Tourism Analysis Vol 15 pp 147ndash165 1083-542310 $6000 + 00Printed in the USA All rights reserved DOI 103727108354210X12724863327605Copyright 2010 Cognizant Comm Corp wwwcognizantcommunicationcom

THE SCALE-ADUSTED LATENT CLASS MODEL

APPLICATION TO MUSEUM VISITATION

PAUL F BURKE CHRISTINE BURTON TWAN HUYBERSdagger TOWHIDUL ISLAMDaggerJORDAN J LOUVIERE and CHELSEA WISE

University of Technology Sydney AustraliadaggerUniversity of New South Wales Australia

DaggerUniversity of Guelph Canada

Preferences of tourists and visitors are varied in a number of markets making it difficult formanagers to understand how underlying segments might respond to changes in service offeringsMarket segments differ in preferences for specific features as well as how consistently they maketheir choices In this article we illustrate recent developments in choice modeling that allows forsimultaneously modeling feature preferences and consistency of choice We use the Scale-AdjustedLatent Class Model (SALCM) to better understand choices in the context of a research projectconducted in collaboration with six major Australian museums involving a sample of 3685 mu-seum visitors We identify three preference classes of museum-goers that explain preferences forlevels of 26 museum attributes Life Force (two thirds of visitors) Educated Thinkers and WealthyAt-Homes Our results indicate sensitivity to general entry prices including preference for freeentry or entry ldquoby donationrdquo Tours are preferred if smaller lengthier and conducted by paidmuseum staff Not unexpectedly the findings suggest that museums should cater for children withsome classes responding positively to providing supervised child areas Most visitors prefer muse-ums that are dynamic offer new experiences and regularly update permanent displays Howeverthe three classes identified have different overall experience preferences for example EducatedThinkers see museums as an educational opportunity but Wealthy At-Homes prefer entertainingexperiences Incentives for return visits and cross-museum promotional offers are valued by theLife Force class but have little effect on Educated Thinkers The SALCM approach may be attrac-tive to other areas of tourism analysis especially where offerings contain many attributes andpotential market segments are difficult to define and understand

Key words Museum visitation Discrete choice experiments Latent class models HeterogeneitySegmentation

Introduction tourists often acquire a better appreciation of whatdefines the destination across widespread dimen-sions like transportation industry technology warMuseums are often a ldquomust-seerdquo attraction for

domestic and international tourists visiting a desti- art and literature Among local residents muse-ums can provide places of remembrance celebra-nation region city or town By visiting museums

Address correspondence to Dr Twan Huybers Senior Lecturer School of Business University of New South Wales AustralianDefence Force Academy Canberra ACT 2600 Australia Tel +61 2 6268 8075 Fax +61 2 6268 8450 Email thuybersadfaeduau

147

148 BURKE ET AL

tion or even escape In general motivations to trol of individual museums such as opening hoursand pricing To do so we consider microeconomicvisit museums vary but largely relate to enhanc-

ing knowledge and social interactions that include models of consumption choice conceptualizingthe decision to visit a particular museum as beingentertaining distractions (Burton Louviere amp

Young 2008 Slater 2007) In turn museums preceded by a series of trade-offs among tangibleand intangible characteristics of museumsrsquo offer-compete with other offerings that appeal to similar

motives such as cinemas theaters home enter- ings and the premium consumers are willing topay in terms of prices times and effort to obtaintainment eating out or attending sporting events

(Johnson amp Thomas 1998) Such diverse competi- these We view existing valuation methods asmeasuring value in absolute rather than relativetion makes attracting visitors difficult and muse-

ums have been criticized for their inability to do terms In contrast we propose and apply discretechoice experiments (DCEs) as a way to deriveso (Burton et al 2008) Arguably such difficul-

ties might be expected when one considers muse- new insights by measuring offerings in terms ofthe relative value that visitors place on attributesumsrsquo historical role as largely not-for-profit orga-

nizations that increasingly are expected to compete features that combine to create an overall museumexperienceagainst ldquofor-profitrdquo organizations Strategic recon-

siderations for museums have evolved out of de- To achieve our wider research objectives webuild on recent advances in discrete choice model-creases in government subsidies (Lee 2005) and

private philanthropic sponsors (Burton 2003) ing to account for differences in preference andchoice variability among museum visitors WhileThus museums increasingly understand that at-

tracting visitors requires them to deliver products many advances have been made to understand dif-ferences in how segments value one feature rela-that are valued by individual customers (Holden

2003 Weil 2002) tive to another (ie preference) accounting fordifferences in choice variability has been moreAddressing museum visitation matters from

both macro- and microeconomic viewpoints Mu- elusive Choice variability refers to how inconsis-tent individuals are in their overall choices andseums can be seen as attractions for visitors exter-

nal to a destination providing economic implica- individuals may be more or less variableconsis-tent due to several reasons including uncertaintytions for other spatially proximate organizations

like tourism operators and hotels (Throsby 2001) or confusion Choice variability is related to thestochastic (random) component and as such re-From a cultural viewpoint the presence of muse-

ums in communities gives a sense of pride and flects any nondeterministic behavior from theviewpoint of researchers In other words individu-identity that act as memorials to the past and sig-

nal a sense of confidence in a regionrsquos future At a als typically do not always choose what appearsto be their most preferred alternative and insteadmicrolevel museums can reconfigure their offering

in many ways including changes to tangible as- choose alternatives that seem to be less than opti-mal Our focus is on groups or segments of choos-pects of their products (eg exhibitions food mer-

chandise) and intangible service aspects (eg staff ers and these groups can be viewed as being moreor less consistent in their choices relative to oneknowledge type of tours interactivity) Museum

managers have limited resources to reconfigure of- anotherDifferences in choice variability are rarelyferings so they would benefit from innovative

methods that allow them to understand what visitors taken into account in empirical choice modelingwork but failure to do so can lead to serious biasvalue such that reconfigurations are strategically

informed as accurately as possible in advance and incorrect conclusions (Louviere 2001) Choicevariability is inversely related to the parameters ofThus the purpose of this article is to apply a

Scale-Adjusted Latent Class Model (SALCM) to choice models that one estimates and so it playsa major role in the quality and interpretation of theaddress the question of what is valued by museum

visitors Our research focuses on what constitutes estimates one derives from choice experiments orother types of choice data For example one mayvalue at a microlevel in terms of attributes or fea-

tures of museums that are directly under the con- conclude that a segment holds very different pref-

SCALE-ADUSTED LATENT CLASS MODEL 149

erences to another but the differences may simply attract visitors but if potential visitors place morebe the result of differences in how consistent the value on staff knowledge of exhibitions it may bechoices are that reveals the preferences in each better to allocate resources to staff training thansegment (Louviere 2001) We take both choice extending opening hoursvariability and preference into account by estimat- Museum managers have been keen observersing a relatively new model the Scale-Adjusted La- of their markets over the past decade Many usetent Class Model (SALCM) proposed by Magid- qualitative methods such as focus groups in-depthson and Vermunt (2005) The SALCM identifies interviews and observational techniques Forlatent (unobservable) segments that simultane- example Bitgood (2006) uses a general valueously differ in two ways 1) in preferences for a principle (ie cost vs benefits) to account for thegiven attributefeature and its levels and 2) in the trade-offs that visitors use in moving around anvariability with which overall choices are made exhibition On a more sensory level Joy andPrevious applications of latent class models did Sherry (2003) use observation and in-depth inter-not take choice variability differences into ac- views to determine visitorsrsquo visceral and embodiedcount characterizing the latent segments (classes) imaginative responses to exhibitions Thyne (2000)only in terms of preference differences As far as uses content analysis of interviews and a ladderingwe are aware this is the first application of this technique to determine factors that influence mu-model in tourism research seum visitation but did not report how influential

they were While these and similar studies (egA Choice Model of Museum Visitation Debenedetti 2003 Goulding 2000 Wiggins 2004)

identified many potential factors that might affectTraditionally marketers assert that successfulmuseum experiences they did not isolate or mea-product differentiation can be achieved by offeringsure key attributes that influence museum visita-consumers something relatively important and val-tion decisions In addition in the absence of well-uable that stands out from what others offer (Ca-conceived external validity tests there are validityhill 1996) The value that consumers derive fromissues noted by Biehal and Chakravarti (1989)products (including services and experiences) de-

Museum managers extensively use survey re-pends on the underlying characteristics or attri-search to learn about visitors For example sur-butes (Lancaster 1966) of the services such asveys are used to try to measure the ldquoimportancerdquomuseums Value to museum visitors can be associ-of various factors in visitation decisions Theated with attributes like guided tours exhibitedmethodologies employed include ranking con-artifacts and interactive displays Visitor choicesstant sum and rating scales to compare institu-involve trade-offs whereby they give up one attri-tions on demographics visitation frequency satis-bute or attribute level to obtain another (Louvierefaction and summative evaluation of exhibitions1988) In the case of museums potential visitors(Kirchberg 1996) Similar to the validity issuesmay visit a museum with interactive exhibits evennoted above with qualitative methods ranking andif it has a high entry price Visitors may considerconstant sum scales can be difficult for respon-many attributes in making decisions or they maydents when they have to rank or allocate pointsconsider only a few hence managers often lackacross many factors (Cohen amp Orme 2004) Sin-clear understanding of the relative importance ofgle-item rating scales can be problematic when re-product attributes Museums face constraints duespondents rate many (or even all) factors as impor-to costs policies supply issues and sustainabletant As every feature is treated equally there iseconomic practices and try to maximize the attri-no real disincentive to choose as noted by Carsonbutes of their ldquoofferingsrdquo to provide an ldquoidealrdquoand Groves (2007) Compounding this are issuesmuseum experience (Sheth Newman amp Grossrelated to differences in individualsrsquo use of such1991) Hence museum managers need to under-scales For example one personrsquos three may be thestand the trade-offs that visitors will make for at-same as anotherrsquos four or vice versa And oftributes that can be changed For example a mu-

seum could consider extending opening hours to course it is widely recognized that rating scale re-

150 BURKE ET AL

sponses are influenced by cultural norms (eg Hui than meets the eye Thus managers of museumsand other tourism service suppliers could benefitamp Triandis 1989)

Other quantitative research methods that have from research methods that provide sharper andbetter discrimination between potential attributesbeen used to study cultural heritage include Kellyrsquos

(2004) Repertory Grid Analysis and Contingent Louviere and Islam (2008) have compared fourdifferent ways of measuring importance weightsValuation Kellyrsquos grid has been used to identify

the strength of museum brands (Caldwell amp Cos- two direct (ie rating scales and bestndashworst scal-ing) and two indirect methods (ie discrete choicehall 2003) by proposing that people make choices

based on analogous associations While repertory experiments and implied willingness to pay mea-sures) Direct approaches measure the importancegrid analysis allows individuals to direct the way

in which associations are determined it does not of a set of attributes by asking respondents to statethe degree of importance or weight on some cate-allow one to make predictions about choices An-

ticipating museum visitation choices requires one gory rating or constant sum scales Indirect ap-to take into account interest tastes and motives proaches generally try to elicit importance by ana-not typically observed in repertory grid analysis lyzing an outcome measure like choices They

Contingent valuation (CV) (Mitchell amp Carson found different results from direct and indirect1989) has been used to measure the economic preference elicitation approaches According tovalue of cultural resources (eg Noonan 2003 Louviere and Islam (2008) indirect measures pro-Santagata amp Signorello 2000) CV methods typi- vide richer insights about trade-offs among attri-cally ask respondents about their willingness to butes and provides more meaningful managerialpay for a well-defined and described option such inputs due its natural link with consumer purchas-as a particular museum exhibition although there ing context Jaccard Brinberg and Lee (1986)is nothing inherent in the CV approach that ex- also found lack of convergence among direct andcludes variation in product features Our research indirect measures of importance In absence of un-approach has many features in common with CV derlying theoretical reasons for such differencesapproaches but is more general than typical CV we use indirect measures that have external valid-applications because we design and implement a ity in the literature (see Louviere amp Islam 2008discrete choice experiment (DCE) that allows us for discussion on external validity of indirect mea-to analyze survey responses with discrete choice sures)models that can be used to predict likely visitorchoices Discrete Choice Experiments

The qualitative and quantitative research re- and Tourism Researchviewed above suggests that visitors may use sev-

While it is safe to assume that various attri-eral attributes of museums to value museum at-buteslevels drive museum visitation decisions ittractiveness These include social factors identityis unclear which ones singly or in combinationdevelopment and consumption experiences (Falkmatter most (Kelly 2004) or to whom Discrete2006) and artisticaesthetic emotional and educa-choice experiments (DCEs) and associated dis-tional factors (Boorsma 2006) Differences in thecrete choice models (DCMs) offer a potential wayattributes that visitors evaluate and care about mayto quantify what drives museum visitor choicesbe associated with psychographic factors differ-(Burton 2003 Burton amp Scott 2003) In the re-ences in visitor segments andor other leisure con-search reported below DCMs are used to analyzesumption differences (Clopton Stoddard amp Davethe outcomes of a DCE in which past or potential2006) As noted by Cohen and Neira (2003) allvisitors were asked to make choices in a series oftoo often when consumers are asked to evaluatechoice sets The sets were constructed to systemat-potential attributes of products or services likeically vary a number of attributes of museums thatmuseums they rate all or many attributes ascan be changed The sets were based on priorequally important providing little discrimination

among attributes and less useful strategic insight qualitative research that identified a range of ele-

SCALE-ADUSTED LATENT CLASS MODEL 151

ments that matter when choosing a museum visit in Greece and Italy respectively to understandmuseum demandDCEs are widely used to represent everyday

choices such as products on shelves in supermar-kets transport mode options for commuting and

Accounting for Individual Differencesrecreational offerings like parks or fishing loca-

in Preference and Variabilitytions (eg Louviere Hensher amp Swait 2000)Choice response data from DCE surveys can be One of the key advantages in using DCMs re-analyzed with choice models to uncover system- volves around modeling decisions to visit muse-atic relationships between choices (dependent vari- ums and how these decisions may be affected byable) and changes in productservice offerings or changes to museum features Early DCMs suchldquoattributesrdquo (independent variables) as McFaddenrsquos (1974) conditional logit model

The choice modeling literature is now fairly restrict the preferences of individuals to be homo-mature and comprehensive reviews can be found geneous hence preferences estimated in thesein Train (2003) or Louviere et al (2000) More models are sample averages In turn this raisesrecently it has been recognized that more complex aggregation bias concerns (Hutchinson Kama-models are required that can take differences in kura amp Lynch 2000) for example when all fe-individual choice variability into account (eg males prefer museums with longer opening hoursAdamowicz et al 2008 Islam Louviere amp but all males are indifferent In such a case oneBurke 2007 Louviere amp Swait 2010) Thus we would underestimate the effects of extended open-make use of relatively new developments in finite ing hours on choices of both genders if they aremixture models known as latent class models We equally represented in a sampletake advantage of a new type of latent class model One way to account for such preference differ-known as a ldquoscale-adjusted latent class modelrdquo ences (or heterogeneity) is to include interactionsthat can simultaneously account for differences in between observable covariate measures (eg de-choice variability and preferences (Magidson amp mographics psychographics) and product featuresVermunt 2005) in DCMs (Gupta amp Chintagunta 1994) Auger

Use of DCMs in tourism and leisure research Burke Devinney and Louviere (2003) report sig-has grown in line with software developments and nificant individual differences in valuation of ethi-applications in other fields Applications of choice cal product features related to several demographicmodels in tourism have typically focused on mod- measures In other cases some of the heterogene-eling tourism demand (eg Huybers amp Bennett ity in preferences may be unrelated to observable2000 Morley 1994) DCMs have also been ap- covariate measures but instead be latent Severalplied to other topics such as resident trade-offs of methods have been proposed to deal with this in-tourism impacts (Lindberg Dellaert amp Rassing cluding random coefficient models (eg Hierar-1999) and understanding decisions relating to lei- chical Bayes and Mixed Logit models) and finitesure experiences (Oh Ditton Gentner amp Riech- mixture models (Kamakura amp Russell 1989ers 2005 Ready Epp amp Delavan 2005 Verma Wedel amp Kamakura 1998 for a recent review ofLouviere amp Burke 2006) This literature demon- models in this area see Train 2003) These meth-strates the flexibility of DCMs beyond traditional ods differ in several ways including whether theconsumption decisions There have been fewer ap- heterogeneity is modeled using several discreteplications of DCMs to cultural heritage research classes (as we do) or is continuous (ie using aBoxall Englin and Adamowicz (2003) and Rolfe distribution as in the case of models like theand Windle (2003) used DCMs to value aboriginal mixed logit) (see Andrews Ainslie amp Currimartifacts in Canada and Australia respectively 2002) In addition there are notable differences in

how researchers conceptualize and identify hetero-Maddison and Foster (2003) used DCEs to studymuseum congestion Apostolakis and Jaffry (2005) geneity with some using a priori theory-based

model development approaches while others useand Mazzanti (2003) applied DCMs to museums

152 BURKE ET AL

a data-driven approach where segments are identi- Differences in preferences for museums werereported by Apostolakis and Jaffry (2005) andfied in a posterior fashion (Dolnicar 2004)

Of course preference differences are not the Mazzanti (2003) who estimated interactions be-tween observable respondent characteristics andonly potential source of differences in individual

choices Individuals also may differ in choice con- museum attributes Colombino and Nese (2009)used a mixed logit model to account for unobserv-sistency For example younger visitors may be

less certain about how much they will enjoy a par- able preference heterogeneity but did not allowfor differences in choice consistency Thus priorticular special exhibition or which particular exhi-

bition they prefer due to less overall experience work suggests differences in choices of museumvisitors but does not pinpoint likely or potentialwith museums Failure to account for differences

in choice variability can lead to biased and mis- sources of heterogeneityOur research makes a methodological contribu-leading conclusions about preferences This is be-

cause each personrsquos mean preference parameter tion to the museum demand literature by using arecent development in latent class estimation thatestimates are inversely related to their choice con-

sistency (Louviere 2001) Yatchew and Griliches allows preference parameters to differ for discrete(1985) showed that error variance or heteroscedas- but unobserved (latent) classes of people whileticity ignored by researchers in a choice model also allowing the underlying variability of the ran-(discussed in the context of a probit model) will dom errors to differ between several discrete latentresult in biased parameter estimates rather than classes Until recently latent class models allowedsimple loss of efficiency as in the case of a linear preferences to differ from class to class but themodel Mroz and Zayats (2008) cautioned inter- error variances were identical over classes Errorpreting coefficient estimates of binary outcomes variance unaccountability confounds true underly-from different samples even if they use identical ing preference variability with individual differ-estimation procedure as these estimates depend ences This means that apparently significant dif-crucially on arbitrary normalization of variance ferences in preference parameters between latentVirtually all random coefficient models assume classes may be due to differences in underlyingthat choice consistency is constant across individ- scale (Louviere 2001)uals but if it is not constant estimates of such Magidson and Vermunt (2005) developed amodels can be highly biased Recent work by Fie- way to deal with the error variance-parameter esti-big Keane Louviere and Wasi (in press) and mate confound embedding the new approach inSalisbury and Feinberg (2010) suggest that it is commercially available software (a beta version ofhighly unlikely that choice consistency is constant Latent GOLD) The estimation approach allowsacross individuals one to relax the restriction that the scale parameter

Addressing the potential confound between is fixed across the whole population effectivelymodel estimates and choice consistency several decomposing class preferences and scale (or errornew types of DCMs have been developed to ac- variability) It also allows latent preference andcount for unobservable error variance heterogene- scale classes to be described as functions of demo-ity such as the Generalized Multinomial Logit graphic and other covariates We now discussmodel (Fiebig et al in press) models for single prior work on DCMs after which we introducepersons (Louviere et al 2008) models that de- and highlight the advantages of the SALCMcompose the random component (Burke amp Reit-zig 2007) and scale-adjusted latent class models Choice Models Derived(Magidson amp Vermunt 2005) Fiebig et al (in From Random Utility Theorypress) have shown unobserved scale heterogeneity

Using the conceptual framework of Randommodels outperform models that only account forUtility Theory (Manski 1977 Thurstone 1927)taste heterogeneity such as MNL and mixed logiteach individual may be described as holding latentin 10 out of 10 data sets Accordingly unobservedpreferences (utilities) associated with all choicescale heterogeneity was more important in datasets

that involve more complex choice objects options being considered Individuals maximize

SCALE-ADUSTED LATENT CLASS MODEL 153

utility by choosing their most preferred option La- individual n depends on the latent preference classq = 1 Q and the unobserved scale parametertent preferences are specified by two components

(observed and unobserved) Different discrete class λdd = 1 D and each individual is ex-pressed in terms of the likelihood that they belongchoice models are derived from different assump-

tions about the unobserved component the best to each class For identification one scale parame-ter is normalized to unity (see Swait amp Louviereknown of which is the conditional multinomial

logit model or MNL (McFadden 1974) MNL as- 1993) the remaining scale parameter estimates areratios of the reference scale class The probabilitysumes the unobserved component is independently

and identically distributed across choices and indi- of choice i by individual n in choice situation tconditioning on preference class q and scale factorviduals In this model the latent utility of option

i judged by respondent n is given by Uni = βprimeXni + class d isεni where Xni is the vector of attributes of option iand β is a vector of parameters representing the

P(ntiqd) = exp(λdβq primexnti)Σjexp(λdβq primexntj)

(2)preference (ie weight) associated with each attri-bute By assuming εni to be iid extreme valuetype I McFadden showed that the choice probabil- We assume that there are Q discrete latentity from a total of J options considered is given by classes and D scale groupings but that class mem-

bership is hidden However we allow for theclasses to be determined using a set of observablecharacteristics or demographics If Hnqd is thePni = exp(βprimeXni)

sumJ

j=1

exp(βprimeXnj)

(1)prior probability for preference class q for individ-ual n and Gnd is the prior probability for scalegroup d the multinomial logit model suggested byHensher and Greene (2002) can be used to esti-It is important to recognize that the MNLmate each prior asmodel parameters in equation (1) βMNL are not the

true underlying β in Uni As is the case for allchoice models the estimates are confounded with

Hnqd = exp(θprimeqzn)Σqexp(θprimeqzn)

(3)σε the standard deviation of the error distribution(Swait amp Louviere 1993 Train 2003) Thus oneactually estimates β = λβ where λ is commonly Gnd = exp(γprimeqzn)

Σdexp(γprimeqzn)(4)

referred to as ldquothe scale parameterrdquo and in the caseof MNL the relationship is λ = π[sqrt(6)σε]

where the vector zi includes k = 1 K relevantTo avoid confounding preference estimatescovariates or sociodemographic characteristics ofwith choice consistency we use the SALCM toindividual n including a constant and θ and γ areestimate relative scale parameters jointly withthe respective vectors of parameter estimates ofpreference classes and associated preference pa-those zi covariates In turn the overall log likeli-rameters Indeed without estimating λ (ie re-hood isstricting λ = 1 for all classes) even in finite mix-

ture models where one allows the preferenceparameters β to vary between latent classes of

ln L = sumN

n=1

Pn = lnsumD

d=1

sumQ

q=1

GndHnqdJT

t=1

Pntiqd (5)consumers it is still possible to misinterpret dif-ferences in the true underlying β between classeswhich may be due only to differences in λ (Lou-

We followed the recommended estimation pro-viere 2001)cedure in Latent GOLD and used 20 randomizedsets of starting values to ensure an identified

Scale-Adjusted Latent Class Models (SALCM)global maximum We choose the number of pref-erence and scale classes by comparing the Bayes-Magidson and Vermunt (2007) proposed a

model in which the random utility of option j for ian Information Criterion (BIC) associated with

154 BURKE ET AL

each class solution which rewards improved fit attributes The collection of all possible combina-tions of these attributes comprises 82 times 49 times 215 =and penalizes additional parameters The posterior

probability that an individual belongs to prefer- 3774873600 combinations We used Street andBurgessrsquo (2007) design theory to construct an effi-ence class q and scale class d can be inferred from

Bayesrsquo theorem cient set of 128 pairs that were blocked into eightchoice sets per respondent This design approachis based on an orthogonal main effects plan (OMEP)Discrete Choice Experiment Surveythat allows estimation of a main effects only

We now describe a DCE that provides choice modeldata that can take advantage of the SALCM by The DCE was embedded in a larger online sur-decomposing choices into preference and scale vey that focused on questions about leisure activi-components and allows both to vary across the ties and museum visitation The first part of theunderlying latent segments The main objective of survey asked respondents about whether they hadthe DCE and associated survey was to quantify the engaged in various leisure activities in the previ-effects of tactical and strategic museum attributes ous year and whether they expected to do so inin order to better align museum offerings tactics the next 12 months Visiting a museum whetherand strategies with visitor choices We first de- past or prospective defined eligibility as a partici-scribe research that preceded the DCE undertaken pant in the study and the study context was re-to enhance external validity and provide strategic vealed to qualifying respondents The second partinformation to the six collaborating museums It is of the survey consisted of two museum exhibitionarguably the case that the most important stage in tasks In the first task respondents were asked toa DCE is the prior identification of salient attri- indicate if they would or would not go to see eachbutes and levels as the accuracy and validity of a of 37 named (and described) exhibitions The sec-DCE depends entirely on it Additionally one ond task focused on a subset of 9 of the 37 exhibi-must balance the number of attributes identified tions respondents were asked to rate each on aagainst the size and scope of the resulting choice five-category rating scale ranging from very unat-tasks for example including many factors leads tractive (one star) to very attractive (five stars)to large DCEs and more complex tasks that have Each subject was randomly assigned to a differentimplications for sample costs task complexity re- set of nine exhibitions based on a balanced incom-port delivery and presentation complexity Yet plete block design The latter design consists of 37omission of salient attributes may lead to model blocks of 9 exhibitions each exhibition appears 9misspecification and omitted variable bias as well times across the 37 sets (blocks) and co-appearsas concerns about external validity with each other exhibition twice

As previously noted we used qualitative inter- In the third section of the survey the DCE re-views with museum visitors (Burton et al 2008) to spondents received instructions and an example ofidentify a long list of potentially salient attributes a DCE scenario that explained how to answer thefor visitors and potential visitors We then con- questions Following that they received eight sce-ducted another round of research to measure the sa- narios or choice sets (see Fig 1) Each attributeliency of 64 attributes using a BestndashWorst Scaling was described in a ldquopop-up boxrdquo displayed whenapproach (Marley amp Louviere 2005) We used the the cursor was held above that feature Respon-results of the BestndashWorst Scaling exercise to dis- dents were instructed to consider star ratings ofcuss which of the attributes in fact could be ldquospecial exhibitionsrdquo as representing their own at-changed with representatives from partner muse- tractiveness ratings exemplified by their responsesums This resulted in a final 26 attributes and dis- to the prior exhibition rating task Respondentscussions with our museum partners allowed us to were asked to indicate their preferred museum of-specify a relevant set of levels for each attribute fer and whether they would actually visit that mu-

The final list of 26 attributes and associated seum for each choice set The survey concludedlevels are in Table 5 There were 2 eight-level at- with questions about preferred days for extended

opening hours and respondent sociodemographictributes 9 four-level attributes and 15 two-level

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 149

erences to another but the differences may simply attract visitors but if potential visitors place morebe the result of differences in how consistent the value on staff knowledge of exhibitions it may bechoices are that reveals the preferences in each better to allocate resources to staff training thansegment (Louviere 2001) We take both choice extending opening hoursvariability and preference into account by estimat- Museum managers have been keen observersing a relatively new model the Scale-Adjusted La- of their markets over the past decade Many usetent Class Model (SALCM) proposed by Magid- qualitative methods such as focus groups in-depthson and Vermunt (2005) The SALCM identifies interviews and observational techniques Forlatent (unobservable) segments that simultane- example Bitgood (2006) uses a general valueously differ in two ways 1) in preferences for a principle (ie cost vs benefits) to account for thegiven attributefeature and its levels and 2) in the trade-offs that visitors use in moving around anvariability with which overall choices are made exhibition On a more sensory level Joy andPrevious applications of latent class models did Sherry (2003) use observation and in-depth inter-not take choice variability differences into ac- views to determine visitorsrsquo visceral and embodiedcount characterizing the latent segments (classes) imaginative responses to exhibitions Thyne (2000)only in terms of preference differences As far as uses content analysis of interviews and a ladderingwe are aware this is the first application of this technique to determine factors that influence mu-model in tourism research seum visitation but did not report how influential

they were While these and similar studies (egA Choice Model of Museum Visitation Debenedetti 2003 Goulding 2000 Wiggins 2004)

identified many potential factors that might affectTraditionally marketers assert that successfulmuseum experiences they did not isolate or mea-product differentiation can be achieved by offeringsure key attributes that influence museum visita-consumers something relatively important and val-tion decisions In addition in the absence of well-uable that stands out from what others offer (Ca-conceived external validity tests there are validityhill 1996) The value that consumers derive fromissues noted by Biehal and Chakravarti (1989)products (including services and experiences) de-

Museum managers extensively use survey re-pends on the underlying characteristics or attri-search to learn about visitors For example sur-butes (Lancaster 1966) of the services such asveys are used to try to measure the ldquoimportancerdquomuseums Value to museum visitors can be associ-of various factors in visitation decisions Theated with attributes like guided tours exhibitedmethodologies employed include ranking con-artifacts and interactive displays Visitor choicesstant sum and rating scales to compare institu-involve trade-offs whereby they give up one attri-tions on demographics visitation frequency satis-bute or attribute level to obtain another (Louvierefaction and summative evaluation of exhibitions1988) In the case of museums potential visitors(Kirchberg 1996) Similar to the validity issuesmay visit a museum with interactive exhibits evennoted above with qualitative methods ranking andif it has a high entry price Visitors may considerconstant sum scales can be difficult for respon-many attributes in making decisions or they maydents when they have to rank or allocate pointsconsider only a few hence managers often lackacross many factors (Cohen amp Orme 2004) Sin-clear understanding of the relative importance ofgle-item rating scales can be problematic when re-product attributes Museums face constraints duespondents rate many (or even all) factors as impor-to costs policies supply issues and sustainabletant As every feature is treated equally there iseconomic practices and try to maximize the attri-no real disincentive to choose as noted by Carsonbutes of their ldquoofferingsrdquo to provide an ldquoidealrdquoand Groves (2007) Compounding this are issuesmuseum experience (Sheth Newman amp Grossrelated to differences in individualsrsquo use of such1991) Hence museum managers need to under-scales For example one personrsquos three may be thestand the trade-offs that visitors will make for at-same as anotherrsquos four or vice versa And oftributes that can be changed For example a mu-

seum could consider extending opening hours to course it is widely recognized that rating scale re-

150 BURKE ET AL

sponses are influenced by cultural norms (eg Hui than meets the eye Thus managers of museumsand other tourism service suppliers could benefitamp Triandis 1989)

Other quantitative research methods that have from research methods that provide sharper andbetter discrimination between potential attributesbeen used to study cultural heritage include Kellyrsquos

(2004) Repertory Grid Analysis and Contingent Louviere and Islam (2008) have compared fourdifferent ways of measuring importance weightsValuation Kellyrsquos grid has been used to identify

the strength of museum brands (Caldwell amp Cos- two direct (ie rating scales and bestndashworst scal-ing) and two indirect methods (ie discrete choicehall 2003) by proposing that people make choices

based on analogous associations While repertory experiments and implied willingness to pay mea-sures) Direct approaches measure the importancegrid analysis allows individuals to direct the way

in which associations are determined it does not of a set of attributes by asking respondents to statethe degree of importance or weight on some cate-allow one to make predictions about choices An-

ticipating museum visitation choices requires one gory rating or constant sum scales Indirect ap-to take into account interest tastes and motives proaches generally try to elicit importance by ana-not typically observed in repertory grid analysis lyzing an outcome measure like choices They

Contingent valuation (CV) (Mitchell amp Carson found different results from direct and indirect1989) has been used to measure the economic preference elicitation approaches According tovalue of cultural resources (eg Noonan 2003 Louviere and Islam (2008) indirect measures pro-Santagata amp Signorello 2000) CV methods typi- vide richer insights about trade-offs among attri-cally ask respondents about their willingness to butes and provides more meaningful managerialpay for a well-defined and described option such inputs due its natural link with consumer purchas-as a particular museum exhibition although there ing context Jaccard Brinberg and Lee (1986)is nothing inherent in the CV approach that ex- also found lack of convergence among direct andcludes variation in product features Our research indirect measures of importance In absence of un-approach has many features in common with CV derlying theoretical reasons for such differencesapproaches but is more general than typical CV we use indirect measures that have external valid-applications because we design and implement a ity in the literature (see Louviere amp Islam 2008discrete choice experiment (DCE) that allows us for discussion on external validity of indirect mea-to analyze survey responses with discrete choice sures)models that can be used to predict likely visitorchoices Discrete Choice Experiments

The qualitative and quantitative research re- and Tourism Researchviewed above suggests that visitors may use sev-

While it is safe to assume that various attri-eral attributes of museums to value museum at-buteslevels drive museum visitation decisions ittractiveness These include social factors identityis unclear which ones singly or in combinationdevelopment and consumption experiences (Falkmatter most (Kelly 2004) or to whom Discrete2006) and artisticaesthetic emotional and educa-choice experiments (DCEs) and associated dis-tional factors (Boorsma 2006) Differences in thecrete choice models (DCMs) offer a potential wayattributes that visitors evaluate and care about mayto quantify what drives museum visitor choicesbe associated with psychographic factors differ-(Burton 2003 Burton amp Scott 2003) In the re-ences in visitor segments andor other leisure con-search reported below DCMs are used to analyzesumption differences (Clopton Stoddard amp Davethe outcomes of a DCE in which past or potential2006) As noted by Cohen and Neira (2003) allvisitors were asked to make choices in a series oftoo often when consumers are asked to evaluatechoice sets The sets were constructed to systemat-potential attributes of products or services likeically vary a number of attributes of museums thatmuseums they rate all or many attributes ascan be changed The sets were based on priorequally important providing little discrimination

among attributes and less useful strategic insight qualitative research that identified a range of ele-

SCALE-ADUSTED LATENT CLASS MODEL 151

ments that matter when choosing a museum visit in Greece and Italy respectively to understandmuseum demandDCEs are widely used to represent everyday

choices such as products on shelves in supermar-kets transport mode options for commuting and

Accounting for Individual Differencesrecreational offerings like parks or fishing loca-

in Preference and Variabilitytions (eg Louviere Hensher amp Swait 2000)Choice response data from DCE surveys can be One of the key advantages in using DCMs re-analyzed with choice models to uncover system- volves around modeling decisions to visit muse-atic relationships between choices (dependent vari- ums and how these decisions may be affected byable) and changes in productservice offerings or changes to museum features Early DCMs suchldquoattributesrdquo (independent variables) as McFaddenrsquos (1974) conditional logit model

The choice modeling literature is now fairly restrict the preferences of individuals to be homo-mature and comprehensive reviews can be found geneous hence preferences estimated in thesein Train (2003) or Louviere et al (2000) More models are sample averages In turn this raisesrecently it has been recognized that more complex aggregation bias concerns (Hutchinson Kama-models are required that can take differences in kura amp Lynch 2000) for example when all fe-individual choice variability into account (eg males prefer museums with longer opening hoursAdamowicz et al 2008 Islam Louviere amp but all males are indifferent In such a case oneBurke 2007 Louviere amp Swait 2010) Thus we would underestimate the effects of extended open-make use of relatively new developments in finite ing hours on choices of both genders if they aremixture models known as latent class models We equally represented in a sampletake advantage of a new type of latent class model One way to account for such preference differ-known as a ldquoscale-adjusted latent class modelrdquo ences (or heterogeneity) is to include interactionsthat can simultaneously account for differences in between observable covariate measures (eg de-choice variability and preferences (Magidson amp mographics psychographics) and product featuresVermunt 2005) in DCMs (Gupta amp Chintagunta 1994) Auger

Use of DCMs in tourism and leisure research Burke Devinney and Louviere (2003) report sig-has grown in line with software developments and nificant individual differences in valuation of ethi-applications in other fields Applications of choice cal product features related to several demographicmodels in tourism have typically focused on mod- measures In other cases some of the heterogene-eling tourism demand (eg Huybers amp Bennett ity in preferences may be unrelated to observable2000 Morley 1994) DCMs have also been ap- covariate measures but instead be latent Severalplied to other topics such as resident trade-offs of methods have been proposed to deal with this in-tourism impacts (Lindberg Dellaert amp Rassing cluding random coefficient models (eg Hierar-1999) and understanding decisions relating to lei- chical Bayes and Mixed Logit models) and finitesure experiences (Oh Ditton Gentner amp Riech- mixture models (Kamakura amp Russell 1989ers 2005 Ready Epp amp Delavan 2005 Verma Wedel amp Kamakura 1998 for a recent review ofLouviere amp Burke 2006) This literature demon- models in this area see Train 2003) These meth-strates the flexibility of DCMs beyond traditional ods differ in several ways including whether theconsumption decisions There have been fewer ap- heterogeneity is modeled using several discreteplications of DCMs to cultural heritage research classes (as we do) or is continuous (ie using aBoxall Englin and Adamowicz (2003) and Rolfe distribution as in the case of models like theand Windle (2003) used DCMs to value aboriginal mixed logit) (see Andrews Ainslie amp Currimartifacts in Canada and Australia respectively 2002) In addition there are notable differences in

how researchers conceptualize and identify hetero-Maddison and Foster (2003) used DCEs to studymuseum congestion Apostolakis and Jaffry (2005) geneity with some using a priori theory-based

model development approaches while others useand Mazzanti (2003) applied DCMs to museums

152 BURKE ET AL

a data-driven approach where segments are identi- Differences in preferences for museums werereported by Apostolakis and Jaffry (2005) andfied in a posterior fashion (Dolnicar 2004)

Of course preference differences are not the Mazzanti (2003) who estimated interactions be-tween observable respondent characteristics andonly potential source of differences in individual

choices Individuals also may differ in choice con- museum attributes Colombino and Nese (2009)used a mixed logit model to account for unobserv-sistency For example younger visitors may be

less certain about how much they will enjoy a par- able preference heterogeneity but did not allowfor differences in choice consistency Thus priorticular special exhibition or which particular exhi-

bition they prefer due to less overall experience work suggests differences in choices of museumvisitors but does not pinpoint likely or potentialwith museums Failure to account for differences

in choice variability can lead to biased and mis- sources of heterogeneityOur research makes a methodological contribu-leading conclusions about preferences This is be-

cause each personrsquos mean preference parameter tion to the museum demand literature by using arecent development in latent class estimation thatestimates are inversely related to their choice con-

sistency (Louviere 2001) Yatchew and Griliches allows preference parameters to differ for discrete(1985) showed that error variance or heteroscedas- but unobserved (latent) classes of people whileticity ignored by researchers in a choice model also allowing the underlying variability of the ran-(discussed in the context of a probit model) will dom errors to differ between several discrete latentresult in biased parameter estimates rather than classes Until recently latent class models allowedsimple loss of efficiency as in the case of a linear preferences to differ from class to class but themodel Mroz and Zayats (2008) cautioned inter- error variances were identical over classes Errorpreting coefficient estimates of binary outcomes variance unaccountability confounds true underly-from different samples even if they use identical ing preference variability with individual differ-estimation procedure as these estimates depend ences This means that apparently significant dif-crucially on arbitrary normalization of variance ferences in preference parameters between latentVirtually all random coefficient models assume classes may be due to differences in underlyingthat choice consistency is constant across individ- scale (Louviere 2001)uals but if it is not constant estimates of such Magidson and Vermunt (2005) developed amodels can be highly biased Recent work by Fie- way to deal with the error variance-parameter esti-big Keane Louviere and Wasi (in press) and mate confound embedding the new approach inSalisbury and Feinberg (2010) suggest that it is commercially available software (a beta version ofhighly unlikely that choice consistency is constant Latent GOLD) The estimation approach allowsacross individuals one to relax the restriction that the scale parameter

Addressing the potential confound between is fixed across the whole population effectivelymodel estimates and choice consistency several decomposing class preferences and scale (or errornew types of DCMs have been developed to ac- variability) It also allows latent preference andcount for unobservable error variance heterogene- scale classes to be described as functions of demo-ity such as the Generalized Multinomial Logit graphic and other covariates We now discussmodel (Fiebig et al in press) models for single prior work on DCMs after which we introducepersons (Louviere et al 2008) models that de- and highlight the advantages of the SALCMcompose the random component (Burke amp Reit-zig 2007) and scale-adjusted latent class models Choice Models Derived(Magidson amp Vermunt 2005) Fiebig et al (in From Random Utility Theorypress) have shown unobserved scale heterogeneity

Using the conceptual framework of Randommodels outperform models that only account forUtility Theory (Manski 1977 Thurstone 1927)taste heterogeneity such as MNL and mixed logiteach individual may be described as holding latentin 10 out of 10 data sets Accordingly unobservedpreferences (utilities) associated with all choicescale heterogeneity was more important in datasets

that involve more complex choice objects options being considered Individuals maximize

SCALE-ADUSTED LATENT CLASS MODEL 153

utility by choosing their most preferred option La- individual n depends on the latent preference classq = 1 Q and the unobserved scale parametertent preferences are specified by two components

(observed and unobserved) Different discrete class λdd = 1 D and each individual is ex-pressed in terms of the likelihood that they belongchoice models are derived from different assump-

tions about the unobserved component the best to each class For identification one scale parame-ter is normalized to unity (see Swait amp Louviereknown of which is the conditional multinomial

logit model or MNL (McFadden 1974) MNL as- 1993) the remaining scale parameter estimates areratios of the reference scale class The probabilitysumes the unobserved component is independently

and identically distributed across choices and indi- of choice i by individual n in choice situation tconditioning on preference class q and scale factorviduals In this model the latent utility of option

i judged by respondent n is given by Uni = βprimeXni + class d isεni where Xni is the vector of attributes of option iand β is a vector of parameters representing the

P(ntiqd) = exp(λdβq primexnti)Σjexp(λdβq primexntj)

(2)preference (ie weight) associated with each attri-bute By assuming εni to be iid extreme valuetype I McFadden showed that the choice probabil- We assume that there are Q discrete latentity from a total of J options considered is given by classes and D scale groupings but that class mem-

bership is hidden However we allow for theclasses to be determined using a set of observablecharacteristics or demographics If Hnqd is thePni = exp(βprimeXni)

sumJ

j=1

exp(βprimeXnj)

(1)prior probability for preference class q for individ-ual n and Gnd is the prior probability for scalegroup d the multinomial logit model suggested byHensher and Greene (2002) can be used to esti-It is important to recognize that the MNLmate each prior asmodel parameters in equation (1) βMNL are not the

true underlying β in Uni As is the case for allchoice models the estimates are confounded with

Hnqd = exp(θprimeqzn)Σqexp(θprimeqzn)

(3)σε the standard deviation of the error distribution(Swait amp Louviere 1993 Train 2003) Thus oneactually estimates β = λβ where λ is commonly Gnd = exp(γprimeqzn)

Σdexp(γprimeqzn)(4)

referred to as ldquothe scale parameterrdquo and in the caseof MNL the relationship is λ = π[sqrt(6)σε]

where the vector zi includes k = 1 K relevantTo avoid confounding preference estimatescovariates or sociodemographic characteristics ofwith choice consistency we use the SALCM toindividual n including a constant and θ and γ areestimate relative scale parameters jointly withthe respective vectors of parameter estimates ofpreference classes and associated preference pa-those zi covariates In turn the overall log likeli-rameters Indeed without estimating λ (ie re-hood isstricting λ = 1 for all classes) even in finite mix-

ture models where one allows the preferenceparameters β to vary between latent classes of

ln L = sumN

n=1

Pn = lnsumD

d=1

sumQ

q=1

GndHnqdJT

t=1

Pntiqd (5)consumers it is still possible to misinterpret dif-ferences in the true underlying β between classeswhich may be due only to differences in λ (Lou-

We followed the recommended estimation pro-viere 2001)cedure in Latent GOLD and used 20 randomizedsets of starting values to ensure an identified

Scale-Adjusted Latent Class Models (SALCM)global maximum We choose the number of pref-erence and scale classes by comparing the Bayes-Magidson and Vermunt (2007) proposed a

model in which the random utility of option j for ian Information Criterion (BIC) associated with

154 BURKE ET AL

each class solution which rewards improved fit attributes The collection of all possible combina-tions of these attributes comprises 82 times 49 times 215 =and penalizes additional parameters The posterior

probability that an individual belongs to prefer- 3774873600 combinations We used Street andBurgessrsquo (2007) design theory to construct an effi-ence class q and scale class d can be inferred from

Bayesrsquo theorem cient set of 128 pairs that were blocked into eightchoice sets per respondent This design approachis based on an orthogonal main effects plan (OMEP)Discrete Choice Experiment Surveythat allows estimation of a main effects only

We now describe a DCE that provides choice modeldata that can take advantage of the SALCM by The DCE was embedded in a larger online sur-decomposing choices into preference and scale vey that focused on questions about leisure activi-components and allows both to vary across the ties and museum visitation The first part of theunderlying latent segments The main objective of survey asked respondents about whether they hadthe DCE and associated survey was to quantify the engaged in various leisure activities in the previ-effects of tactical and strategic museum attributes ous year and whether they expected to do so inin order to better align museum offerings tactics the next 12 months Visiting a museum whetherand strategies with visitor choices We first de- past or prospective defined eligibility as a partici-scribe research that preceded the DCE undertaken pant in the study and the study context was re-to enhance external validity and provide strategic vealed to qualifying respondents The second partinformation to the six collaborating museums It is of the survey consisted of two museum exhibitionarguably the case that the most important stage in tasks In the first task respondents were asked toa DCE is the prior identification of salient attri- indicate if they would or would not go to see eachbutes and levels as the accuracy and validity of a of 37 named (and described) exhibitions The sec-DCE depends entirely on it Additionally one ond task focused on a subset of 9 of the 37 exhibi-must balance the number of attributes identified tions respondents were asked to rate each on aagainst the size and scope of the resulting choice five-category rating scale ranging from very unat-tasks for example including many factors leads tractive (one star) to very attractive (five stars)to large DCEs and more complex tasks that have Each subject was randomly assigned to a differentimplications for sample costs task complexity re- set of nine exhibitions based on a balanced incom-port delivery and presentation complexity Yet plete block design The latter design consists of 37omission of salient attributes may lead to model blocks of 9 exhibitions each exhibition appears 9misspecification and omitted variable bias as well times across the 37 sets (blocks) and co-appearsas concerns about external validity with each other exhibition twice

As previously noted we used qualitative inter- In the third section of the survey the DCE re-views with museum visitors (Burton et al 2008) to spondents received instructions and an example ofidentify a long list of potentially salient attributes a DCE scenario that explained how to answer thefor visitors and potential visitors We then con- questions Following that they received eight sce-ducted another round of research to measure the sa- narios or choice sets (see Fig 1) Each attributeliency of 64 attributes using a BestndashWorst Scaling was described in a ldquopop-up boxrdquo displayed whenapproach (Marley amp Louviere 2005) We used the the cursor was held above that feature Respon-results of the BestndashWorst Scaling exercise to dis- dents were instructed to consider star ratings ofcuss which of the attributes in fact could be ldquospecial exhibitionsrdquo as representing their own at-changed with representatives from partner muse- tractiveness ratings exemplified by their responsesums This resulted in a final 26 attributes and dis- to the prior exhibition rating task Respondentscussions with our museum partners allowed us to were asked to indicate their preferred museum of-specify a relevant set of levels for each attribute fer and whether they would actually visit that mu-

The final list of 26 attributes and associated seum for each choice set The survey concludedlevels are in Table 5 There were 2 eight-level at- with questions about preferred days for extended

opening hours and respondent sociodemographictributes 9 four-level attributes and 15 two-level

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

150 BURKE ET AL

sponses are influenced by cultural norms (eg Hui than meets the eye Thus managers of museumsand other tourism service suppliers could benefitamp Triandis 1989)

Other quantitative research methods that have from research methods that provide sharper andbetter discrimination between potential attributesbeen used to study cultural heritage include Kellyrsquos

(2004) Repertory Grid Analysis and Contingent Louviere and Islam (2008) have compared fourdifferent ways of measuring importance weightsValuation Kellyrsquos grid has been used to identify

the strength of museum brands (Caldwell amp Cos- two direct (ie rating scales and bestndashworst scal-ing) and two indirect methods (ie discrete choicehall 2003) by proposing that people make choices

based on analogous associations While repertory experiments and implied willingness to pay mea-sures) Direct approaches measure the importancegrid analysis allows individuals to direct the way

in which associations are determined it does not of a set of attributes by asking respondents to statethe degree of importance or weight on some cate-allow one to make predictions about choices An-

ticipating museum visitation choices requires one gory rating or constant sum scales Indirect ap-to take into account interest tastes and motives proaches generally try to elicit importance by ana-not typically observed in repertory grid analysis lyzing an outcome measure like choices They

Contingent valuation (CV) (Mitchell amp Carson found different results from direct and indirect1989) has been used to measure the economic preference elicitation approaches According tovalue of cultural resources (eg Noonan 2003 Louviere and Islam (2008) indirect measures pro-Santagata amp Signorello 2000) CV methods typi- vide richer insights about trade-offs among attri-cally ask respondents about their willingness to butes and provides more meaningful managerialpay for a well-defined and described option such inputs due its natural link with consumer purchas-as a particular museum exhibition although there ing context Jaccard Brinberg and Lee (1986)is nothing inherent in the CV approach that ex- also found lack of convergence among direct andcludes variation in product features Our research indirect measures of importance In absence of un-approach has many features in common with CV derlying theoretical reasons for such differencesapproaches but is more general than typical CV we use indirect measures that have external valid-applications because we design and implement a ity in the literature (see Louviere amp Islam 2008discrete choice experiment (DCE) that allows us for discussion on external validity of indirect mea-to analyze survey responses with discrete choice sures)models that can be used to predict likely visitorchoices Discrete Choice Experiments

The qualitative and quantitative research re- and Tourism Researchviewed above suggests that visitors may use sev-

While it is safe to assume that various attri-eral attributes of museums to value museum at-buteslevels drive museum visitation decisions ittractiveness These include social factors identityis unclear which ones singly or in combinationdevelopment and consumption experiences (Falkmatter most (Kelly 2004) or to whom Discrete2006) and artisticaesthetic emotional and educa-choice experiments (DCEs) and associated dis-tional factors (Boorsma 2006) Differences in thecrete choice models (DCMs) offer a potential wayattributes that visitors evaluate and care about mayto quantify what drives museum visitor choicesbe associated with psychographic factors differ-(Burton 2003 Burton amp Scott 2003) In the re-ences in visitor segments andor other leisure con-search reported below DCMs are used to analyzesumption differences (Clopton Stoddard amp Davethe outcomes of a DCE in which past or potential2006) As noted by Cohen and Neira (2003) allvisitors were asked to make choices in a series oftoo often when consumers are asked to evaluatechoice sets The sets were constructed to systemat-potential attributes of products or services likeically vary a number of attributes of museums thatmuseums they rate all or many attributes ascan be changed The sets were based on priorequally important providing little discrimination

among attributes and less useful strategic insight qualitative research that identified a range of ele-

SCALE-ADUSTED LATENT CLASS MODEL 151

ments that matter when choosing a museum visit in Greece and Italy respectively to understandmuseum demandDCEs are widely used to represent everyday

choices such as products on shelves in supermar-kets transport mode options for commuting and

Accounting for Individual Differencesrecreational offerings like parks or fishing loca-

in Preference and Variabilitytions (eg Louviere Hensher amp Swait 2000)Choice response data from DCE surveys can be One of the key advantages in using DCMs re-analyzed with choice models to uncover system- volves around modeling decisions to visit muse-atic relationships between choices (dependent vari- ums and how these decisions may be affected byable) and changes in productservice offerings or changes to museum features Early DCMs suchldquoattributesrdquo (independent variables) as McFaddenrsquos (1974) conditional logit model

The choice modeling literature is now fairly restrict the preferences of individuals to be homo-mature and comprehensive reviews can be found geneous hence preferences estimated in thesein Train (2003) or Louviere et al (2000) More models are sample averages In turn this raisesrecently it has been recognized that more complex aggregation bias concerns (Hutchinson Kama-models are required that can take differences in kura amp Lynch 2000) for example when all fe-individual choice variability into account (eg males prefer museums with longer opening hoursAdamowicz et al 2008 Islam Louviere amp but all males are indifferent In such a case oneBurke 2007 Louviere amp Swait 2010) Thus we would underestimate the effects of extended open-make use of relatively new developments in finite ing hours on choices of both genders if they aremixture models known as latent class models We equally represented in a sampletake advantage of a new type of latent class model One way to account for such preference differ-known as a ldquoscale-adjusted latent class modelrdquo ences (or heterogeneity) is to include interactionsthat can simultaneously account for differences in between observable covariate measures (eg de-choice variability and preferences (Magidson amp mographics psychographics) and product featuresVermunt 2005) in DCMs (Gupta amp Chintagunta 1994) Auger

Use of DCMs in tourism and leisure research Burke Devinney and Louviere (2003) report sig-has grown in line with software developments and nificant individual differences in valuation of ethi-applications in other fields Applications of choice cal product features related to several demographicmodels in tourism have typically focused on mod- measures In other cases some of the heterogene-eling tourism demand (eg Huybers amp Bennett ity in preferences may be unrelated to observable2000 Morley 1994) DCMs have also been ap- covariate measures but instead be latent Severalplied to other topics such as resident trade-offs of methods have been proposed to deal with this in-tourism impacts (Lindberg Dellaert amp Rassing cluding random coefficient models (eg Hierar-1999) and understanding decisions relating to lei- chical Bayes and Mixed Logit models) and finitesure experiences (Oh Ditton Gentner amp Riech- mixture models (Kamakura amp Russell 1989ers 2005 Ready Epp amp Delavan 2005 Verma Wedel amp Kamakura 1998 for a recent review ofLouviere amp Burke 2006) This literature demon- models in this area see Train 2003) These meth-strates the flexibility of DCMs beyond traditional ods differ in several ways including whether theconsumption decisions There have been fewer ap- heterogeneity is modeled using several discreteplications of DCMs to cultural heritage research classes (as we do) or is continuous (ie using aBoxall Englin and Adamowicz (2003) and Rolfe distribution as in the case of models like theand Windle (2003) used DCMs to value aboriginal mixed logit) (see Andrews Ainslie amp Currimartifacts in Canada and Australia respectively 2002) In addition there are notable differences in

how researchers conceptualize and identify hetero-Maddison and Foster (2003) used DCEs to studymuseum congestion Apostolakis and Jaffry (2005) geneity with some using a priori theory-based

model development approaches while others useand Mazzanti (2003) applied DCMs to museums

152 BURKE ET AL

a data-driven approach where segments are identi- Differences in preferences for museums werereported by Apostolakis and Jaffry (2005) andfied in a posterior fashion (Dolnicar 2004)

Of course preference differences are not the Mazzanti (2003) who estimated interactions be-tween observable respondent characteristics andonly potential source of differences in individual

choices Individuals also may differ in choice con- museum attributes Colombino and Nese (2009)used a mixed logit model to account for unobserv-sistency For example younger visitors may be

less certain about how much they will enjoy a par- able preference heterogeneity but did not allowfor differences in choice consistency Thus priorticular special exhibition or which particular exhi-

bition they prefer due to less overall experience work suggests differences in choices of museumvisitors but does not pinpoint likely or potentialwith museums Failure to account for differences

in choice variability can lead to biased and mis- sources of heterogeneityOur research makes a methodological contribu-leading conclusions about preferences This is be-

cause each personrsquos mean preference parameter tion to the museum demand literature by using arecent development in latent class estimation thatestimates are inversely related to their choice con-

sistency (Louviere 2001) Yatchew and Griliches allows preference parameters to differ for discrete(1985) showed that error variance or heteroscedas- but unobserved (latent) classes of people whileticity ignored by researchers in a choice model also allowing the underlying variability of the ran-(discussed in the context of a probit model) will dom errors to differ between several discrete latentresult in biased parameter estimates rather than classes Until recently latent class models allowedsimple loss of efficiency as in the case of a linear preferences to differ from class to class but themodel Mroz and Zayats (2008) cautioned inter- error variances were identical over classes Errorpreting coefficient estimates of binary outcomes variance unaccountability confounds true underly-from different samples even if they use identical ing preference variability with individual differ-estimation procedure as these estimates depend ences This means that apparently significant dif-crucially on arbitrary normalization of variance ferences in preference parameters between latentVirtually all random coefficient models assume classes may be due to differences in underlyingthat choice consistency is constant across individ- scale (Louviere 2001)uals but if it is not constant estimates of such Magidson and Vermunt (2005) developed amodels can be highly biased Recent work by Fie- way to deal with the error variance-parameter esti-big Keane Louviere and Wasi (in press) and mate confound embedding the new approach inSalisbury and Feinberg (2010) suggest that it is commercially available software (a beta version ofhighly unlikely that choice consistency is constant Latent GOLD) The estimation approach allowsacross individuals one to relax the restriction that the scale parameter

Addressing the potential confound between is fixed across the whole population effectivelymodel estimates and choice consistency several decomposing class preferences and scale (or errornew types of DCMs have been developed to ac- variability) It also allows latent preference andcount for unobservable error variance heterogene- scale classes to be described as functions of demo-ity such as the Generalized Multinomial Logit graphic and other covariates We now discussmodel (Fiebig et al in press) models for single prior work on DCMs after which we introducepersons (Louviere et al 2008) models that de- and highlight the advantages of the SALCMcompose the random component (Burke amp Reit-zig 2007) and scale-adjusted latent class models Choice Models Derived(Magidson amp Vermunt 2005) Fiebig et al (in From Random Utility Theorypress) have shown unobserved scale heterogeneity

Using the conceptual framework of Randommodels outperform models that only account forUtility Theory (Manski 1977 Thurstone 1927)taste heterogeneity such as MNL and mixed logiteach individual may be described as holding latentin 10 out of 10 data sets Accordingly unobservedpreferences (utilities) associated with all choicescale heterogeneity was more important in datasets

that involve more complex choice objects options being considered Individuals maximize

SCALE-ADUSTED LATENT CLASS MODEL 153

utility by choosing their most preferred option La- individual n depends on the latent preference classq = 1 Q and the unobserved scale parametertent preferences are specified by two components

(observed and unobserved) Different discrete class λdd = 1 D and each individual is ex-pressed in terms of the likelihood that they belongchoice models are derived from different assump-

tions about the unobserved component the best to each class For identification one scale parame-ter is normalized to unity (see Swait amp Louviereknown of which is the conditional multinomial

logit model or MNL (McFadden 1974) MNL as- 1993) the remaining scale parameter estimates areratios of the reference scale class The probabilitysumes the unobserved component is independently

and identically distributed across choices and indi- of choice i by individual n in choice situation tconditioning on preference class q and scale factorviduals In this model the latent utility of option

i judged by respondent n is given by Uni = βprimeXni + class d isεni where Xni is the vector of attributes of option iand β is a vector of parameters representing the

P(ntiqd) = exp(λdβq primexnti)Σjexp(λdβq primexntj)

(2)preference (ie weight) associated with each attri-bute By assuming εni to be iid extreme valuetype I McFadden showed that the choice probabil- We assume that there are Q discrete latentity from a total of J options considered is given by classes and D scale groupings but that class mem-

bership is hidden However we allow for theclasses to be determined using a set of observablecharacteristics or demographics If Hnqd is thePni = exp(βprimeXni)

sumJ

j=1

exp(βprimeXnj)

(1)prior probability for preference class q for individ-ual n and Gnd is the prior probability for scalegroup d the multinomial logit model suggested byHensher and Greene (2002) can be used to esti-It is important to recognize that the MNLmate each prior asmodel parameters in equation (1) βMNL are not the

true underlying β in Uni As is the case for allchoice models the estimates are confounded with

Hnqd = exp(θprimeqzn)Σqexp(θprimeqzn)

(3)σε the standard deviation of the error distribution(Swait amp Louviere 1993 Train 2003) Thus oneactually estimates β = λβ where λ is commonly Gnd = exp(γprimeqzn)

Σdexp(γprimeqzn)(4)

referred to as ldquothe scale parameterrdquo and in the caseof MNL the relationship is λ = π[sqrt(6)σε]

where the vector zi includes k = 1 K relevantTo avoid confounding preference estimatescovariates or sociodemographic characteristics ofwith choice consistency we use the SALCM toindividual n including a constant and θ and γ areestimate relative scale parameters jointly withthe respective vectors of parameter estimates ofpreference classes and associated preference pa-those zi covariates In turn the overall log likeli-rameters Indeed without estimating λ (ie re-hood isstricting λ = 1 for all classes) even in finite mix-

ture models where one allows the preferenceparameters β to vary between latent classes of

ln L = sumN

n=1

Pn = lnsumD

d=1

sumQ

q=1

GndHnqdJT

t=1

Pntiqd (5)consumers it is still possible to misinterpret dif-ferences in the true underlying β between classeswhich may be due only to differences in λ (Lou-

We followed the recommended estimation pro-viere 2001)cedure in Latent GOLD and used 20 randomizedsets of starting values to ensure an identified

Scale-Adjusted Latent Class Models (SALCM)global maximum We choose the number of pref-erence and scale classes by comparing the Bayes-Magidson and Vermunt (2007) proposed a

model in which the random utility of option j for ian Information Criterion (BIC) associated with

154 BURKE ET AL

each class solution which rewards improved fit attributes The collection of all possible combina-tions of these attributes comprises 82 times 49 times 215 =and penalizes additional parameters The posterior

probability that an individual belongs to prefer- 3774873600 combinations We used Street andBurgessrsquo (2007) design theory to construct an effi-ence class q and scale class d can be inferred from

Bayesrsquo theorem cient set of 128 pairs that were blocked into eightchoice sets per respondent This design approachis based on an orthogonal main effects plan (OMEP)Discrete Choice Experiment Surveythat allows estimation of a main effects only

We now describe a DCE that provides choice modeldata that can take advantage of the SALCM by The DCE was embedded in a larger online sur-decomposing choices into preference and scale vey that focused on questions about leisure activi-components and allows both to vary across the ties and museum visitation The first part of theunderlying latent segments The main objective of survey asked respondents about whether they hadthe DCE and associated survey was to quantify the engaged in various leisure activities in the previ-effects of tactical and strategic museum attributes ous year and whether they expected to do so inin order to better align museum offerings tactics the next 12 months Visiting a museum whetherand strategies with visitor choices We first de- past or prospective defined eligibility as a partici-scribe research that preceded the DCE undertaken pant in the study and the study context was re-to enhance external validity and provide strategic vealed to qualifying respondents The second partinformation to the six collaborating museums It is of the survey consisted of two museum exhibitionarguably the case that the most important stage in tasks In the first task respondents were asked toa DCE is the prior identification of salient attri- indicate if they would or would not go to see eachbutes and levels as the accuracy and validity of a of 37 named (and described) exhibitions The sec-DCE depends entirely on it Additionally one ond task focused on a subset of 9 of the 37 exhibi-must balance the number of attributes identified tions respondents were asked to rate each on aagainst the size and scope of the resulting choice five-category rating scale ranging from very unat-tasks for example including many factors leads tractive (one star) to very attractive (five stars)to large DCEs and more complex tasks that have Each subject was randomly assigned to a differentimplications for sample costs task complexity re- set of nine exhibitions based on a balanced incom-port delivery and presentation complexity Yet plete block design The latter design consists of 37omission of salient attributes may lead to model blocks of 9 exhibitions each exhibition appears 9misspecification and omitted variable bias as well times across the 37 sets (blocks) and co-appearsas concerns about external validity with each other exhibition twice

As previously noted we used qualitative inter- In the third section of the survey the DCE re-views with museum visitors (Burton et al 2008) to spondents received instructions and an example ofidentify a long list of potentially salient attributes a DCE scenario that explained how to answer thefor visitors and potential visitors We then con- questions Following that they received eight sce-ducted another round of research to measure the sa- narios or choice sets (see Fig 1) Each attributeliency of 64 attributes using a BestndashWorst Scaling was described in a ldquopop-up boxrdquo displayed whenapproach (Marley amp Louviere 2005) We used the the cursor was held above that feature Respon-results of the BestndashWorst Scaling exercise to dis- dents were instructed to consider star ratings ofcuss which of the attributes in fact could be ldquospecial exhibitionsrdquo as representing their own at-changed with representatives from partner muse- tractiveness ratings exemplified by their responsesums This resulted in a final 26 attributes and dis- to the prior exhibition rating task Respondentscussions with our museum partners allowed us to were asked to indicate their preferred museum of-specify a relevant set of levels for each attribute fer and whether they would actually visit that mu-

The final list of 26 attributes and associated seum for each choice set The survey concludedlevels are in Table 5 There were 2 eight-level at- with questions about preferred days for extended

opening hours and respondent sociodemographictributes 9 four-level attributes and 15 two-level

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 151

ments that matter when choosing a museum visit in Greece and Italy respectively to understandmuseum demandDCEs are widely used to represent everyday

choices such as products on shelves in supermar-kets transport mode options for commuting and

Accounting for Individual Differencesrecreational offerings like parks or fishing loca-

in Preference and Variabilitytions (eg Louviere Hensher amp Swait 2000)Choice response data from DCE surveys can be One of the key advantages in using DCMs re-analyzed with choice models to uncover system- volves around modeling decisions to visit muse-atic relationships between choices (dependent vari- ums and how these decisions may be affected byable) and changes in productservice offerings or changes to museum features Early DCMs suchldquoattributesrdquo (independent variables) as McFaddenrsquos (1974) conditional logit model

The choice modeling literature is now fairly restrict the preferences of individuals to be homo-mature and comprehensive reviews can be found geneous hence preferences estimated in thesein Train (2003) or Louviere et al (2000) More models are sample averages In turn this raisesrecently it has been recognized that more complex aggregation bias concerns (Hutchinson Kama-models are required that can take differences in kura amp Lynch 2000) for example when all fe-individual choice variability into account (eg males prefer museums with longer opening hoursAdamowicz et al 2008 Islam Louviere amp but all males are indifferent In such a case oneBurke 2007 Louviere amp Swait 2010) Thus we would underestimate the effects of extended open-make use of relatively new developments in finite ing hours on choices of both genders if they aremixture models known as latent class models We equally represented in a sampletake advantage of a new type of latent class model One way to account for such preference differ-known as a ldquoscale-adjusted latent class modelrdquo ences (or heterogeneity) is to include interactionsthat can simultaneously account for differences in between observable covariate measures (eg de-choice variability and preferences (Magidson amp mographics psychographics) and product featuresVermunt 2005) in DCMs (Gupta amp Chintagunta 1994) Auger

Use of DCMs in tourism and leisure research Burke Devinney and Louviere (2003) report sig-has grown in line with software developments and nificant individual differences in valuation of ethi-applications in other fields Applications of choice cal product features related to several demographicmodels in tourism have typically focused on mod- measures In other cases some of the heterogene-eling tourism demand (eg Huybers amp Bennett ity in preferences may be unrelated to observable2000 Morley 1994) DCMs have also been ap- covariate measures but instead be latent Severalplied to other topics such as resident trade-offs of methods have been proposed to deal with this in-tourism impacts (Lindberg Dellaert amp Rassing cluding random coefficient models (eg Hierar-1999) and understanding decisions relating to lei- chical Bayes and Mixed Logit models) and finitesure experiences (Oh Ditton Gentner amp Riech- mixture models (Kamakura amp Russell 1989ers 2005 Ready Epp amp Delavan 2005 Verma Wedel amp Kamakura 1998 for a recent review ofLouviere amp Burke 2006) This literature demon- models in this area see Train 2003) These meth-strates the flexibility of DCMs beyond traditional ods differ in several ways including whether theconsumption decisions There have been fewer ap- heterogeneity is modeled using several discreteplications of DCMs to cultural heritage research classes (as we do) or is continuous (ie using aBoxall Englin and Adamowicz (2003) and Rolfe distribution as in the case of models like theand Windle (2003) used DCMs to value aboriginal mixed logit) (see Andrews Ainslie amp Currimartifacts in Canada and Australia respectively 2002) In addition there are notable differences in

how researchers conceptualize and identify hetero-Maddison and Foster (2003) used DCEs to studymuseum congestion Apostolakis and Jaffry (2005) geneity with some using a priori theory-based

model development approaches while others useand Mazzanti (2003) applied DCMs to museums

152 BURKE ET AL

a data-driven approach where segments are identi- Differences in preferences for museums werereported by Apostolakis and Jaffry (2005) andfied in a posterior fashion (Dolnicar 2004)

Of course preference differences are not the Mazzanti (2003) who estimated interactions be-tween observable respondent characteristics andonly potential source of differences in individual

choices Individuals also may differ in choice con- museum attributes Colombino and Nese (2009)used a mixed logit model to account for unobserv-sistency For example younger visitors may be

less certain about how much they will enjoy a par- able preference heterogeneity but did not allowfor differences in choice consistency Thus priorticular special exhibition or which particular exhi-

bition they prefer due to less overall experience work suggests differences in choices of museumvisitors but does not pinpoint likely or potentialwith museums Failure to account for differences

in choice variability can lead to biased and mis- sources of heterogeneityOur research makes a methodological contribu-leading conclusions about preferences This is be-

cause each personrsquos mean preference parameter tion to the museum demand literature by using arecent development in latent class estimation thatestimates are inversely related to their choice con-

sistency (Louviere 2001) Yatchew and Griliches allows preference parameters to differ for discrete(1985) showed that error variance or heteroscedas- but unobserved (latent) classes of people whileticity ignored by researchers in a choice model also allowing the underlying variability of the ran-(discussed in the context of a probit model) will dom errors to differ between several discrete latentresult in biased parameter estimates rather than classes Until recently latent class models allowedsimple loss of efficiency as in the case of a linear preferences to differ from class to class but themodel Mroz and Zayats (2008) cautioned inter- error variances were identical over classes Errorpreting coefficient estimates of binary outcomes variance unaccountability confounds true underly-from different samples even if they use identical ing preference variability with individual differ-estimation procedure as these estimates depend ences This means that apparently significant dif-crucially on arbitrary normalization of variance ferences in preference parameters between latentVirtually all random coefficient models assume classes may be due to differences in underlyingthat choice consistency is constant across individ- scale (Louviere 2001)uals but if it is not constant estimates of such Magidson and Vermunt (2005) developed amodels can be highly biased Recent work by Fie- way to deal with the error variance-parameter esti-big Keane Louviere and Wasi (in press) and mate confound embedding the new approach inSalisbury and Feinberg (2010) suggest that it is commercially available software (a beta version ofhighly unlikely that choice consistency is constant Latent GOLD) The estimation approach allowsacross individuals one to relax the restriction that the scale parameter

Addressing the potential confound between is fixed across the whole population effectivelymodel estimates and choice consistency several decomposing class preferences and scale (or errornew types of DCMs have been developed to ac- variability) It also allows latent preference andcount for unobservable error variance heterogene- scale classes to be described as functions of demo-ity such as the Generalized Multinomial Logit graphic and other covariates We now discussmodel (Fiebig et al in press) models for single prior work on DCMs after which we introducepersons (Louviere et al 2008) models that de- and highlight the advantages of the SALCMcompose the random component (Burke amp Reit-zig 2007) and scale-adjusted latent class models Choice Models Derived(Magidson amp Vermunt 2005) Fiebig et al (in From Random Utility Theorypress) have shown unobserved scale heterogeneity

Using the conceptual framework of Randommodels outperform models that only account forUtility Theory (Manski 1977 Thurstone 1927)taste heterogeneity such as MNL and mixed logiteach individual may be described as holding latentin 10 out of 10 data sets Accordingly unobservedpreferences (utilities) associated with all choicescale heterogeneity was more important in datasets

that involve more complex choice objects options being considered Individuals maximize

SCALE-ADUSTED LATENT CLASS MODEL 153

utility by choosing their most preferred option La- individual n depends on the latent preference classq = 1 Q and the unobserved scale parametertent preferences are specified by two components

(observed and unobserved) Different discrete class λdd = 1 D and each individual is ex-pressed in terms of the likelihood that they belongchoice models are derived from different assump-

tions about the unobserved component the best to each class For identification one scale parame-ter is normalized to unity (see Swait amp Louviereknown of which is the conditional multinomial

logit model or MNL (McFadden 1974) MNL as- 1993) the remaining scale parameter estimates areratios of the reference scale class The probabilitysumes the unobserved component is independently

and identically distributed across choices and indi- of choice i by individual n in choice situation tconditioning on preference class q and scale factorviduals In this model the latent utility of option

i judged by respondent n is given by Uni = βprimeXni + class d isεni where Xni is the vector of attributes of option iand β is a vector of parameters representing the

P(ntiqd) = exp(λdβq primexnti)Σjexp(λdβq primexntj)

(2)preference (ie weight) associated with each attri-bute By assuming εni to be iid extreme valuetype I McFadden showed that the choice probabil- We assume that there are Q discrete latentity from a total of J options considered is given by classes and D scale groupings but that class mem-

bership is hidden However we allow for theclasses to be determined using a set of observablecharacteristics or demographics If Hnqd is thePni = exp(βprimeXni)

sumJ

j=1

exp(βprimeXnj)

(1)prior probability for preference class q for individ-ual n and Gnd is the prior probability for scalegroup d the multinomial logit model suggested byHensher and Greene (2002) can be used to esti-It is important to recognize that the MNLmate each prior asmodel parameters in equation (1) βMNL are not the

true underlying β in Uni As is the case for allchoice models the estimates are confounded with

Hnqd = exp(θprimeqzn)Σqexp(θprimeqzn)

(3)σε the standard deviation of the error distribution(Swait amp Louviere 1993 Train 2003) Thus oneactually estimates β = λβ where λ is commonly Gnd = exp(γprimeqzn)

Σdexp(γprimeqzn)(4)

referred to as ldquothe scale parameterrdquo and in the caseof MNL the relationship is λ = π[sqrt(6)σε]

where the vector zi includes k = 1 K relevantTo avoid confounding preference estimatescovariates or sociodemographic characteristics ofwith choice consistency we use the SALCM toindividual n including a constant and θ and γ areestimate relative scale parameters jointly withthe respective vectors of parameter estimates ofpreference classes and associated preference pa-those zi covariates In turn the overall log likeli-rameters Indeed without estimating λ (ie re-hood isstricting λ = 1 for all classes) even in finite mix-

ture models where one allows the preferenceparameters β to vary between latent classes of

ln L = sumN

n=1

Pn = lnsumD

d=1

sumQ

q=1

GndHnqdJT

t=1

Pntiqd (5)consumers it is still possible to misinterpret dif-ferences in the true underlying β between classeswhich may be due only to differences in λ (Lou-

We followed the recommended estimation pro-viere 2001)cedure in Latent GOLD and used 20 randomizedsets of starting values to ensure an identified

Scale-Adjusted Latent Class Models (SALCM)global maximum We choose the number of pref-erence and scale classes by comparing the Bayes-Magidson and Vermunt (2007) proposed a

model in which the random utility of option j for ian Information Criterion (BIC) associated with

154 BURKE ET AL

each class solution which rewards improved fit attributes The collection of all possible combina-tions of these attributes comprises 82 times 49 times 215 =and penalizes additional parameters The posterior

probability that an individual belongs to prefer- 3774873600 combinations We used Street andBurgessrsquo (2007) design theory to construct an effi-ence class q and scale class d can be inferred from

Bayesrsquo theorem cient set of 128 pairs that were blocked into eightchoice sets per respondent This design approachis based on an orthogonal main effects plan (OMEP)Discrete Choice Experiment Surveythat allows estimation of a main effects only

We now describe a DCE that provides choice modeldata that can take advantage of the SALCM by The DCE was embedded in a larger online sur-decomposing choices into preference and scale vey that focused on questions about leisure activi-components and allows both to vary across the ties and museum visitation The first part of theunderlying latent segments The main objective of survey asked respondents about whether they hadthe DCE and associated survey was to quantify the engaged in various leisure activities in the previ-effects of tactical and strategic museum attributes ous year and whether they expected to do so inin order to better align museum offerings tactics the next 12 months Visiting a museum whetherand strategies with visitor choices We first de- past or prospective defined eligibility as a partici-scribe research that preceded the DCE undertaken pant in the study and the study context was re-to enhance external validity and provide strategic vealed to qualifying respondents The second partinformation to the six collaborating museums It is of the survey consisted of two museum exhibitionarguably the case that the most important stage in tasks In the first task respondents were asked toa DCE is the prior identification of salient attri- indicate if they would or would not go to see eachbutes and levels as the accuracy and validity of a of 37 named (and described) exhibitions The sec-DCE depends entirely on it Additionally one ond task focused on a subset of 9 of the 37 exhibi-must balance the number of attributes identified tions respondents were asked to rate each on aagainst the size and scope of the resulting choice five-category rating scale ranging from very unat-tasks for example including many factors leads tractive (one star) to very attractive (five stars)to large DCEs and more complex tasks that have Each subject was randomly assigned to a differentimplications for sample costs task complexity re- set of nine exhibitions based on a balanced incom-port delivery and presentation complexity Yet plete block design The latter design consists of 37omission of salient attributes may lead to model blocks of 9 exhibitions each exhibition appears 9misspecification and omitted variable bias as well times across the 37 sets (blocks) and co-appearsas concerns about external validity with each other exhibition twice

As previously noted we used qualitative inter- In the third section of the survey the DCE re-views with museum visitors (Burton et al 2008) to spondents received instructions and an example ofidentify a long list of potentially salient attributes a DCE scenario that explained how to answer thefor visitors and potential visitors We then con- questions Following that they received eight sce-ducted another round of research to measure the sa- narios or choice sets (see Fig 1) Each attributeliency of 64 attributes using a BestndashWorst Scaling was described in a ldquopop-up boxrdquo displayed whenapproach (Marley amp Louviere 2005) We used the the cursor was held above that feature Respon-results of the BestndashWorst Scaling exercise to dis- dents were instructed to consider star ratings ofcuss which of the attributes in fact could be ldquospecial exhibitionsrdquo as representing their own at-changed with representatives from partner muse- tractiveness ratings exemplified by their responsesums This resulted in a final 26 attributes and dis- to the prior exhibition rating task Respondentscussions with our museum partners allowed us to were asked to indicate their preferred museum of-specify a relevant set of levels for each attribute fer and whether they would actually visit that mu-

The final list of 26 attributes and associated seum for each choice set The survey concludedlevels are in Table 5 There were 2 eight-level at- with questions about preferred days for extended

opening hours and respondent sociodemographictributes 9 four-level attributes and 15 two-level

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

152 BURKE ET AL

a data-driven approach where segments are identi- Differences in preferences for museums werereported by Apostolakis and Jaffry (2005) andfied in a posterior fashion (Dolnicar 2004)

Of course preference differences are not the Mazzanti (2003) who estimated interactions be-tween observable respondent characteristics andonly potential source of differences in individual

choices Individuals also may differ in choice con- museum attributes Colombino and Nese (2009)used a mixed logit model to account for unobserv-sistency For example younger visitors may be

less certain about how much they will enjoy a par- able preference heterogeneity but did not allowfor differences in choice consistency Thus priorticular special exhibition or which particular exhi-

bition they prefer due to less overall experience work suggests differences in choices of museumvisitors but does not pinpoint likely or potentialwith museums Failure to account for differences

in choice variability can lead to biased and mis- sources of heterogeneityOur research makes a methodological contribu-leading conclusions about preferences This is be-

cause each personrsquos mean preference parameter tion to the museum demand literature by using arecent development in latent class estimation thatestimates are inversely related to their choice con-

sistency (Louviere 2001) Yatchew and Griliches allows preference parameters to differ for discrete(1985) showed that error variance or heteroscedas- but unobserved (latent) classes of people whileticity ignored by researchers in a choice model also allowing the underlying variability of the ran-(discussed in the context of a probit model) will dom errors to differ between several discrete latentresult in biased parameter estimates rather than classes Until recently latent class models allowedsimple loss of efficiency as in the case of a linear preferences to differ from class to class but themodel Mroz and Zayats (2008) cautioned inter- error variances were identical over classes Errorpreting coefficient estimates of binary outcomes variance unaccountability confounds true underly-from different samples even if they use identical ing preference variability with individual differ-estimation procedure as these estimates depend ences This means that apparently significant dif-crucially on arbitrary normalization of variance ferences in preference parameters between latentVirtually all random coefficient models assume classes may be due to differences in underlyingthat choice consistency is constant across individ- scale (Louviere 2001)uals but if it is not constant estimates of such Magidson and Vermunt (2005) developed amodels can be highly biased Recent work by Fie- way to deal with the error variance-parameter esti-big Keane Louviere and Wasi (in press) and mate confound embedding the new approach inSalisbury and Feinberg (2010) suggest that it is commercially available software (a beta version ofhighly unlikely that choice consistency is constant Latent GOLD) The estimation approach allowsacross individuals one to relax the restriction that the scale parameter

Addressing the potential confound between is fixed across the whole population effectivelymodel estimates and choice consistency several decomposing class preferences and scale (or errornew types of DCMs have been developed to ac- variability) It also allows latent preference andcount for unobservable error variance heterogene- scale classes to be described as functions of demo-ity such as the Generalized Multinomial Logit graphic and other covariates We now discussmodel (Fiebig et al in press) models for single prior work on DCMs after which we introducepersons (Louviere et al 2008) models that de- and highlight the advantages of the SALCMcompose the random component (Burke amp Reit-zig 2007) and scale-adjusted latent class models Choice Models Derived(Magidson amp Vermunt 2005) Fiebig et al (in From Random Utility Theorypress) have shown unobserved scale heterogeneity

Using the conceptual framework of Randommodels outperform models that only account forUtility Theory (Manski 1977 Thurstone 1927)taste heterogeneity such as MNL and mixed logiteach individual may be described as holding latentin 10 out of 10 data sets Accordingly unobservedpreferences (utilities) associated with all choicescale heterogeneity was more important in datasets

that involve more complex choice objects options being considered Individuals maximize

SCALE-ADUSTED LATENT CLASS MODEL 153

utility by choosing their most preferred option La- individual n depends on the latent preference classq = 1 Q and the unobserved scale parametertent preferences are specified by two components

(observed and unobserved) Different discrete class λdd = 1 D and each individual is ex-pressed in terms of the likelihood that they belongchoice models are derived from different assump-

tions about the unobserved component the best to each class For identification one scale parame-ter is normalized to unity (see Swait amp Louviereknown of which is the conditional multinomial

logit model or MNL (McFadden 1974) MNL as- 1993) the remaining scale parameter estimates areratios of the reference scale class The probabilitysumes the unobserved component is independently

and identically distributed across choices and indi- of choice i by individual n in choice situation tconditioning on preference class q and scale factorviduals In this model the latent utility of option

i judged by respondent n is given by Uni = βprimeXni + class d isεni where Xni is the vector of attributes of option iand β is a vector of parameters representing the

P(ntiqd) = exp(λdβq primexnti)Σjexp(λdβq primexntj)

(2)preference (ie weight) associated with each attri-bute By assuming εni to be iid extreme valuetype I McFadden showed that the choice probabil- We assume that there are Q discrete latentity from a total of J options considered is given by classes and D scale groupings but that class mem-

bership is hidden However we allow for theclasses to be determined using a set of observablecharacteristics or demographics If Hnqd is thePni = exp(βprimeXni)

sumJ

j=1

exp(βprimeXnj)

(1)prior probability for preference class q for individ-ual n and Gnd is the prior probability for scalegroup d the multinomial logit model suggested byHensher and Greene (2002) can be used to esti-It is important to recognize that the MNLmate each prior asmodel parameters in equation (1) βMNL are not the

true underlying β in Uni As is the case for allchoice models the estimates are confounded with

Hnqd = exp(θprimeqzn)Σqexp(θprimeqzn)

(3)σε the standard deviation of the error distribution(Swait amp Louviere 1993 Train 2003) Thus oneactually estimates β = λβ where λ is commonly Gnd = exp(γprimeqzn)

Σdexp(γprimeqzn)(4)

referred to as ldquothe scale parameterrdquo and in the caseof MNL the relationship is λ = π[sqrt(6)σε]

where the vector zi includes k = 1 K relevantTo avoid confounding preference estimatescovariates or sociodemographic characteristics ofwith choice consistency we use the SALCM toindividual n including a constant and θ and γ areestimate relative scale parameters jointly withthe respective vectors of parameter estimates ofpreference classes and associated preference pa-those zi covariates In turn the overall log likeli-rameters Indeed without estimating λ (ie re-hood isstricting λ = 1 for all classes) even in finite mix-

ture models where one allows the preferenceparameters β to vary between latent classes of

ln L = sumN

n=1

Pn = lnsumD

d=1

sumQ

q=1

GndHnqdJT

t=1

Pntiqd (5)consumers it is still possible to misinterpret dif-ferences in the true underlying β between classeswhich may be due only to differences in λ (Lou-

We followed the recommended estimation pro-viere 2001)cedure in Latent GOLD and used 20 randomizedsets of starting values to ensure an identified

Scale-Adjusted Latent Class Models (SALCM)global maximum We choose the number of pref-erence and scale classes by comparing the Bayes-Magidson and Vermunt (2007) proposed a

model in which the random utility of option j for ian Information Criterion (BIC) associated with

154 BURKE ET AL

each class solution which rewards improved fit attributes The collection of all possible combina-tions of these attributes comprises 82 times 49 times 215 =and penalizes additional parameters The posterior

probability that an individual belongs to prefer- 3774873600 combinations We used Street andBurgessrsquo (2007) design theory to construct an effi-ence class q and scale class d can be inferred from

Bayesrsquo theorem cient set of 128 pairs that were blocked into eightchoice sets per respondent This design approachis based on an orthogonal main effects plan (OMEP)Discrete Choice Experiment Surveythat allows estimation of a main effects only

We now describe a DCE that provides choice modeldata that can take advantage of the SALCM by The DCE was embedded in a larger online sur-decomposing choices into preference and scale vey that focused on questions about leisure activi-components and allows both to vary across the ties and museum visitation The first part of theunderlying latent segments The main objective of survey asked respondents about whether they hadthe DCE and associated survey was to quantify the engaged in various leisure activities in the previ-effects of tactical and strategic museum attributes ous year and whether they expected to do so inin order to better align museum offerings tactics the next 12 months Visiting a museum whetherand strategies with visitor choices We first de- past or prospective defined eligibility as a partici-scribe research that preceded the DCE undertaken pant in the study and the study context was re-to enhance external validity and provide strategic vealed to qualifying respondents The second partinformation to the six collaborating museums It is of the survey consisted of two museum exhibitionarguably the case that the most important stage in tasks In the first task respondents were asked toa DCE is the prior identification of salient attri- indicate if they would or would not go to see eachbutes and levels as the accuracy and validity of a of 37 named (and described) exhibitions The sec-DCE depends entirely on it Additionally one ond task focused on a subset of 9 of the 37 exhibi-must balance the number of attributes identified tions respondents were asked to rate each on aagainst the size and scope of the resulting choice five-category rating scale ranging from very unat-tasks for example including many factors leads tractive (one star) to very attractive (five stars)to large DCEs and more complex tasks that have Each subject was randomly assigned to a differentimplications for sample costs task complexity re- set of nine exhibitions based on a balanced incom-port delivery and presentation complexity Yet plete block design The latter design consists of 37omission of salient attributes may lead to model blocks of 9 exhibitions each exhibition appears 9misspecification and omitted variable bias as well times across the 37 sets (blocks) and co-appearsas concerns about external validity with each other exhibition twice

As previously noted we used qualitative inter- In the third section of the survey the DCE re-views with museum visitors (Burton et al 2008) to spondents received instructions and an example ofidentify a long list of potentially salient attributes a DCE scenario that explained how to answer thefor visitors and potential visitors We then con- questions Following that they received eight sce-ducted another round of research to measure the sa- narios or choice sets (see Fig 1) Each attributeliency of 64 attributes using a BestndashWorst Scaling was described in a ldquopop-up boxrdquo displayed whenapproach (Marley amp Louviere 2005) We used the the cursor was held above that feature Respon-results of the BestndashWorst Scaling exercise to dis- dents were instructed to consider star ratings ofcuss which of the attributes in fact could be ldquospecial exhibitionsrdquo as representing their own at-changed with representatives from partner muse- tractiveness ratings exemplified by their responsesums This resulted in a final 26 attributes and dis- to the prior exhibition rating task Respondentscussions with our museum partners allowed us to were asked to indicate their preferred museum of-specify a relevant set of levels for each attribute fer and whether they would actually visit that mu-

The final list of 26 attributes and associated seum for each choice set The survey concludedlevels are in Table 5 There were 2 eight-level at- with questions about preferred days for extended

opening hours and respondent sociodemographictributes 9 four-level attributes and 15 two-level

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 153

utility by choosing their most preferred option La- individual n depends on the latent preference classq = 1 Q and the unobserved scale parametertent preferences are specified by two components

(observed and unobserved) Different discrete class λdd = 1 D and each individual is ex-pressed in terms of the likelihood that they belongchoice models are derived from different assump-

tions about the unobserved component the best to each class For identification one scale parame-ter is normalized to unity (see Swait amp Louviereknown of which is the conditional multinomial

logit model or MNL (McFadden 1974) MNL as- 1993) the remaining scale parameter estimates areratios of the reference scale class The probabilitysumes the unobserved component is independently

and identically distributed across choices and indi- of choice i by individual n in choice situation tconditioning on preference class q and scale factorviduals In this model the latent utility of option

i judged by respondent n is given by Uni = βprimeXni + class d isεni where Xni is the vector of attributes of option iand β is a vector of parameters representing the

P(ntiqd) = exp(λdβq primexnti)Σjexp(λdβq primexntj)

(2)preference (ie weight) associated with each attri-bute By assuming εni to be iid extreme valuetype I McFadden showed that the choice probabil- We assume that there are Q discrete latentity from a total of J options considered is given by classes and D scale groupings but that class mem-

bership is hidden However we allow for theclasses to be determined using a set of observablecharacteristics or demographics If Hnqd is thePni = exp(βprimeXni)

sumJ

j=1

exp(βprimeXnj)

(1)prior probability for preference class q for individ-ual n and Gnd is the prior probability for scalegroup d the multinomial logit model suggested byHensher and Greene (2002) can be used to esti-It is important to recognize that the MNLmate each prior asmodel parameters in equation (1) βMNL are not the

true underlying β in Uni As is the case for allchoice models the estimates are confounded with

Hnqd = exp(θprimeqzn)Σqexp(θprimeqzn)

(3)σε the standard deviation of the error distribution(Swait amp Louviere 1993 Train 2003) Thus oneactually estimates β = λβ where λ is commonly Gnd = exp(γprimeqzn)

Σdexp(γprimeqzn)(4)

referred to as ldquothe scale parameterrdquo and in the caseof MNL the relationship is λ = π[sqrt(6)σε]

where the vector zi includes k = 1 K relevantTo avoid confounding preference estimatescovariates or sociodemographic characteristics ofwith choice consistency we use the SALCM toindividual n including a constant and θ and γ areestimate relative scale parameters jointly withthe respective vectors of parameter estimates ofpreference classes and associated preference pa-those zi covariates In turn the overall log likeli-rameters Indeed without estimating λ (ie re-hood isstricting λ = 1 for all classes) even in finite mix-

ture models where one allows the preferenceparameters β to vary between latent classes of

ln L = sumN

n=1

Pn = lnsumD

d=1

sumQ

q=1

GndHnqdJT

t=1

Pntiqd (5)consumers it is still possible to misinterpret dif-ferences in the true underlying β between classeswhich may be due only to differences in λ (Lou-

We followed the recommended estimation pro-viere 2001)cedure in Latent GOLD and used 20 randomizedsets of starting values to ensure an identified

Scale-Adjusted Latent Class Models (SALCM)global maximum We choose the number of pref-erence and scale classes by comparing the Bayes-Magidson and Vermunt (2007) proposed a

model in which the random utility of option j for ian Information Criterion (BIC) associated with

154 BURKE ET AL

each class solution which rewards improved fit attributes The collection of all possible combina-tions of these attributes comprises 82 times 49 times 215 =and penalizes additional parameters The posterior

probability that an individual belongs to prefer- 3774873600 combinations We used Street andBurgessrsquo (2007) design theory to construct an effi-ence class q and scale class d can be inferred from

Bayesrsquo theorem cient set of 128 pairs that were blocked into eightchoice sets per respondent This design approachis based on an orthogonal main effects plan (OMEP)Discrete Choice Experiment Surveythat allows estimation of a main effects only

We now describe a DCE that provides choice modeldata that can take advantage of the SALCM by The DCE was embedded in a larger online sur-decomposing choices into preference and scale vey that focused on questions about leisure activi-components and allows both to vary across the ties and museum visitation The first part of theunderlying latent segments The main objective of survey asked respondents about whether they hadthe DCE and associated survey was to quantify the engaged in various leisure activities in the previ-effects of tactical and strategic museum attributes ous year and whether they expected to do so inin order to better align museum offerings tactics the next 12 months Visiting a museum whetherand strategies with visitor choices We first de- past or prospective defined eligibility as a partici-scribe research that preceded the DCE undertaken pant in the study and the study context was re-to enhance external validity and provide strategic vealed to qualifying respondents The second partinformation to the six collaborating museums It is of the survey consisted of two museum exhibitionarguably the case that the most important stage in tasks In the first task respondents were asked toa DCE is the prior identification of salient attri- indicate if they would or would not go to see eachbutes and levels as the accuracy and validity of a of 37 named (and described) exhibitions The sec-DCE depends entirely on it Additionally one ond task focused on a subset of 9 of the 37 exhibi-must balance the number of attributes identified tions respondents were asked to rate each on aagainst the size and scope of the resulting choice five-category rating scale ranging from very unat-tasks for example including many factors leads tractive (one star) to very attractive (five stars)to large DCEs and more complex tasks that have Each subject was randomly assigned to a differentimplications for sample costs task complexity re- set of nine exhibitions based on a balanced incom-port delivery and presentation complexity Yet plete block design The latter design consists of 37omission of salient attributes may lead to model blocks of 9 exhibitions each exhibition appears 9misspecification and omitted variable bias as well times across the 37 sets (blocks) and co-appearsas concerns about external validity with each other exhibition twice

As previously noted we used qualitative inter- In the third section of the survey the DCE re-views with museum visitors (Burton et al 2008) to spondents received instructions and an example ofidentify a long list of potentially salient attributes a DCE scenario that explained how to answer thefor visitors and potential visitors We then con- questions Following that they received eight sce-ducted another round of research to measure the sa- narios or choice sets (see Fig 1) Each attributeliency of 64 attributes using a BestndashWorst Scaling was described in a ldquopop-up boxrdquo displayed whenapproach (Marley amp Louviere 2005) We used the the cursor was held above that feature Respon-results of the BestndashWorst Scaling exercise to dis- dents were instructed to consider star ratings ofcuss which of the attributes in fact could be ldquospecial exhibitionsrdquo as representing their own at-changed with representatives from partner muse- tractiveness ratings exemplified by their responsesums This resulted in a final 26 attributes and dis- to the prior exhibition rating task Respondentscussions with our museum partners allowed us to were asked to indicate their preferred museum of-specify a relevant set of levels for each attribute fer and whether they would actually visit that mu-

The final list of 26 attributes and associated seum for each choice set The survey concludedlevels are in Table 5 There were 2 eight-level at- with questions about preferred days for extended

opening hours and respondent sociodemographictributes 9 four-level attributes and 15 two-level

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

154 BURKE ET AL

each class solution which rewards improved fit attributes The collection of all possible combina-tions of these attributes comprises 82 times 49 times 215 =and penalizes additional parameters The posterior

probability that an individual belongs to prefer- 3774873600 combinations We used Street andBurgessrsquo (2007) design theory to construct an effi-ence class q and scale class d can be inferred from

Bayesrsquo theorem cient set of 128 pairs that were blocked into eightchoice sets per respondent This design approachis based on an orthogonal main effects plan (OMEP)Discrete Choice Experiment Surveythat allows estimation of a main effects only

We now describe a DCE that provides choice modeldata that can take advantage of the SALCM by The DCE was embedded in a larger online sur-decomposing choices into preference and scale vey that focused on questions about leisure activi-components and allows both to vary across the ties and museum visitation The first part of theunderlying latent segments The main objective of survey asked respondents about whether they hadthe DCE and associated survey was to quantify the engaged in various leisure activities in the previ-effects of tactical and strategic museum attributes ous year and whether they expected to do so inin order to better align museum offerings tactics the next 12 months Visiting a museum whetherand strategies with visitor choices We first de- past or prospective defined eligibility as a partici-scribe research that preceded the DCE undertaken pant in the study and the study context was re-to enhance external validity and provide strategic vealed to qualifying respondents The second partinformation to the six collaborating museums It is of the survey consisted of two museum exhibitionarguably the case that the most important stage in tasks In the first task respondents were asked toa DCE is the prior identification of salient attri- indicate if they would or would not go to see eachbutes and levels as the accuracy and validity of a of 37 named (and described) exhibitions The sec-DCE depends entirely on it Additionally one ond task focused on a subset of 9 of the 37 exhibi-must balance the number of attributes identified tions respondents were asked to rate each on aagainst the size and scope of the resulting choice five-category rating scale ranging from very unat-tasks for example including many factors leads tractive (one star) to very attractive (five stars)to large DCEs and more complex tasks that have Each subject was randomly assigned to a differentimplications for sample costs task complexity re- set of nine exhibitions based on a balanced incom-port delivery and presentation complexity Yet plete block design The latter design consists of 37omission of salient attributes may lead to model blocks of 9 exhibitions each exhibition appears 9misspecification and omitted variable bias as well times across the 37 sets (blocks) and co-appearsas concerns about external validity with each other exhibition twice

As previously noted we used qualitative inter- In the third section of the survey the DCE re-views with museum visitors (Burton et al 2008) to spondents received instructions and an example ofidentify a long list of potentially salient attributes a DCE scenario that explained how to answer thefor visitors and potential visitors We then con- questions Following that they received eight sce-ducted another round of research to measure the sa- narios or choice sets (see Fig 1) Each attributeliency of 64 attributes using a BestndashWorst Scaling was described in a ldquopop-up boxrdquo displayed whenapproach (Marley amp Louviere 2005) We used the the cursor was held above that feature Respon-results of the BestndashWorst Scaling exercise to dis- dents were instructed to consider star ratings ofcuss which of the attributes in fact could be ldquospecial exhibitionsrdquo as representing their own at-changed with representatives from partner muse- tractiveness ratings exemplified by their responsesums This resulted in a final 26 attributes and dis- to the prior exhibition rating task Respondentscussions with our museum partners allowed us to were asked to indicate their preferred museum of-specify a relevant set of levels for each attribute fer and whether they would actually visit that mu-

The final list of 26 attributes and associated seum for each choice set The survey concludedlevels are in Table 5 There were 2 eight-level at- with questions about preferred days for extended

opening hours and respondent sociodemographictributes 9 four-level attributes and 15 two-level

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 155

Figure 1 Example choice set

characteristics consistent with Census Bureau ques- sponses and withdrawals the final sample for thistions study consisted of 3685 respondents Table 1 dis-

Survey responses were collected over 3 months plays the sociodemographic characteristics of the(beginning June 2007) using PureProfile an on- sampleline panel providing access to a sampling framebroadly representative of the Australian popula- Resultstion Visiting a museum (the screening criterion

We fitted a SALCM using sociodemographicsfor the survey) was ranked 12th in the past activi-types of activities engaged in by individuals andties list and fifth in the intended activities listrated preferences for different real exhibitions toThere were 643 of respondents who had visitedcapture preference heterogeneity (ie latent classes)a museum in the last year and 892 of respon-and variance heterogeneity (ie scale classes) Wedents reported that they intended to visit one in

the next year After accounting for incomplete re- iteratively estimated a number of models to ensure

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

156 BURKE ET AL

Table 1 BICs are 408886 (one class) 397109 (two class)Sociodemographic Profile of Respondents 38288 (three class) and 384534 (four class)

The BIC indicated two scale groupings withCharacteristic

λ1 = 1 (by normalization see Swait amp LouviereAge in years 1993) and λ2 = 0186 implying that the second

18ndash24 134 class shows 54 times (10186) more variability in25ndash29 116

underlying responses than the first group Vari-30ndash34 10735ndash39 120 ance heterogeneity is explained by age (γ = 0081040ndash45 134 p = 00175) indicating that younger people were46ndash49 75

more variable in their museum visitation choices50ndash55 11556ndash59 55 than older people Of the total DCE participants60ndash65 94 73 and 27 were most likely to be members of66ndash69 28

scale class one and two respectively70+ 20Hosehold income To minimize multicollinearity between mea-

$20000 or less 57 sures of activities engaged in during the past year$20001ndash$30000 61

and stated interest in various museum exhibitions$30001ndash$40000 69$40001ndash$50000 88 we specified the SALCM using constructs in-$50001ndash$60000 88 formed by the results of a factor analysis That is$60001ndash$70000 80

we used principal components analysis with Vari-$70001ndash$80000 89$80001ndash$90000 74 max rotation retaining components with eigenval-$90001ndash$100000 89 ues larger than 1 to derive a more parsimonious$100000+ 304

set of covariates This produced seven activity-HouseholdSingle living alone 142 related factors (Table 2) and seven exhibition-Single living in a shared house 104 related factors (Table 3) We labeled factors bySingle living with children 51

considering exhibitions or activities with the high-Single living with parents 91Marriedde facto with no children 205 est factor loadings and identifying a consistentMarriedde facto with children 287 theme In both cases the seven factors explainMarriedde facto children left home 120

about 60 of the variance in the dataEducationPrimary school 07 We estimated the preference classes using theHigh school 190 multinomial logit model of equation (3) and re-TAFEcollegediploma 302

tained three latent classes using the BIC criterionUndergraduate degree 305Postgraduate degree 196 The vector of covariates zi consisted of age house-

Stateterritory hold income education and the 14 rotated factorAustralian Capital Territory 19

scores discussed previously Table 4 shows theNew South Wales 336Northern Territory 05 characteristics that predict membership of theQueensland 191 three resulting preference classesSouth Australia 78

The largest visitor segment (61 of the sam-Tasmania 17Victoria 260 ple) is distinguished from other segments by beingWestern Australia 94 more likely to be lower to middle in income with

Areaa basic education and no tertiary qualificationsMetropolitan 788

Ruralregional 195 They prefer activities associated with betting andgambling are less likely to prefer self-expressiveactivities (eg playing music and writing) and areattracted to exhibitions with themes related tothat we had reliable estimates of the number of

classes and final model specification The final fashion and romance Museum managers whosponsored the research confirm that such peoplemodel (three preference classes) was selected on

the basis of minimizing BIC = minus 2L + mLog(n) comprise most of their museum visitors By sheerweight of numbers these people represent the Lifewhere L is the model log-likelihood with m pa-

rameters and n observations The corresponding Force that flows through museums

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 157

Table 2Factor Analysis Interest in Exhibitions

Factor Exhibitions Loading Onto Factor

Australia Since Australian Journeys How people traveled to Australia over time and traveled back The River LifeBritish Settlement on the Murray-Darling Great railway journeys of Australia Surviving Australia Celebrating the

unique diverse and surprising aspects Your Cityrsquos History The social cultural and physical charac-ter of your city over time

Wartime Great wartime inventions Lucky escapes Untold stories of wartime escapes In Flanders Fields 1917Fighting around Ypres in WWI Lawrence of Arabia and the Light Horse Objects and artifacts

Animals Bugs Alive Up close to live spiders and insects Jellyfish Beautiful awe inspiring and dangerous toknow Marine life Exploring our seasmdashdeep sea to shallow waters

Ancient Worlds Ancient civilizations Civilizations including Egypt Greece and China Gladiators Combining ar-chaeology popular culture and forensic science Dinosaurs of Gondwana Discoveries in AustraliaAntarctica and South America

Fashion and Tiffany Bejeweled A collection of superbly crafted jewelry and accessories Diana An exhibitionRomance celebrating the life of HRH Princess Diana Love and war Pinups sweethearts boyfriends amp love

letters during the war

The Modern Mind Drugs amp culture The use and (mis)use of drugs across cultures The Mind Exploring the workingsof the human mind and our emotions Modern Times Celebrates ldquomodernismrdquo

Indigenous Cultures Deeds Large Aboriginal canvases of the Papunya art movement of the 1970s Vaka Moana Voyagesof the ancestors exhibition of Maori seafaring and culture

The smallest of the three visitor groups has ad- trade or diploma qualifications than the othergroupsvanced education qualifications and enjoy self-

expressive pursuits They are less likely to engage The SALCM allows us to quantify preferencesfor museum features for each underlying class inin activities relating to betting gaming or home

entertainment They avoid fashion and romanti- the sample conditional on variance and preferenceheterogeneity These results are in Table 5 whichcally themed exhibitions but prefer indigenous

culture and Australian history This group of Edu- is structured to reflect the attribute groupings inthe survey (Fig 1) Significant results are denotedcated Thinkers comprises 14 of the sample

The third visitor group (comprising a quarter of by two symbols the first () denotes whether afactor is significantly different from zero (ie β nerespondents) were labeled as Wealthy At-Homes

They have higher incomes and engage in activi- 0) and therefore significant in explaining choicewithin a class the second (dagger) denotes whether aties at home including helping children with

homework and computer use They show little factor is significantly different (ie βq ne βl) fromthe preferences exhibited by the Life Force seg-interest in Australian-themed exhibitions They

completed high school but are less likely to hold ment

Table 3Factor Analysis Activities Engaged in Over Previous Year

Factor Activities Loading Onto Factor

Self-expressive studying writing playing musical instrumentsPassive Arts theatercultural events visiting art gallery gourmet food and wineDo-it-Yourself home renovationsimprovements gardening vehicle maintenanceGambling gamblingbettingGames (with children) board gamespuzzles helping children with homework computer gamesGrandchildren spending time with grandchildrenExercise personal exercise participation in organized sports

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

158 BURKE ET AL

Table 4Multinomial Logit Estimation of Preference Class Characteristics

1 Life Force 2 Educated Thinkers 3 Wealthy At-HomesCovariate (61 of Sample) (14 of Sample) (25 of Sample)

Intercept 0973 (0106) minus0700 (0143) minus0273 (0102)

Age minus0015 (0039) minus0077 (0052) minus0062 (0039)Quadratic minus0006 (0024) minus0052 (0029) minus0046 (0024)

Income minus0021 (0017) 0000 (0000) 0021 (0017)Quadratic minus0013 (0005) minus0005 (0007) 0018 (0006)

High school 0175 (0090) minus0367 (0138) 0191 (0094)TAFEcollegediploma 0181 (0070) minus0145 (0101) minus0036 (0078)Undergraduate degree minus0066 (0068) 0214 (0090) minus0148 (0074)Postgraduate degree minus0290 (0079) 0298 (0096) minus0008 (0081)

ActivitiesSelf-expressive minus0178 (0045) 0253 (0060) minus0076 (0047)Passive arts minus0014 (0043) minus0064 (0056) 0078 (0045)Do-it-Yourself minus0001 (0048) 0082 (0054) minus0082 (0042)Gambling 0098 (0043) minus0123 (0056) 0025 (0043)Games (with children) 0008 (0040) minus0158 (0054) 0149 (0043)Grandchildren 0025 (0045) 0040 (0058) minus0066 (0049)Exercise minus0012 (0042) minus0035 (0058) 0047 (0044)

ExhibitionsAustralia Since British Settlement 0028 (0043) 0102 (0059) minus0130 (0045)Wartime minus0009 (0039) 0016 (0052) minus0007 (0045)Animals 0040 (0040) minus0050 (0056) 0009 (0042)Ancient Worlds minus0010 (0045) 0075 (0057) minus0066 (0042)Fashion and Romance 0247 (0044) minus0324 (0063) 0077 (0046)The Modern Mind 0023 (0042) minus0053 (0057) 0029 (0043)Indigenous Cultures minus0057 (0040) 0092 (0053) minus0034 (0042)

Values are estimated β with SE in parenthesesCoefficient significant at 01 005 and 001 level

Special Exhibitions However the importance of highly rated exhibi-tions is less important among older visitors

Special Exhibitions were also described with anSpecial exhibitions were described by a one- toextra entrance fee The Life Force class prefers ex-five-star rating (five is ldquonot to be missedrdquo) andhibitions that are free or by donation preferencescorresponded to respondentsrsquo own attractivenesslinearly decline with increases in the adult entryrating in the prior exhibition rating task in the sur-price Both other classes also prefer free specialvey Across all segments attractiveness of specialexhibitions and entry by donation and this effectexhibitions is significant in explaining museumis significantly larger than for Life Force There isvisitation choices (Table 5a) However there aregreater sensitivity among Wealthy At-Homes anddifferences in each segment For example for theEducated Thinkers to entrance charges Despite be-Life Force class a highly rated exhibition is moreing labeled as Wealthy At-Homes results suggestlikely to produce visits from higher educated per-a high level of price sensitivity which may be be-sons but the opposite occurs for Educated Think-cause family museum visits can be expensive iners Highly rated exhibitions have a large impactrelation to a familyrsquos disposable incomeon visitation choices in the Wealthy At-Homes

class Thus managing themes of exhibitions shouldGeneral Entry Pricebe carefully considered for all classes especially

Wealthy At-Homes who have little interest in ex- The pattern of results in relation to general en-try prices is similar to special exhibition priceshibitions like ldquoAustralia Since British Settlementrdquo

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 159

Table 5Attributes and Associated Levels

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

a Special ExhibitionsStar rating

Intercept (aggregate) 0232 (0026) 0307 (0112) 1791 (0193)daggerdaggerdaggertimesAge (linear) minus0018 (0010) minus0041 (0041) minus0022 (0030)timesAge (quadratic) minus0012 (0007) minus0036 (0028) minus0055 (0019)daggerdaggertimesIncome (linear) minus0004 (0005) 0005 (0019) minus0006 (0015)timesIncome (quadratic) minus0003 (0002) 0002 (0007) minus0009 (0006)timesGender (male) 0031 (0013) minus0004 (0049) 0003 (0045)timesEducation (high school) minus0028 (0025) minus0074 (0078) 0023 (0070)timesEducation (college) minus0048 (0020) 0118 (0079)daggerdagger minus0059 (0069)timesEducation (undergraduate) 0009 (0023) 0131 (0074) minus0009 (0070)timesEducation (postgraduate) 0066 (0031) minus0175 (0080)daggerdaggerdagger 0044 (0077)

PriceFree 0214 (0066) 0745 (0245)daggerdaggerdagger 1063 (0246)daggerdaggerdaggerFree donation appreciated 0306 (0062) 0759 (0218)daggerdaggerdagger 1411 (0268)daggerdaggerdaggerAdult $6 children Free 0025 (0060) 0346 (0213) 0355 (0221)Adult $6 children $3 minus0010 (0061) minus0020 (0247) minus0333 (0217)Adult $6 children $3 family $15 minus0057 (0062) minus0369 (0229) minus0105 (0255)Adult $10 children Free minus0171 (0064) minus0614 (0211)daggerdagger minus0590 (0256)Adult $10 children $3 minus0197 (0062) minus0445 (0238) minus0980 (0211)daggerdaggerdaggerAdult $10 children $3 family $25 minus0110 (0064) minus0402 (0246) minus0821 (0230)daggerdaggerdagger

b General entry priceFree 0246 (0064) 0340 (0205) 1347 (0252)daggerdaggerdaggerFree donation appreciated 0479 (0068) 1015 (0260)daggerdagger 1536 (0292)daggerdaggerdaggerAdult $6 children Free 0188 (0062) minus0198 (0255) minus0369 (0226)daggerdaggerAdult $6 children $3 minus0019 (0060) minus0413 (0242) minus0588 (0247)daggerdaggerAdult $6 children $3 family $15 minus0070 (0060) minus0070 (0208) minus0429 (0198)daggerAdult $10 children Free minus0235 (0061) 0037 (0218) minus0381 (0199)Adult $10 children $3 minus0306 (0063) minus0089 (0248) minus0851 (0253)daggerdaggerAdult $10 children $3 family $25 minus0284 (0063) minus0622 (0244) minus0266 (0210)

c StaffingStaff knowledge

Museum facilities only minus0100 (0030) minus0062 (0100) minus0574 (0123)daggerdaggerdaggerFacilities amp surround area 0010 (0026) minus0048 (0103) 0113 (0107)Facilities and exhibitions minus0074 (0029) 0052 (0107) 0487 (0141)daggerdaggerdaggerFacilities amp expert on exhibitions 0163 (0032) 0058 (0091) minus0025 (0102)dagger

Location of staffLocated at entrances minus0088 (0016) 0092 (0050)daggerdaggerdagger 0293 (0067)daggerdaggerdaggerThroughout museum 0088 (0016) minus0092 (0050)daggerdaggerdagger minus0293 (0067)daggerdaggerdagger

Approach to inquiriesStaff approach you 0063 (0029) 0069 (0120) 0174 (0104)You seeklittle waiting time 0046 (0029) 0131 (0091) minus0021 (0100)You seekwait up to 5 minutes minus0048 (0029) minus0072 (0108) 0044 (0112)You seekwait up to 10 minutes minus0062 (0029) minus0128 (0109) minus0197 (0100)

d ToursSize of tour group

Exclusive or small groups 0076 (0014) 0075 (0055) 0195 (0059)daggerLarge groups (20ndash30 people) minus0076 (0014) minus0075 (0055) minus0195 (0059)

Frequency of toursDepart two times a day minus0036 (0014) minus0045 (0048) 0035 (0054)Depart four times a day 0036 (0014) 0045 (0048) minus0035 (0054)

Duration of tour15 minutes minus0105 (0030) minus0044 (0109) minus0161 (0113)30 minutes minus0005 (0032) 0013 (0102) 0020 (0112)1 hour 0101 (0030) 0099 (0111) minus0530 (0114)daggerdaggerdagger2 hours 0008 (0029) minus0068 (0117) 0670 (0140)daggerdaggerdagger

(continued)

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

160 BURKE ET AL

Table 5Continued

Class 1 Class 2 Class 3Life Force Educated Thinkers Wealthy At-Homes

Tour guideVolunteers minus0002 (0011) minus0103 (0051)dagger minus0056 (0047)Paid museum staff 0002 (0011) 0103 (0051)dagger 0056 (0047)

ExclusivitySee ldquobehind the scenesrdquo 0104 (0016) 0157 (0061) minus0244 (0068)daggerdaggerdaggerSee things seen by all visitors minus0104 (0016) minus0157 (0061) 0244 (0068)daggerdaggerdagger

InteractivityldquoTouchrdquo with supervision 0176 (0021) minus0035 (0054)daggerdaggerdagger minus0307 (0071)daggerdaggerdaggerCannot ldquotouch thingsrdquo minus0176 (0021) 0035 (0054)daggerdaggerdagger 0307 (0071)daggerdaggerdagger

Tour costFree 0360 (0037) 0275 (0111) 0152 (0096)daggerdagger$5 for tour 0177 (0031) 0088 (0101) minus0255 (0116)daggerdaggerdagger$10 minus0130 (0029) minus0131 (0104) minus0250 (0093)

$15 minus0407 (0042) minus0232 (0110) 0353 (0133)daggerdaggerdaggere Interactivity

Education focused minus0018 (0018) 1006 (0133)daggerdaggerdagger minus0153 (0063)daggerdaggerEntertainment focused 0018 (0018) minus1006 (0133)daggerdaggerdagger 0153 (0063)daggerdagger

f ChildrenShow bag

(Blank implies no show bag) minus0058 (0014) minus0053 (0048) minus0049 (0048)Free ldquoshow bagrdquo 0058 (0014) 0053 (0048) 0049 (0048)

Supervision(Blank No designated area) minus0047 (0013) minus0032 (0047) minus0188 (0047)daggerdaggerdaggerSupervised kids area 0047 (0013) 0032 (0047) 0188 (0047)daggerdaggerdagger

g Closing timesSummer

Closing at 6 pm minus0087 (0030) minus0344 (0122)daggerdagger 0318 (0114)daggerdaggerdagger8 pm minus0048 (0029) 0199 (0120)daggerdagger 0001 (0500)10 pm 0075 (0032) 0158 (0097) minus0373 (0111)daggerdaggerdaggerMidnight 0060 (0029) minus0013 (0108) 0054 (0110)

WinterClosing at 6 pm minus0004 (0027) minus0244 (0106)daggerdagger 0186 (0106)dagger8 pm minus0015 (0029) minus0040 (0112) minus0012 (0095)10 pm minus0002 (0027) 0188 (0104)dagger minus0145 (0104)Midnight 0022 (0030) 0096 (0095) minus0028 (0093)

h Special offersReturn offer

20 off next visit minus0092 (0030) minus0094 (0123) 0191 (0105)daggerdaggerdagger30 off next visit 0037 (0030) 0033 (0114) minus0214 (0095)daggerdagger50 off next visit 0107 (0031) 0109 (0116) minus0394 (0130)daggerdaggerdaggerBring back for free minus0052 (0031) minus0048 (0107) 0417 (0117)daggerdaggerdagger

Discount (50 off shopentry)a

Museum 1 0063 (0014) 0045 (0046) minus0064 (0043)daggerdaggerdaggerMuseum 2 0045 (0013) minus0009 (0049) 0100 (0049)Museum 3 0041 (0013) minus0096 (0049)daggerdaggerdagger 0163 (0049)daggerdaggerMuseum 4 0053 (0015) 0080 (0057) minus0232 (0055)daggerdaggerdaggerMuseum 5 0050 (0013) 0079 (0053) minus0113 (0051)daggerdaggerdaggerMuseum 6 0037 (0014) minus0024 (0052) 0035 (0043)

i Permanent displaysFrequency displays updated

Every 2 months 0342 (0036) 0075 (0124)daggerdagger 0584 (0118)daggerdaggerEvery 6 months 0268 (0035) minus0008 (0118)daggerdagger 0492 (0118)daggerEvery year 0025 (0028) 0105 (0095) minus0474 (0133)daggerdaggerdaggerRarely minus0635 (0052) minus0172 (0141)daggerdaggerdagger minus0603 (0116)

The values are estimated β with SE in parenthesesaThe museum discounts were modeled using six two-level attributes Unlike other tables values of the base level (no

discount) are not explicitly reported but given by the negative value listed (eg marginal utility of museum 1 with nodiscount is minus0063)

Coefficient significant at 01 05 and 001 level significant from Class 1 at dagger01 daggerdagger05 and daggerdaggerdagger001 level

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 161

with sensitivity to free entry significantly higher three classes Members of Wealthy At-Homes pre-fer tours of 2-hour duration that give access toamong Wealthy At-Homes followed by Educated

Thinkers (Table 5b) When fees apply the Life items accessible by all visitors and which areldquolook but donrsquot touchrdquo Wealthy-At-Homes areForce class is sensitive to any adult pricing at the

$10 level while at the adult price level of $6 any willing to pay $15 for such tours which may sug-gest that price signals quality or they hope thatfee for children has a strong negative impact In

contrast Educated Thinkers are largely indifferent higher prices lead to more exclusivity and beingwith people just like them There is a strong non-to fee options except for the highest fee family

bundled option Wealthy-At-Homes generally are linear relationship between fees and preferences inthis class These results suggest that it may not besensitive to both adult and children fees family

options are a potentially competitive pricing strat- straightforward for museums to set tour fees sim-ply based on length or tour exclusivityegy providing value for money to this class

In comparison with the senses of ldquotouchingrdquo(white gloved tours) and ldquoseeingrdquo (behind-the-Staffingscenes tours) the ldquolisteningrdquo aspect (the kind of

Three staff-related attributes were included intour guide) is less important (except in the case of

the choice scenarios staff knowledge their physi-the Educated Thinkers) This is an important in-

cal location in the museum and their method ofsight for museum managers who might contem-

interacting with visitors (Table 5c) The largest ofplate using paid museum staff instead of volun-

the three classes (Life Force) are attracted to staffteers in an effort to attract visitors Results suggest

that have expert knowledge of exhibitions whoa relatively minor effect in terms of higher visitor

are located throughout the museum and who ap-numbers

proach visitors rather than waiting for visitors toask for assistance Hence these visitors appear to

Interactivitybe keen to be assisted and directed throughouttheir museum experience Educated Thinkers on The interactivity attribute (Table 5e) applies tothe other hand show little sensitivity to any staff- two generic types of museum foci education anding characteristics one exception being a margin- entertainment The Life Force class is indifferentally significant preference for staff to be located to either type of interactivity The Educatedat entrances In contrast the Wealthy At-Homes Thinkers class on the other hand prefers educa-strongly prefer to see staff at the museum entrance tion-focused interactivity while the Wealthy At-and to provide them with general knowledge about Homes prefer an entertainment focusthe museum facilities and the exhibitions In otherwords they are satisfied to be given a general in- Childrentroduction to the museumrsquos collection and re-

Children-related museum attributes (Table 5f)sources but then perhaps obtain more specific in-matter to the Life Force class They prefer freeformation from place cards brochures or asmuseum show bags (with mementos and otherdiscussed shortly guided toursgoodies) and the ability to put children in a super-vised area The Wealthy At-Homes class has evenToursstronger supervision preferences Educated Think-

The DCE varied seven attributes related to mu- ers are indifferent to child-relevant attributesseum tours (Table 5d) The Life Force is attractedto frequent tours with limited numbers longer du-

Closing Timeration tours (preferably 1 hour) conducted by paidmuseum staff options to access things behind the Significant closing time effects (Table 5g) are

mainly associated with summer The Life Forcescenes and opportunities to touch items They pre-fer tours to be free with preferences declining ap- class prefers museums that open until very late

including midnight closing which implies thatproximately linearly in feesThere are some interesting differences in the they see a museum visit as an alternative to other

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

162 BURKE ET AL

after-hours leisure activities like theatrrs restau- tify the value of numerous tangible and intangiblefactors that might be used to affect visitor experi-rants or bars Educated Thinkers also prefer mu-

seums that are open until after 600 pm in summer ences and visitation choices A second objectivewas to show that one can simultaneously accountand prefer an 800 pm closing time Wealthy At-

Homes do not respond to closing times beyond 6 for differences in a sample of respondents (in thiscase museum visitors) in terms of their underlying00 pmpreferences and choice consistency

Special Offers To achieve our objectives we designed and im-plemented an online DCE that asked respondentsThe DCE varied two types of special offersto choose their preferred museum offering where(Table 5h) inducements for visitors to return tothe offerings were described by 26 attributesthe museum and cross-promotional offers to visitwhose levels were varied by an underlying experi-other museums The DCE used specific namedmental design Attributes and levels were informedmuseums (not identified here to honor the nondis-by qualitative research and a bestndashworst scalingclosure agreement with the collaborating partnerstudy was used to determine the relative impor-museums) The Life Force class responds posi-tance of 64 attributes The list of 64 attributes wastively to all cross-promotional inducements whereasfurther refined in consultation with six AustralianEducated Thinkers are indifferent to all offersmuseums who collaborated in the research pro-Wealthy At-Homes are more selective respondingcessboth positively and negatively to the offers We

The impacts of changes in attributes of museumnote in passing that difference in geographic loca-offerings and choices were estimated using a re-tions and associated travel costs of cross-visitationcent development in DCMs called a SALCM Thislikely explain some of these resultsnew approach allowed us to identify two latentThe Life Force class responds positively to ascale classes that differed in choice consistencydiscounted return visit whereas the EducatedPrevious approaches to latent class modeling haveThinkers are indifferent to this offer Discountingfocused on accounting for differences in prefer-more than 20 has a negative effect on prefer-ence only However because we have applied aences of Wealthy At-Homes but this class re-model that accounts for additional differences insponds positively to the option to return with achoice consistency the confounding between errorfamily member or friend at a later timevariability and parameter estimates has beenavoided In turn our conclusions and interpreta-Permanent Displaystion of any preference classes are free from any

The final attribute for a museum (re)visit was bias relating to ignoring differences in variancethe effect of temporal changes on the museum ex- With this in mind the model simultaneously iden-perience (Table 5i) Visitors may delay returning tified and estimated parameters for three latentuntil they think that the permanent exhibits and preference classes and estimated relationships be-displays change sufficiently which typically is as- tween several sociodemographic covariates activi-sociated with rotating collections and permanent ties and interests in exhibitions and class member-displays The Life Force class responds negatively ship Three underlying classes were profiled andto infrequent changes in permanent displays and described as being the Life Force of the museumpositively to more frequent rotations Wealthy At- visitors (around two thirds of visitors) EducatedHomes also respond positively to shorter rotations Thinkers and Wealthy At-Homes(less than 6 months) they dislike annual updating Our results suggest many ways that museumand rarely changing offerings In contrast Edu- managers can attract more visitors by changing of-cated Thinkers are indifferent to updating ferings associated with several attributeslevels

The model allows one to predict the impact thatSummary and Conclusions small or large changes (within the ranges of the

levels studied) will have both in terms of generalThe objective of this article was to introduceand describe a new way to understand and quan- demand and among each segment identified Such

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

SCALE-ADUSTED LATENT CLASS MODEL 163

predictions can be made over a number of differ- of audience heterogeneity The modeling approachdiscussed in this article provides a way to identifyent areas of operation (eg pricing tours) and the

impact can be compared in relative terms This is and target various visitor segmentsbecause the research method embedded the im-plicit trade-offs that respondents made between Acknowledgmentsvarious attributes in the context of the discrete

This research was supported by an Australianchoice scenariosResearch Council Linkage Project Grant The au-Not surprisingly all visitors are attracted tothors gratefully acknowledge the contributionsmuseums with special exhibitions that are compet-made by six major Australian museums that par-itively priced and rated as ldquonot to be missedrdquoticipated The authors would also like to thankPrice sensitivity also applies to general entrythree anonymous referees for their useful sugges-prices and confirms that museums that offer freetionsentry or entry ldquoby donationrdquo are the most pre-

ferred Tours are preferable if smaller in numberReferenceslengthier and conducted by paid museum staff

with some indifference to tour attributes (eg Adamowicz W Bunch D Cameron T Dellaert Bamong Educated Thinkers) Museums should cater Hanneman M Keane M Louviere J Meyer R

Steenburgh T amp Swait J (2008) Behavioral frontiersfor children with some classes responding stronglyin choice modeling Marketing Letters 19(3ndash4) 215ndashand positively to providing supervised child areas228Most visitors (Educated Thinkers an exception)

Andrews R L Ainslie A amp Currim I S (2002) Anprefer museums that change often offering new empirical comparison of logit choice models with dis-experiences and some level of updating to perma- crete versus continuous representations of heterogene-

ity Journal of Marketing Research 39 479ndash487nent displaysApostolakis A amp Jaffry S (2005) A choice modelingMuseums should consider the underlying pref-

application for Greek heritage attractions Journal oference classes in making strategic decisions aboutTravel Research 43 309ndash318

segmentation and targeting For example the Auger P Burke P Devinney T amp Louviere JJ (2003)classes differ in preferences for types of tours in- What will consumers pay for social product features

Journal of Business Ethics 42(3) 281ndash304cluding the aspects of what they will see and whatBiehal G amp Chakravarti D (1989) The effects of con-it costs Educated Thinkers see museums as educa-

current verbalization on choice processing Journal oftional opportunities but Wealthy At-Homes preferMarketing Research 26(2) 84ndash96

museums that offer entertaining experiences The Bitgood S (2006) An analysis of visitor circulationLife Force class is indifferent to types of interacti- Movement patterns and the general value principle Cu-vity which may be true indifference or additional rator 49(4) 463ndash475

Boorsma M (2006) A strategic logic for arts marketingpreference heterogeneity not accounted for in thisInternational Journal of Cultural Policy 12(1) 73ndash92class (Hutchinson et al 2000) Another class dif-

Boxall P Englin J amp Adamowicz W L (2003) Valu-ference is associated with extended opening hours ing aboriginal artefacts A combined revealed-statedmost prefer longer opening hours in summer but preference approach Journal of Environmental Eco-are indifferent towards longer winter hours Fi- nomics and Management 45 213ndash30

Burke P F amp Reitzig M (2007) Measuring patent as-nally incentives to return visits were valued bysessment qualitymdashAnalyzing the degree and kind ofthe Life Force class but have little effect on Edu-(in)consistency in patent officesrsquo decision making Re-cated Thinkers with the latter class also unaf-search Policy 36(9) 1404ndash30

fected by cross-museum promotional offers Burton C (2003) Scoping the challenge EntrepreneurialOur findings suggest many different ways that arts management in times of uncertainty Journal of

Arts Management Law and Society 3(3) 185ndash195a museum experience can be enjoyed An experi-Burton C amp Scott C (2003) Museums Challenges forence comprises various attributes that can be

the 21st century International Journal of Arts Manage-changed by museum managers to align to the pref-ment 5(2) 56ndash68

erences of their audiences However to make Burton C Louviere J amp Young L (2008) Retaining thegood decisions about the offerings and adopt bet- visitor enhancing the experience Identifying attributes

of choice in repeat museum visitation Internationalter positions museum managers need to be aware

164 BURKE ET AL

Journal of Nonprofit and Voluntary Sector Marketing planation for ldquoreversalrdquo effects in behavioral researchJournal of Consumer Research 27(3) 324ndash4414(1) 21ndash34

Huybers T amp Bennett J (2000) Impact of the environ-Cahill D J (1996) Pioneer advantage Is it real Does itment on holiday destination choices of prospective UKmatter Marketing Intelligence amp Planning 14(4) 5ndash8tourists Implications for Tropical North QueenslandCaldwell N amp Coshall J (2003) Touristsrsquo preferenceTourism Economics 6(1) 21ndash46structures for Londonrsquos Tate Modern Gallery The im-

Islam T Louviere J J amp Burke P (2007) Modelingplications for strategic marketing Journal of Travel ampthe effects of includingexcluding attributes in choiceTourism Marketing 14(2) 23ndash45experiments on systematic and random components In-Carson R T amp Groves T (2007) Incentive and informa-ternational Journal of Research in Marketing 24 289ndashtional properties of preference questions Environmental300and Resource Economics 37 181ndash210

Jaccard J Brinberg D amp Lee J A (1986) AssessingClopton S Stoddard J E amp Dave D (2006) Event pref-attribute importance A comparison of six methodserences among arts patrons Implications for marketJournal of Consumer Research 12(4) 463ndash468segmentation and arts management International Jour-

Johnson P amp Thomas B (1998) The economics of muse-nal of Arts Management 9(1) 48ndash77ums A research perspective Journal of Cultural Eco-Cohen S amp Orme B (2004) Whatrsquos your preferencesnomics 22 75ndash85Asking survey respondents about their preferences cre-

Joy A amp Sherry J F (2003) Speaking of art as embod-ates new scaling decisions Marketing Research 16(2)ied imagination A multisensory approach to under-32ndash37standing aesthetic experience Journal of Consumer Re-Cohen S amp Neira L (2003) Measuring preferences forsearch 30(2) 59ndash82product benefits across countries Overcoming scale us-

Kamakura W A amp Russell G J (1989) A probabilisticage bias with maximum difference scaling Paper pre-choice model for market segmentation and elasticitysented at the Latin American Conference of the Euro-structuring Journal of Marketing Research 26 379ndash90pean Society for Opinion and Marketing Research

Kelly L (2004) Evaluation research and communities ofPunta del Este Uruguay

practice Program evaluation in museums Archival Sci-Colombino U amp Nese A (2009) Preference heterogene-

ence 4 45ndash69ity in relation to museum services Tourism Economics

Kirchberg V (1996) Museum visitors and non-visitors in15(2) 381ndash395 Germany A representative survey Poetics 24 239ndash

Debenedetti S (2003) Investigating the role of compan- 258ions in the art museum experience International Jour- Lancaster K (1966) A new approach to consumer theorynal of Arts Management 5(3) 52ndash63 Journal of Political Economy 74(2) 132ndash157

Dolnicar S (2004) Beyond ldquocommonsense segmentation Lee H-K (2005) Rethinking arts marketing in a changingrdquo A systematics of segmentation approaches in tourism cultural policy context International Journal of Non-Journal of Travel Research 42(3) 244ndash250 profit and Voluntary Sector Marketing 10 151ndash164

Falk J H (2006) An identity-centred approach to under- Lindberg K Dellaert B amp Rassing C R (1999) Resi-standing museum learning Curator 49(2) 151ndash166 dent trade-offsmdashA choice modeling approach Annals

Fiebig D Keane M Louviere J J amp Wasi N (in of Tourism Research 26(3) 554ndash569press) The generalized multinomial logit model Mar- Louviere J J (1988) Analyzing decision makingmdashmetricketing Science conjoint analysis Newbury Park CA Sage

Goulding C (2000) The museum environment and the Louviere J J (2001) What if consumer experiments im-visitor experience European Journal of Marketing pact variances as well as means Response variability34(3ndash4) 261ndash272 as a behavioral phenomenon Journal of Consumer Re-

Gupta S amp Chintagunta P K (1994) On using demo- search 28(3) 506ndash11graphic variables to determine segment membership in Louviere J J Hensher D A amp Swait J D (2000)logit mixture models Journal of Marketing Research Stated choice methods Cambridge Cambridge Univer-31(1) 128ndash36 sity Press

Hensher D A amp Greene W (2002) Specification and Louviere J J amp Islam T (2008) A comparison of impor-estimation of the nested logit model Alternative nor- tance weights and willingness-to-pay measures derivedmalisations Transportation Research BmdashMethodologi- from choice-based conjoint constant sum scales andcal 36(1) 1ndash17 bestndashworst scaling Journal of Business Research 61(9)

Holden (2003) Capturing cultural value How culture has 903ndash911become a tool of government London Demos Louviere J J Street D Burgess L Wasi N Islam T

Hui C H amp Triandis H C (1989) Effects of culture and amp Marley A A J (2008) Modeling the choices ofresponse format on extreme response style Journal of individual decision-makers by combining efficientCross-Cultural Psychology 20(3) 296ndash309 choice experiment designs with extra preference infor-

Hutchinson J W Kamakura W A amp Lynch Jr J G mation The Journal of Choice Modeling 1(1) 128ndash163(2000) Unobserved heterogeneity as an alternative ex-

SCALE-ADUSTED LATENT CLASS MODEL 165

Louviere J J amp Swait J (2010) Commentary Discus- stant stochastic variance assumption in decision re-search Theory measurement and experimental testsion of lsquoAlleviating the constant stochastic variance as-

sumption in decision research Theory measurement Marketing Science 29(1) 1ndash17Santagata W amp Signorello G (2000) Contingent valua-and experimental testrsquo Marketing Science 29(1)

18ndash22 tion of a cultural public good and policy design Thecase of ldquoNapoli Musei Apertirdquo Journal of CulturalMaddison D amp Foster T (2003) Valuing congestion

costs in the British Museum Oxford Economic Papers Economics 24 181ndash204Slater A (2007) ldquoEscaping to the galleryrdquo Understanding55 173ndash90

Magidson J amp Vermunt J (2005) Technical guide to La- the motivations of visitors to galleries InternationalJournal of Nonprofit Voluntary Sector Marketing 12tent Gold Software 45 Statistical Innovations

Manski C (1977) The structure of random utility models 149ndash162Sheth J N Newman B I amp Gross B L (1991) WhyTheory and Decision 8 229ndash54

Marley A amp Louviere J J (2005) Some probabilistic we buy what we buy A theory of consumption valuesJournal of Business Research 22(2) 159ndash70models of best worst and best-worst choices Journal

of Mathematical Psychology 19 464ndash480 Street D amp Burgess L (2007) The construction of opti-mal stated choice experiments Theory and methodsMazzanti M (2003) Valuing cultural heritage in a multi-

attribute framework microeconomic perspectives and Hoboken NJ John Wiley and SonsSwait J amp Louviere J J (1993) The role of the scalepolicy implications Journal of Socio-Economics 32

549ndash569 parameter in the estimation and comparison of multi-nomial logit models Journal of Marketing ResearchMcFadden D (1974) Conditional logit analysis of qualita-

tive choice behavior In P Zarembka (Ed) Frontiers 30(3) 305ndash14Throsby D (2001) Economics and culture Cambridgein econometrics (pp 105ndash142) New York Academic

Press Cambridge University PressThyne M (2000) The importance of values research forMitchell R C amp Carson R T (1989) Using surveys to

value public goods The contingent valuation method nonprofit organisations The motivation-based values ofmuseum visitors International Journal of NonprofitWashington DC Resources for the Future

Morley C L (1994) Experimental destination choice and Voluntary Sector Marketing 6(2) 116ndash130Thurstone L L (1927) A law of comparative judgmentanalysis Annals of Tourism Research 21(4) 780ndash791

Mroz A T amp Zayats V Y (2008) Arbitrarily normal- Psychological Review 34 273ndash86Train K (2003) Discrete choice methods with simulationized coefficients information sets and false reports of

lsquobiasesrsquo in binary outcome models Review of Econom- Cambridge Cambridge University PressVerma R Louviere J amp Burke P (2006) Using a mar-ics and Statistics 90(3) 406ndash413

Noonan D (2003) Contingent valuation and cultural re- ket-utility-based approach to designing public servicesA case illustration from United States Forest Servicesources A meta-analytic review of the literature Jour-

nal of Cultural Economics 27 159ndash176 Journal of Operations Management 24(4) 407ndash16Wedel M amp Kamakura W A (1998) Marketing seg-Oh C-O Ditton R Gentner B amp Riechers R (2005)

A stated preference choice approach to understanding mentation Conceptual and methodological founda-tions Boston Kluwerangler preferences for management options Human Di-

mensions of Wildlife 10(3) 173ndash86 Weil S E (2002) Making museums matter WashingtonDC Smithsonian Institute PressReady R Epp D amp Delavan W (2005) A comparison

of revealed stated and actual behavior in response to a Wiggins J (2004) Motivation ability and opportunity toparticipate A reconceptualization of the RAND modelchange in fishing quality Human Dimensions of Wild-

life 10(1) 39ndash52 of audience development International Journal of ArtsManagement 7(1) 22ndash33Rolfe J amp Windle J (2003) Valuing the protection of

aboriginal cultural heritage sites Economic Record 79 Yatchew A amp Griliches Z (1985) Specification error inprobit models The Review of Economics and Statisticss85ndashs95

Salisbury L amp Feinberg F (2010) Alleviating the con- 67(1) 134ndash139