INTERNATIONAL JOURNAL OF TOURISM RESEARCH Int J Tourism Res 12, 647–664 (2010) Published online 27 April 2010 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/jtr.780 Differences in Travel Objectives between First-time and Repeat Tourists: An Empirical Analysis for the Kansai Area in Japan Kaoru Okamura1 and Mototsugu Fukushige2,* Competition Policy Research Center, Japan Fair Trade Commission, Tokyo, Japan Graduate School of Economics, Osaka University, Osaka, Japan ABSTRACT This study investigates the differences in travel objectives between first-time and repeat tourists We conduct a questionnaire survey of travel agencies, which asked about specific tour plan for target tourists, their experiences and travel objectives in the Kansai area in Japan We estimate a logit model for the relationship between travel objectives and visiting experiences The results indicate that the first-time tourists’ main objective is to enjoy looking around sightseeing spots, while the repeat tourists’ objective is simply to enjoy the stay, including the hotel visit and participating in events Copyright © 2010 John Wiley & Sons, Ltd Received 19 September 2009; Revised 16 February 2010; Accepted March 2010 Keywords: first-time tourists; repeat tourists; visiting experiences; logit model INTRODUCTION B utler (1999, 2004) pointed out that to preserve sightseeing spots, it is important not to increase the number of tourists but rather to encourage previous tourists to visit an area again This paper examines how repeat tourists differ from first-time tourists in terms *Correspondence to: Mototsugu Fukushige, Professor, Graduate School of Economics, Osaka University 1-7, Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan E-mail: mfuku@econ.osaka-u.ac.jp of their travel objectives This analysis provides an important clue to understanding why a first-time visitor becomes a repeat visitor or what type of first-time visitor becomes a repeat visitor We also distinguish between the sightseeing spot’s attractiveness for first-time tourists and for repeat tourists by investigating the changes in each type of visitor’s objectives This research is also useful when promoting a sightseeing spot that corresponds to a particular visit frequency We conducted a questionnaire survey of travel agencies to create a tourist database of visitor characteristics, frequency of visits and travel objectives for the Kansai area, which includes Kyoto, the most popular sightseeing spot in Japan (see Figure 1) Our approach, which conducts questionnaire survey to travel agencies, is an unique approach in this type researches But travel agencies are always facing travel demand market and analyzing tourists’ tastes We consider that they know the tourists’ tastes exactly, so to conduct a questionnaire survey to the travel agencies is more comprehensive than to conduct a questionnaire survey to tourists themselves The latter is also expensive because we have to collect a large number of answer sheets to cover most of the sightseeing spots in Kansai area Exactly speaking, our research is to analyse the travel agencies’ viewpoint for the differences between first-time visitors and repeat visitors However, to conduct a questionnaire survey for tourists’ characteristics or objectives in a wide area like Kansai, this type survey is inexpensive and useful for the purpose of our research The results of our paper will also make clear the advantages of our approach in the rest of the paper Copyright © 2010 John Wiley & Sons, Ltd 648 K Okamura and M Fukushige Figure Kansai area We estimate a logit model to analyse the relationship between the objectives of the tourists and their visit frequency The results indicate that there are two types of tourists: one whose main objective is sightseeing and another whose objective is to enjoy the stay The former is characteristic of first-time tourists and the latter is observed in repeat tourists This finding is supported by three results Copyright © 2010 John Wiley & Sons, Ltd First, first-time tourists are likely to have the travel objective of visiting sightseeing spots such as ‘historical buildings or streets’ Second, a visitor who has visited four times or more chooses with a high probability ‘accommodation facilities (including spas)’ In this paper, we call this type of tourists ‘repeat tourist’ Third, a visitor who has visited two or three times lies in the middle of the first and repeat Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr Differences in Travel Objectives tourists, displaying a mixture of characteristics of both first-time and repeat tourists In this paper, we call this type of tourists, who has visited two or three times, ‘second time tourist’ for convenience sake Of course, we can classify tourists who has visited once or more into second time, third time, forth time and so on, but such a classification seems complicated and troublesome, so we categorise the tourists into three categories for simplicity: first-time, second-time and repeat tourists The remainder of the paper is organised as follows Section surveys the existing literature Section describes the questionnaire and target of the survey, and provides a summary of the survey results Section conducts an empirical study of the choice probabilities of travel objectives Section analyses the relationships between visit frequency and travel objectives Section provides conclusions and discusses the remaining problems LITERATURE SURVEY Before going into the analysis, we conduct a literature survey to place this paper in context Our analysis relates the co-called destination loyalty Several papers analyse the characteristics of the destination from this point of view, e.g Clottey and Lennon (2003) analysed the relationships between frequency of visits and types of information received for German tourists to Lithuania However, in this paper, we focus on the relationships between the tourists’ objectives and frequencies with controlling tourists’ characteristics for segmenting tourists into first-time tourists and repeat tourists To understand what encourages a tourist to visit a place again, some previous studies investigated the types of sightseeing spots likely to be visited or the types of tourists that are likely to visit a place again Some researchers focused on the characteristics of first-time tourists and repeat tourists For example, Kozak (2001) found the importance of the experience of previous visit and the satisfaction at the previous visit for revisit; Ledesma et al (2005) also found that the information obtained from previous visit and/or relatives and friends is important; Truong and King (2009) showed that tourists who are highly satisfied with their previous visit tend to visit again Correia et al (2007), Copyright © 2010 John Wiley & Sons, Ltd 649 using the random parameter logit model, found that the upkeep is important for golfplaying repeaters These researches imply that the satisfaction, including costs, is important to revisit However, our analysis focuses on the changes of objectives when the tourist becomes a repeat tourist In other words, we focus on what kind of properties in sightseeing spots satisfies the repeat tourists Light (1996) and Law (2002) reported that a sightseeing spot could easily attract a visitor who has visited previously by introducing a new event McWilliams and Crompton (1997) also found that tour promotions, such as advertising or direct mail, which introduce a sightseeing spot, also show that these are more effective for tourists who have been there previously than for those who have not These previous studies mainly analysed tourism marketing to reveal what encourages tourists who have already visited to visit again; however, these studies also did not investigate tourists’ objectives when visiting or whether they change their objectives according to their visiting experience Stewart and Vogt (1999), in an analysis of the city of Branson, Missouri, pointed out that repeat tourists tend to reduce their time spent on sightseeing activities; however, they did not analyse why travel objectives changed Lam and Hsu (2006) also examined repeat tourists’ behaviour by breaking down the frequency of visits; however, they did not analyse the relationship between the frequency of visits and travel objectives As for the analyses about the differences between first time and repeat visitors, there exist several researches For example, Litvin (2007) focused on the fact that the attendance on the visitor attraction activity is important for making repeat visitors Vassiliadis (2008), using CHAID and CRT model, analyse the repeat visiting and recommendation behaviours Additionally, Tiefenbacher et al (2000), Correia et al (2008) and Fallon and Schofield (2004) analysed the differences between firsttime tourists’ and repeat tourists’ images of a destination or perceptions of its attractive attributes; Hughes and Allen (2008) compared the images of resorts held by visitors and nonvisitors; Beaman et al (2001) estimated a Markov matrix for visitors moving from firsttime tourists to repeat tourists; and Darnell Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr 650 and Johnson (2001) analysed repeat visits to attractions However, these researches seem to assumed that the taste of the tourists not change with their experience, implicitly In the present paper, we consider that tourists’ taste change with their visiting experience or frequencies From this point of view, Wang et al (2006) might be closely related to our investigation They analysed the changes in consumers’ expenditure patterns The changes in objectives between first-time tourists and repeat tourists and these changes might be revealed in changes in consumption patterns METHODOLOGY We needed a specific data set indicating tourists’ frequency of visits to sightseeing spots and travel objectives to investigate whether travel objectives change according to visit experience The Japan Tourism Association (JTA) and the Japan National Tourist Organization (JNTO) have conducted and published surveys of visitor behaviour in Japan JNTO (2006) conducted questionnaire survey only for the foreign tourists who visit Japan, but reported the tourists’ objectives by countries JTA (2006) reported the frequencies, objectives and other characteristics of the tourists, but most of them are reported in a simple aggregated data or cross tabulated We cannot obtain any information about the repeater for specific sightseeing spot from neither of them Therefore, using these data, we cannot examine the relationship between visit frequency and travel objectives for a specific sightseeing spot In conducting an original questionnaire survey, it is necessary to consider its cost and method Past studies have used several survey methods for examining visitor behaviour One is to directly ask tourists questions relating to a specific sightseeing spot and another is to send a questionnaire to households (Tiefenbacher et al., 2000) The former possibly has seasonal or site bias depending on the survey site and period, and the latter involves significant cost because of the need to distribute an enormous number of surveys to assemble responses To avoid these problems, we elected to use a survey asking tour operators and tourism authorities about visitor behaviour at sightseeing spots Because travel agencies have Copyright © 2010 John Wiley & Sons, Ltd K Okamura and M Fukushige the most extensive information, this enabled us to investigate visitors’ behaviour with no seasonal bias and at a much lower cost In evaluating the appropriateness of the questionnaire asking travel agencies or municipal tourism authorities their observations of the behaviour of tourists, the questionnaire conducted by the JTA (2006) provided helpful information According to the results of this survey, a visitor may use information sent by travel agencies or tourism authorities extensively when he or she travels, which shows there is little difference between what the travel agency thinks and what the actual visitor wants In this paper, we use a questionnaire survey for four categories of businesses: domestic travel agencies (registered travel agencies approved by the Minister of Land, Infrastructure and Transport), land operators, national government (or municipal) tourism offices and hotels affiliated with the Japan Hotel Association This survey was designed by Fukushige and conducted through the Kansai Institute of Social and Economic Research (KISER) We are grateful to the KISER for their kind permission to use and analyse the survey As this questionnaire was sent to all registered travel agencies and tourism authorities in the Kansai area, this is not a randomly gathered sample This paper investigates the relationship between visit frequency and travel objectives for Japanese tourists because very few foreign tourists frequently visit Japanese sightseeing spots and they may have different perceptions of sightseeing spots in Japan compared with the Japanese In the questionnaire, we asked the respondents to develop a tour plan for Japanese tourists and for foreign tourists However, this paper focuses on the analysis of Japanese visitor behaviour; therefore, we not discuss tour plans for foreigners We provide an English translation of the questionnaire in the appendix Here, we summarise the questionnaire and explain its objectives The main question was: ‘If you were to plan and market travel or tours to the Kansai area in Japan targeting Japanese customers, what would they be? Please provide two different plans’ In response to this question, the respondent provides answers to three parts of the questionnaire: A details of the tour; B Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr Differences in Travel Objectives destination; and C objective of the tour Each part contained essential details for developing a tour plan, and the respondent created his or her ideal tour plan by selecting the appropriate answers from the choices provided Part A of the survey contains detailed characteristics of the plans, the length of tour, season, numbers of tourists and visitor characteristics (age, sex, visit frequency and residence) In this part, the respondent may develop a suitable tour by choosing one aspect of the travel plan to attract customers In part B, the respondent selects between two and 10 visitor destinations from 77 listed sightseeing spots and events In part C, the respondent chooses his or her travel objective depending on the answers in parts A and B The response sheet provides 15 options for travel objectives and the respondent may choose multiple items as long as they match his or her tour plan The target area of this analysis is an area called Kansai, which contains Kyoto, one of the most popular cities in Japan Kyoto was the capital of Japan from 794 to 1868 A.D., and is famous both in Japan and worldwide for its historical buildings or streets and beautiful natural scenery throughout the year Nara, located south of Kyoto, which was the capital of Japan before it moved to Kyoto, is also famous for the historical buildings or streets in the old town Tourists to the area can not only visit the historical buildings or streets but also participate in traditional cultural events such as festivals As for urban tourism in the Kansai area, the cities of Osaka and Kobe have modern buildings or streets There is also a range of other sightseeing spots, such as Lake Biwa, the largest lake in Japan; Wakasa Bay, on the Rias coast; Kumano-Kodou, a sacred site and pilgrimage routes in the Kii Mountain Range, also listed as a World Heritage site; and Universal Studios Japan, a theme park opened in 2001 As this list shows, the Kansai area contains almost every kind of sightseeing spot that may interest tourists, including natural scenery, historical buildings or streets, modern buildings or streets, and cultural events The diversity of resources for sightseeing in the Kansai area is confirmed by the survey conducted by the JTA called ‘A questionnaire survey for individual tourists’ This survey lists travel Copyright © 2010 John Wiley & Sons, Ltd 651 objectives, almost all of which are met in this area Furthermore, these sightseeing resources are located within a 200-km radius of Osaka, which is the central transportation terminal As tourists can therefore easily move between individual sightseeing locations by car or train, we can consider the Kansai area one big sightseeing spot We assume that we can observe whether tourists change their travel objectives according to their visiting experience We mailed the questionnaires to domestic travel agencies on 19 December 2004, and asked them to return their responses within two weeks Of the 953 surveys sent out, 140 were returned (response rate of 14.7%); the total number of travel plans completed on the answer sheet was 231 because some agencies responded with only one plan in spite of asking two plans Tables and show the results of questionnaire part A and part C, which are used for our econometric analysis in the next section Each item selected by a respondent from the questionnaire is assigned a value of 1; items not chosen are assigned a value of The aggregated results are shown in Table The most common answers are: ‘2 or days’ for the length of the tour; ‘autumn’ for the season of travel; ‘group travelers’ for the type of traveler; and ‘sixties’, ‘both sexes’, a ‘Kanto’ resident and someone who has been to Kansai ‘2 or times’ for visitor characteristics In part C, as in part A, we aggregate the answers if a respondent chooses travel objectives from the 15 listed objectives Each objective chosen is assigned a value of 1, and otherwise We show the results in Table The most popular travel objective according to travel agencies is ‘to see the historical buildings or streets’ This result contrasts with that of the questionnaire survey of the JTA (2006), in which ‘natural scenery’ is the most popular The result of our survey seems to reflect the fact that the Kansai area has many historical buildings or streets as sightseeing spots Other objectives such as ‘natural scenery’, ‘cuisine’, ‘theme parks’ and ‘festivals or special events’ follow these two major objectives: ‘to see the historical buildings or streets’ and ‘natural scenery’ These results are similar to the results of the JTA (2006) survey except for the reversal of the first and second objectives Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr 652 K Okamura and M Fukushige Table Summary statistics of part A Abbreviations A1 A2 A3 A4 A21 A22 A23 A24 A25 A41 A42 A43 A44 A45 A51 A52 A53 A54 A55 A56 A57 A58 A61 A62 A63 A71 A72 A73 A81 A82 A83 A84 A85 A86 A87 A88 A89 A810 A811 Variables Sum Std Dev Mean One day 2–3 days 4–6 days More than one week Spring Summer Autumn Winter Throughout the year Solo travelers Couples Families Group travelers Other Teens Twenties Thirties Forties Fifties Sixties Seventies All age groups Female Male Both First time Two or three times Four times or more Hokkaido Tohoku Kanto Hokuriku Chubu Kinki Chugoku Shikoku Kyushu Okinawa Any area 20 154 24 63 16 84 14 88 20 96 43 134 18 10 26 38 45 94 109 42 25 53 141 59 110 34 19 19 69 29 11 13 21 12 38 0.299 0.429 0.324 0.155 0.464 0.270 0.494 0.254 0.497 0.299 0.501 0.410 0.475 0.285 0.217 0.335 0.391 0.416 0.500 0.500 0.406 0.329 0.440 0.195 0.462 0.455 0.499 0.374 0.292 0.292 0.475 0.170 0.351 0.227 0.183 0.245 0.305 0.236 0.391 0.099 0.759 0.118 0.025 0.310 0.079 0.414 0.069 0.433 0.099 0.473 0.212 0.660 0.089 0.049 0.128 0.187 0.222 0.463 0.537 0.207 0.123 0.261 0.039 0.695 0.291 0.542 0.167 0.094 0.094 0.340 0.030 0.143 0.054 0.034 0.064 0.103 0.059 0.187 EMPIRICAL ANALYSIS To investigate the relationships between visit frequency and travel objective, using the results from our questionnaire, a multinomial choice model should be suitable However, as there are 15 objects in a binary form, it is very hard to estimate a multinomial choice model and obtain stable estimates In the present Copyright © 2010 John Wiley & Sons, Ltd paper, for simplicity and efficiency, we adopt a binary choice model where a visitor chooses a specific objective or not This model explains choice behaviour with a latent index For example, when a tourist visits a sightseeing spot with his or her travel objective ‘being seeing the natural scenery’, his or her latent index y* of ‘seeing the natural scenery’ has a positive number Additionally, we set y* to be Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr Differences in Travel Objectives 653 Table Summary statistics of part C Abbreviations C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 Variables Sum Variance Natural scenery Historical buildings or streets Modern buildings or streets Cuisine Shopping Parks or wandering Accommodation facilities (including spas) Art galleries and museums (including treasure houses in temples and shrines) Concerts or performances at music halls or theaters Festivals or special events Nightlife Night views Theme parks Sports (such as golf and skiing) Industrial facilities 135 158 20 73 52 17 51 42 0.224 0.173 0.089 0.231 0.191 0.077 0.189 0.165 57 18 16 59 13 0.038 0.203 0.081 0.073 0.207 0.033 0.060 a single index, written as y* = x′β + ε (note that x is a vector of non-stochastic independent variables, β is a vector of unknown parameters and ε is an error term) We cannot estimate this model directly, because y* is an unobserved variable However, we can observe a stochastic variable (y) If y* > 0, the visitor chooses ‘seeing the natural scenery’ as a travel objective (y = 1) If y* ≤ 0, he or she does not choose it (y = 0) The observed stochastic variable y is written as follows: Pr [ y = x ] = Pr [ y* > ] , This equation means that the event to choose ‘seeing the natural scenery’ coincides with the event that y* is positive, so both the probabilities for these events are mutually equal Then, we rewrite y* with x′β + ε: Pr [ y = x ] = Pr [ y* > ] = Pr [ x ′β + ε > x ] The inequality rewritten: in parentheses can be Pr [ y = x ] = Pr [ y* > ] = Pr [ x ′β + ε > x ] = Pr [ − ε < x ′ β x ] Replacing the probability function of error term (−ε) Pr[*] with the cumulative distribution function F(•), the above equation becomes: Pr [ y = x ] = F ( x ′β ) (1) If we assume the cumulative distribution function F(•) of the error term −ε to be a logistic Copyright © 2010 John Wiley & Sons, Ltd distribution function, then F(•) can be written using x′β as: F ( x ′β ) = e x ′β (1 + ex′β ) Therefore, we can construct a simultaneous probability density function for all the observations with assuming mutual independence of each observation We can consider this density function to be a likelihood function for unknown parameters, so that we can obtain the maximum likelihood estimator to maximise it We apply a logit model for each of the 15 objectives in the questionnaire We adopt visitor attributes as the independent variables These variables are obtained from the answers in part A of the questionnaire As additional independent variables, we also adopt some cross products of independent variables relating to the visit frequency, such as age, residence, length of stay and visitor type, but we omit cross products with all the observations taking the values of or because most of the independent variables are dummy variables As a result, the number of candidates for the independent variables including cross products of the attributes of tourists and tour plans is 40 The summary statistics of the cross products are given in Table We select a model by minimizing Akaike (1973)’s information criteria because, in the estimation results with all independent variables, many variables with insignificant Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr Copyright © 2010 John Wiley & Sons, Ltd Four times or more Two or three times First time Four times or more Two or three times First time × × × × × × × × × × × × × × × × × × × × × × × × × × × × × × TYEX11 TYEX21 TYEX31 TYEX51 TYEX12 TYEX22 TYEX32 TYEX52 TYEX13 TYEX23 TYEX33 TYEX53 AEX11 AEX21 AEX31 AEX41 AEX51 AEX71 AEX81 AEX12 AEX22 AEX32 AEX42 AEX52 AEX72 AEX82 AEX23 AEX33 AEX43 AEX53 AEX73 Solo travelers Couples Families Other Solo travelers Couples Families Other Solo travelers Couples Families Other Teens Twenties Thirties Forties Fifties Seventies All age groups Teens Twenties Thirties Forties Fifties Seventies All age groups Twenties Thirties Forties Fifties Seventies Variable Abbreviations Table Summary statistics of cross products 26 15 10 53 26 17 2 12 15 12 22 11 10 17 27 57 21 15 6 15 10 Sum 0.139 0.335 0.262 0.195 0.217 0.440 0.335 0.195 0.170 0.278 0.099 0.099 0.206 0.236 0.262 0.236 0.312 0.227 0.195 0.070 0.217 0.278 0.340 0.451 0.305 0.262 0.139 0.170 0.170 0.262 0.217 Std Dev DEX11 DEX31 DEX41 DEX12 DEX32 DEX42 DEX13 DEX33 LEX11 LEX12 LEX14 LEX15 LEX16 LEX19 LEX110 LEX21 LEX22 LEX24 LEX25 LEX26 LEX27 LEX28 LEX29 LEX210 LEX211 LEX31 LEX36 LEX37 LEX38 LEX311 Abbreviations One day 4–6 days More than one week One day 4–6 days More than one week One day 4–6 days Hokkaido Tohoku Hokuriku Chubu Kinki Kyusyu Okinawa Hokkaido Tohoku Hokuriku Chubu Kinki Chugoku Shikoku Kyushu Okinawa Any area Hokkaido Kinki Chugoku Shikoku Any area × × × × × × × × × × × × × × × × × × × × × × × × × × × × × × Variable Four times or more Two or three times First time Four times or more Two or three times First time 12 5 13 18 14 14 3 6 Sum 0.155 0.206 0.121 0.195 0.236 0.099 0.183 0.121 0.195 0.197 0.099 0.155 0.139 0.155 0.155 0.206 0.245 0.139 0.270 0.139 0.121 0.170 0.254 0.170 0.254 0.099 0.121 0.121 0.170 0.170 Std Dev 654 K Okamura and M Fukushige Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr Differences in Travel Objectives 655 coefficients are included, which may cause inefficient estimation The estimation results are given in Table As shown in Table 4, there are some cases where the dummy variables representing visiting experience to Kansai are not chosen as independent variables However, if the cross products are considered, the visit frequency affects the choice probability of the travel objectives in all cases The cross products selected in each model are different, and include frequency of traveling, age or type of travel These results make the effects of the frequency on the choice probability complex We compared the choice ratios of the surveys with the estimated choice probability model, and calculated the ‘hit or lose ratio’ as a goodness of fit measure of the selected model in Table According to the results, the prediction accuracies of the models are low in the model that sets travel objectives such as ‘modern buildings or streets’, ‘parks or wandering’, ‘concerts or performances at music halls or theaters’, ‘night views’ and ‘industrial facilities’ We should pay attention to these low explanatory power when we interpret the estimation results in the next section Table Results of estimation Natural scenery: Observations = 208, C1 = 0.569 +1.649 ∗ A (1.974) (1.702) −2.894 ∗ DEX 32 −2.618 ∗ AEX11 ( −2.441) ( −2.310) Log-likelihood = +1.115 ∗ A 21 ( 2.872) +2.217 ∗ AEX 41 ( 2.372) −106.483, AIC = 117.483 −1.423 ∗A 54 +1.224 ∗ A 55 −1.122 ∗ A71 ( −2.961) ( 3.278) ( −2.342) −2.570 ∗ TYX52 +1.823 ∗ LEX19 ( −2.788) (1.500) Historical buildings or streets: Observations = 208, Log-likelihood = −82.702, AIC = 93.702 C2 = 0.886 +1.746 ∗ A 23 −1.646 ∗ A 24 +1.698 ∗ A 41 +1.382 ∗ A 42 −1.033 ∗ A71 ( 2.172) ( 3.672) ( −2.258) (1.962) ( 2.567 ) ( −1.838) −1.053 ∗ A73 +1.053 ∗ A 85 −1.737 ∗ AEX 42 +2.626 ∗ AEX 51 −1.960 ∗ TYEX 21 ( −1.757 ) (1.478) ( −2.831) ( 2.693) ( −2.108) Modern building or streets: Observations = 208, Log-likelihood = −45.514, AIC = 56.514 C3 = −3.494 +2.400 ∗ A 41 +1.977 ∗ A 43 +2.793 ∗ A621 −2.202 ∗ A 811 +2.345 ∗ DEX 32 ( −5.979) ( 3.672) ( 2.922) ( 2.930) (1.962) ( 2.229) +1.587 ∗ AEX 31 +2.436 ∗ TYEX 51 −2.617 ∗ TYEX 52 +4.097 ∗ LEX 26 (1.978) ( 2.230) ( −2.068) ( −2.487) Cuisine: Observations = 208, Log-likelihood = −121.614, AIC = 128.614 C = −0.655 −1.743 ∗ A1 +2.723 ∗ A −2.272 ∗ A 51 +2.398 ∗ DEX12 +0.992 ∗ DEX 32 +3.322 ∗ LEX19 ( −3.865) (1.647) ( 2.029) ( −1.788) (1.901) (1.627) ( 2.161) Shopping: Observations = 208, Log-likelihood = −103.542, AIC = 111.542 C5 = −1.710 +0.846 ∗ A +1.145 ∗ A 22 −0.536 ∗ A 55 +0.929 ∗ A 85 +1.372 ∗ A 811 ( −5.341) (1.690) (1.969) ( −1.475) (1.823) ( 3.164) +1.844 ∗AEX 22 +1.910 ∗ LEX11 (1.844) ( 2.479) Parks or wandering: Observations = 208, Log-likelihood = −47.132, AIC = 56.132 C6 = −.3.335 −1.956 ∗ A 43 +1.560 ∗ A 45 +0.988 ∗ A 54 −1.132 ∗ A61 +1.764 ∗ DEX 32 ( −5.780) ( −1.779) ( 2.083) (1.527) ( −1.310) (1.806) +1.971∗ AEX72 +1.982 ∗ TYEX 21 +2.730 ∗ LEX19 ( 2.533) ( 2.558) ( 2.064) Accommodation facilities (including spas): Observations = 208, Log-likelihood = −97.741, AIC = 108.741 C7 = −1.497 +1.935 ∗ A 55 −3.722 ∗ A71 +0.887 ∗ A 811 +3.856 ∗ DEX 31 +3.303 ∗ DEX41 ( −4.792) ( 3.505) ( −2.990) (1.900) ( 2.796) (1.856) +3.963 ∗ AEX 81 −1.732 ∗ AEX 52 +2.930 ∗ TYEX 21 −3.395 ∗ TYEX 31 +3.583 ∗ LEX19 ( 2.882) ( −2.966) ( 2.377) ( −2.385) ( 2.020) Copyright © 2010 John Wiley & Sons, Ltd Int J Tourism Res 12, 647–664 (2010) DOI: 10.1002/jtr Ecotourism in the Nicoya Peninsula, Costa Rica setting and benchmarking (Wöber, 2002; Font and Harris, 2004) This is due to the fact that information is not very meaningful if quantified beyond basic statistics for measuring community participation with a tourism enterprise (e.g perceptions towards crowdedness, income and employment) (Moore et al., 2003) When assessing environmental impacts, however, remote sensing data provide information on the differences in land-cover characteristics on spatial and temporal levels and have been used on a wide range of analyses, one of which is forest change detection (DiFiore, 2002; Southworth et al., 2004) Ecotourism will have the best chance of maintaining responsible actions when backed by clear consistent standards The Certification for Sustainable Tourism (CST) developed in Costa Rica, a good example of such a system, monitors a variety of social and environmental impacts including emissions, conservation and protection of fauna and flora, and cultural and economic impacts Objectives The objective of this study was to evaluate the environmental and social impacts of the Punta Islita (PI) eco-lodge, located on Costa Rica’s Nicoya Peninsula We tested the value of ecotourism using PI as a model case study, as a conservation and development tool, and also sought to test the utility and value of the supporting certification system (PI is top-ranked by Costa Rica’s CST) The relevance of certification is that by operationalising definitions of ecotourism, it will endeavour to improve industry performance and influence markets (Font, 2001; Buckley, 2002) Specifically, this study strived to understand residents’ opinions regarding socio-cultural, environmental and economic costs and benefits In addition, this study evaluated the local environmental changes experienced since PI began its operations in 1994, using remote sensing The primary questions that guided our investigation were as follows: What have been the main social, economic and environmental impacts — positive and negative — of ecotourism at the PI Eco-lodge on the Nicoya Peninsula? Copyright © 2010 John Wiley & Sons, Ltd 805 Have conservation efforts at the lodge been of sufficient magnitude and duration to reduce deforestation? Has PI had an identifiable impact on local environmental awareness, and specifically has it contributed to the spread of conservation ethics in the area? METHODS Study site Costa Rica is a forerunner in the development and certification of sustainable tourism businesses and a model for nations seeking to manage tourism responsibly Within Costa Rica, Hotel PI, is acclaimed for its dedication to community development and environmental conservation PI eco-lodge is privately owned and situated amidst secondary forest on the Pacific-facing side of the Nicoya Peninsula (Figure 1) PI eco-lodge developed from a traditional cattle and timber ranch operation This rural area was initially developed for timber extraction in the 1940s, but once the precious timbers were exhausted, the area tuned towards cattle ranching and later to agriculture until government agricultural incentives ended By the mid-1990s, most families had migrated towards urban areas for employment The land was later divided into three independent but interlinked entities — PI eco-lodge, Forestales and Lomas de Islita PI has invested in the local communities through art education, microenterprise development, local workforce training and promotion, economic equality for women and children through handicraft production and general infrastructural improvements Fostering private and public collaboration, PI has combined a for-profit hotel with a community-based foundation Both entities are dedicated to sustainable tourism and adhere to the Costa Rican-based certification system, CST Data collection Data collection used an interdisciplinary ‘nested-scale’ methodology (see Almeyda et al., 2010 for detailed description) combining spatial analyses of forest cover and change using remote sensing with extensive in depth interviews with local households, semiInt J Tourism Res 12, 803–819 (2010) DOI: 10.1002/jtr 806 A M Almeyda et al Figure Close up of Punta Islita properties Red = Lomas, white = Punta Islita Hotel, blue = Forestales (A) A color composite 2001 Landsat image (B) Courtesy of Google Earth Green areas represent forest cover; pink areas are developed, pasture or agricultural areas Figure Nested scales of analysis used in this study, showing key methods used at each level to assess the impact of ecotourism on the Nicoya Peninsula, Costa Rica structured interviews with community leaders and self administered questionnaires for hotel guests Each scale includes opportunities to assess and monitor multi-temporal environmental and socio-cultural changes, as well as building up from information derived from finer scales (Figure 2) The largest scale, ‘landscape’, is equivalent to the entire Nicoya Peninsula and provided the context within which the finer scale levels were evaluated Second largest in extent is the ‘community’ scale, which included five communities within PI’s spatial and cultural areas of influence Third is the ‘household’ scale, which included both households influenced by PI indirectly (such as Copyright © 2010 John Wiley & Sons, Ltd proximity) or directly (through personal employment) The final scale, ‘Punta Islita’, spatially includes the PI property, as well as two additional adjacent properties managed by PI ownership, and socially includes individuals associated with the eco-lodge (owners, operators, tourists and employees) The complementary scales and methods used at multiple locations provided a better understanding of the diverse ecotourism impacts Spatial analyses Landsat satellite imagery spanning from 1975 until 2008 was acquired from online databases Images were georeferenced to a base image generated through a NASA Int J Tourism Res 12, 803–819 (2010) DOI: 10.1002/jtr Ecotourism in the Nicoya Peninsula, Costa Rica directive to generate a global database of orthorectified Landsat imagery covering all terrestrial areas (http://glcf.umiacs.umd.edu/ research/portal/geocover/) Root mean square errors of the warp models used to georeference images to the base image was less than 0.5 pixel (or 15 m) Radiance imagery was converted to pseudo-reflectance using log residuals Cloud and cloud shadow areas within each image were manually removed Multiple image dates were merged to minimise land area not visible due to clouds or striping due to scan line correction malfunction in the Landsat ETM+ 2007– 2008 images Following collection of forest, pasture and water spectral endmembers unique to each merged image, the spectral angle mapper algorithm was used to classify each 30 m × 30 m Landsat pixel per satellite image into forest and non-forest classes Areas outside the Nicoya Peninsula were removed from the study area An accuracy assessment was performed using 126 ground control points spread throughout agriculture, pasture, secondary forest and forest plantations within the Nicoya Peninsula Geographic coordinates of these locations were collected during September 2008 using a handheld geographic positioning system This assessment was performed against the 2008 image classification as rapid land cover transformations have been occurring throughout the peninsula We employed a classification approach independent of field points and designed to encompass the variability due to satellite sensor differences Analyses of the 2008 forest/non-forest classification calculated a user’s accuracy and kappa coefficient of 92% and 0.83, respectively Forest plantations were classified as forests 88% of the time Analyses of land cover changes were performed at multiple scales (Figure 3): (1) the PI eco-lodge, Lomas and Forestales properties; (2) a 1.5-km buffer surrounding the five main communities within which PI employees lived; (3) a 1.5-km buffer surrounding PI property; (4) and the Pacific and interior sections of the Nicoya Peninsula Temporal analysis included only pixels that were not obstructed by atmospheric issues in any study year These scales assessed land cover changes in the PI property compared with surrounding areas and the entire peninsula Comparison between change Copyright © 2010 John Wiley & Sons, Ltd 807 trajectories at the community scale assessed general impacts of PI employees and non-tourism-affiliated neighbours on the peninsula’s forest cover Socio-economic analyses Socio-economic data were collected using questionnaires, related data, and formal and informal interviews Questionnaires were applied, and field visits took place in September 2008 Interviews were conducted with PI owners, operators and managers, as well as with locals involved in PI’s community projects In-depth questionnairebased surveys were conducted with a sample of PI employees and neighbours not working at PI eco-lodge At the landscape level, PI staff and neighbours were interviewed about the development of the Nicoya Peninsula At the community scale, semi-structured interviews were conducted with community elders and participants of PI-supported development projects (N = 15) We asked elders about economic activities and cultural values that pre-date PI We asked art group participants on the impacts the projects have had on their well-being At the household level, researchers conducted in-depth questionnaire-based surveys with household heads, including both husband and wife whenever possible, for a total of 63 households (45 had at least one member employed by PI and 17 not employed by PI, but may receive income from tourism-related activities) In-depth surveys included household demography and education, land use practices, income and expense sources, perceived tourism impacts and knowledge of key concepts in ecology and ecotourism Contingency tables, Pearson coefficients and Wilconxon/ Kruskal–Wallis rank sum analyses statistically compared PI employees and their neighbours Several questions required quantitative interpretation or transformation prior to statistical comparison, presented side by side with the results Non-parametric statistics were used to avoid skewness and non-normality, as many of the variables were ordinal or categorical At the PI scale, semi-structured interviews were conducted with management on PI’s past, present and future; the relationship between PI and local communities and institutions in the Peninsula; and PI staff PI guests Int J Tourism Res 12, 803–819 (2010) DOI: 10.1002/jtr 808 A M Almeyda et al Figure Study areas and community names addressed in this study The Punta Islita hotel property (red), Lomas (light blue) and Forestales (orange) properties, shown in the close up image, were compared with land cover changes in areas surrounding these properties, areas surrounding the five principal towns within which Punta Islita employees lived and with Pacific and interior portions of the entire Nicoya Peninsula (divisions shown in red) Green areas on the peninsula are forested areas, whereas pink areas are developed, pasture or agricultural areas took short, self-administered questionnairebased surveys on their trip, trip expenses and tourism development in PI A random sample of 45 employees was selected for in-depth surveys, representing 32% of total employment (N = 140) Employees identified their closest neighbours, if applicable, not employed by a tourism-related company The same indepth survey was then conducted with these neighbours, providing a control for issues of Copyright © 2010 John Wiley & Sons, Ltd spatial auto-correlation of access and environmental variables RESULTS Spatial analyses An increase in forest cover occurred from 1975 to 1987 (Table 1) and from 1975 to 2008 for all study scales (Figure 4) From 1975 to 1987, the Int J Tourism Res 12, 803–819 (2010) DOI: 10.1002/jtr Ecotourism in the Nicoya Peninsula, Costa Rica 809 Table Proportion forest cover at all scales during study years Proportion forest in study year Study area PI property PI Lomas PI Forestales PI, all 1.5-km buffer Colonia del Valle, 1.5-km buffer Corozalito, 1.5-km buffer Islita, 1.5-km buffer Pilas de Bejuco, 1.5-km buffer Pueblo Nuevo, 1.5-km buffer Nicoya Peninsula Nicoya Pacific Coast Nicoya Interior Area (ha) 2008 2001 1987 1975 27 233 236 1870 468 713 667 783 625 502 353 202 639 299 714 0.76 0.56 0.82 0.7 0.45 0.35 0.66 0.41 0.36 0.36 0.51 0.26 0.66 0.61 0.81 0.71 0.32 0.35 0.64 0.49 0.44 0.43 0.57 0.34 0.78 0.88 0.88 0.88 0.57 0.81 0.88 0.8 0.86 0.51 0.68 0.39 0.04 0.21 0.14 0.3 0.08 0.08 0.16 0.19 0.22 0.19 0.27 0.14 All pixels having either cloud issues or water present in any study year were removed prior to calculating spatial statistics PI, Punta Islita Figure All categories ranked according to the proportion increase in forest cover from 1975 to 1987 and from 1975 to 2008 Copyright © 2010 John Wiley & Sons, Ltd Int J Tourism Res 12, 803–819 (2010) DOI: 10.1002/jtr 810 A M Almeyda et al entire peninsula increased from 19% to 51% forest cover However, from 1975 to 2008, peninsula trends are split; the interior portion decreased to 26%, while the Pacific coast region decreased to 36% forest cover The PI eco-lodge property increased in forest cover from 4% to 76% during this same time period, with only a small decrease of 2% from 1987 to 2008 The PI property remains the scale most reforested in both forest cover change and total forest cover The surrounding communities have experienced similar changes in forest cover from 1975 to 1987, with all increasing from almost entirely pastoral or agricultural areas in 1975 to 50–80% forest cover by 1987 Of the five study communities, the community of Islita, followed by Colonia del Valle, experienced the greatest reforestation between 1975 and 2008 Socio-economic surveys: guests In total, 39 tourists filled out questionnaires Demographic information was obtained but is not presented in this paper Of relevance to this study, median visit to PI was days of a total median 9-day trip to Costa Rica Almost all guests were visiting PI for the first time, were completely satisfied with their stay (mean 4.5/5) and were very likely to return (mean 4.5/5) On average, tourists spent US$1815 while at PI, including travel there from within Costa Rica, with 55% of tourists using a tour package Guests, on average, would be willing to spend an additional US$138 to make the trip possible and US$25 to support the natural and cultural patrimony of the area In general, outdoor beauty and luxury were of greatest importance and met or exceeded expectations Two exceptions occurred in food, dining and general affordability Of less importance were local customs, architecture and sustainability, although the quality of these categories exceeded expectations Of no importance to guests were medical and dental services, likely as few required them during their stay, and entertainment (Table 2) Table Guest perceptions on importance and quality of natural and societal categories at PI Category Scenic landscapes Lodgings Food and dining Friendly people Cleanliness/waste disposal Personal safety Outdoor recreation General affordability Lack of crowds Information availability Climate Guide services Sustainability/responsibility Roads and transport Communications (Internet, telephone) Local arts and crafts Interesting architecture Local music, dance or customs Medical/dental services Entertainment/nightlife Importance Quality Delta 4.64 4.57 4.46 4.45 4.42 4.38 4.24 4.20 4.17 4.14 4.05 3.89 3.86 3.50 3.30 3.23 3.11 3.00 2.64 2.42 4.67 4.61 4.30 4.79 4.38 4.53 4.27 3.55 4.67 4.24 3.97 4.24 4.52 2.94 3.88 3.97 3.53 3.13 3.03 2.84 −0.03 −0.04 0.16 −0.34 0.04 −0.15 −0.03 0.65 −0.50 −0.10 0.08 −0.35 −0.66 0.56 −0.58 −0.74 −0.42 −0.13 −0.39 −0.42 Mean value of categories (all with sample size greater than 35) PI, Punta Islita Copyright © 2010 John Wiley & Sons, Ltd Int J Tourism Res 12, 803–819 (2010) DOI: 10.1002/jtr Ecotourism in the Nicoya Peninsula, Costa Rica 811 Table Comparison between PI employees and their neighbours on average values of socio-demographic variables Mean (SD) N Socio-demographic variables Neighbour Employee Neighbour Employee p-value Testa Female head of household % that was born within the Peninsula Years living in current community Years of education Age 66.7 37.3 (20.3) 6.9 (3.2) 49.5 (17.5) 54.8 14.2 (13.2) 8.4 (3.5) 30.1 (8.5) 15 16 16 16 31 31 31 31 0.45 0.0002 0.13