Port Econ J (2001) 1: ♣–♣
002 c Springer-Verlag 2002
Measuring the economic benefits of protecting the Parque Natural do Sudoeste Alentejano e Costa Vicentina from commercial tourism development: results from a contingent valuation survey Paulo A. L. D. Nunes Department of Spatial and Environmental Economics, Free University, Amsterdam, The Netherlands (e-mail:
[email protected]) Received: ♣ ♣, ♣; Accepted: ♣ ♣, ♣
Abstract. This paper focus on a contingent valuation application as to measure the value that the Portuguese households are willing to pay for the recreation and biodiversity benefits provided by a Natural Park in Portugal. A national survey was designed and implemented. It gathered 1678 respondents and focus on three tourism development policy options for the Park. We refer to (1) the WildernessArea commercial tourism development scenario; (2) the Recreational-Area commercial tourism development scenario and, finally, a scenario version which is characterized by the tourism development of both Wilderness and Recreational Areas of the Natural Park. The results show that the Portuguese evaluate the Wilderness-Area and the Recreational-Area differently. In particular, the Portuguese show a higher willingness to pay for preserving the Wilderness-Area free from any commercial tourism development. However, we find no statistical difference between the willingness to pay for the Wilderness-Area and the willingness to pay for the Wilderness-Area jointly with the Recreational-Area. The last result can be interpreted as a consequence of the lack of credibility of the protection scenarios or, more interestingly, as pointing to the eventual presence of warm glow in the stated willingness to pay responses. Introduction This paper focus on a contingent valuation application as to measure the value that the Portuguese are willing to pay for the recreation and biodiversity bene The author wishes to thank John B. Loomis, Colorado State University, St¨ ale Navrud, Agriculture University of Norway, Jeroen van den Bergh, Free University Amsterdam, and two anonymous reviewers for their valuable suggestions on earlier drafts of the manuscript; Olvar Bergland, Agriculture University of Norway, and Richard Carson, University of California-San Diego, for sharing their expert software know-how. The author also acknowledges the financial support of the Calouste Gulbenkian Foundation, the European Commission, and the Portuguese Institute for the Conservation of Nature.
2
P. A. L. D. Nunes
fits provided by a Natural Park in Portugal. Combining the use of maps, photos and computer generated photos, a set of three survey instruments are designed, corresponding to three policy protection programs: the Wilderness Area the Recreational-Area, and the jointly Wilderness-Recreational-Areas protection program. In each questionnaire, the 1678 respondents are asked directly about their maximum willingness to pay to prevent the tourism development scenario as described in the survey instrument. The paper is divided into five major sections. The first section describes the current situation concerning the management of the Natural Park, identifies the alternative survey protection scenarios, and presents particular aspects of the research design formulation and data collection. The second section presents the econometric model that is underpinning the selected elicitation question format, explores the choice of the underlying distribution prior, and identifies possible “no-no” protest arguments. The third section presents some socio-economic descriptive statistics of the questionnaire and the willingness to pay estimates across the different survey described protection programs. In the fourth section formal testing procedures are performed as to examine the ex ante presumptions on the magnitude of the WTP estimates. Section five concludes by discussing the implications of the WTP test results in terms of policy guidance.
1. Statement of natural resource problem 1.1. The Alentejo coastline The Alentejo coastline constitutes one of the least urbanized littoral areas of Portugal. Like in many other countries, the Portuguese authorities have been observing an intensification of the conflicts and disputes over the alternative use possibilities of such natural site. On one hand, the Portuguese Governmental Agency for Nature Protection,1 claims the preservation of the natural area. Recently, more specifically in 1995, ICN established the Parque Natural do Sudoeste Alentejano e Costa Vicentina, shortly, the Alentejo Natural Park. In this protected area roads, commercial and tourism development, mechanical equipment and other improvements are prohibited [13]. On the other hand, tourism industry, together with the local municipalities, has been claiming for the development of the tourism potential of the Natural Park and respective creation of employment in the area. The question that the present paper proposes to address is to determine the value that the Portuguese households place on the different tourism policy protection programs that recently have been discussed between the tourism industry and the Portuguese Government. Since the major values in dispute are typically referred to as non-use or existence values [11], the contingent valuation method as been selected as the measurement approach since it is the valuation technique capable of including the non-use value component when measuring the total value of the natural resource [16]. 1
Instituto de Conservac¸a˜ o da Natureza (ICN).
Measuring the economic benefits
3
1.2. Hypothetical protection scenarios Given the present zoning [14], it is possible to characterize the Alentejo Natural Park in terms of two major zones: the Wilderness- Area and the RecreationalArea. The first refers to the geographical area of Park that is allocated to the protection of the local biodiversity: the visitors’ access is here restricted and roads, commercial development, mechanical equipment and other improvements are prohibited - the Park’s nonuse value component. The second category refers to the geographical area of the Park that is allocated to recreational human use: it is open to all visitors where they are able to enjoy a set of recreational activities in a natural environment - the Park’s use value component. Given the present zoning, a set of three protection scenarios is designed, which corresponds to three policy management programs. One policy option is characterized by the partial development of the Alentejo Natural Park and the urbanization of the WA for commercial tourism; a second policy option is characterized by the partial development of the Alentejo Natural Park and the urbanization of the RA for commercial tourism; finally, a policy option that describes full development of the Alentejo Natural Park and the urbanization of both WA and RA for tourism commercial development (see Fig. 1).
Version 1
Interpretation Version 2
Interpretation Version 3
Interpretation
Policy On WA: protected RA: Protected WTP to prevent Policy On WA: protected RA: Protected WTP to prevent Policy On WA: protected RA: Protected WTP to prevent
Policy Off WA: for tourism development RA: Protected Tourism Development in WA Policy Off WA: protected RA: for tourism development Tourism Development in RA Policy Off WA: for tourism development RA: for tourism development Tourism Development in WA and RA
Fig. 1. Survey protection scenarios
The survey formulation of each policy protection program combines the use of narrative and visual material, including the use of maps, photos and computer generated photos, in order to maintain respondent interest and enhance the comprehension of the hypothetical scenario description. The narrative material was based in a multidisciplinary work, involving the participation of ICN biologists with solid experience in the field, and making the use of all available scientific information.
4
P. A. L. D. Nunes
1.3. The contingent valuation survey Before putting the final survey instrument into the field, an overall up-grading and extensive fine-tuning process was engaged during a ten-month period. It was used state-of-the-art techniques in developing questionnaires and followed closely the guidelines recommended by the NOAA panel [1], [5] including the use of focus groups, field pre-testing and in-home personal interviews. The national survey was implemented in September 1997 and executed by the Survey Department of the Portuguese Catholic University.2 The selected sampling selection mechanism is characterized by a two-stage area probability sample [18]. In the first stage a set of Primary Sampling Units (PSUs) is drawn, which corresponds to the selection of a sample of parishes across Continental Portugal. In the second stage a set of Residential Dwelling Units (RDUs) is drawn from the selected PSUs, which corresponds to the households that will be visited by the interviewer. The operationalization of the two-stage area probability sampling mechanism is done according to a statistical software package developed by the CESOP, which guarantees the quality of the collected data and the socio-economic representativeness of the total 3898 parishes as registered in the 1991 Decennial Census – see [15] for more details. In the execution of the contingent valuation (CV), the interviewer teams paid in-home personal visits to 3597 households but 21% of them were not reachable since nobody answered the door. From households that were successfully contacted, we received a total of 1678 completed interviews.3 To better mimic price taking in market behavior, the respondents are asked whether they are willing to pay a given monetary amount as to continue the protection of the Alentejo Natural Park. The monetary amount is stated in the instrument survey and varies randomly from respondent to respondent. This question format is referred to as take-it-or-leave-it, i.e., a dichotomous-choice (DC) format question [2], [3]. As to improve the statistical efficiency, a follow-up valuation question was included [8]. Figure 2 describes the complete elicitation procedure. The bid design used across the three survey versions is presented in Table 1. For example, if the respondent gets a bid card such that she will be asked if her household “would agree to pay 2 400 escudos” for the described protection program; if she answers “yes” then she faces a follow-up question with a higher amount: “would your household still agree to pay 4 800 escudos”. If she refuses the initial bid then in the second round she will be asked for a smaller amount “would your household still agree to pay 1 200 escudos”. Furthermore, before answering to the WTP questions, the respondents are informed that expressed monetary amounts are an annual-once-for-all-payment, which will be deposit in a National Conservation Fund exclusively allocated to the protection of the Alentejo Natural Park. The 2
Centro de Estudos e Sondagens de Opini˜ao P´ublica (CESOP). For a personal interview, the present study reveals a rather low participation response, around 60%. The CV in-person interviews are characterized by higher response rates than, for example, mail surveys. The latter typically range between 20% and 60% [19]. This low participation response goes, however, hand-in-hand with Deaton findings that state that survey researchers have been facing an increasing non-cooperative trend over time [6]. 3
Measuring the economic benefits
5
sequence of responses to the double dichotomous choice response model will be used to infer the respondent’s maximum willingness to pay (WTP). This will be discussed in the following section. +PKVKCN $KF ;'5
01
01
;'5
KPETGCUGF DKF
FGETGCUGF DKF
;'5
01
Fig. 2. Elicitation question format
Table 1. Bid amounts used in the CV (in PTE4 ) Bid card 1 2 3 4
Initial bid (bi ) 1200 2400 4800 9600
Higher bid (bh ) 3600 4800 9600 24000
Lower bid (bl ) 600 1200 2400 4800
2. Analysis of the stated WTP responses 2.1. The econometric model The respondent’s stated WTP responses are assumed to be based on an unobserved continuous variable. Therefore, the respondent’s decisions upon the bid amounts are used as a proxy variable for the unobserved WTP. The underlying idea is that the respondent evaluates her utility in two stages: with and without the survey described protection plan. If she thinks that her own willingness to pay for the scenario exceeds the stated bid, then she would accept to pay or else she refuses it. Consequently, we use a limited dependent variable choice model and explore the variation of the bid values across the sample as to assess the underlying valuation function [12]. Since the double bounded was adopted, for each j respondent four possible response outcomes are possible: “no/no”, j j j j “no/yes”, “yes/no” and “yes/yes” respectively coded as rnn , rny , ryn and ryy binary indicators variables. The contribution to the (log)likelihood function from one observation is then, j j rnn ln F blj ; θ + rny ln F bij ; θ − F blj ; θ j ln F bhj ; θ − F bij ; θ + ryn j ln 1 − F bhj ; θ (2.1) + ryy 4
189 PTE ∼ = 1USD (Sept. 97).
6
P. A. L. D. Nunes
where F (.) is a statistical distribution function with parameter vector θ. Here it is interpreted as the cumulative distribution function of the respondent’s. The sum of these contributions to the likelihood function over the sample is then maximized, Max L (θ) with L (θ) defined as, {θ}
j j N rnn ln F blj ; θ + rny ln F bij ; θ − F blj ; θ + +r j ln F b j ; θ − F b j ; θ + r j ln 1 − F b j ; θ j =1 yn yy i h h ˆ is the solution to the equation the ML estimator for the double-bounded model, θ, ∂L θˆ ∂θ = 0. To come up with such estimates is necessary to assume that the stated WTP responses are distributed according to a particular distribution family. However the “true” underlying distribution of the WTP is unknown. So the choice of the distribution function is an important step for the analysis of the stated WTP responses.
2.2. The choice of the distributional prior For each survey version, the double bounded data was fitted to different family distributions. We use the Weibull distribution, the lognormal distribution, the exponential distribution and the log-logistic. Table 2 contains the Log Likelihood statistic for each of the distributions across the different survey versions. The task is now to choose one of the specifications. For the Weibull and the exponential, the choice is straightforward because these are nested distributions. Using the restricted and unrestricted log-likelihood, the test statistic consents the rejection the exponential distribution in favor of the Weibull. Table 2. Log likelihood statistics for all survey versions Survey version WA RA (WA + RA)
Lognormal −515.25 −360.39 −807.78
Weibull −504.14 −357.79 −796.72
Exponential −522.81 −373.17 −841.17
Loglogistic −517.09 −361.14 −810.67
This test cannot be extended to the lognormal and loglogistic because these distributions are not nested with the Weibull. The choice can be then be based on the Akaike Information Criteria (AIC) for each parametric specification [17]. Table 3 reports the AIC values by survey version and distribution family. According to the AIC, the Weibull and the lognormal distribution functions provide the best fit to our data. Therefore, we explore two parametric models, one using the Weibull and the other the lognormal, in the estimation exercises.
Measuring the economic benefits
7
Table 3. Akaike Information Criteria (AIC) Survey version WA RA (WA + RA)
Lognormal 1035 724 1620
Weibull 1012 720 1597
Loglogistic 1038 727 1625
2.3. The zero-response protests The estimation of the WTP is performed in a sub-sample in which we control for the zero-response protests. The underlying idea is to exclude zero bid responses that are based on protesting arguments rather than a “true” zero valuation. Direct survey information given by the “no-no” respondents is used as to identify such response pattern. In fact, the instrument survey includes a section containing a list of arguments that could possibly justify such answering behavior. The complete list presented in the instrument survey is given in Figure 3. Without having the ambition of being exhaustive, the list tries to capture the possible reasons for “no-no” refusals. From this list, the respondent is asked to choose the most important motive that she thinks is responsible for being unwilling to pay any positive amount of money. For example, the respondent can argue that she has not sufficient money to contribute, reason i), or simply be unwilling to contribute because she thinks that the protection plan constitutes a break to the development of the region, reason vi). As to identify zero protest bids, reasons ii), v), vii) and ix) are considered as the underlying protest arguments. i) ii) iii) iv) v) vi) vii) viii) ix)
I can not afford to contribute with so much money I do not believe in the described payment scheme I prefer to spend that amount of money in other things The proposed protection plan does not worth so much money I do not agree with this type of questions The proposed plan is a break to the development of the region This questionnaire is not the best way to approach the topic Nature has no price and therefore I do not accept any tourism development of the natural area I do not accept any voluntary contribution, the protection of natural areas is a responsibility of the government Fig. 3. Possible reasons for not being willing to pay
Unlike the others, these arguments are not associated with budget constraint situations neither with weak preference motivations towards the protection programs. On the contrary, they reflect a set of respondent’s objections concerning the lack of seriousness of the proposed payment mechanism or credibility in public institutions to guarantee the preservation of the Alentejo Natural Park (reason ii), or the respondent’s reluctance to the questionnaire as an approach to deal with the topic, respectively reasons v) and vii). Furthermore, the zero
8
P. A. L. D. Nunes
response protest may underline the respondent’s explicitly disapproval towards the payment mechanism – reason ix). 3. Estimation the WTP responses 3.1. Some basic statistics of the questionnaire The questionnaire’s demographics and socio-economic characteristics indicate that the median respondent is between 40 and 49 years old, married, and having one child. It was also verified that the majority of the respondents, about 56.2 percent of the sample, are women and have completed 7 to 9 schooling years. Furthermore, respondents were asked to report their net household income. We got an overall response rate of approximately 83 percent. The median household income was in the 150.000$00 − 299.000$00 category. Nevertheless, around 20 percent of the sample reported a household income lower than 75.000 escudos per month (see results in Table 4). Table 4. Questionnaire demographic data
20–29 years 19.5 One 7.5 75 Six 21.8 26.9 29.1 14.0 0.0 Monthly net income (in 103 PTE) >150 >300 >451 >600 >800 < 1200
