Sie sind auf Seite 1von 21

This article was originally published in a journal published by Elsevier, and the attached copy is provided by Elsevier for

the authors benet and for the benet of the authors institution, for non-commercial research and educational use including without limitation use in instruction at your institution, sending it to specic colleagues that you know, and providing a copy to your institutions administrator. All other uses, reproduction and distribution, including without limitation commercial reprints, selling or licensing copies or access, or posting on open internet sites, your personal or institutions website or repository, are prohibited. For exceptions, permission may be sought for such use through Elseviers permissions site at: http://www.elsevier.com/locate/permissionusematerial

Annals of Tourism Research, Vol. 34, No. 3, pp. 610629, 2007 0160-7383/$ - see front matter 2007 Elsevier Ltd. All rights reserved. Printed in Great Britain

www.elsevier.com/locate/atoures

doi:10.1016/j.annals.2007.01.007

Antonia Correia University of the Algarve, Portugal Carlos M. Santos University of the Azores, Portugal Carlos Pestana Barros Technical University of Lisbon, Portugal
Abstract: The main aim of this paper is to estimate a mixed logit model in order to analyze the decision of Portuguese tourists traveling to Latin America in charter ights. The model estimates the probability of choosing this type of destination, by means of explanatory variables, disentangling homogeneous and heterogeneous variables. The conclusion is that strategies by clusters are needed in order to account for the heterogeneity identied. Moreover, the combination of all former variables, which contribute substantially to a better understanding of tourism choice behavior, with the latter variables is needed to dene a comprehensive strategy. Keywords: Latin America, mixed logit model, destination choice, market segmentation. 2007 Elsevier Ltd. All rights reserved. Resume: Tourisme en Amerique latine : une analyse de choix. Le but principal de ce travail est destimer un modele logit multinomial an danalyser la decision des touristes portugais ` qui voyagent vers lAmerique latine en vol charter. Le modele estime la probabilite de choisir ` ce genre de destination par moyen de variables explicatives en denouant les variables homogenes et heterogenes. On conclut que, pour rendre compte de lheterogeneite identi` ` ee, il faut des strategies de groupement. En outre, il faut combiner toutes les variables homogenes, qui contribuent considerablement a une meilleure comprehension du comport` ` ement des choix de tourisme, et les variables heterogenes an de denir une strategie ` comprehensive. Mots-cles: Amerique Latine, modele logit multinomial, choix de destination, ` segmentation du marche. 2007 Elsevier Ltd. All rights reserved.

INTRODUCTION

Antonia Correia is Assistant Professor of tourism economics at the University of the Algarve (Campus de Gambelas, 8005-139 Faro, Portugal. Email <acorreia@ualg.pt>). Carlos Santos is Professor of tourism economics and Carlos Pestana Barros is Assistant Professor of microeconomics. Their research interests, respectively, are on consumer behavior; island tourism; and hotel and travel agency productivity. 610

Au

This paper analyzes the probability that a Portuguese tourist traveling on Air Luxor will choose Latin America as a vacation destination. This is a privately-owned Portuguese airline which, besides scheduled ights in Europe, to the tropical Atlantic island of Sao Tome, and to Guinea-Bissau in West Africa, also has a separate charter operation. These ights operate on routes between Portugal and various Latin

th o

r's

pe

rs

on

al

co

py

TOURISM IN LATIN AMERICA A Choice Analysis

CORREIA, SANTOS AND BARROS

611

American destinations, particularly in Brazil and Mexico, and to certain African markets. For the purpose of this study, a detailed survey was conducted in 2004 on board Air Luxor charter ights, with a sample of 1,097 individuals. Given human complexity and diversity, it was anticipated that research into what led tourists to choose a certain destination would generate different perceptions meriting further analysis in order to develop more effective marketing strategies. Previous ndings suggest that a destinations image is what drives the behavioral intentions (Gnoth 1997). Moreover, perceptions of the destination are a function of internal and external motivations that are determined by a set of psychological and socioeconomic variables and expectations (Baloglu and McCleary 1999). This paper estimates an econometric model testing the multi-attribute models of Fishbein and Ajzen (1980) and the destination choice model of Mathieson and Wall (1984), together with the evidence from the literature review. According to Sirakaya and Woodside (2005), the development of tourist-decision models that incorporate a wide array of variables and ll the gap between behavioral and choice-set approaches by means of probability theory are required in order to obtain deeper insights into the decision process. This paper contributes to the literature in the two ways highlighted by Sirakaya and Woodside (2005). First, it demonstrates the usefulness in tourism research of the logit regression technique, more precisely, the mixed logit regression, to dene clusters statistically signicant in a sample, overcoming the limitations of traditional cluster analysis. Second, it offers an integrated approach to understanding how the different attributes related to tourist proles, touristic awareness, destination characteristics, and holiday features affect the probability of choice. Third, this paper uses a mixed logit to determine the probability that Portuguese tourists will choose Latin America rather than Africa for their vacation. This is similar to Nicolau and Mas (2005, 2006), who analyzed the determinants of the tourism decisions made by Spanish families, and Kemperman, Ponje and Timmermans (2005), who analyzed preferences for urban parks. The mixed logit is considered to be the most promising state-of-theart discrete choice model currently available to analyze questionnaire data (Hensher and Greene 2003). Its advantage over alternative others derives from two improvements. First, it allows for the error term to combine different statistical distributions, which is an improvement on alternative specications that rely on one specic distribution. Second, it allows for random preference parameters (parameters that describe characteristics not linked to observed characteristics), whereas the traditional logit permits preference variations related to observed characteristics. This procedure may be more effective in achieving results than the traditional procedure, which considers all the individuals as homogeneous. Therefore, the aim of the paper is to estimate a mixed logit model disentangling heterogeneous and homogeneous explanatory variables to identify those variables which can be managed in a homogeneous way and those that must be managed by clusters.

Au

th o

r's

pe

rs

on

al

co

py

612

CHOICE ANALYSIS

TOURISM CHOICE ANALYSIS Discrete choice models in tourism are based on multivariate attributes of decision (Lancaster 1966), with the data obtained by means of questionnaires, and involve the use of either binomial or multinomial logit models. The nested logit (McFadden 1974, 1981) has proven very successful in identifying the determinants of choice in tourism models (Domencich and McFadden 1975; McFadden 1981). Examples of studies that use the binomial logit include those carried out by Fleischer and Pizam (2002), who determine the constraints of senior Israeli tourists, and De la Vina and Ford (2001), who describe the demographic and vacation factors of potential cruise passengers, based on a sample of individuals who previously requested tourism information. Costa and Manente (1995) investigate the characteristics of tourists to the city of Venice with respect to their origin and socioeconomic prole, their preferences and their holiday decisions. Sheldon (1995) examines the travel incentive among US corporations, and Stynes and Peterson (1984) propose the logit model to estimate recreational choices. Kockelman and Krishnamurthy (2004) propose a microeconomically rigorous method to characterize demand across a great variety of choice dimensions, including vacation generation. Their study applies a multivariate negative binomial model for demand functions derived from an underlying trans-logarithmic indirect utility one. Both time and money budgets were incorporated into the structure via an effective or generalized budget constraint. A nested logit model of vacation mode and destination was used to calculate the effective prices for each journey purpose via nested logsum expressions. Time-of-day and other choice attributes could be added. The binomial logit model is also used by Hay and McConnel (1979), Miller and Hay (1981), and Walsh, McKean and Hof (1992) who present a logit relationship between several explanatory variables and the propensity to take a vacation, concluding that there exists among the variables a positive marginal effect up to a certain point and a negative beyond a certain point. Multinomial and nested logit models can be developed from a rigorous behavioral theory of utility maximization. However, standard multinomial ones require discrimination of choices, such as peak vs. off-peak seasons, and travel vs. no travel vacation. This causes a loss of cardinality and continuity, which determine many choices, such as time of day and the number of journeys made (Kockelman and Krishnamurthy 2004). Examples of studies which use the multinomial logit model are those by Luzar, Diagne, EcGan and Henning (1998), who investigate the socioeconomic and psychographic factors that inuence Louisiana tourists decisions to participate in nature-based tourism, and Morley (1994), who assesses the independent effects of tourists from Kuala Lumpur to Australia with eight contexts, in which the prices of the Sydney alternative vary. The mixed logit model adopted in this paper is more exible than the standard one or the multinomial and nested types, since it allows for heteroskedasticity in the error term, depending on the explanatory

Au

th o

r's

pe

rs

on

al

co

py

CORREIA, SANTOS AND BARROS

613

variables. The mixed model has been used previously in various elds of research, such as terrorism (Barros and Proenca 2005), agriculture (Alfnes 2004), transportation (Bath 1996; Brownstone, Bunch and Train 2000; Brownstone and Train 1999), recreation (Train 1998), energy (Revelt and Train 1998), and marketing (Bonnet and Simioni 2001). In tourism, it appears that the mixed logit has only been applied by Nicolau and Mas (2005, 2006) and Kemperman et al (2005). Theoretical Argument

Most of the studies conducted on behavioral intentions are based on Fishbein and Ajzens theory of reasoned action (Baker and Crompton 2000). This constitutes an extension of Fishbeins (1967) original model. The purchase intention is a function of the attitude towards behavior, as well as social norms. The former consists of perceived expectations in terms of the possibility of adopting a certain form of behavior, and the evaluation of how the consumer feels towards it. The subjective norms constitute a measure of inuence of the social environment on behavior. Its evaluation is carried out in terms of the motivation of the consumer to adopt the attitude that social groups consider as being correct. The behavioral intention is dened as a subjective probability to either adopt, or not adopt, a certain form of behavior. The models studied, which are commonly referred to in the specialist literature as multi-attribute, since they consider that a product possesses several self-compensating attributes (compensatory), nd their basis in the value-expectancy theory (Baker and Crompton 2000). This denes expectation as the probability of a certain attitude leading to positive or negative benets, thus allowing the isolation of determining factors of behavior and, furthermore, specifying how expectation and value can be combined in order to make choices. There are numerous possibilities for combining expectation with value. The expectancy value theory is a way of measuring the subjective utility (Edwards 1954). The decisionmaking process of Mathieson and Wall (1984) is also adopted as a theoretical reference of this paper. According to their model, tourism is the temporary movement of people to destinations outside their normal place of work and residence, the activities undertaken during their stays, and the facilities created to cater for their needs. The decision process is explained by a multidisciplinary perspective represented by four factors: tourist proles, vacation features, touristic awareness, and destination characteristics. These factors are represented in the variables applied in the present research with the aim of explaining the probability of a Portuguese consumer choosing a Latin American destination. The Mixed Model An example is the tourist ying with Air Luxor from Portugal to foreign destinations, with a choice between Latin America and Africa. The

Au

th o

r's

pe

rs

on

al

co

py

614

CHOICE ANALYSIS

Ben-Akiva and Lerman (1985), McFadden (1974), and Train (1986) used the logit model to relate the probability of making a choice to a set of variables reecting decisionmaker preferences. The logit model uses as an endogenous variable a dummy variable dened as zero and one, which denes the dichotomic choices made. Multiple choice is allowed with the multinomial logit model. The dichotomic choice variable is regressed in the exogenous variables that explain the selection. Despite its popularity, the logit model exhibits the so-called independence from irrelevant alternatives property, according to which the choice probabilities are independent, which is a hypothesis that may be untenable. In most applications, a better specication for the latent variable is y b0 x i mi ei where mi is a random term that accounts for i unobserved heterogeneity among tourists, with zero mean and possibly heteroskedasticity with distribution that frequently depends on the explanatory variables and underlying parameters. In the case of the present study, it represents characteristics that inuence the probability of choosing Latin America, which are not measured or observed, nor are there measurement errors in the variables. Unobserved heterogeneity has been a subject of concern and analysis in Chesher (1984), Chesher and Santos-Silva (2002), Gonul and Srinivasan (1993), and McFadden and Train (2000). Heterogeneous behavior is commonly observed in individuals; not to take it into account is likely to lead to inconsistent parameter estimates or more importantly, inconsistent tted choice probabilities. This paper has adopted the random coefcients (also named mixed logit) of McFadden and Train (2000). The main characteristics of the random or mixed logit are the following. First, it allows the parameter associated with each variable to vary randomly across individuals, often referred to as unobserved heterogeneity (Revelt and Train 1998). In the present study, this implies that different individuals can have different preferences relative to the probability of choosing Latin America.

Au

th o

r's

pe

rs

on

al

e b xi P yi 1jx i P b; x i 0 1 e b xi

co

main goal is to determine the probability of choosing to vacation in Latin America rather than Africa, in view of a number of given characteristics, denoted by the vector xi. A binary random variable yi, that veries yi = 1 if the tourist chooses Latin America and yi = 0 otherwise, yields the aimed probability of P(yi = 1jxi). Models to determine the probability of an event given a set of characteristics, xi, can be derived based on a latent variable, y , that is not observed and veries i y b0 x i ei , where b is a vector of unknown parameters, and ei is i an unobserved random variable allowing that individuals with the same characteristics xi have different outcomes. To use the general framework of binary dependent models, it is supposed that yi = 1 if y > 0 i and yi = 0 otherwise. Then P(yi = 1jxi) = P(ei > b 0 xi) and the desired probability depends on the statistical assumptions regarding ei. When ei is independent and identically distributed as extreme value type I, the above probability is given by the highly popular logit model:

py

CORREIA, SANTOS AND BARROS

615

Research Design

The tourist who travels on an Air Luxor charter ight, for example, has chosen Latin America as a destination rather than Africa. This declaration is based on the utility that the tourist receives from choosing a Latin American spot compared with the utility received from an alternative location. The choice underlying the utility denes the following null hypotheses: Hypothesis 1 (Budget): A tourist who chooses a Latin American destination does not have a high vacation budget. This is a traditional hypothesis in demand models, in which the price and budget constraints determine the destination choice. While De la Vina and Ford (2001) use price to explain the choice, others consider income (Hay and McConnel 1979; Nicolau and Mas 2005). The budget is more appropri ate for evaluating the cost, since it includes different bundles of prices of products and services. Hypothesis 2 (Destination Attributes): A tourist who chooses to visit a Latin American country is not attracted by destination attributes such as nature, culture, nightlife, climate, gastronomy, distance and ethnic values. This is also a traditional hypothesis in tourism demand models (Costa and Manente 1995). Goodrich (1980), Woodside and Lysonski (1989) and Woodside and MacDonald (1994) argued that a

Au

th o

r's

pe

rs

Second, the variance of unobserved individual specic parameters induces correlation across the alternatives in the choice. Hence, the mixed logit does not exhibit the so-called independence from irrelevant alternatives property that is implied by the assumption of independently- and identically-distributed error terms and the independence of choice probabilities from the characteristics of the other alternatives in the choice set. A pattern of correlation in unobserved factors inuencing the preferences for Latin America can be obtained. Third, correlation in the unobserved factor over time can also be examined. Therefore, the mixed logit model allows for efcient estimation when there are choices available to the individuals and heterogeneity is present. However, the model requires sophisticated calculations, namely, Bayesian procedures and some assumptions in the form of the distribution of mi in order to estimate the parameters and the choice probabilities consistently, provided that the distribution assumptions are correctly stated. The procedure is derived from the fact that the integral has no closed form and thus exact maximum likelihood is not possible. The integral has an expected value and can hence be approximated by simulation, the simulated log-likelihood function being maximized (Hajivassiliou and Ruud 1994; Revelt and Train 1998). Train (2003) suggests the use of the Halton sequence of draws instead of random draws as the simulation procedure. The assumptions on the distribution of the error terms can be standard normal distribution, the triangular distribution, or the standard uniform distribution.

on

al

co

py

616

CHOICE ANALYSIS

destinations image and its choice is inuenced by associated attributes. The latter are ex-ante the main drivers of tourism, since at least a part of the attraction of exotic destinations is the curiosity aroused by the people and their environment. Hypothesis 3 (Sociodemographic Characteristics): A tourist who chooses a Latin American vacation is not characterized by sociodemographic characteristics such as age, higher level of education, and higher social class, and is likely to be married and have children. This is also a traditional hypothesis of demand models based on questionnaire data (Goodall and Ashworth 1988; Weaver, McCleary, Lepisto and Damonte 1994; Woodside and Lysonski 1989; Zimmer, Brayley and Searle 1995). This hypothesis validates the socioeconomic characterization of the questionnaire respondents. Hypothesis 4 (Information): A tourist who chooses a Latin American destination is not one of those who have been adequately informed. The perception formation derives from information previously obtained, which helps the consumer to clarify and to evaluate alternative options (Um and Crompton 1990). The information processed and stored from an image can be the combination between the cognitive and the affective component. Therefore, it is the tourists perceptions which inuence his/her behavior, rather than the real characteristics of the destination (Dann 1981; Pearce 1982). Hypothesis 5 (Hotel Attributes and Facilities): A tourist who chooses a Latin American destination is not one attracted by the destinations accommodation attributes and facilities. Shih (1986) and Taplin and McGinley (2000) identied hotel and climate as important attributes of destination choice. Gibson and Gray (2005) identied a cluster in which accommodation, restaurants, and environmental quality were the main attributes, naming this cluster the discerning tourist type. Hypothesis 6 (Return): A tourist who chooses a Latin American country is not one who has previously visited the region, and would consequently be returning to their preferred destination. Festinger (1954) stated that satisfaction in relation to the destination inuences future choices. Beerli and Martn (2004) established that sun-and-sand desti nations with a good image enjoy a high level of repeaters. Kozak (2001) demonstrated that overall satisfaction and the number of previous visits considerably inuence the intention to return, especially in mature destinations. Kozak (2003) also concluded that destination attributes inuence future behavioral intentions and satisfaction. Hypothesis 7 (Temporal Constraints): A tourist who chooses a Latin American destination is not among those without temporal constraints and therefore, the less time he/she has committed, the higher the probability of choosing the region. Crompton (1979) suggests that the choice of destination may be framed in the contextual setting as a function among money, time, experience, and image. Additionally, recent trends in tourism behavior reveal the last-minute buyer, who looks for bargain prices, as a signicant player. Hypothesis 8 (Frequent Flyer): One who is a frequent yer does not tend to be a tourist in Latin America. This is the rst time that tourism frequency has been related to the probability of choice. According to

Au

th o

r's

pe

rs

on

al

co

py

CORREIA, SANTOS AND BARROS

617

b13 education b14 brochure b15 advertizing b16 mail b17 typelodge b18 regime b19 previous b20 tempmot b21 treserve b22 ftrip b23 exp ectations ei

pe

v i b0 b1 budget b2 culture b3 nightlife b4 c lim e b5 gastro b6 dis tan ce b7 ethnic b8 nature b9 age b10 gender b11 civilstate b12 class

rs

where N() is the normal distribution, and:

on

Pearce (1982), the individual develops a tourism career, similar to the working career. This implies that the individual starts by traveling abroad to the nearest and cheapest destinations. Then as he/she climbs the occupational ladder, the individual starts become progressively more demanding with regard to vacations. Hypothesis 9 (Expectations): Expectations are not a major attribute in destination choices. They appear in the literature as the probability that a certain attitude will lead to positive or negative benets, thus allowing the isolation of determining factors of behavior and, furthermore, specifying how expectations and values can be combined in order for choices to be made (Fishbein 1967). Further research into expectations is presented by Dalen (1989), Iso-Ahola and Mannell (1987), Muller (1991), Pitts and Woodside (1986), and Shih (1986). The hypotheses outlined above were tested by means of the adoption of the mixed logit representation that assumes that the probability of choosing Latin America instead of Africa can be described by a cumulative logit-probability function of the exogenous variables Xi, Prob (choice/type). On the basis of this denition, the above-mentioned probability was estimated for choice i as: Z 1 PrChoice i jv i P b; v i N b23 jl23 ; r23 db5 2

al

co

py

The variables were selected from a questionnaire distributed among this population. The vi was measured by the probability that the respondents will declare that they chose Latin America instead of Africa (Yes = 1, No = 0), labeled (latinamerica) and measured Xi as observed characteristics. This signies that the same questionnaire was completed by passengers bound for Latin America and Africa, and transported by the same airline. The questions are displayed in the variables listed in Eq. (3). Some variables were discarded from the model because they were statistically insignicant or because too few passengers replied to the question, so that too many zeros were displayed. To estimate the mixed logit model, the paper used a simulator type in the Gauss programming language (Train 2006). For comparative purposes, the standard logit model is estimated using the TSP software. Survey Methods The empirical study was carried out by means of the previouslymentioned questionnaire in August and September 2004, which was

Au

th o

r's

618

CHOICE ANALYSIS

presented to a stratied, random sample of Air Luxor passengers, with the central aim of determining their reasons for choosing a specic destination. The denition of the sample was based on the number of charter departures from Lisbon Airport in 2004, the countrys only international airport from which there are charter ights operated by Portuguese airlines. The departures totaled 114 in August and 81 in September. Charter ights represented 39.25% of total ights and 49.65% of total passengers, amounting to 36,652 persons. Air Luxor was the leading Portuguese charter company, ying tourists on behalf of almost all the tour operators, with a market share of 36.68% of total charters, corresponding to 13,080 passengers. The sample was randomly stratied by destination. On the chosen ights, the attendants approached the passenger seated in the randomly-chosen seat with the questionnaire, after the meals had been served. Because of budgetary restrictions and the limited time available, it was decided to collect data from 1,097 questionnaires. As each instrument distributed cost a set amount and the funds available were limited, it was necessary to restrict the survey to the maximum allowed by the funds available, corresponding to 8.3% of the Air Luxor charter passengers. The questionnaires returned totaled 792, from which 442 completed ones were retained for the present analysis, (a response rate of 40.3%). This corresponds to a sampling error of 2.7% with a condence interval of 95%. The remaining questionnaires not considered for the present research were discarded because of uncompleted elds and incorrect completion. The respondents who had chosen Brazil accounted for 19% of the total who traveled to Latin America. To ensure the validity and reliability of the data, several steps were taken. First, the point of departure was a questionnaire already applied in tourism (Correia, Barros and Silvestre 2007) which was adapted for the present purpose, ensuring that prior research in the eld was considered and face validity established. Second, all relevant literature was taken into consideration. Third, the questionnaire was pre-tested on students of tourism economics at the University of the Algarve. Following the administration of the nal survey, a stratied random subset of 50 respondents was contacted by phone a second time to check if any problem persisted, but none were revealed. These procedures ensure the validity of the questionnaire, meaning that it computes what it was intended to measure. Fourth, the questionnaire opted for a random sample, with a response rate of 40.3%, which was considered an acceptable sample of respondents (Dillman 1978). This procedure ensures the generalizability of the data, meaning that the ndings are applicable to a more general population. Fifth, the reliability of the data was examined, analyzing it extensively with alternative methods and reaching the same conclusions (Correia, Barros and Silvestre 2007). The extensive examination of the survey validity, reliability, and generalizability leads to the inference that there is nothing in the evaluation to suggest that it is either invalid or unreliable. The 40.3% response rate raises the question of non-response, for which a testing procedure based on Dillman (1978) was adopted. A rst test for this problem involved dening a subsample random

Au

th o

r's

pe

rs

on

al

co

py

CORREIA, SANTOS AND BARROS

619

Study Results The ndings point to a signicant correlation between the probability of choosing Latin America with the exogenous variables (Table 2). Taking the evidence of the heterogeneity in the sample, the mixed logit model is chosen as the adequate representation of the data. The dependent variable is a dummy variable: To choose a Latin American country for vacation=1, and 0 elsewhere. Statistical tests that detect mis-specication in the standard logit detect evidence of heterogeneity in the model, concluding that the mixed logit better ts the data. What do the results mean for the hypotheses proposed? The null of Hypothesis 1 is rejected, due to the statistically positive but insignicant results for budget. This signies that it is around US$1,600 for an average stay of 8 days. For comparative purposes, this value is similar to the monthly wage received by recent graduates from Portuguese universities, and is thus considered a low wage by the standards of older graduates. Given their average age, the average cost does not exhaust the family budget. Moreover, this result conrms previous research relative to this issue (Hay and McConnel 1979; Nicolau and Mas 2005),

Au

th o

r's

pe

choice group of respondents, contacting them by phone again and testing the answers. These maintained the declared values, ensuring the accuracy of the responses. A second test involved contacting a random subsample of those who had not answered, to understand the reasons for their nonresponse. As a result of this, several explanations were given: the individuals declared secrecy policy, which is a common obstacle to questionnaires; a lack of time available to complete the questionnaire during the ight; and saturation associated with completing too many questionnaires. From these three reasons, it can be asserted that the nonresponses have the same characteristics as those who did respond, establishing the representativeness of the questionnaires that were fully completed. The rate response does not differ signicantly from the sample in terms of age (chi-square = 8.53, p = 0.05), or gender (chi-square=7.55, p = 0.05). Hence, it can be asserted that the 442 individuals who completed the questionnaire are representative of Air Luxor charter passengers and thus of Portuguese tourists to Latin America, since they mostly travel on these ights. Their general characteristics were that 52% were male, with an average age of 33. This prole leads to an overall denition of the respondent as male, early middle-aged and middleclass, with a family that includes one child (Table 1). The objective was to evaluate the Portuguese tourists who choose to travel to Latin America rather than to Africa. To pursue this objective, the questionnaire was structured according to this table. The survey has three types of variables: dichotomous, continuous, and qualitative (7-item Likert scale). The set of explanatory variables considered in this study attempts to capture the key determinants of the decision process, based on the theoretical framework and the literature review.

rs

on

al

co

py

620

CHOICE ANALYSIS

Table 1. Characterization of the Variables


Variable Description Dependent Variable Dummy variable Mina Maxb Mean Std. Dev

Latin America Budget

.490

Budget Hypothesis Travel budget (1-lower of US$1268,5; 5-equal or higher 1 than US$3171,3) Destination Attributes Hypothesis Culture What was the importance of cultural attractions in your 1 decision? (1-without importance; 7-extremely important) Nightlife What was the importance of night-life in your decision? 1 (1-without importance; 7-extremely important) Clime What was the importance of climate in your decision? 1 (1-without importance; 7-extremely important) Gastro What was the importance of gastronomy in your decision? 1 (1-without importance; 7-extremely important) Distance What was the importance of the distance from home in 1 your decision? (1-without importance; 7-extremely important) Ethnic What was the importance of ethnic composition in your 1 decision? (1-without importance; 7-extremely important) Nature What was the importance of natural environment in your 1 decision? (1-without importance; 7-extremely important) Sociodemographic Characteristics Hypothesis Age The age 19 Gender The gender (0-male, 1female) 0 Civilstate Marital status (1-single; 2-married; 3-with children) 1 Class Social class (1-lower; 2-middle; 3-upper-middle) 1 Education Education (number of years of education) (4-primary 4 school , 24PhD) Information Hypothesis Brochure What was the importance of brochures and tourist guides 1 in your decision? (1-without importance; 7-extremely important) Advertizing What was the importance of advertizing in your decision? 1 (1-without importance; 7-extremely important) Mail What was the importance of direct mail in your decision? 1 (1-without importance; 7-extremely important) Accommodation Attributes and Facilities Hypothesis Typelodge What type of lodging did you book? (1-ve star hotel; 1 2-four star hotel; 3-three star hotel; 4two star hotel, 5-city aparthotel; 6-beach aparthotel; 7others) Regime Which accommodation regime did you choose? (1-all 1 meals included; 2-half. board; 3-only breakfast) Returning Hypothesis Previous Have you visited this destination before? 0 Temporal Constraints Hypothesis 1 Tempmot Why did you choose this period for a vacation? (1-school vacations; 2-work vacations not related to school vacations; 3-other type of vacations not related to school neither to work; 4-price incentive; 5-weather; 6-other) Treserve How long before the vacation did you make the booking? 1 (1-less than 15 days; 2-15 days or more, but less than a month; 3-1 month or more; 4-three months or more) Frequent Flier Hypothesis Ftrip Frequency of traveling 1 Expectations Hypothesis Expectations Were your expectations fullled in relation to this 1 destination? (1-worse than expected, 7-much better than expected)

1.981

1.140

7 7 7 7

co
7 7 7 69 1 3 3 24 7 7 7 7 3 1 6 4 4 7

al

on

pe

rs

r's

th o

Au

Min: Minimum; b Max: Maximum.

py
5.454 4.789 5.923 5.470 4.685 4.542 5.776 33.271 .520 2.260 2.124 14.013 4.766 4.038 4.106 3.104 1.794 .083 4.266 2.036 2.457 5.305

1.465

1.567

1.368 1.532 1.703 1.728 1.379

10.342 1.451 .911 3.461

1.524 1.483 1.535

1.745 .753

1.972

1.018

.796 1.152

CORREIA, SANTOS AND BARROS

621

Table 2. Parameter Estimates and t-statistics


Standard Logit Variables Homogeneous Variables Intercept Budget Culture Nightlife Clime Gastro Distance Ethnic Age Gender Civilstate Class Education Brochure Advertising Mail Regime Previous Tempmot Treserve Ftrip Expectations Coefcients t-stat

co
Coefcients 4.874 .036 .437 .160 .281 .458 .187 .128 .016 .050 .252 .280 .003 .163 .294 .327 2.475 1.139 .318 .515 .636 .176 .525 .118 442 148.253

Heterogeneous Variables Nature Typelodge Observations LogLikelihood Specication tests: RESET on stand. Logit HAL on standard Logit LR: standard vs Mixed

th o

4.607 .070 .477 .289 .296 .605 .207 .023 .020 .160 .314 .304 .004 .149 .223 .333 2.640 1.042 .329 .597 .591 .175

2.548a .481 2.932a 1.852 1.653 3.373b 1.753 1.284 1.158 .486 2.586a 1.644 .915 1.251 1.583 2.450a 9.665b 1.686 4.212b 3.739b 2.634a 1.325 3.270 2.782

r's

pe

rs

on
p-val. .01000 .00120 .00015

al

.597 .125 442 146.231 Statistic 3.132 22.842 42.31

Au

Dependent variable: To choose a Latin American country for vacation = 1, and 0 elsewhere. ^ RESET: Detects mis-specication in the logit model and was performed with b0 x 2 ; HAL: i Detects evidence of heterogeneity in the logit model depending on nature and typelodge. LR: Likelihood ratio test. a Statistically signicant at 5%; b Statistically signicant at 1%.

py
Mixed Logit

which found that prices did not have a dissuasive effect on destination choice. Therefore, the insignicant statistic means that the afuence of the Portuguese who choose Latin America for their vacations is dispersed, with an average value of $1,600. The null of Hypothesis 2 is rejected, as nightlife and gastro are statistically signicant positive motives, conrming previous research (Costa and Manente 1995). However, other destination characteristics are statistically negative (culture, distance and ethnic). Portuguese tourists who take Latin American vacations could be described as hedonists. This is because

t-stat

5.330b .489 4.978b 2.409a 3.280b 5.536b 3.318b 2.385a 2.767a .316 4.888b 2.850a .175 2.715a 4.116b 4.561b 13.993b 3.749b 7.818b 6.219b 5.526b 2.955b 5.463b 3.128b

622

CHOICE ANALYSIS

they value nightlife and gastro positively. However, they do not value the clime, since they inhabit a relatively warm country, nor do they value the destination for its culture or ethnic. The distance does not attract them, since it is the most inconvenient part of traveling. An average ight from Portugal to South America takes approximately 10 hours. The conclusion relative to this issue is that they are primarily motivated by gastronomy and nightlife. The characterization is of a hedonistic, little-sophisticated tourist, with no great interest in cultural ethnic themes, which might be expected to form part of the attraction of exotic vacations. With regard to the direct impact of socioeconomic variables on the probability of choosing Latin America, the null of Hypothesis 3 is rejected, since the social class and civilstate are positive and statistically signicant, conrming previous research (Nicolau and Mas 2005). How ever, gender is positive but not statistically signicant, and education is negative and statistically insignicant. Therefore, it emerges from the research that the average respondents are married men and women, subjectively belonging to a middle social class, with an average education level (14 years on average). An explanation for this result stems from the Portuguese education system. However, since schooling is insignicant, it follows that the average is not signicant. Relative to the subjective social class, this is a common result in questionnaire data. It means that there are signicant subjective feelings in this sample relative to this issue. However, this result is not supported by the education level. Moreover, in the Portuguese market, human capital is an important element in earnings which dene a subjective social class. Gender is also statistically insignicant, which means that the sample is comprised of couples. Turning now to information issues, the null of Hypothesis 4 is rejected, since advertising is positively and statistically signicant, conrming previous research (Um and Crompton 1990). Thus, advertising is fundamental to attracting Portuguese consumers to Latin America. However, brochure and mail are negative and statistically signicant. This nding suggests that decisions on destinations are taken on the basis of advertising. Brochures and mails are used as a way to collect information that helps the consumer to decide. Unfortunately, it is also evident that the consumers do not seem to nd the information required in the brochures. This nding is intriguing, since it is evident that prospective tourists collect all the brochures available from the travel agencies. A possible explanation is that they may simply like to browse through them for the photographs, but do not pay much or any attention to the text. Moreover, mails are not used for this purpose. This conclusion does not point to great scope yet for the expansion of internet sales in respect of this type of consumer, booking the best deal on web. Focusing on accommodation conditions at the destination, the null of Hypothesis 5 is rejected, since the lodging conditions are important, as was veried by the regime and by the variable, typelodge. However, typelodge randomly signies that opinion relative to the type of lodging is mixed. This in turn implies that preferences for high-quality hotels are a dominant characteristic of the sample, meaning that they seem to conform to the discerning tourist type, identied by Gibson and Gray (2005). However, the heterogeneous nature of typelodge signies that the average

Au

th o

r's

pe

rs

on

al

co

py

CORREIA, SANTOS AND BARROS

623

Au

tourist prefers high-quality hotels, whereas a proportion of others favor other types of lodging, for example, economic hotels. Thus, different selling strategies have to be dened by the travel agencies for the different consumer types, according to this variable. However, regime is homogeneous and the average tourist prefers either full-board or half-board accommodation. So, it is concluded that accommodation is a fundamental issue for the Portuguese taking vacations in Latin America. The analysis of the relationship of the previous experience with the probability of choosing the destination results in the rejection of the null of Hypothesis 6. The returning effect is important in this market, given that previous is found to be positive and statistically signicant (Aguilo Alegre and Sard 2005). Therefore, a Latin American destina, tion is attractive to Portuguese tourists. While the attraction of Brazil is understandable, given the common language and sociocultural ties between the two countries, the attractions of other destinations such as Mexico are not so clear. Addressing the role of temporal restrictions on touristic choice, the null of Hypothesis 7 is rejected, since these restrictions are important, as the variable tempmot and treserve are positively and statistically significant. Therefore, Portuguese tourists bound for Latin American resorts book their holidays relatively early, and do it during their winter/ Christmas vacations (treserve) and are mainly motivated to make their choice on the basis of price discounts. The analysis of the frequent-yer effect on tourist behavior enables the acceptance of the null Hypothesis 8, since ftrip is negatively and statistically signicant. Therefore, those who choose Latin America are not frequent iers, but rather those who travel simply for the purpose of a vacation. Finally, analyzing the role of expectations in destination choice, the null of Hypothesis 9 is rejected, since expectations are high among the Portuguese traveling to Latin America (Iso-Ahola and Mannell 1987). Such expectations are common in choice analysis. Prior to the selection, the consumer establishes an expected result of the choice, which may or may not be fullled, depending on the reality of the vacation. In the present case, expectations were clearly met, since the data was compiled during the tourists return ights. The overall outcome is that Portuguese who decide to vacation in Latin America are relatively young, not very highly educated, with young families, who consider themselves as belonging to the middle class, with a relatively low vacation budget (between $1,268.5 and $1,902.8 per adult on average). These are people who do not travel frequently, choosing to take their vacations on the basis of the discount prices offered, who plan their trip in advance and choose the halfboard hotel option, or alternatively, full-board.

CONCLUSION In this paper, a mixed logit model was used to gain a deeper insight into what leads Portuguese tourists to exotic destinations in Latin

th o

r's

pe

rs

on

al

co

py

624

CHOICE ANALYSIS

America, validating Fishbein and Ajzens (1980) reasoned action theory and Mathieson and Walls (1984) tourist decisionmaking theory. The variables that affect the probability that Portuguese individuals will choose to vacation in Latin America are destination attributes, travel arrangements, and tourist socioeconomic proles, validating previous research (Nicolau and Mas 2005). The identied factors leverage the natural afnity and shared characteristics of the population with Latin American culture and societies. Furthermore, with particular regard to Brazil, the common language is clearly a factor that has a powerful pull effect. The Spanish language of Mexico is a positive factor in attracting Portuguese tourists, given the proximity between the two languages. Lastly, the feel-good and self-esteem factors that are related to a holiday in another continent should not be discounted as leveraging factors. The paper results reject the null of the majority of the hypotheses and the determinants of the probability of choosing Latin America are identied. This result has implications for an active marketing strategy, namely, for the denition of the selling of Latin America to Portuguese families. Single homogeneous marketing strategies do not correspond to the preferences of all individuals in a sample. Diversied strategies are needed to comply with the heterogeneity identied in the sample data. However, as the sample is homogeneous for some variable, and thus different strategies will have common components, so it will be easily implemented. Segmentation should be carried out based in the heterogeneous variables identied. With regard to the managerial implications of this research, it is of value for tour operators to have a deeper insight into the driving motives behind Portuguese tourists choice of Latin America, and, with greater awareness of what these consumers require from a vacation, they can pay particular attention to the statistically signicant variables estimated in the model when dening their supply. In addition, they can investigate the reasons behind some of the negative coefcients, such as those related to brochures. A possible explanation is that this advertisement medium lacks dynamism and is not sufciently attractive or informative. Alternatively, it is possible that respondents to the questionnaire undervalued the role of the brochures in their decision process. Furthermore, the industry should tailor special programs for the distinct clusters related to nature and types of lodging. On the basis of this papers ndings, homogeneous tourism offers will not satisfy demand with regard to these two heterogeneous variables. In relation to other papers analyzing destination choice, this paper cannot be directly compared with the studies referred to, since it uses an innovative approach that enables the identication of heterogeneous variables in the data permitting the statistical denition of clusters. However, this investigation is directly comparable with Nicolau and Mas (2005), who use a multinomial logit model with random coef cients to identify the probability of choosing ve vacation alternatives being driven by personal restrictions, and sociodemographic and psychographic characteristics of the individuals who respond to the questionnaire. This paper is also comparable to Nicolau and Mas (2006),

Au

th o

r's

pe

rs

on

al

co

py

CORREIA, SANTOS AND BARROS

625

who adopt a multinomial logit model with random coefcients to identify the probability of choosing intra-country destinations being dependent on distance, their prices, and motivations. Moreover, it is directly comparable with Kemperman et al (2005), who similarly adopt a multinomial mixed logit to estimate the probability of visiting a public garden in Eindhoven, Holland, as a function of socioeconomic variables and the time of the visit. Relative to all these papers, a dichotomous mixed logit is used in the present paper, ensuring a clear focus on a specic destination in comparison with the multinomial model. This study uses a larger set of explanatory variables those mentioned above. What is more, this paper clearly identies the variables with heterogeneity and separates them from the homogeneous ones (Train 2003). This separation is a primary outcome of the mixed logit, but is not clearly identied in those mentioned. As a result, the estimated parameters are either homogeneous or randomly distributed. Relative to the modeling procedure adopted by Nicolau and Mas (2006) and Kemperman et al (2005), this paper does not adopt interactions among explanatory variables, due to the large number of variables used, so that the role of interactions in the estimated model was not clear. Two main limitations relate to the data set of this investigation. First, the data span is relatively short. Second, the sample procedure adopted was restricted to a single airline charter company. Moreover, since this research is an exploratory study, the intention is not to obtain denitive results for direct use by tourism enterprises. Rather, it calls their attention to the value of identifying heterogeneity among their clients, and dening business strategies for the clusters of customers in order to satisfy the preferences of all. In order to draw more generalized conclusions, a larger data set would be necessary, while more air charter companies operating in the market under analysis should be included. The limitations of the study suggest directions for new research. Additional research is needed to conrm the results reported here, as well as to clarify the noted issues. Further, research concerning choice must be expanded to consider other destinations, including newly emerging resorts and regions. It is also necessary to consider motivations and learning processes within the set of determinants of the tourists decision. From the social science perspective, the present results highlights the importance of cultural afnity as driving factor attracting Portuguese tourists for Latin America ( Jackson 1989). Although not the focus of the study reported here, the cultural and linguist relationship between Portugal and Brazil underlies all the analysis. Moreover, relative to Mexico the proximity between the Portuguese and Spanish language also serves as an attractor, signifying that the paper is analyzing a specic context: tourism on countries sharing common cultural heritage. This form of tourism co-exists with other types on alien countries and is characterized by common ethical ideologies and shared perceptions on the reality (Yaman and Gurel 2006), past colonial relationships (Teo and Leong 2005), and hence such shared or common heritage that has to be taken into consideration when interpreting

Au

th o

r's

pe

rs

on

al

co

py

626

CHOICE ANALYSIS

the results. More investigation is needed to conrm the present ndings and to insert in the analysis neglected attributes, such the cultural issues noted in this conclusion.
AcknowledgementsThe authors acknowledge grant support from Air Luxor, SA for this study.

REFERENCES

Aguilo E., J. Alegre, and M. Sard , 2005 The Persistence of the Sun and Sand Tourism Model. Tourism Management 26:219231. Alfnes, F. 2004 State Preferences for Imported and Hormone-Treated Beef: Application of a Mixed Logit Model. European Review of Agricultural Economics 31:1937. Baker, D., and J. Crompton 2000 Quality, Satisfaction and Behavioral Intentions. Annals of Tourism Research 27:785804. Baloglu, S., and K. McCleary 1999 A Model of Destination Image Formation. Annals of Tourism Research 26:868897. Barros, C., and I. Proenca 2005 Mixed Logit Estimation of Radical Islamic Terrorism in Europe and North America. The Journal of Conict Resolution 49:298314. Bath, C. 1996 An Endogenous Mode Choice Model with an Application to Intercity Travel. Transportation Science 31:3448. Beerli, A., and J. Martn 2004 Tourists Characteristics and the Perceived Image of Tourist Destinations: A Quantitative AnalysisA Case Study of Lanzarote, Spain. Tourism Management 25:623636. Ben-Akiva, M. and S. Lerman, eds. 1985 Discrete Choice Analysis: Theory and Application to Travel Demand. Cambridge: MIT Press. Bonnet, C., and M. Simioni 2001 Assessing Consumer Response to Protected Designation of Origin Labeling: a Mixed Multinomial Logit Approach. European Review of Agricultural Economics 28:433449. Brownstone, D., and K. Train 1999 Forecasting New Product Penetration with Flexible Substitution Patterns. Journal of Econometrics 89:109129. Brownstone, D., D. Bunch, and K. Train 2000 Joint Mixed Logit Models of State and Revealed Preferences for Alternative-fuel Vehicles. Transportation Research Part B 34:315338. Chesher, A. 1984 Testing for Neglected Heterogeneity. Econometrica 52:865872. Chesher, A., and J. Santos-Silva 2002 Taste Variation in Discrete Choice Models. Review of Economic Studies 69:147168. Correia, A., C. Barros, and A. Silvestre 2007 Golf Tourism Repeat Choice Behaviour in the Algarve: A Mixed Logit Approach. Tourism Economics 11:111127. Costa, P., and M. Manente 1995 Venice and its Visitors: A Survey and a Model of Qualitative Choice. Journal of Travel Research and Marketing 4:4569.

Au

th o

r's

pe

rs

on

al

co

py

CORREIA, SANTOS AND BARROS

627

Crompton, J. 1979 Motivations for Pleasure Vacations. Annals of Tourism Research 6:408424. Dalen, E. 1989 Research into Values and Consumer Trends in Norway. Tourism Management 10:183186. Dann, G. 1981 Tourist Motivation: An Appraisal. Annals of Tourism Research 7:187219. De la Vina, L., and J. Ford 2001 Logistic Regression Analysis of Cruise Vacation Market Potential: Demographic and Trip Attribute Perception Factors. Journal of Travel Research 39:406410. Dillman, D., ed. 1978 Mail and Telephone Surveys: The Total Design Method. New York: Wiley. Domencich, T. and D. McFadden, eds. 1975 Urban Travel Demand: A Behavioral Analysis. Amsterdam: North Holland. Edwards, W. 1954 The Theory of Decision Making. Psychology Bulletin 51:380417. Festinger, L. 1954 A Theory of Social Comparison Processes. Human Relations 7:117140. Fishbein, M., ed. 1967 Readings in Attitude Theory and Measurement. New York: Wiley. Fishbein, M. and I. Ajzen, eds. 1980 Predicting and Understanding Consumer Behavior: Attitude Behavior Correspondence. Upper Sadle River: Prentice Hall. Fleischer, A., and A. Pizam 2002 Tourism Constraints among Israeli Seniors. Annals of Tourism Research 29:106123. Gibson, H., and L. Gray 2005 Insights from Role Theory: Understanding Golf Tourism. European Sport Management Quarterly 5(4):443468. Gnoth, J. 1997 Tourism Motivation and Expectation Formation. Annals of Tourism Research 24:283304. Gonul, F., and K. Srinivasan 1993 Modeling Multiple Sources of Heterogeneity in Multinomial Logit Models: Methodological and Managerial Implications. Marketing Science 12:213229. Goodall, B. and G. Ashworth, eds. 1988 Marketing in the Tourism Industry. The Promotion of Destination Regions. London: Routledge. Goodrich, J. 1980 Segmentation of US International Travelers: An Empirical Study with American Express. In Tourism Marketing and Management Issues, L. Donald, S. Elwood and M. James, eds., pp. 133147. Washington DC: George Washington University. Hajivassiliou, V., and P. Ruud 1994 Classical Estimation Methods for LDV Models Using Simulation. In Handbook of Econometrics, R. Engle and D. McFadden, eds., pp. 23842438. Amsterdam: North Holland. Hay, M., and K. McConnel 1979 An Analysis of Participation in Non-Consumptive Wildlife Recreation. Land Economics 55:460471. Hensher, D., and W. Greene 2003 The Mixed Logit Model: The State of Practice. Transportation 30:133176. Iso-Ahola, S., and R. Mannell 1987 Psychological Nature of Leisure and Tourism Experience. Annals of Tourism Research 14:314331. Jackson, P. 1989 Maps of Meaning: An Introduction to Cultural Geography. London: Allen.

Au

th o

r's

pe

rs

on

al

co

py

628

CHOICE ANALYSIS

Kemperman, A., W. Ponje, and P. Timmermans 2005 Analyzing Heterogeneity and Substitution in Trip Making Propensity to Urban Parks: A Mixed Logit Model. Tourism Analysis 10:223232. Kockelman, K., and S. Krishnamurthy 2004 A New Approach for Travel Demand Modelling: Linking Roys Identity to Discrete Choice. Transportation Research Part B 38:459475. Kozak, M. 2001 Repeaters Behavior at Two Distinct Destinations. Annals of Tourism Research 28:784807. 2003 Measuring Tourist Satisfaction with Multiple Destination Attributes. Tourism Analysis 7:229240. Lancaster, K. 1966 A New Approach to Consumer Theory. Journal of Political Economy 74:132157. Luzar, E., A. Diagne, C. EcGan, and B. Henning 1998 Proling the Nature-based Tourist: A Multinational Logit Approach. Journal of Travel Research 37:4855. Mathieson, A. and G. Wall, eds. 1984 Tourism: Economic, Physical and Social Impacts. New York: Longman. McFadden, D. 1974 The Measurement of Urban Demand. Journal of Public Economics 3:303328. 1981 Conditional Logit Analysis of Qualitative Choice Behavior. In Frontiers in Econometrics, P. Zarembka, ed., pp. 105142. New York: Academic Press. McFadden, D., and K. Train 2000 Mixed MNL Models of Discrete Response. Journal of Applied Econometrics 15:447470. Miller, J., and M. Hay 1981 Determinants of Hunter Participation: Duck Hunting in the Mississippi Flyway. American Journal of Agricultural Economics 63:677684. Morley, C. 1994 Experimental Destination Choice Analysis. Journal of Tourism Research 21:780791. Muller, T. 1991 Using Personal Values to Dene Segments in an International Tourism Market. International Marketing Review 8:5770. Nicolau, J., and F. Mas 2005 Stochastic Modelling: a Three-stage Tourist Choice Process. Annals of Tourism Research 32:4969. 2006 The Inuence of Distance and Prices on the Choice of Tourist Destinations: The Moderating Role of Motivations. Tourism Management 27:982 996. Pearce, P. 1982 Perceived Changes in Holiday Destinations. Annals of Tourism Research 9:45164. Pitts, R., and A. Woodside 1986 Personal Values and Travel Decisions. Journal of Travel Research 25:2025. Revelt, D., and K. Train 1998 Mixed Logit with Repeated Choices: Households Choices of Appliance Efciency Level. Review of Economics and Statistics 80:647657. Sheldon, P. 1995 The Demand for Incentive Travel: An Empirical Study. Journal of Travel Research Spring:2328. Shih, D. 1986 Demand for Package Tours: A Mode Choice Model. Journal of Travel Research 15:317. Sirakaya, A., and G. Woodside 2005 Building and Testing Theories of Decision Making by Travelers. Tourism Management 26:815832.

Au

th o

r's

pe

rs

on

al

co

py

CORREIA, SANTOS AND BARROS

629

Au

th o

Submitted 10 November 2005. Resubmitted 11 January 2006. Resubmitted 31 May 2006. Resubmitted 14 July 2006. Resubmitted 10 September. Resubmitted 22 September 2006. Final version 22 September 2006. Accepted 16 December 2006. Refereed anonymously. Coordinating Editor: Metin Kozak

r's

Stynes, D., and G. Peterson 1984 A Review of Logit Models with Implications for Modelling Recreational Choices. Journal of Leisure Research 16:295310. Taplin, J., and C. McGinley 2000 A Linear Program to Model: Daily Car Touring Choices. Annals of Tourism Research 27:451467. Teo, P., and S. Leong 2005 A Postcolonial Analysis of Backpacking. Annals of Tourism Research 33:109131. Train, K. 1986 Qualitative Choice Analysis. Cambridge: MIT Press. 1998 Recreational Demand Models with Taste Variation. Land Economics 74:230239. 2003 Discrete Choice Methods with Simulation. Cambridge: Cambridge University Press. Cambridge University Press, <http://elsa.berkeley.edu/ ~train>. Um, S., and J. Crompton 1990 Attitude Determinants in Tourism Destination Choice. Annals of Tourism Research 17:432448. Walsh, R., J. McKean, and J. Hof 1992 Effect of Price on Forecasts of Participating in Fish and Wildlife Recreation: An Aggregate Demand Model. Journal of Leisure Research 24:140156. Weaver, P., K. McCleary, L. Lepisto, and L. Damonte 1994 The Relationship of Destination Selection Attributes to Psychological, Behavioral and Demographic Variables. Journal of Hospitality and Leisure Marketing 2:93109. Woodside, A., and S. Lysonski 1989 A General Model of Travel Destination Choice. Journal of Travel Research 27(4):814. Woodside, A., and R. MacDonald 1994 General System Framework of Customer Choice Processes of Tourism Services. In Spoilt Lop Choice, R. Gasser and K. Weiermair, eds., pp. 3059. Thaur: Kulturverl. Yaman, H., and E. Gurel 2006 Ethical Ideologies of Tourism Marketers. Annals of Tourism Research 33:470489. Zimmer, Z., R. Brayley, and M. Searle 1995 Whether to Go and Where to Go: Identication of Important Inuences . on Seniors Decisions to Travel. Journal of Travel Research 33(3):310

pe

rs

on

al

co

py

Das könnte Ihnen auch gefallen