: 10.1002/jtr

Understanding Tourist Decisions

evaluation of product or service options offeredby various companies and examination ofproduct attributes such as price and brand.(Gupta et al., 2004). Zhou (1997) evaluatedthe effectiveness of destination brochuresrequested by travellers, suggesting that ofthose who read the brochure, 50% visited thedestination and most consulted the brochureduring the visit. Andereck (2005) found thattourists who have no knowledge about a tour-ist destination demonstrate greater interest ininformation contained in brochures than thosewho visited the destination previously.This study focuses on the evaluation of alter-

natives, the third stage of the buying process. Itis important to recognize that there exist vari-ous market segments with differing tastes andpreferences; experiments require working withgroups that are as homogeneous as possible sodifferences in responses demonstrate combina-tions of factors rather than socio-demographicrespondent characteristics (Haaijer and Wedel,2007). In this case, the target market chosenwas students in their final year of universitywho wished to take a graduation trip for 4 daysduring the Easter holiday. The decision aboutdestination and duration, 4 days prior to Easterday, was static, and participants searched foraccommodations in travel agency brochuresand on the Internet.

PERCEIVED VALUE AND CONSUMERDECISIONS

Choosing a hotel room or a trip from a brochuredepends on the perceived utility people expectfrom the room or trip and the price. The market-ing literature defines the relationship betweenperceived utility and price as perceived value.Zeithaml (1988, p.14) proposed a definition: ‘per-ceived value is the consumer’s overall assessmentof the utility of a product based on perceptions ofwhat is received and what is given’. What isreceived and what is given represent trade-offsbetween give and get components. Benefitcomponents include intrinsic and extrinsic attri-butes, perceived quality and other relevanthigh-level abstractions, and sacrifice componentsinclude price and other non-monetary costs.Managers and academicians recognize a stronginfluence of perceived value on consumer behav-iour, making it a priority for researchers (Hauser


and Urban 1986, Zeithaml 1988, Dodds et al.,1991, Ostrom and Iacobucci 1995, Teas andAgarwal 2000, Gallarza and Gil, 2006).Price is often a conflictive element because it

plays a double role as an extrinsic productquality cue and a product choice monetaryconstraint (Erickson and Johansson, 1985). Theseprice roles can be labelled as informational andallocation roles (Rao and Sattler, 2007). Althoughthe two roles of price are clear and derive fromdisparate theoretical conceptions, measure-ments are difficult and tend to confuse thetwo when measured directly. In tourismliterature, Naylor and Frank (2001) suggestedthat price bundling increases perceptions ofvalue for first-time holidaymakers. A study byRewtrakunphaiboon andOppewal (2008) exam-ines whether package information influencesintentions to visit and choice of beach holidaydestinations. On the basis of extant research,the authors assumed that the importance of anattribute increases when products displayaccording to the attribute (Rewtrakunphaiboonand Oppewal, 2008, p.129). Accordingly, find-ings suggest that presenting price as a packageheading (i.e. presenting packages with onlyprice) increases intentions to visit and choice ofbeach holiday destinations among students.Services such as travel products are perceived

as riskier purchases than goods (Zeithaml, 1988).Because services are more difficult to evaluateand riskier to purchase, consumers use diverseprocesses and cues for evaluation. Searchingfor information is one heuristic used to reducerisks and help travellers make decisions. Travel-lers who have little information about a destin-ation from internal sources (i.e. memory,friends and relatives) use external sources suchas brochures and the Internet (Andereck, 2005).For purchasing services, extrinsic attributesbecome cues when information on intrinsic attri-butes is unavailable (Zeithaml 1988). Severalresearchers develop and test models of percep-tions of value with particular emphasis on abuyer’s extrinsic cues such as price and brandname as indicators of quality and value. Doddset al. (1991) conducted a study to examine directand indirect relationships between price, brandname and store name, and perceived quality,product value and willingness to buy. Theyargue that price alone is the most importantcue of perceived quality; adding store and brand



name information hasmoderate effects. If a pricechanges from low to high, perceived valueincreases and at a point decreases significantly,impacting willingness to buy negatively. Thenegative effect of price on value perceptionsreduces if brand or store name is provided. Pricehas a positive influence on perceived quality butnegative effects on perceived value and willing-ness to buy.Perceived value also depends on the frame-

work under which customers conduct evalua-tions (Hempel and Daniel, 1993), defined as adynamic variable. Perceptions of value changeif an assessment is made before or after purchase(Woodruff 1997) and whether it is assessedbefore sale, at themoment of sale, at themomentof use or after use (Moliner et al., 2005). Someauthors consider other factors that alterperceived value. Lee et al. (2007) reported thatvalue depends on a consumer’s characteristicsand the type of product under consideration.Value is a latent construct not observed directly(Teichert and Shehu, 2007); most studies in thisarea use scales to estimate the latent component.We use stated preferences to estimate the valueor utility generated by varying profiles in thepre-purchase stage.

METHODOLOGY

Because perceived value and utility are unob-servable, they require scales or proxy variablesto estimate them (Teichert and Shehu, 2007).According to Klein (1990), there are two meth-ods to identify and measure consumer needsand desires: asking directly or deducing themotives from other kinds of data. The advan-tages of indirect methods include establishingrobust causal relationships between attributesand customer evaluations. A popular device,choice-based conjoint analysis, is an indirectmethod used to obtain measurement of prefer-ences (Haaijer andWedel, 2007). During conjointanalysis’ evolution over the last 40 years, therewas a shift in the types of responses used foranalysis from ranking, to ratings, and finally tochoosing the best profiles (Elrod and Chrzan,2007). The most frequent criterion for assessinga choice setwas ratings of full profiles, evaluatedindividually (Wittink and Cattin, 1989). Sincethe late 1980s, the choice prevailed as the


dominant criterion for evaluating profiles, esti-mated by aggregate multi-nomial logit models(Louviere et al., 2000; Street and Burgess, 2007).Implicit models are compensatory and decom-positional, enabling analyses of the importanceof each product’s characteristics, starting with acustomer’s stated preferences. In this study, wedemonstrate the process of designing a choice-based conjoint analysis and how it determinesthe value of extrinsic cues that, for a market seg-ment, are used to choose a hotel room from atourist brochure. We use one of the least com-mon models, ranking, based on a one-sixth frac-tion from a 28 full factorial design in eight two-size blocks in which the second profile of eachblock is a mirror image of the first.

Choice-based conjoint analysis

The process of evaluating the choice of accom-modation from a brochure operates withindiscrete choice models (CMs). CM is a familyof survey-based methodologies for modellingpreferences for goods in which goods aredescribed by their attribute levels. Respondentsare presented with alternative descriptions ofproducts differentiated by attributes and levelsand rank the various alternatives, rating themor choosing those they prefer (Hanley et al.,2001). The form of evaluation (e.g. lexicographic,elimination by aspect, economic and attributescreening) that predominates among respon-dents in driving such selections remains elusive(Scarpa and Rose, 2008).

Choice model based on choosing the bestoption has several advantages over traditionalratings and rankings. Choice is often the behav-iour of ultimate interest in a decision processbecause it estimates behaviours accurately.Models estimated from choice allow direct pre-diction of choices, avoiding the need for conjointsimulators (Elrod et al., 1992). Its primary disad-vantage is that it works poorly with smallsamples because the process of choice containsminimal information concerning consumerpreferences. A chosen option indicates whichalternative is the most preferred, but it does notprovide information about other alternativesnot selected. A choice process implicitly entailsconsideration of multiple alternatives, but onlyone option is chosen; in a ranking process, allalternatives are considered (Elrod and Chrzan,



2007). We propose a ranking model adjustedby a proportional odds ordinal logistic model(Train, 2009). The decision to use this modelwas based on two objectives. Estimations madeby using all ranking information are more effi-cient, and efficiency is greatest when a full pro-file design is used. This improvement inefficiency allows more acceptable results withreduced samples. Ranking is consistent withconsumer behaviour in both economy theoryand with respect to the nature of preferencesrepresenting an ordered relationship (Frank,2009). A typical choice-based conjoint analysisis characterized by several stages:

• Identification of relevant attributes of goodsto be valued. The use of literature reviews,consultation with experts and qualitativetechniques such as focus groups establishesthese attributes.

• Assignment of a feasible, realistic range ofplausible levels for attributes. Qualitativeresearch or literature searches establish ap-propriate levels.

• Choose an experimental design for gatheringdata. The experimental design is based onstatistical theory to combine levels of the attri-butes into alternative profiles presented torespondents. Whereas full factorial designsoften generate an impractically large numberof combinations for evaluation, fractionalfactorial designs and block designs reducethe number of combinations but with a con-comitant loss of power.

• Profiles identified by the experimental designare grouped into choice sets and presented torespondents. Profiles can be presented torespondents individually, in respondent pairsor in groups.

• Measurement of preferences. Individual pre-ferences are discovered in choice-modellingsurveys by asking respondents to rankoptions, score them or choose those theyprefer most. These methods of measuringpreferences correspond with variations ofmodelling approaches.

• Estimation procedure. The model is adjustedusing Ordinary Least Square (OLS) regressionor maximum likelihood estimation.

In this study, preference differences betweenhotel listings indicated differences in value


and tendency to rent a room. The theory ofreasoned action posits behaviour intentionsinfluences behaviours (Ajzen and Madden,1986). Social psychology research suggestsintentions are the best predictor of behavioursbecause they allow individuals to incorporateall relevant factors independently that influ-ence behaviours (Fishbein and Ajzen, 1975).Several studies examine the relationshipbetween purchase intentions and purchasebehaviours for non-durable goods (Dowlingand Staelin, 1994). The degree to which peopleexpress preferences is a reasonable predictor ofpurchase behaviours.

Qualitative research. Investigating the first twostages of the experimental process of conjointselection, we began with qualitative researchto identify the attributes considered relevantin the perceived value of choosing a hoteldisplayed in a travel agent’s brochure.We evalu-ated aggregate utility generated by the sample ofstudents from varying combinations of attri-butes offered by advertisements in a brochure.The fieldwork for the qualitative research wascarried out in Segovia, Spain from November2006 to February 2007. The use of university stu-dent samples for experimental research attractedthe attention of many researchers recently(Gallarza and Gil, 2006; Rewtrakunphaiboonand Oppewal, 2008). We used a focus groupwith three smaller groups of eight students eachbased on a chosen destination of Punta Cana,Santo Domingo. The sessions were conductedwith two travel agency brochures, asking open-ended questions and recording answers. Afterthe sessions, we coded and classified answersto reveal relevant attributes. The data werecoded using inductive category coding, consist-ing of labelling factors repeatedly found in text(Spiggle 1994). Similar processes are used incontent analysis in the service literature (Taxet al., 1998), and the method is used extensivelyin studies of consumer behaviour to identifyrelationships within text (Spiggle, 1994).The attributes gathered in the focus groups

were as follows: the image and aesthetics ofthe illustrations in the advertisement were animportant factor; aerial or panoramic imageswere preferred, although well-lit, nocturnalphotos were also viewed favourably; the sizeof the illustrations and the hotel’s starred rating



were determining factors; five-star establish-ments displayed in large photographs werepreferred; the structure of the buildings wasvery important; low, open buildings werepreferred to high, square buildings, althoughthere was no consensus concerning the size ofthe hotel; accessibility and proximity to thebeach and restaurants were important; thestarred rating was indicative of quality, but noimportance was given to the ICTE’s Q (Spanishtourist quality symbol); subjects preferred theprice to be on the same page as the advertise-ment that described the hotel, although pricerange was not one of the most importantattributes. A wide range of hotel services waspreferred but not essential. No remarks weremade or attention paid to room amenities (e.g.air conditioning) or a hotel’s surroundings (e.g.topography).Not all attributes must or can be considered;

they must be chosen realistically and in a waythat is appropriate to the situation. Gustafssonet al. (1999) suggested a list of rules for choosingattributes when evaluating service quality:choose attributes that are important when theinterviewees are buying, ones that can be modi-fied and ones used to comparewith competitors.Because there is no consensus in the literatureabout the elements encompassing each of thepositive and negative dimensions of value(Woodruff, 1997), eight relevant attributes werechosen for evaluation:

• Name (one short name ‘Majestic Punta Cana’and one long name ‘Majestic Colonial PuntaCana Beach Resort Golf Casino & Spa’);

• Hotel’s starred rating (four or five stars);• Phrase used to describe the hotel size (24

buildings with three floors, suites and juniorsuites; 659 suites and junior suites);

• Picture in the brochure (a general view of thetourist building or a view of the swimmingpool);

• Size of the hotel listing (quarter or half page);• Position in the brochure (right or left page);• Position on the page (top or bottom);• Prices (a low price of €277 or a high price

of €499)

The first five variables were gathered fromthe qualitative research, and the last three wereadded by the researchers. Although subjects


reported that the hotel listing’s position wasirrelevant, we incorporated it anyway giventhe influence product placement has on prefer-ences (Cox, 1970; Chevalier, 1975). To deter-mine high and low prices, we calculated theaverage ranks of high and low prices from asample of Santo Domingo hotels published inthree brochures.

Quantitative research. For quantitative research,we propose a choice-based conjoint experi-ment of a sample of undergraduate studentswho travel in groups. The underlying choicesurveys were the statistical design of theexperiment, used to allocate all combinations offactors and their levels to form a set of alterna-tives to be used on the survey. An experimentaldesign is the systematic arrangement of profilesin matrices of coded values researchers use todescribe attribute levels representing hypothet-ical alternatives of marketing options in thechoice set. It is not an easy task; Kuhfeld et al.(1994) pointed out that the best designs arediscovered when the researcher uses bothhuman design skills and computerized searches.

To design the experiment,we considered eightfactors: three quantitative variables (starredratings, advertisement size and prices) and fivequalitative variables (names, descriptions ofhotel size, brochure illustrations, positions inthe pamphlet and positions on the page). Allvariables were coded as vectors (�1, 1); forfactor 1, name, a positive sign indicated the longname ‘Majestic Colonial Punta Cana BeachResort Golf Casino & Spa’, and a negative signindicated the short name ‘Majestic Punta Cana’and so on. Table 1 shows the factors and theircodes. A full factorial design would haverequired a 28 factorial experiment (i.e. 256profiles in the choice set). To avoid saturationof information to respondents, we used a IVresolution design, 28�4

IV , a fraction one-sixth ofthe 28 full design. Resolution levels determinethe degree of confusion generated by the experi-ment among factors and interactions betweenthem. A IV resolution design avoids confusionbetween main effects and second-order interac-tions, but two-factor interactions are mutuallyconfused. A complete overview of the designof the experiment can be found in Box et al.(2005) and Myers and Montgomery (2002).


Table 1. Proposed variables and their codification

Variable �1 +1

A Name Majestic Punta Cana Majestic Colonial Punta CanaBeach Resort Golf Casino & Spa

B Hotel rating(number of starts)

4 5

C Hotel size 24 buildings of three floorswith suites and junior suites

659 suites and junior suites

D Picture of advertisement View of building View of swimming poolE Hotel listing size 1/4 of page 1/2 of pageF Right or left location Left RightG Position on page On the bottom On the topH Price 3 days, full-board, service

and airport; transfers included€277 €499


The experimental design required a designgenerator. A 2k�p fractional factorial designnecessitates the selection of p independentdesign generators. A reasonable criterion toselect generators is the one with the highest pos-sible resolution. Myers and Montgomery (2002)presented a selection of tables of 2k�p fractionalfactorial designs with the highest resolution forfactors below 11 (k≤ 11). For an experiment of28�4IV , the generators are E=�BCD, F=�ACD,G=�ABC and H=�ABD. The complete designis shown in Table 2.

Table 2. Fractional factorial design

Variables NameNumberof stars

Hotelsize

Pictureadvertisement

Adv

Experiment A B C D

1 �1 �1 �1 �12 1 �1 �1 �13 �1 1 �1 �14 1 1 �1 �15 �1 �1 1 �16 1 �1 1 �17 �1 1 1 �18 1 1 1 �19 �1 �1 �1 110 1 �1 �1 111 �1 1 �1 112 1 1 �1 113 �1 �1 1 114 1 �1 1 115 �1 1 1 116 1 1 1 1


In many cases, it is inappropriate to proposea choice set of 16 profiles either because of theload of work involved in assessing the rank of16 profiles or because of the difficulties withpractical application. It is possible to dividethe profile set proposed by the researcher intosmaller blocks. Confusion is a design techniquefor arranging a full or fractional factorial intoblocks; the researcher multiplies two or morefactors (A�B=AB) as a criterion to separatethe design into smaller blocks. When the modelis estimated, the researcher does not know

ertisementsize

Right or leftlocation

Position onpage

Price3 days

E=BCD F=ACD G=ABC H=ABD

�1 �1 �1 �1�1 1 1 11 �1 1 11 1 �1 �11 1 1 �11 �1 �1 1

�1 1 �1 1�1 �1 1 �11 1 �1 11 �1 1 �1

�1 1 1 �1�1 �1 �1 1�1 �1 1 1�1 1 �1 �11 �1 �1 �11 1 1 1



whether values correspond to the combinedeffects of two factors or to individual compo-nents. Another design that produces econom-ical and more precise designs for main factorsis the two-size block arrangement. We used adesign called mirroring where the secondprofile is arranged with all factors with the signchanged. This design was chosen because itfacilitates practical application of the experi-ment. For building this arrangement, Box et al.(2005) proposed a three-block generator basedon two-factor interactions. For example,B1=AB, B2=AC and B3=AD. Table 3 showsthe reassignment of the 16 profiles into blocksof two.

Field work

Once the eight-block experiment was config-ured, it was shown to a number of potentialclients to evaluate each experiment by using aranking scale. The profiles were constructedfollowing the design of the hotel listing shownin the tourist brochures, combining images andtexts (Vriens et al., 1998). The scenario thatframes the experiment was the choice of a hotelin Santo Domingo for a graduation trip. Anexample scenario with seven profiles is shownin Appendix A.Some authors such as Hempel and Daniel

(1993) criticize laboratory experiments employedfor drawing inferences from experimental situa-tions inwhich individual choices aremanipulatedby frames imposed by a researcher. Subjects areoften charged with solving problems in a labora-tory in which respondents intensify sensitivity toa researcher’s instructions, making biassed deci-sions in comparison with the decisions subjectsmake on their own in a real environment. Toreduce the risks posed by a laboratory, we usedfamiliarity formats similar to a travel agent’sbrochure. Interviews were conducted during thepre-purchase process when students were organ-izing trips. We used a mix of profiles, someaffected and others placebo. To administer thequestionnaire within a frame of reference thatfacilitates interviewee decisions, each question-naire contained seven hotel listings: two experi-mental and five non-exposed advertisements.The dimensions of the frame of reference wereseven prices with a range of 285 Euros-235 Euroswas the minimum price and 520 Euros the


maximum-a standard deviation of €112.6, anaverage trim of €349.6 and a median of €304.

Regarding sample size, we used the resourceequation proposed by Mead (1988). In thisequation, we required at least n runs in blocksof size nb:

n ¼ nnb

þ q qþ 3ð Þ2

þ nlof þ npe (1)

where q is the number of variables, nlof is asmall number of degrees of freedom (typically5 to 10) for estimating higher-order terms andto check for lack of fit and npe is a small numberof degrees of freedom (typically 5 to 15) forestimating pure error. In our case, nb = 2, q= 8,taken as nlof + npe = 15, the same value used byGilmour and Trinca (2006). The resource equa-tion suggested a minimum sample of 118profiles. However, instead of the minimum,we used eight blocks in which two profileswere replicated eight times (i.e. 128 profiles).Each block of the two profiles was shown toeight students. A sample of 64 students fromValladolid University (Spain) who were choos-ing a destination for their end-of-studies tripparticipated in the experiment by answeringquestionnaires during class time. Fieldworkwas conducted between February and March2007. Descriptive statistics show that 58.5%were female and 41.5% were male. More than38.4% had previous international travel experi-ence. Respondents reported an average age of22.3 years, and the destinations preferred werethe Caribbean (76.2%), the Canary Islands(13.3%) and European cultural cities (9.5%).

The random utility model provides the eco-nomic theory framework for analysing the datafrom a ranking exercise. This model is derivedunder the assumption of utility-maximizingbehaviour by the decision maker. Under theassumption of independent and identicallydistributed random error with an extreme valuedistribution, Beggs et al. (1981) developed arank-order logit model capable of using allinformation contained in a survey where alter-natives are fully ranked by respondents. Speci-fication is based on repeated application of theprobability expression given in Equation (2)until a full ranking of all the alternatives isobtained. This model involves consideringeach of the choices made by the respondents


Table3.

Blocked

design

Variables

Nam

eNum

berof

stars

Hotel

size

Picture

adve

rtisem

entAdve

rtisem

ent

size

Right

orleftlocation

Position

onpa

gePrice3day

sBlock

variab

les

Block

Exp

erim

ent

AB

CD

E=BCD

F=ACD

G=ABC

H=ABD

B1=AB

B2=AC

B3=AD

12

1�1

�1�1

�11

11

�1�1

�115

�11

11

1�1

�1�1

�1�1

�12

13�1

�11

1�1

�11

11

�1�1

41

1�1

�11

1�1

�11

�1�1

36

1�1

1�1

1�1

�11

�11

�111

�11

�11

�11

1�1

�11

�14

9�1

�1�1

11

1�1

11

1�1

81

11

�1�1

�11

�11

1�1

57

�11

1�1

�11

�11

�1�1

110

1�1

�11

1�1

1�1

�1�1

16

121

1�1

1�1

�1�1

11

�11

5�1

�11

�11

11

�11

�11

73

�11

�1�1

1�1

11

�11

114

1�1

11

�11

�1� 1

�11

18

161

11

11

11

11

11

1�1

�1�1

�1�1

�1�1

�11

11


Copyright © 2012 John Wiley & Sons, Ltd. Int. J. Tourism Res. (2012)DOI: 10.1002/jtr


as independent. Each of the choices is called apseudo-observation because each completeranking is only one observation formed bymultiple pseudo-observations depending onthe number of choice sets (Train, 2009). Themodel considers that the first option is chosenamong seven profiles; afterwards, the decisionmaker chooses the second option among theremains six profiles and so on. The probabilityof any ranking of an alternative made by indi-vidual i can be expressed as a sequence ofprobability multiplications:

Pi Ui1 > Ui2 > ⋯ > Uijð Þ ¼YJ

j¼1

exp Vijð ÞPJk¼j

exp Vikð Þ

266664

377775

(2)

where Vij ¼ b0 þPq

i¼1bixi þ e , Pi (Uij)measures the probability of this chosen order,bi are the values of the slope of the vector foreach main factor and e is an error term. Anordinal logistic regression model was used toadjust the full profile data in the proportional

Table 4. Estimated effects from the experimental design

Effect Estimation

Order = 1 �1.004Order = 2 0.106Order = 3 1.140Order = 4 2.211A 0.175B 0.917C �0.456D 0.038E=BCD 0.924F=ACD 0.363G=ABC �0.324H=ABD �0.661B1 =AB=CG=DH=EF �0.476B2 =AC=BG=DF=EH 0.803B3 =AD=BH=CF=EG 0.174

Pseudo R-squareCox and Snell 0.220Nagelkerke 0.225McFadden 0.066Diagnostic �2 log-likelihood

160.768


odds model (or cumulative logit model) byusing the PLUM process from SPSS.

RESULTS AND DISCUSSION

The results of estimation are shown in Table 4.The ordinal logistic regression model has a de-gree of adjustment usual in this type of study(Cox and Snell’s pseudoR2 = 0.220;�2 log-likeli-hood=160.768, p< 0.001) (Louviere et al., 2000).To determinewhether themodel offers adequatepredictions, we used two log-likelihood values,comparing two log-likelihood values for theintercept-only model and the final model withthe predictors. The chi-square statistic indicateswhether the model offers significant improve-ment over the baseline intercept-only model.The dependent variable had only five of theseven possible values, meaning the two exposedhotel listings were chosen between the first andfifth orders. Hence, the ordinal model estimatesonly four cut-off values that separate the fiveresponse categories. Parameter estimates areshown in Table 4, where four of the 11 effectsare significant. These are B (starred ratings),H (size of hotel listing), E (prices) and B2 (theblocking variable). Whereas a variable in

28�4IV

Wald Significance

3.562 0.0590.041 0.8394.632 0.03115.842 0.0000.333 0.5648.842 0.0032.254 0.1330.015 0.9018.950 0.0031.431 0.2321.143 0.2854.682 0.0302.453 0.1176.826 0.0090.332 0.565

0.001

Int. J. Tourism Res. (2012DOI: 10.1002/jtr

)


statistical design experiment can be inter-preted singly if there is no evidence that it isconfused with another, B2 (the blocking vari-able) is confused with several second-orderinteractions, B2 =AC=VG=DF=HE; inter-pretation is, therefore, imprecise. We suspectthat the value of B2 responds significantly tothe interaction between H (size of hotel listing)and E (prices).To verify these results, we considered the

analysis of only three factors, those whichdemonstrated significance after the first experi-ment of the three main effects (B, H and E).Because only three factors were significant, afull factorial of 23 required eight profiles drawnfrom the previous design using 16 profiles. Thisis possible because any fractional factorialdesign of resolution R may contain completefactorials in any sub-group of R� 1 variables.Because the design is resolution R IV, it waspossible to configure a complete factorialdesign with these three variables. This designallows estimation of main effects and interac-tionswithout any confusion patterns and, conse-quently, allowed direct interpretation. Ordinallogistic regression was used to evaluatesignificance of the effects. Table 5 shows themodel-fitting information and the estimated

Table 5. Estimated effects from the experimentaldesign 23

Effect Estimation Wald Significance

Order = 1 �0.850 3.771 0.052Order = 2 0.247 0.331 0.565Order = 3 1.249 7.962 0.005Order = 4 2.281 23.423 0.000B 0.907 8.756 0.003E 0.859 7.894 0.005H �0.663 4.760 0.029BH 0.149 0.246 0.620BE �0.110 0.133 0.716HE 0.827 7.314 0.007BHE �0.298 0.979 0.322

Pseudo R-squareCox and Snell 0.170Nagelkerke 0.177McFadden 0.058Diagnostic �2log-likelihood

100.845 0.000


effects from the derivative model. Values aresimilar to the previous analysis.H (prices) had a negative impact on evalu-

ation by the sample. This result is in accordwith the interpretation of price as sacrifice; thelower the price, the greater the customer’sevaluation. Other more highly valued effectswere as follows: B (starred ratings), E (hotellisting size) and HE (interaction between priceand size of hotel listing). B (starred ratings)and E (advertisement size) had a positive influ-ence on the subjects’ evaluation. Respondentsused both factors as external value cues; thehigher the starred ratings and the bigger the sizeof the advertisement, the greater its evaluation.Moreover, the joint presence of HE (interactionbetween price and size of hotel listing) increasedpositive valuation. The dual role of price assacrifice and extrinsic cue of service quality isappreciated when price interacts with a largeadvertisement. Thus, a negative evaluation repre-sented by high price accommodation is offset bya large advertisement. Likewise, the positiveinfluence generated by a large advertisementincreases with price.

CONCLUSIONS

Customer evaluation and decision making areprimary targets in all areas of business, particu-larly in tourism management. To understandcustomer needs and determine the primaryfactors of perceived value, a researcher can asksubjects directly or deduce desires throughindirect methods such as experimentation. Inthis study, we demonstrate the process of imple-menting a choice-based conjoint analysis andhow this analysis can be used as a powerful toolfor evaluating preferences. These designs areparticularly suitable for subjects related todemand for introduction of new products, lineextensions (Louviere et al., 2000) or design of anew hotel listings in a tourist brochure. Inaddition to providing an evaluation of theweight of main attributes, it also shows theweight of interactions of several attributes.Value perceptions depend on a frame of

reference in which consumers make evalua-tions. Employing an experimental design in acontext of pre-purchase with a static frame ofreference, we examined the main factors of



value perception in the process of choosing ahotel from a travel agent brochure. Subjectsused a reduced number of attributes to inferperceived quality and the sacrifice necessaryto acquire it. In this case, only three of the eightattributes were significant: advertisement size(quarter or half page) and the hotel’s starredrating (4 or 5) as quality attributes, and priceindicating sacrifice. For this market segmentof students in their final year of university, bro-chure illustration was not significant in spite ofthe results from qualitative research, neitherwere position on page, nor name ofestablishment, nor left versus right location,although results are in line with the qualitativeresearch. Hotel size occupies an intermediateposition of importance; an increase in samplesize is necessary to verify the variable’s influ-ence on selection.Attributable to the second analysis using a

full factorial design, it was possible to detectthe second inductive factor of perceived qualityconsisting of the joint presence of two vari-ables, advertisement size (quarter or half page)and price. The two joint variables demonstratethe dual role of price (Dodds et al., 1991), as asacrifice people must make to buy goods or asan extrinsic cue of service quality, when it inter-acts with a large advertisement. Interactionbetween price and hotel listing size showed apositive result when negative was expected,suggesting some kind of compensatory effect.These findings have important implications formanagers, specifically that a higher priceexpected to generate negative utility is offset bya larger advertisement to produce a positiveeffect on a tourist’s final decision. The increasedcost of a larger advertisement could be recoupedby charging a higher price for the trip andhotel room.This study considers perceived value attributes

in a pre-purchase stage, a stage traditionallyunderstudied. Knowing which aspects or attri-butes determine perceived value of the product


before purchasing is extremely useful for busi-nesses and managers. A travel agent brochuremay be giving more prominence to elements thatdo not determine perceived value of a tourist,making communications ineffective and notcontributing to a consumer’s decision. However,the validity of the attributes of perceived valuedisappears with purchase. Fisher et al. (1994)argued that when consumers evaluate perceivedservice quality after purchase, they rarelymention the criteria used for evaluation beforepurchase;when they do, such criteria are relegatedto a lower hierarchic level to the one occupiedbefore purchase.

This study has some limitations. Participantswere from one university, disallowing statis-tical inferences from the sample to a popula-tion. Some variables such as room quality andhotel location were ignored as potentially rele-vant attributes. The laboratory environmentthat afforded internal validity because of envir-onment control compromised external validitybecause of an artificial venue. The sample sizewas also a limiting factor because it did notallow us to determine the significance of someprimary variables (e.g. hotel size), although themodel using ordinal logistic regression wasproposed because of its suitability to fit smallsamples. Designing this experiment, wetransformed a full fractional model into afractional one; the resulting design was orga-nized into blocks of two. This design onlyallows the estimation of main factors withoutinteractions. The design could be improvedby taking advantage of multiple runs andusing a fractional factorial with higher reso-lution such as R V, although there is thedisadvantage of the cost of designing 32 bro-chures instead of eight. The model helped usget closer to customers, identifying opinionsand preferences; it was a simple procedure thatfacilitated sequential research that enabled usto continuously improve knowledge of thesubject.



REFERENCES

Ajzen I, Madden TJ. 1986. Prediction of goal-directedbehavior: attitudes, intentions, and perceivedbehavioral control. Journal of Experimental SocialPsychology 22(5): 453–474.

Andereck KL. 2005. Evaluation of a tourist brochure.Journal of Travel & Tourism Marketing 18(2): 1–13.

Beggs S, Cardell S, Hausman J. 1981. Assessing thepotential demand for electric cars. Journal ofEconometrics 17(1): 1–19.

Bieger T, Laesser C, Gallen S. 2000. Segmentingtravel situations on the basis of motivation andinformation-collection by the traveller. Revue deTourisme 2: 54–64.

Box EP, Hunter JS, Hunter WG. 2005. Statistics forexperimenters: design, innovation, and discovery.2ndedn . Wiley: Hoboken.

Chevalier M. 1975. Increase in sales due to in-storedisplay. Journal of Marketing Research 12(4): 426–431.

Chiou WB, Wan CS, Lee HY. 2008. Virtual experi-ence vs brochures in the advertisement of scenicspots: How cognitive preferences and ordereffects influence advertising effects on consumers.Tourism Management 29(1): 146–150.

Cox KK. 1970. The effect of shelf space upon salesof branded products. Journal of Marketing Research7(1): 55–58.

Dellaert B, Ettema D, Lindh C. 1998. Multi-facetedtourist travel decisions. a constraint-based concep-tual framework to describe tourist’s sequentialchoices of travel components. Tourism Management19(4): 313–320.

Dodds WB, Monroe, KB, Grewal D. 1991. Effects ofprice, brand and store information on buyers’product evaluations. Journal of Marketing Research28(4): 307–319.

Dowling G, Staelin R. 1994. A model of perceivedrisk and intended risk-handling activity. Journalof Consumer Research 21(1): 119–134.

Elrod T, Chrzan K. 2007: The value of extent-of-preference information in choice-based conjointanalysis. In Conjoint Measurement. Methods andApplications, 4thedn , Gustafsson A, HerrmannA, Huber F (eds). Springer: Berlin; 133–144.

Elrod T, Louviere JJ, Davey KS. 1992. An empiricalcomparison of ratins-based and choice-basedconjoint models. Journal of Marketing Research 29(3):368–377

Erickson, GM, Johansson, JK. 1985. The role of pricein multi-attribute product evaluations. Journal ofConsumer Research 17: 195–199.

Fishbein M, Ajzen I. 1975. Belief, attitude, intentionand behavior: an introduction to theory and research.Addison-Wesley: New York.


Fisher S, Clemons D, Scott, R,Woodruff D, SchumannW, Burns MJ.1994. Comparing consumers’ recall ofprepurchase and postpurchase product evaluationexperiences. Journal of Consumer Research 20(2):548–560.

Fodness D, Murray B. 1999. A model of tourist infor-mation search behaviour. Journal of Travel Research37(3): 220–230.

Frank RH. 2009. Microeconomics and behavior. McGraw-Hill: New York

Gallarza MG, Gil I. 2006. Value dimensions,perceived value, satisfaction and loyalty: an inves-tigation of university students’ travel behaviour.Tourism Management 27(3): 437–452.

Gilmour SG, Trinca LA. 2006. Response surfaceexperiments on processeswith high variation. InRe-sponse Surface Methodology and Related Topics, KhuriAI (ed.). World Scientific Publishing Co: New York;19–46.

Gupta A, Bo-Chiuan S, Walter Z. 2004: An empiricalstudy of consumer switching from traditional toelectronic channels: a purchase-decision processperspective. International Journal of ElectronicCommerce 8(3): 131–161.

Gustafsson A, Ekdahl F, Bergman B. 1999. Conjointanalysis: a useful tool in the design process. TotalQuality Management 10(3): 327–343.

Haaijer R, Wedel M. 2007. Conjoint choice experi-ments: general characteristics and alternativemodel specifications. In Conjoint Measurement.Methods and Applications, 4thedn , Gustafsson A,Herrmann A, Huber F (eds). Springer: Berlin;199–229.

Hanley N, Mourato S, Wright RE. 2001. Choicemodelling approaches: a superior alternative forenvironmental valuation?. Journal of EconomicSurveys 15(3): 435–462

Hauser JA, Urban GL. 1986. The value priorityhypotheses for consumer budget plans Journal ofConsumer Research 12(4): 446–462.

Hempel DJ, Daniel HZ. 1993. Framing dynamics:measurement issues and perspectives. Advancesin Consumer Research 20: 273–279.

Hodgson P. 1993. Tour operator brochure designresearch revisited. Journal of Travel Research 32(1):50–52.

Hsieh S, O’Leary L. 1993. Communication channelsto segment pleasure travelers. Journal of Travel &Tourism Marketing 2(2/3): 57–76.

Klein R. 1990. New techniques for listening to thevoice of the customer. Applied Marketing Science,Inc., Second Symposium on Quality Function Deploy-ment, June 18–19, 1990. Cited by Cohen L. 1995.Quality FunctionDeployment: How toMakeQFDWorkfor You. Addison-Wesley Publishing Company,Reading, Massachusetts, USA.



Kotler P. 2000. Marketing Management. The Millen-nium Edition. Prentice Hall Inc: New Jersey.

Kuhfeld WF, Tobias RD, Garratt M. 1994. Efficientexperimental design with marketing applications.Journal of Marketing Research 31(4): 545–557

Lee C-K, Yoon Y-S, Lee S-K. 2007. Investigating therelationships among perceived value, satisfactionand recommendations: the case of Korean DMZ.Tourism Management 28(1): 204–214.

Louviere JJ, Hensher, DA, Swait JD. 2000. Stated ChoiceMethods. Cambridge University Press: Cambridge.

MeadR. 1988. The design of experiments. Cambridge:Cambridge University Press; cited by Gilmour SG,Trinca L A. 2006. Response surface experiments onprocesses with high variation. In Response SurfaceMethodology and Related Topics, Khuri AI (ed.).WorldScientific Publishing Co: New York.

Molina A, Esteban A. 2006. Tourism brochures:usefulness and image. Annals of Tourism Research33(4): 1036–1056.

Moliner MA, Rodríguez RM, Callarisa Ll J, SánchezJ. 2005. Dimensionalidad del valor percibidoglobal de una compra. Revista Española de Investiga-ción de Marketing 9(2): 135–157.

Munger JL, Grewal D. 2001. Effects of alternativeprice promotional methods on consumers’ evalu-ation and purchase intentions. Journal of Product& Bran Management 10(3): 185–197.

Myers RH, Montgomery DC. 2002. Response SurfaceMethodology. John Wiley & Sons, Inc: New Jersey.

Naylor G, Frank, KE. 2001. The effect of price bundlingon consumer perceptions of value. Journal of ServicesMarketing 15(4): 270 – 281.

Nicolau JL, Mas FJ. 2006. Elección de la duraciónvacacional: una aproximación con modelos derecuento. Revista Europea de Dirección y Economíade la Empresa 15(2): 99–116.

Ortega E, Rodriguez B. 2007. Information at tourismdestinations. Importance and cross-cultural differ-ences between international and domestic tourist.Journal of Business Research 60(2): 146–152.

Ostrom A, Iacobucci D. 1995. Consumer trade-offsand the evaluation of services. Journal of Marketing59(1): 17–28.

Rao, VR, Sattler, H. 2007. Measurement of priceeffects with conjoint analysis: separating informa-tional and allocative effects of price. In ConjointMeasurement. Methods and Applications, 4thedn ,Gustafsson A, Herrmann A, Huber F (eds).Springer: Berlin; 31–46.

RewtrakunphaiboonW, Oppewal H. 2008. Effects ofpackage holiday information presentation ondestination choice. Journal of Travel Research 47(2):127–136.


Roehl W, Fesenmaier D. 1992. Risk perceptions andpleasure travel: an exploratory analysis. Journal ofTravel Research 30(4): 17–26.

Scarpa R, Rose JM. 2008. Design efficiency for non-market valuation with choice modeling: how tomeasure it, what to report and why. The AustralianJournal of Agricultural and Resource Economics 52(3):253–282.

Spiggle S. 1994. Analysis and interpretation of qualita-tive data in consumer research. Journal of ConsumerResearch 21(3): 491–503.

Street D, Burgess L. 2007. The Construction of OptimalStated Choice Experiments: Theory and Methods.Wiley: Hoboken, NJ.

Tax S, Brown SW, Chandrashekaran M. 1998.Customer evaluation of service complaint experi-ences: implications for relationship marketing.Journal of Marketing 62(April): 60–76.

Teas RK, Agarwal, S. 2000. The effects of extrinsicproduct cues on consumer’s perceptions ofquality, sacrifice and value. Journal of the Academyof Marketing Science 28(2): 278–290.

Teichert T, Shehu, E. 2007. Evolutionary conjoint. InConjoint Measurement. Methods and Applications,4thedn , Gustafsson A, Herrmann A, Huber F(eds). Springer: Berlin; 295–320.

Train K. 2009. Discrete Choice Methods withSimulation. 2ndedn , Cambridge University Press:Cambridge.

Vriens M, Loosschilder, G, Rosbergen E, Wittink D.1998. Verbal versus realistic pictorial representa-tions in conjoint analysis with design attributes.Journal of Product Innovation Management 15(5):455–467.

Wicks BE, Schuett MA. 1991. Using travel brochuresto target frequent travelers and “big-spenders”.Journal of Travel & Tourism Marketing 2(3): 77–90.

Wittink DR, Cattin P. 1989. Commercial use ofconjoint analysis: an update. Journal of Marketing53(3): 91–96.

Woodruff BR. 1997. Customer value: the next sourcefor competitive advantage. Journal of the Academyof Marketing Science 25(2): 139–153.

Yamamoto D, Gill AM. 1999. Emerging trends inJapanese package tourism. Journal of Travel Research38(2): 134–143.

Zeithaml VA. 1988. Consumer perceptions of price,quality and value: a means-end model and synthe-sis of evidence. Journal of Marketing 52(3): 2–22.

Zhou Z. 1997. Destination marketing: measuring theeffectiveness of brochures. Journal of Travel & TourismMarketing 6(3): 143–158.


DOI: Conjoint Analysis of Tourist Choice of Hotel ...Conjoint Analysis of Tourist Choice of Hotel...

Documents

Transcript of DOI: Conjoint Analysis of Tourist Choice of Hotel ...Conjoint Analysis of Tourist Choice of Hotel...