Dealing with preference uncertainty in contingent ...

Ecological Economics 88 (2013) 76–85

Contents lists available at SciVerse ScienceDirect

Ecological Economics journal homepage: www.elsevier.com/locate/ecolecon

Analysis

Dealing with preference uncertainty in contingent willingness to pay for a nature protection program: A new approach Louinord Voltaire a,⁎, Claudio Pirrone b, Denis Bailly b a b

UMR-AMURE, Université de Brest, 12, rue de Kergoat, CS 93837, 29238 Brest Cedex 3, France UMR-AMURE, Université de Brest, 12, rue de Kergoat, CS 93837, 29200 Brest Cedex 3, France

a r t i c l e

i n f o

Article history: Received 8 December 2011 Received in revised form 11 January 2013 Accepted 18 January 2013 Available online 24 February 2013 JEL Classification: Q20 Q26 Keywords: Contingent valuation Preference uncertainty Payment card Uncertainty calibration Willingness to pay Nature protection

a b s t r a c t In this paper, we propose an alternative preference uncertainty measurement approach where respondents have the option to indicate their willingness to pay (WTP) for a nature protection program either as exact values or intervals from a payment card, depending on whether they are uncertain about their valuation. On the basis of their responses, we then estimate their degree of uncertainty. New within this study is that the respondent's degree of uncertainty is “revealed”, while it is “stated” in those using existing measurement methods. Three statistical models are used to explore the sources of respondent uncertainty. We also present a simple way of calculating the uncertainty adjusted mean WTP, and compare this to the one obtained from an interval regression. Our findings in terms of determinants of preference uncertainty are broadly consistent with a priori expectations. In addition, the uncertainty adjusted mean WTP is quite similar to the one derived from an interval regression. We conclude that our method is promising in accounting for preference uncertainty in WTP answers at little cost to interviewees in terms of time and cognitive effort, on the one hand, and without researcher assumptions regarding the interpretation of degrees of uncertainty reported by respondents, on the other. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Researchers have developed a variety of individual preference-based methods for estimating the demand for environmental goods in order to cope with the absence of market prices (for a review see Champ et al., 2003). One of the most widely used methods is contingent valuation (CV). Through surveys, CV presents individuals with a hypothetical market for a change in the quantity and/or quality of a particular non-market good and elicits preferences by asking them about their willingness to pay (WTP). The amount of money they are willing to pay thus provides a monetary indication of their preferences for the good. The microeconomic theory of consumer behavior is based on the assumption that individuals have well-defined preferences for any choice they are faced with (the assumption of completeness) (Pindyck and Rubinfeld, 2005). In the context of non-market valuation, this implies that for any change in the provision of a particular good, each individual is capable of expressing his preferences in monetary terms by stating an exact WTP (Hanemann et al., 1996). Empirical CV studies show, however, that some people feel uncertain about their answers to valuation questions, and therefore are unable to name a specific sum (e.g. Bateman et al., 2005; Håkansson, 2008; Hanley et al., 2009). Although a single ⁎ Corresponding author. Tel.: +33 2 98 01 70 87; fax: +33 2 98 01 69 35. E-mail addresses: [email protected] (L. Voltaire), [email protected] (C. Pirrone), [email protected] (D. Bailly). 0921-8009/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ecolecon.2013.01.009

unifying theoretical model explaining such uncertainty has not yet emerged (Brouwer, 2012), a consensus seems to exist among researchers on a number of underlying hypotheses. In Shaikh et al. (2007) some of these are listed, including: (1) lack of experience or unfamiliarity with the good being valuated; (2) prices of both substitutes and complementary goods; (3) insufficient information about the hypothetical market presented in the questionnaire; (4) inability to make a tradeoff between the commodity offered and their money, apart from the hypothetical nature of the exercise; (5) difficulty of understanding the policy proposed and the way in which it would be achieved. Another argument is provided by Svedsater (2007) who argues that respondent (or preference) uncertainty is related to the fact that interviewees have insufficient time to think about the valuation task. Different approaches have been developed and applied for allowing people to express their uncertainty when answering WTP questions. The information collected is often used to correct the disparity sometimes observed between actual and stated payments (e.g. Blumenschein et al., 1998; Champ and Bishop, 2001; Champ et al., 1997; Ethier et al., 2000; Poe et al., 2002), a phenomenon known as hypothetical bias (Schulze et al., 1981). Overall, it was found that these approaches can be effective at mitigating this bias (see Champ et al., 2009; Morrison and Brown, 2009). However, current preference uncertainty elicitation approaches are not without their problems (Akter et al., 2008; Loomis and Ekstrand, 1998), which suggests the need for further development. In this paper, we make an attempt to

L. Voltaire et al. / Ecological Economics 88 (2013) 76–85

77

overcome these problems by proposing an alternative way for capturing respondent uncertainty. In addition, we empirically explore the determinants of respondent uncertainty. According to Akter et al. (2008) very little is known about this issue. Finally, we develop a calibration technique of collected uncertainty information for estimating the mean WTP. The good being valuated is a nature reserve project in the gulf of Morbihan (France). As is well known, prior to any public decisionmaking, an economic valuation is recommended to justify the public intervention in question. To this end, beneficiaries have to be clearly identified, and asked about their preferences. In general, when considering nature protected areas, nature-based tourists are regarded as one of the main beneficiaries (Secretariat of the Convention on Biological Diversity, 2008), and as such, are asked about their WTP for those areas (e.g. Baral et al., 2008; Dharmaratne et al., 2000; Mmopelwa et al., 2007; Walpole et al., 2001). Therefore, within an ex-ante valuation framework of new nature reserves, we are interested in determining the amount of money that tourists would be willing to pay to benefit from these areas.1 The remainder of the paper is structured as follows. In Section 2, we present the main preference uncertainty elicitation approaches applied and their weaknesses. This is followed by an introduction to our approach. In Section 3, the empirical attempts to explain preference uncertainty are explored. Section 4 describes the case study and data collection methods. The statistical results are presented in Section 5, and the determinants of preference uncertainty are discussed in Section 6. The calibration technique is presented in Section 7, and concluding comments are provided in Section 8.

Despite the popularity of these approaches, they present a number of problems. First, with regard to the DCU approach, it is implicitly assumed that all interviewees interpret the numerical or percentage certainty scale in the same way (Loomis and Ekstrand, 1998). That is, if two individuals A and B each mark a rating of 4 or 40%, then they are considered to have the same level of uncertainty. This assumption works against some empirical findings in the stated preference literature, which show that respondents sometimes exhibit a “scale preference” in which certain individuals tend to be low raters or high raters (Roe et al., 1996; MacKenzie, 1993 cited by Loomis and Ekstrand, 1998). Third, with regard to the MBU approach, while researchers usually assume a uniform interpretation of the verbal certainty scale by respondents (e.g. Alberini et al., 2003), the inverse phenomenon appears to be a more reasonable assumption due to the subjective nature of some used statements (e.g. “probably yes”, “maybe yes” or “very uncertain”, “not unlikely”) (Hanley et al., 2009). 2 Fourth, regarding the MBU and TWPL approaches, the valuation process encourages respondents to report WTP as a range, rather than as a single point (Hanley et al., 2009; Vossler and McKee, 2006). Finally, these approaches, in particular the MBU and TWPL, might prove burdensome and cognitively challenging (Mentzakis et al., 2010) because they require respondents to both understand the logic of the contingent market 3 and think about the level of uncertainty related to their choice to pay or not each proposed amount. Some people, at the moment of interview, for whatever reasons, might be unwilling to invest the time and effort needed to fully exert the valuation task. Consequently, they might express what Alberini et al. (2003) call “false uncertainty”.

2. Respondent Uncertainty Elicitation Approaches

2.2. An Alternative Approach

2.1. Previous Research and Limitations

In this paper we adopt an alternative way to allow for expressions of uncertainty in contingent WTP answers. Respondents are presented with two separate, similar amounts and asked to choose between two WTP answer options: (1) state an exact maximum WTP; and (2) state an interval of WTP. This valuation question bears some resemblance to the classic and interval open-ended (CIOE) elicitation format developed by Håkansson (2008). Similar to this author, it is assumed that all individuals have a true WTP, but some of them cannot state it because they are uncertain. However, they are able to indicate a range in which it certainly falls. Thus, it is expected that people who are certain of their WTP would choose the first option, whereas those who are uncertain would select the second one. To the best of our knowledge, this kind of PC design was first used by Mentzakis et al. (2010). However, their valuation question is completely different from ours: respondents have to indicate their minimum and maximum willingness to accept (WTA); then, they have to assess their degree of uncertainty related to each of these amounts on a scale from 0 (not sure at all) to 10 (absolutely sure). Thus, this approach has the same problems as those presented above. Moreover, respondents are pressured to give uncertain valuation responses, since they have to assess how sure they are about their answers regardless of whether the same amount was chosen as the both minimum WTA and maximum WTA. From our point of view, it can be seen as a variation of the TWPL approach introduced by Mahieu et al. (2012) as it aims to identify individuals' bound amounts more precisely by asking them to rate the certainty of their lower and upper values. Our approach has several advantages. First, no verbal or numerical/ percentage certainty scales are needed. Hence, on the one hand, the

Currently, the most widely used preference uncertainty elicitation approaches are the dichotomous choice uncertainty (DCU), multiple bounded uncertainty (MBU) and two-way payment ladder (TWPL). In the DCU approach, the dichotomous choice “Yes/No” WTP question is followed up by either a numerical certainty scale from 1 (very uncertain) to 10 (very certain) (e.g. Champ and Bishop, 2001; Champ et al., 1997; Loomis and Ekstrand, 1998; Lyssenko and Martínez-Espiñeira, 2012) or a percentage certainty scale of 0% (absolutely uncertain) to 100% (absolutely certain) (e.g. Brouwer, 2012; Chang et al., 2007; Li and Mattsson, 1995; Li et al., 2009). Under the MBU approach, a combination of a payment card (PC) and the polychotomous choice question (Broberg and Brännlund, 2008), the individual faces k bids and is asked to indicate whether he would pay by marking one of multiple responses associated with each bid amount: “definitely yes”, “probably yes”, “not sure”, “probably no” or “definitely no” (e.g. Akter et al., 2009; Alberini et al., 2003; Boman, 2009; Kobayashi et al., 2010; Welsh and Poe, 1998). In the TWPL approach, the respondent is presented with a series of values and asked to tick amounts he would definitely pay, cross off amounts he would definitely not pay, and leave blank amounts for which he cannot say either “definitely yes” or “ definitely no” (e.g. Bateman et al., 2005; Hanley et al., 2009). Recently, Mahieu et al. (2012) developed a modified version of the TWPL approach, arguing that the range of the willingness to pay is not observed in the ordinary approach because the endpoints are not elicited. A third step is thus added to the valuation task in order to identify “more precisely” these endpoints: respondents are required to specify their bound amounts from ones located between the highest amount for which they say “definitely yes” and the lowest amount for which they say “definitely no”. 1 Of course, local people are also concerned by the establishment of a nature protected area. Hence, it would be interesting for future research to investigate their preferences.

2 Combining a graphic rating scale with the verbal certainty scale may help respondents to better interpret these statements. However, this does not solve the problem entirely (Wang and Whittington, 2005). 3 Broberg and Brännlund (2008) argue that valuation questions are likely perceived as difficult by many people due to their hypothetical nature.

78


valuation task is less expensive for respondents in terms of time and cognitive effort, and on the other, this avoids researchers having to make strong assumptions about how respondents interpret those concepts. Second, the cognitive burden is also lower than in the CIOE elicitation format because respondents are proposed a series of amounts. Third, it provides more information than the both MBU and TWPL approaches as it allows people to state an exact amount. Fourth, respondents are not influenced in their decision to report a point or an interval of WTP. 2.2.1. Estimating the Degree of Uncertainty from WTP Responses From WTP responses it is possible to determine each individual's level of uncertainty applying the following reasoning. If an agent A prefers stating his WTP as an interval rather than a precise value, this suggests that he is uncertain regarding the economic value he places on the good of interest. The economic value of a good is measured by the maximum amount of money that someone is willing to pay. Hence, it can be considered that agent A is only uncertain about whether he would pay the reported upper bound of the interval, which is his maximum WTP, i.e. the amount beyond which he is not willing to pay. Beyond the definition of economic value, the implicit assumption that the individual expressing a range has no, or a marginal, uncertainty about the lower bound makes sense. Indeed, since both of these values, upper and lower bounds, are freely chosen by the respondent, he has no obvious reason to report a range within which he is not sure that his true WTP falls. Thus, the stated lower bound of the range should logically be not higher than the actual WTP. As a consequence, it can be reasonably assumed that the individual would certainly pay the amount stated as the lower bound of his range of WTP, if placed in a situation to do so. On the other hand, if an agent B chooses to formulate a precise value rather than an interval, this means that he bears no more than marginal uncertainty about his maximum WTP, since his lower and upper amounts coincide. The degree of uncertainty associated with the maximum WTP reported by each agent can be denoted as: UNCERTAINTY ¼

U i −Li Ui

100

ð1Þ

where Ui and Li are the upper and lower amounts indicated by the subject i respectively. This is consistent with the literature suggesting that the width of the range (Ui–Li) reflects the respondent uncertainty (e.g. Broberg and Brännlund, 2008; Håkansson, 2008; Hanley et al., 2009; Mahieu and Riera, 2010). Note that in Mahieu and Riera (2010) the expression U iU−Li is used as a measure of the width of i

the WTP interval. In the present paper, however, it is taken as a measure of the respondent's degree of uncertainty, i.e. as the likelihood of paying the maximum amount ticked on the card. Finally, it should be noted that our approach is not comparable with all of those presented above in terms of kind of measured uncertainty. Indeed, with the DCU approach, the uncertainty concerns the individual's decision to buy (or not) the good at a given amount, while with ours, it is associated with the amount chosen by the individual from a menu of amounts as his maximum WTP. On the other hand, it can be compared with MBU and TWPL approaches as they deal with the uncertainty relating to different possible amounts that the individual would have to pay in case of purchase. However, it is worth noting that in both MBU and TWPL approaches, the respondent degree of uncertainty is explicitly embedded in each WTP answer, while in ours, it is implicitly embedded in the picked WTP answer. 3. Review of Recent Attempts to Explain Respondent Uncertainty To the best of our knowledge, eight studies have empirically investigated the determinants of respondent uncertainty. They can be

classified into three groups according to the type of econometric model used. The first group estimates an OLS regression of the respondents' rating of their certainty in their both Yes and No responses (Loomis and Ekstrand, 1998) or the size of stated WTP (Ui–Li) (Hanley et al., 2009; Mahieu et al., 2012). Loomis and Ekstrand (1998) find a significant quadratic effect of bid price, implying that “respondents are more certain of their responses at extremely low and high bids, and less certain at intermediate bid levels closest to their maximum WTP”. They also find a significant positive relationship between the self-reported certainty and both respondents' prior knowledge about the good of interest and their visiting the area proposed for protection. In Hanley et al. (2009), respondent uncertainty shrinks with the increasing experience and the coastal water quality differences between study areas (spatial heterogeneity effect). It grows, however, with the degree of respondents' interest in the interview and household income. In Mahieu et al. (2012), women appear to be more certain about their valuation, and respondent uncertainty decreases with age. The second group uses a binary logit model (Samnaliev et al., 2006). Two separate binary logit models for Yes and No answers to the WTP question are estimated where the dependent variable takes the value 1 if the certainty score for the Yes or No response equals 10 (very certain). It is found that people who protest against the user fees in principle are more certain in rejecting the bid levels than others, implying that high certainty to a No response is a way to assert objection towards some component of the contingent market. The third group employs an ordered probit (Akter et al., 2009; Brouwer, 2012; Champ and Bishop, 2001) or logit model (Lyssenko and Martínez-Espiñeira, 2012). Champ and Bishop (2001) find that respondents who are in favor of the contingent wind energy program and willing to pay the extra cost, express the high certainty levels. In Akter et al. (2009), some findings are consistent with that of Champ and Bishop (2001) and Samnaliev et al. (2006). For example, a higher sense of responsibility for contributing to climate change results in a higher likelihood of actually paying for a tree plantation program to store the CO2 emitted into the atmosphere by airplanes. Passengers who believe that the environment should be protected irrespective of the costs are more certain in paying the bid price than others. Other findings partly support that of Loomis and Ekstrand (1998): a negative relationship is detected between bid amount and the stated likelihood of paying, but a quadratic effect is not observed. Contrary to Akter et al. (2009), Brouwer (2012)'s findings fully support that of Loomis and Ekstrand (1998) regarding the quadratic influence of bid price. They are also consistent with that of Samnaliev et al. (2006) regarding the relationship between protest belief and uncertainty levels. In addition, it is found that the familiarity with the information in the questionnaire, and the respondent's degree of confidence in the valuation scenario reduce the uncertainty. Finally, the author highlights the impact of demographic and socio-economic characteristics on preference uncertainty. Lyssenko and Martínez-Espiñeira (2012) find evidence that supports the negative relationship between bid price and degree of uncertainty regarding the WTP for a whale conservation program. However, as in Akter et al. (2009), a quadratic effect is not detected. Results also show that people who say “yes” to a dichotomous Yes/No WTP question are more certain about their answers, and that men are less certain than women. 4. Case Study and Data Collection Method 4.1. Study Area Our study was carried out in the gulf of Morbihan (Fig. 1). This covers a land and sea area of about 750 km 2 and 170 km2 respectively. The resident population amounts to 165,000. From an ecological viewpoint, the gulf is extremely important because of its natural heritage. Indeed, it is rich in vegetal and animal species, some populations of which


account for 46–82% of the total in Brittany, and 10–50% in France (Schéma de mise en valeur de la mer, 2005). Moreover, it has extensive eelgrass beds, which provide shelter for hundreds of species of invertebrates and millions of young fish (Schéma de mise en valeur de la mer, 2005). It is also among the major sites of Western Europe for the reception of migrating birds (Queffelec and Philippe, 2008). At the same time, the gulf of Morbihan is a multi-use area (e.g. agriculture, shellfish farming, fishing, tourism, maritime transport, leisure shipping), which sometimes generates strong tensions in resource or space use (Schéma de mise en valeur de la mer, 2005; Carlisle and Green, 2008). The natural heritage conservation and regulation of the use of space and resources are currently major issues in the gulf. Tourism activity in the gulf of Morbihan has developed considerably over recent decades. With 1.2 million tourists and 25 million nights spent at accommodation establishments a year, this area is the fourth most commonly visited département in France (Chambre de commerce et d'industrie du Morbihan, 2008). According to this source, the total turnover due to tourism currently accounts for 10% of GDP in the département of Morbihan, and the tourist sector provides jobs for about 11,000 people, which is 5.5% of the total employment. The success of tourism depends in large part on the landscape and natural heritage, which are, according to MORGOAT Enquête Tourisme (2005), the second major reason given by tourists (39%) for staying there. Moreover, 96% of tourists surveyed indicate that natural sites are a valuable asset for the gulf of Morbihan, and 33% say they have already visited at least one of those sites. This implies that tourists place a value on natural sites in the gulf, and are therefore concerned with any nature protection program. To conciliate economic development and nature conservation, a regional nature park (PNR) project has been launched in 1999 by an inter-communal organization called “Syndicate inter-communal for the planning of the gulf of Morbihan” (SIAGM). According to the Federation of the regional nature Parks of France,4 a PNR is “a fragile rural area, with a remarkable heritage, organized around a project designed to ensure its permanent protection with respect to its management, and its economics and social development”. Recently, in 2009, the gulf of Morbihan's PNR project was given a favorable opinion by public authorities.5 One of its key objectives is to make the natural heritage an asset for the territory by “preserving and safeguarding the biodiversity” (Syndicat inter-communal d'aménagement du golfe du Morbihan, 2009). Article 6 of the park charter project states that PNR authorities are empowered to undertake actions to protect the diversity of landscape and natural heritage (Syndicat inter-communal d'aménagement du golfe du Morbihan, 2009).6 Among possible actions, the creation of protected areas, such as nature reserves, would be highly envisaged.

4.2. Questionnaire Design The questionnaire consists of four parts. In the first part, information for the visit is collected. In the second part, a series of questions is asked regarding respondents' perception of, and attitude towards nature conservation, and respondents' pro-environment behavioral choice. A four-point Likert scale is used for almost all responses. In the third part the nature protection program is described. It was drawn up in close

4

See its web site: www.parcs-naturels-regionaux.tm.fr. It should be noticed that the Gulf of Morbihan is also protected by a number of designations from national to international level, namely Ramsar site, Natura 2000 site, Nature reserve, Special Protection Areas (SPAs), Special Areas of Conservation (SACs), “Arrêté de protection de biotope” etc. 6 The park charter determines the guidelines for protection, improvements and development, and the measures enabling them to be implemented for the park area. Its signatories are committed for 10 years. The gulf of Morbihan's park charter project is available in the SIAGM's web site: www.golfe-morbihan.fr/siagm.htlm. 5

79

cooperation with some members of SIAGM in order to ensure its realism and credibility. Respondents are first informed about the existence of natural sites (both unprotected and protected areas) within the gulf of Morbihan. Then, they are provided with information that the quality of those natural sites is highly appreciated by nearly 95% of tourists, according to the results of the tourist survey carried out by MORGOAT Enquête Tourisme (2005). At this stage, they are told that several unprotected natural areas are unfortunately under increasing pressure from tourism, threatening not only the biodiversity, but also the continued success of tourism. Subjects are asked to think about a public intervention, which would designate most of these areas as nature reserves. This decision would aim to conserve biodiversity, while at the same time allowing the use of areas in question for recreation and tourism. To reduce a possible scope effect, respondents are presented with a map depicting the exact geographical area covered by the contingent scenario. All concerned sites have been selected due to their importance both from an ecological and tourist point of view. Specifically, we inform interviewees that three nature reserves would be created: the first would be established at the entrance of the gulf of Morbihan, and would include several islands (Berder, Gavrinis, Ile Longue) and islets; the second would be implemented at the southern end of Ile-aux-Moines, one of most famous island destinations in the gulf; the third would be an extension of the current nature reserve, and would include the entire Noyalo river, an important bird nesting area. As protected areas have a central role to play in educating people about biodiversity and nature conservation (Booth et al., 2009), respondents are informed that free public awareness activities of nature conservation would be provided within the new nature reserves. Precisely, these activities would consist of both organizing guided tours and producing information booklets on the gulf of Morbihan's natural features. After describing the hypothetical project, people are offered the opportunity to express their support for it. A 4-point Likert scale allows them to rate the degree to which they are in favor of its implementation. Then, regardless of their response, they are presented with the payment method. As is well known, choosing a payment format is crucial in CV studies since it provides the context for payment (Morrison et al., 2000) and can impact on WTP (for a review of studies comparing payment vehicles, see Campos et al., 2007). A conclusion that seems to emerge from these studies is that one payment vehicle is not unequivocally better than any other. In fact, when selecting a payment format, the simple guideline is to choose the one which is likely to be employed in the real world decision (Bateman et al., 2002). Following this suggestion, we selected two payment formats: an accommodation tax like Jones et al. (2011), because it is widely applied in commercial accommodation to provide additional financing for tourist activities (Bonham et al., 1992), and an entrance fee like Dharmaratne et al. (2000) and Baral et al. (2008), because the Séné nature reserve is partly financed by this. As the purpose of this paper is not to investigate the effect of payment vehicles on preference uncertainty, we decide to show only results related to the entrance fee. Thus, respondents are provided with information that, if the project was implemented, it could be financed via an individual entrance fee, which would provide visitors with a weekly pass valid at all nature reserves in question. Environmental and financial management would be the responsibility of local authorities. At this stage, we introduce our valuation question where subjects are given the option to state their WTP as either exact values or intervals (Box 1). Both the interviewer and the respondent read the valuation question. The respondent is told that, if he bears an exact amount in mind, he has to tick this amount from the two ladders. The interviewer ensures that the amount ticked is the same in the two columns. The respondent is also informed about the option “other” included on the card, which allows him stating his own amounts if he does not wish to

80


Fig. 1. The gulf of Morbihan map.

choose from the proposed ones. Finally, before answering the question, he is verbally encouraged to fully think about the project, his income and expenses during the stay. As may be seen from Box 1, we designed a PC with uniform increments (5€), instead of employing an exponential response scale, as Rowe et al. (1996) suggest. The reason for this choice, which was already made in the literature (e.g. Blaine et al., 2005; O'Garra and Mourato, 2007; Vossler and McKee, 2006), is based on the results of two careful pilot surveys. These were administered to 30 (in November 2006) and 150 (from May to June 2007) tourists respectively, under the conditions to be followed in the final survey. One purpose of these pilot surveys was to determine the most suitable number and level of bids for use in the PC. This process helps to avoid the use of inappropriate PC intervals for the full sample in the main survey (Cameron and Huppert, 1989). Following Boyle and Bishop (1988) cited by Bateman et al. (1995), we employed an open-ended approach to elicit tourists' WTP. The most frequently stated WTP obtained from both pilot surveys were 0€, 5€, 10€, 15€ and 20€. Given these findings, it was decided to design a PC ranging from 0€ to 60€ with steps of 5€. Finally, in the fourth part of the questionnaire, respondents' demographic and socio-economic characteristics are investigated. 4.3. Sampling and Data Collection After the pre-testing step, followed by minor changes in the wording of some questions, the final survey was administered in French and English from July to August 2007 through face to face interviews by six Master students, who were thoroughly instructed about both the objective of the survey questionnaire and the survey protocol. As detailed data regarding the tourist population in the gulf of Morbihan were not available, probabilistic sampling methods could not be applied. Tourists were thus randomly approached, according to common practice (e.g. Jones et al., 2011; Togridou et al., 2006). In accordance with the definition of “tourist” (see Cuvélier, 1998), only people who are not residents in the gulf of

Morbihan and spend at least one night there answered the questionnaire. To sample a wide range of tourists, different categories of survey locations were selected with the help of members of both SIAGM and the Morbihan Tourist Board. Individuals approached were very cooperative with few refusals. In total, 504 French and foreign tourists answered

Box 1 The WTP valuation question. If you were to purchase the entrance fee, could you pick the maximum amount from those listed below that you would be willing to pay? (Option 1). If you cannot state a single amount, could you indicate a range that describes the amount you would be willing to pay? (Option 2) Lower amount

Upper amount

□ €0 □ €5 □ €10 □ €15 □ €20 □ €25 □ €30 □ €35 □ €40 □ €45 □ €50 □ €55 □ €60 □Other:

□ €0 □ €5 □ €10 □ €15 □ €20 □ €25 □ €30 □ €35 □ €40 □ €45 □ €50 □ €55 □ €60 □Other:


the questionnaire. This sample size was, however, reduced to 498 respondents due to inconsistent responses (after giving a positive WTP, six respondents stated they have no intention to visit the nature reserves). 5. Statistical Results 5.1. Sample Characteristics The final sample consists of as many women as men, ranging in age between 18 and 80 years, with an average age of 44 years. Respondents are highly educated; over 67% have university degrees (precisely, 23.9% and 43.3% complete 2 years and more than 2 years of higher education respectively), while the rest has a secondary school education. All people surveyed mentioned their income, which averages around 3418€ a month. French tourists comprise the bulk of respondents (87%). Among them, the largest share comes from Ile de France (26.6%). The tourists' full profile is summarized in Table 1. From this table, it appears that respondents are familiar with the gulf of Morbihan's natural amenities, since the majority of them are not first-time tourists, spending more than 4 days in the area and visiting natural sites (both protected and unprotected areas) during the stay. Moreover, a high level of concern emerges regarding nature conservation, as seen by the average scores of CONCERN, CONTRIBUTE and IRRESPONSABILITY. Another interesting lesson is that tourists are overwhelmingly in favor of the implementation of the contingent project (average score of FAVOR is 3.41), which leads us to expect that a high proportion of them would be willing to pay for it. 5.2. Intention to Pay and Reason of Refusal As expected, few people did not want to pay anything at all (8%). The reasons explaining their major motivations for expressing zero maximum WTP are given in Table 2. These arguments are classified as protest responses as they reflect a dissension with the payment vehicle (II), are related to fairness aspects (III and IV), and a lack of information about the project (I). After comparing protest and non-protest respondents in terms of characteristics described in Table 1, it was determined that the removal of the former does not bias the sample (for a further discussion of true zero values and protest responses, see Strazzera et al., 2003; Meyerhoff and Liebe, 2006). The rest of our analysis is thus based on a sample of 459 subjects. 5.3. Uncertainty About WTP Answers Respondents are predominantly uncertain about their valuation: 80.8% give an interval of amounts, while 19.2% report an exact value. The arithmetical mean, median, mode, minimum and maximum of the lower and upper amounts chosen by subjects are provided in Table 3. An important observation is that none of tourists stated WTP above the maximum of the payment scale (60€). This means that the length of scale proposed adequately covered their range of responses. It is also interesting to note the difference between the mean and median WTP values, indicating a positively skewed distribution of both the lower and upper amounts (coefficients of skewness are 1.39 and 1.50 for the lower and upper amounts respectively). Respondent uncertainty (henceforth UNCERTAINTY) is treated here in three ways. First, it is interpreted as a binary variable, where it takes the value 1 if the subject gives an exact WTP (certain respondent) and 0 otherwise (uncertain respondent). Comparing these two groups of respondents, it appears that uncertain respondents are younger (Mann Whitney Z = − 2.850; p ≤ 0.004) and have lower monthly income (Mann Whitney Z = − 2.714; p ≤ 0.007) than certain respondents. They are also more likely to stay with relatives and friends as compared to the latter (Mann Whitney Z = − 2.104; p ≤ 0.035). Finally, they are less likely to have visited the Séné nature

81

Table 1 Variables constructed from tourists' responses and descriptive statistics. Variables descriptions Demographic and socio-economic characteristics Gender 1 if a male (0 if female) Age Age in years Education 1 if the respondent has a secondary school education (0 if he has university degrees) Income Monthly household income, in euros Nationality 1 if foreign tourist (0 if French tourist) Region 1 if the respondent lives in Ile de France (0 otherwise) Visit characteristics F_VISITOR SHORT_STAY LONG_STAY VLONG_STAY SECOND_HOME RF_HOME COM_ACCOM NATURE_VISIT CULTURE_VISIT

1 if the respondent is a first-time tourist (0 if occasional or repeat tourist) 1 if the length of stay is less than 4 days (0 otherwise) 1 if the length of stay is from 4 days to 1 month (0 otherwise) 1 if the length of stay is more than 1 month (0 otherwise) 1 if the respondent stays in his own second home (0 otherwise) 1 the respondent stays with relatives and friends (0 otherwise) 1 if the respondent stays in commercial accommodation (0 otherwise) 1 if the respondent visits natural sites during the stay (0 otherwise) 1 if the respondent visits monuments or museums during the stay (0 otherwise)

Mean (Std. error) 0.51 (0.50) 44.03 (13.97) 0.34 (0.47) 3418 (1010) 0.14 (0.27) 0.27 (0.44)

0.31 (0.46) 0.13 (0.34) 0.75 (0.43) 0.12 (0.32) 0.15 (0.35) 0.23 (0.43) 0.63 (0.48) 0.76 (0.58) 0.52 (0.50)

Perception of and attitude towards nature conservation, and pro-environment behavioral choice characteristics 3.45 (0.58) CONCERN Rating of the degree to which the respondent is concerned with nature conservation (1 = very unconcerned, 4 = very concerned) 2.88 (0.85) CONTRIBUTE Rating of the degree to which the respondent agrees with the argument “tourists should directly contribute to the financing of nature conservation initiatives” (1 = completely disagree, 4 = completely agree) 2.72 (0.99) IRRESPONSABILITY Rating of the degree to which the respondent agrees with the argument “refusing to financially support nature conservation initiatives is in itself an individual act of irresponsibility” (1 = completely disagree, 4 = completely agree) VIS_SENE 1 if the respondent has already visited the 0.32 (0.46) Séné nature reserve (0 otherwise) Support for the program characteristics FAVOR Rating of the degree to which the respondent is favorable to the implementation of the program (1=very unfavorable, 4=very favorable)

3.41 (0.54)

reserve (Mann Whitney Z = −1.820; p ≤ 0.069), and they rate the level to which they agree with “tourists should directly contribute to the financing of nature conservation initiatives” lower than certain respondents (Mann Whitney Z = −1.748; p ≤ 0.081).

Table 2 Reasons for expressing zero maximum WTP for the program. Arguments I II III IV V VI

Percentage

I have too little information about the project 7.7 The payment vehicle is inappropriate 23.1 The entry to nature protected areas should be free of charge for all 35.9 Only rich people would benefit from these protected areas 33.3 My income does not allow me to pay anything 0 Others 0

82


Table 3 Mean of the lower and upper stated WTP and other associate statistics.

Lower amount Upper amount

30,0%

Arithmetical mean (std. error)

Median

Mode

Minimum

Maximum

8.07€ (7.37) 13.92€ (8.14)

5€ 10€

5€ 10€

0€ 5€

40€ 50€

25,0%

25,6%

23,7%

23,7% 19,2%

20,0% 15,0% 10,0%

Second, respondent uncertainty is interpreted as an ordinal variable, which consists of an uncertainty scale ranging from completely uncertain to completely certain. To design this scale, we estimate each tourist's degree of uncertainty about the stated maximum WTP using Eq. (1). The degrees of uncertainty obtained are then coded as 1 (completely uncertain: UNCERTAINTY = 100%), 2 (highly uncertain: 50% b UNCERTAINTY b 100%), 3 (uncertain: UNCERTAINTY = 50%), 4 (highly certain: 0% b UNCERTAINTY b 50%) and 5 (completely certain: UNCERTAINTY = 0%). 7 Fig. 2 illustrates the corresponding distribution. As shown, a significant proportion of respondents (42.9%) has a degree of uncertainty under 50%. Finally, respondent uncertainty is interpreted as a quantitative variable. It is given by Eq. (1), and thus ranges from 0 (or 0%) to 1 (or 100%). The determinants of uncertainty are explored in the next section. 6. Determinants of Respondent Uncertainty Based on the characteristics of the dependent variable (UNCERTAINTY), three statistical models are estimated: a binary probit model, an ordered probit model and an OLS regression model. UNCERTAINTY is expected to be a function of all variables presented in Table 1. Explanatory variables finally retained in the models are selected using the stepwise elimination technique, and after controlling for the problem of multicollinearity (Table 4).8 We begin the data analysis with the binary probit model. From this, it appears that, ceteris paribus, the certainty among respondents is a form of behavior related to their maturity as older subjects are more likely to express an exact value. Income is found to influence respondent uncertainty; the higher the income level, the higher the probability of being certain about the maximum stated WTP. An explanation is that, as income is a proxy for people's ability to pay, the more it increases, the less uncertain respondents are as to whether they can really pay every amount proposed in the questionnaire. In other words, an increase in income reduces what Petrolia and Kim (2010) call “budgetary uncertainty”. As a result, when faced with the choice between stating either a single value or an interval, subjects with higher incomes are more likely to report exactly the maximum amount they are willing to pay than those with lower incomes. In addition, respondent uncertainty depends significantly on nationality (foreign tourists tend to give exact values as compared to French tourists, suggesting cultural differences) and type of tourism accommodation. Relative to respondents staying in commercial accommodation, those who stay with relatives and friends are more likely to be uncertain about their WTP. The results from the ordered probit and OLS regression models confirm the effects of these explanatory variables with one exception: age is not significant. However, five other variables become significant. This means that uncertainty is best explained here through both these models. These models suggest that subjects who are first-time tourists are less uncertain about their maximum WTP than occasional and repeat tourists.9 This is not in accordance with our expectations, as we assumed 7 Like the example presented in Wang and Whittington (2005), an alternative classification might be: 1 (definitely no to pay the stated upper amount), 2 (probably no), 3 (not sure), 4 (probably yes) and 5 (definitely yes). 8 No presence of collinearity is detected as each tolerance value is above 0.2; the lowest value is 0.80. 9 Other ordered probit and OLS regression models were estimated in which, for example, both first-time and occasional tourists, and occasional tourists were chosen as the reference category. However, no statistically significant effect was detected.

7,8%

5,0% 0,0%

Completely Highly uncertain uncertain

Uncertain

Highly certain

Completely certain

Fig. 2. Histogram of respondents' degrees of uncertainty about their maximum WTP.

that occasional and repeat tourists are, more or less, familiar with the gulf of Morbihan's natural features, and thus have well-established preferences, contrary to first-time tourists. One reason for this unexpected result might arise from the fact that, due to their familiarity with the gulf, occasional and repeat tourists know many natural areas free of charge that would be substitutes for nature reserves offered in the survey. Because of this knowledge, they are then less certain to pay the reported maximum amounts as compared to first-time tourists. 10 Respondents who visit monuments or museums are found to be more uncertain about their maximum WTP. As this type of recreational activity is usually not free in the gulf of Morbihan, other things remaining the same, the amount of money spent on it reduces expenditures for other leisure activities. As a result, people who visit monuments or museums are more likely to face uncertainties about whether they would actually pay the stated maximum amount than those who practice free leisure activities, such as walking, hiking, going to the beach and kayaking (the reference). Another interesting finding concerns the significant relationship between respondent uncertainty and attitudes towards paying for nature conservation actions, which are measured by the variable CONTRIBUTE. The more a subject agrees with the statement “tourists should directly contribute to the financing of nature conservation initiatives”, the less uncertain they are regarding their maximum WTP. That is, the less they object to user fees in principle, the less uncertain they are regarding their maximum WTP. Respondent uncertainty is also found to be related to individuals' pro-environment behavioral choices. Specifically, tourists who have already visited the Séné nature reserve are less uncertain about their valuation. In other words, previous use experience with the type of the good being valuated reduces uncertainty concerning the maximum WTP. Finally, the degree to which people are favorable to the implementation of the contingent program has a significant impact on preference uncertainty. This is interpreted here as respondent trust in the program. The higher the trust level, the less uncertain subjects are about their valuation.

7. Estimation of the Uncertainty Adjusted Mean WTP Having established that uncertainty exists in tourists' valuations, the question that now arises relates to the way to take it into account in the welfare estimation procedure. For subjects stating their WTP as intervals, the upper amount cannot be taken as their true maximum WTP. Indeed, such a choice might overestimate the welfare measure, 10 This argument can be used to explain the effect of the variable RF_HOME on respondent uncertainty, since our statistical results reveal that people who stay with relatives and friends are less likely to be first-time tourists than those who stay in commercial accommodation (Mann Whitney Z = −5.490; p ≤ 0.001). This implies that the former are, more or less, familiar with the gulf of Morbihan's natural amenities as compared to the latter.


83

Table 4 Explaining respondent uncertainty about their maximum WTP. Binary probit model

Ordered probit model

OLS regression model

UNCERTAINTY (1 = certain or exact WTP, 0 uncertain or interval of WTP)

UNCERTAINTY (from 1 = completely uncertain to 5 = completely certain)

UNCERTAINTY = Ui − Li / Ui

Variable

Coefficient (standard error)

Coefficient (standard error)

Constant Age Log income Nationality F_VISITOR SECOND_HOME RF_HOME COM_ACCOM CULTURE_VISIT CONTRIBUTE VIS_SENE FAVOR Log likelihood LR chi square Mc Fadden Pseudo-R2 F-statistic Adjusted-R2 N

−2.848*** 0.009* 0.205* 0.445* 0.116 ns −0.088 ns −0.329* Reference −0.067 ns 0.007 ns 0.220 ns −0.047 ns −215.835 21.97*** 0.048

(1.012) (0.005) (0.108) (0.233) (0.161) (0.213) (0.190)

0.0002 ns 0.337*** 0.385** 0.236** −0.089 ns −0.235* Reference −0.207** 0.119* 0.203* 0.152* −694.345 46.49*** 0.032

(0.145) (0.084) (0.149) (0.129)

459

(0.003) (0.077) (0.187) (0.119) (0.158) (0.127) (0.105) (0.062) (0.110) (0.093)

459

Coefficient (standard error) 1.574*** 0.0002 ns −0.102*** −0.138** −0.066* 0.023 ns 0.074** Reference 0.065** −0.041** −0.066* −0.046*

(0.217) (0.001) (0.023) (0.057) (0.036) (0.052) (0.037) (0.032) (0.019) (0.034) (0.027)

5.04*** 0.090 459

Standard errors are in parentheses: ***, **, and * indicate statistically significant at the 1%, 5% and 10%, respectively, while ns indicates statistically nonsignificant.

since the empirical evidence shows that people expressing higher degrees of uncertainty in their WTP responses are responsible for hypothetical bias. It seems also reasonable to consider that the lower amount cannot be interpreted as the true maximum WTP, as respondents freely chosen to state a range of WTP. On the other hand, it can be seen as what Harrison and Kriström (1995) call “Minimum legal WTP (MLW)”. 11 In the context of our PC format, this would mean that the highest amount that public authorities could expect to receive from tourists is the low end of their stated range of WTP. However, the MLW approach is very conservative as it treats equivalently in terms of degree of uncertainty two subjects A and B expressing intervals of amounts with different sizes (e.g. A: 5€–10€ and B: 5€–40€) or similar size (e.g. A: 10€–15€ and B: 5€–10€). Yet, in both cases, B is more uncertain than A. To solve this problem, we propose a new calibration technique which consists of weighting the size of every interval of WTP by the corresponding degree of uncertainty. This allows estimation of an individual uncertainty adjusted WTP (henceforth adjUWTP). Formally, this is defined as: AdjUWTP ¼ U i −½ðU i −Li Þ UNCERTAINTY:

ð2Þ

It should be noted that the idea of weighting WTP responses by the degree of uncertainty has already been proposed in the literature, although our calibration technique is new. For example, Li and Mattsson (1995) and Evans et al. (2004), using the DCU and MBU approaches respectively, developed a weighted likelihood function model, which consists in incorporating directly the degrees of uncertainty stated by respondents into a log-likelihood function. We test the performance of this approach (Calibration or AdjUWTP Approach) by comparing the mean WTP obtained with that derived from an interval regression (IR), in which the respondent's true WTP (treated as the AdjUWTP here) is assumed to lie somewhere in the reported interval (see Cameron and Huppert, 1989). For respondents stating exact WTP, their responses are converted to a bounded interval data set by creating a switching interval with a lower bound of “Exact WTP− 0.001” and an upper bound of “Exact WTP+ 0.001”. This 11

Harrison and Kriström (1995)'s suggestion initially referred to the doublebounded dichotomous choice format.

procedure was first used by Welsh and Poe (1998) to compare results from open-ended question and closed WTP question. It was recently transposed by Mahieu et al. (2010) in the context of response uncertainty by creating tight intervals for people giving exact WTP. As distributions of WTP amounts are right-skewed, these are transformed into their natural log form. The mean and median are thus calculated by: ′^ 2 E½WTP ¼ exp x β þ σ^ =2

ð3Þ

′^ Median ¼ exp x β

ð4Þ

^ the vecwhere x' is the vector of mean values of explanatory variables, β tor of estimated coefficients and σ^ is the estimated σ.12 The results are reported in Table 5. As expected, the MLW approach produces the lowest mean WTP. Interestingly, the mean WTP estimated from our calibration technique is very close to the one derived from the IR approach. This seems to suggest that, at least for this study, the adjUWTP given by Eq. (2) can be seen as a good proxy for the individual's WTP obtained from the interval regression. 8. Conclusions and Suggestions for Further Research In this paper, we proposed an alternative approach to elicit preference uncertainty in order to overcome concerns associated with current ones. Respondents were faced with two separate sets of amounts, and given the opportunity to report their WTP either as a single value (their maximum WTP) or an interval. Consistent with the literature, it is assumed that respondents who are certain of their valuation would pick an exact WTP, while those who are uncertain would express an interval. On the basis of WTP answers, we estimated respondent's degrees of uncertainty about the economic value (i.e. the maximum amount they are willing to pay) they place on the good in question. We also empirically explored the determinants of preference uncertainty. Finally, we introduced a “direct” technique to adjust for respondent uncertainty in WTP 12 As the aim of this paper is not to explore the determinants of WTP responses, we do not report the estimated models. These are available from the corresponding author on request.

84


Table 5 Expected mean WTP and other associated statistics.

Mean Median N Models estimated

MLW approach

Calibration or AdjUWTP approach

IR. approach

8.76 (3.569) 5.06 459 OLS regression

12.40 (5.160) 7.02 459 OLS Regression

11.00 (2.400) 9.12 459 Interval regression

Standard errors are in parentheses.

answers. This allowed us to estimate the mean and median WTP through an OLS regression. These estimations were compared with those derived from an “indirect” adjustment technique, an interval regression. Despite the particular nature of our uncertainty measurement approach, some results obtained concerning the determinants of preference uncertainty are consistent with those of the few studies with which this approach is comparable in terms of the type of measured uncertainty. In particular, subjects who are familiar with the type of the non-market under consideration are found to be more certain about their WTP, as in Hanley et al. (2009). The other results generally conform to our expectations. For example, the degree to which people object to user fees in principle is positively correlated with their level of uncertainty. The degree to which they are favorable to the nature reserves project reduces their uncertainty. These findings highlight the importance of factors describing the respondent's perception of, and attitude towards the good being valuated on preference uncertainty. Finally, socio-economic and demographic factors also impact on respondent uncertainty, as in Brouwer (2012). Unfortunately, most of our results cannot be compared with those of studies presented in Section 3 due to the difference in type of measured uncertainty. Thus, more research is needed to confirm or refute the direction of the effect of explanatory variables retained here on respondent uncertainty. Moreover, comparing the mean WTP estimated from our calibration technique with the one coming from the well-known interval regression, we find that these values are quite similar. We interpret this finding as lending support for this new technique to account for respondent uncertainty in welfare estimation. Taken together, both lessons suggest that the preference uncertainty measurement approach presented in this paper is a promising one. Still, our approach has limitations, suggesting that much remains to be done before its validity and ability to measure response certainty are established. First, it is possible that some respondents report an interval of amounts, not because they are uncertain about their exact WTP, as assumed here, but because it is not included on the PC. However, we believe that this problem is mitigated in the present case, given that the amounts proposed were chosen from two careful pre-test surveys and respondents were provided the opportunity to state other ones, if necessary. Nonetheless, to avoid any doubt, our variant of PC elicitation format might be combined with the CIOE format. Three sets of similar amounts would be presented to respondents: a single set for respondents who bear an exact amount in mind, and two separate ones for those who have an interval of WTP. For respondents wishing to state amounts that are not indicated on the card, the CIOE valuation question should be explicitly posed in the questionnaire. Second, the degree of uncertainty given by Eq. (1) is sensitive to the bid design, making it possible that an individual A who is considered as more uncertain than B becomes less uncertain when modifying one or several amounts initially shown on a card. That is why the bid design must be based upon rigorous pre-tests, and the option “other” has to be explicitly included on the card. Third, although the assumption underlying the measure of the degree of uncertainty that the subject has no, or a marginal, uncertainty about the lower bound of his stated interval of WTP is reasonable, it would be interesting to test it. For this purpose, a split sample procedure might be applied. For instance, a group of respondents would receive our valuation question, while another group would be faced with that of Mentzakis et al. (2010). Third, the minimum WTP and maximum

WTP are provided by individuals, as in Håkansson (2008). It would be good to investigate whether this freedom leads a hypothetical uncertainty bias, i.e. whether it encourages people to overestimate their uncertainty. Again, a split sample design might be used where a group of subjects would be placed in a hypothetical purchase situation, and another group would receive a real purchase question. However, it should be noted that the calibration technique given by Eq. (3) allows mitigating the effect of a possible overestimation of the uncertainty on the welfare measure. Finally, besides the issue of preference uncertainty, our results provide decision makers with information in the current context of increasing demand for nature conservation in the gulf of Morbihan. It is well known that successful implementation of any nature conservation program should overcome two major obstacles: (1) to show that benefits from this are greater than total costs and (2) if this is the case, to find funding sources (Dharmaratne et al., 2000). In this paper, we demonstrated that tourism derives benefits from the establishment of nature reserves. A future analysis could focus on appropriate mechanisms to recover these benefits.

Acknowledgements We would like to thank Jean Boncœur and three anonymous referees for many insightful comments. We would also like to thank Ben Tomlinson and Daniel Pomerleano for the English language review. Errors remain our responsibility.

References Akter, S., Bennett, J., Akhter, S., 2008. Preference uncertainty in contingent valuation. Ecological Economics 67, 345–351. Akter, S., Brouwer, R., Brander, L., van Beukering, P., 2009. Respondent uncertainty in a contingent market for carbon offsets. Ecological Economics 68, 1858–1863. Alberini, A., Boyle, K., Welsh, M., 2003. Analysis of contingent valuation data with multiple bids and response options allowing respondents to express uncertainty. Journal of Environmental Economics and Management 45, 40–62. Baral, N., Stern, M.J., Bhattarai, R., 2008. Contingent valuation of ecotourism in Annapurna conservation area, Nepal: implications for sustainable park finance and local development. Ecological Economics 66, 218–227. Bateman, I.J., Langford, I.H., Turner, R.K., Willis, K.G., Garrod, G.D., 1995. Elicitation and truncation effects in contingent valuation studies. Ecological Economics 12, 161–179. Bateman, I.J., Carson, R.T., Day, B., Hanemann, M., Hanley, N., Hett, T., Jones-Lee, M., Loomes, G., Mourato, S., Özdemiroglu, E., Pearce, D.W., Sugden, R., Swanson, J., 2002. Economic Valuation with Stated Preference Techniques: A manual. Edward Elgar Publishing Ltd, Cheltenham, UK. Bateman, I., Cooper, P., Georgiou, S., Navrud, S., Poe, G., Ready, R., Riera, P., Ryan, M., Vossler, C., 2005. Economic valuation of policies for managing acidity in remote mountain lakes: examining validity through scope sensitivity testing. Aquatic Sciences—Research Across Boundaries 67, 274–291. Blaine, T.W., Lichtkoppler, F.R., Jones, K.R., Zondag, R.H., 2005. An assessment of household willingness to pay for curbside recycling: a comparison of payment card and referendum approaches. Journal of Environmental Management 76, 15–22. Blumenschein, K., Johannesson, M., Blomquist, G.C., Liljas, B., O'Conor, R.M., 1998. Experimental results on expressed certainty and hypothetical bias in contingent valuation. Southern Economic Journal 65, 169–184. Boman, M., 2009. To pay or not to pay for biodiversity in forests—what scale determines responses to willingness to pay questions with uncertain response options? Journal of Forest Economics 15, 79–91. Bonham, C., Fujii, E., Im, E., Mak, J., 1992. The impact of the hotel room tax: an interrupted time series approach. National Tax Journal 45, 433–441. Booth, J.E., Gaston, K.J., Armsworth, P.R., 2009. Public understanding of protected area designation. Biological Conservation 142, 3196–3200. Broberg, T., Brännlund, R., 2008. An alternative interpretation of multiple bounded WTP data—certainty dependent payment card intervals. Resource and Energy Economics 30, 555–567. Brouwer, R., 2012. A mixed approach to payment certainty calibration in discrete choice welfare estimation. Applied Economics 43, 2129–2142. Cameron, T.A., Huppert, D.D., 1989. OLS versus ML estimation of non-market resource values with payment card interval data. Journal of Environmental Economics and Management 17, 230–246. Campos, P., Caparros, A., Oviedo, J.L., 2007. Comparing payment-vehicle effects in contingent valuation studies for recreational use in two protected Spanish forests. Journal of Leisure Research 39, 60–85. Carlisle, M., Green, D.R., 2008. A Template for Human Impacts Evaluation: Compiled as part of the Corepoint Project. University of Aberdeen, p. 70.

L. Voltaire et al. / Ecological Economics 88 (2013) 76–85 Chambre de commerce et d'industrie du Morbihan, 2008. Profils et tendances du tourisme dans le Morbihan: Un impact économique décisif. Chambre de commerce et d'industrie du Morbihan, Morbihan, p. 24. Champ, P.A., Bishop, R.C., 2001. Donation payment mechanisms and contingent valuation: an empirical study of hypothetical bias. Environmental and Resource Economics 19, 383–402. Champ, P.A., Bishop, R.C., Brown, T.C., McCollum, D.W., 1997. Using donation mechanisms to value nonuse benefits from public goods. Journal of Environmental Economics and Management 33, 151–162. Champ, P.A., Boyle, K.J., Brown, T.C., 2003. A Primer on Nonmarket Valuation. The economics of non-market goods and resources. Kluwer Academic Publishers, Dordrecht. Champ, P.A., Moore, R., Bishop, R.C., 2009. A comparison of approaches to mitigate hypothetical bias. Agricultural and Resource Economics Review 38, 166–180. Chang, J.-I., Yoo, S.-H., Kwak, S.-J., 2007. An investigation of preference uncertainty in the contingent valuation study. Applied Economics Letters 14, 691–695. Cuvélier, P., 1998. Anciennes et nouvelles formes de tourisme: une approche socioéconomique Montréal. Dharmaratne, G.S., Yee Sang, F., Walling, L.J., 2000. Tourism potentials for financing protected areas. Annals of Tourism Research 27, 590–610. Ethier, R.G., Poe, G.L., Schulze, W.D., Clark, J., 2000. A comparison of hypothetical phone and mail contingent valuation responses for green-pricing electricity programs. Land Economics 76, 54–67. Evans, M., Flores, N., Boyle, K., 2004. A decision theoretic approach to modeling multiplebounded uncertainty choice data. Working Paper, Center for Economic Analysis, Department of Economics, University of Colorado, Boulder. Håkansson, C., 2008. A new valuation question: analysis of and insights from interval open-ended data in contingent valuation. Environmental and Resource Economics 39, 175–188. Hanemann, W.M., Kristrom, B., Li, C.Z., 1996. Nonmarket valuation under preference uncertainty: econometric models and estimation. Hanley, N., Kriström, B., Shogren, J.F., 2009. Coherent arbitrariness: on value uncertainty for environmental goods. Land Economics 85, 41–50. Harrison, G.W., Kriström, B., 1995. On the interpretation of responses in contingent valuation surveys. In: Johansson, P.-O., Mäler, K.-G. (Eds.), Current Issues in Environmental Economics. Manchester University Press, Manchester. Jones, N., Panagiotidou, K., Spilanis, I., Evangelinos, K.I., Dimitrakopoulos, P.G., 2011. Visitors' perceptions on the management of an important nesting site for loggerhead sea turtle (Caretta caretta L.): the case of Rethymno coastal area in Greece. Ocean & Coastal Management 54, 577–584. Kobayashi, M., Rollins, K., Evans, M.D.R., 2010. Sensitivity of WTP estimates to definition of ‘Yes’: reinterpreting expressed response intensity. Agricultural and Resource Economics Review 39, 37–55. Li, C.-Z., Mattsson, L., 1995. Discrete choice under preference uncertainty: an improved structural model for contingent valuation. Journal of Environmental Economics and Management 28, 256–269. Li, H., Jenkins-Smith, H.C., Silva, C.L., Berrens, R.P., Herron, K.G., 2009. Public support for reducing US reliance on fossil fuels: investigating household willingness-to-pay for energy research and development. Ecological Economics 68, 731–742. Loomis, J., Ekstrand, E., 1998. Alternative approaches for incorporating respondent uncertainty when estimating willingness to pay: the case of the Mexican spotted owl. Ecological Economics 27, 29–41. Lyssenko, N., Martínez-Espiñeira, R., 2012. Respondent uncertainty in contingent valuation: the case of whale conservation in Newfoundland and Labrador. Applied Economics 44, 1911–1930. Mahieu, P.A., Riera, P., 2010. The influence of cheap talk on willingness to pay ranges: an empirical application. Paper Presented at the 59th Congress of the French Economic Association (AFSE), Paris, September 2010. Mahieu, P.A., Riera, P., Giergiczny, M., 2010. An Approach to Estimate the Valuation Function in Contingent Valuation Studies Mixing Point and Range Elicitation Questions. University of Rouen. Mahieu, P.-A., Riera, P., Giergiczny, M., 2012. The influence of cheap talk on willingness-to-pay ranges: some empirical evidence from a contingent valuation study. Journal of Environmental Planning and Management 55, 753–763. Mentzakis, E., McNamee, P., Ryan, M., 2010. Incorporating uncertainty into payment card contingent valuation experiments. Working Paper. McMaster University, Hamilton. Meyerhoff, J., Liebe, U., 2006. Protest beliefs in contingent valuation: explaining their motivation. Ecological Economics 57, 583–594.

85

Mmopelwa, G., Kgathi, D.L., Molefhe, L., 2007. Tourists' perceptions and their willingness to pay for park fees: a case study of self-drive tourists and clients for mobile tour operators in Moremi Game Reserve, Botswana. Tourism Management 28, 1044–1056. MORGOAT Enquête Tourisme, 2005. Les fréquentations touristiques étrangères et de proximité en Bretagne en 2005. Observatoire Régional du Tourisme de Bretagne, Morbihan, p. 51. Morrison, M., Brown, T., 2009. Testing the effectiveness of certainty scales, cheap talk, and dissonance-minimization in reducing hypothetical bias in contingent valuation studies. Environmental and Resource Economics 44, 307–326. Morrison, M.D., Blamey, R.K., Bennett, J.W., 2000. Minimising payment vehicle bias in contingent valuation studies. Environmental and Resource Economics 16, 407–422. O'Garra, T., Mourato, S., 2007. Public preferences for hydrogen buses: comparing interval data, OLS and quantile regression approaches. Environmental and Resource Economics 36, 389–411. Petrolia, D.R., Kim, T.-G., 2010. Contingent valuation with heterogeneous reasons for uncertainty. Resource and Energy Economics 33, 515–526. Pindyck, R., Rubinfeld, D., 2005. Microeconomics (6th edn). Pearson Prentice Hall, Upper Saddle River, NJ. Poe, G.L., Clark, J.E., Rondeau, D., Schulze, W.D., 2002. Provision point mechanisms and field validity tests of contingent valuation. Environmental and Resource Economics 23, 105–131. Queffelec, B., Philippe, M., 2008. La gestion des zones côtières dans le Golfe du Morbihan. UMR-AMURE, Brest, p. 148. Roe, B., Boyle, K.J., Teisl, M.F., 1996. Using conjoint analysis to derive estimates of compensating variation. Journal of Environmental Economics and Management 31, 145–159. Rowe, R.D., Schulze, W.D., Breffle, W.S., 1996. A test for payment card biases. Journal of Environmental Economics and Management 31, 178–185. Samnaliev, M., Stevens, T.H., More, T., 2006. A comparison of alternative certainty calibration techniques in contingent valuation. Ecological Economics 57, 507–519. Schéma de mise en valeur de la mer, 2005. Rapport soumis à enquête publique, Schéma de mise en valeur de la mer. Préfecture du Morbihan, Morbihan, p. 87. Schulze, W.D., D'Arge, R.C., Brookshire, D.S., 1981. Valuing environmental commodities: some recent experiments. Land Economics 57, 151. Secretariat of the Convention on Biological Diversity, 2008. In: Secretariat of the Convention on Biological Diversity (Ed.), Protected Areas in Todays' World: Their Values and Benefits for the Welfare of the Planet. Shaikh, S.L., Sun, L., Cornelis van Kooten, G., 2007. Treating respondent uncertainty in contingent valuation: a comparison of empirical treatments. Ecological Economics 62, 115–125. Strazzera, E., Genius, M., Scarpa, R., Hutchinson, G., 2003. The effect of protest votes on the estimates of WTP for use values of recreational sites. Environmental and Resource Economics 25, 461–476. Svedsater, H., 2007. Ambivalent statements in contingent valuation studies: inclusive response formats and giving respondents time to think. Australian Journal of Agricultural and Resource Economics 51, 91–107. Syndicat inter-communal d'aménagement du golfe du Morbihan, 2009. Parc naturel régional du Golfe du Morbihan, Horizon 2021: Rapport de charte, version 2.4. Syndicat Intercommunal d'Aménagement du Golfe du Morbihan, Golf du Morbihan, p. 132. Togridou, A., Hovardas, T., Pantis, J.D., 2006. Determinants of visitors' willingness to pay for the National Marine Park of Zakynthos, Greece. Ecological Economics 60, 308–319. Vossler, C., McKee, M., 2006. Induced-value tests of contingent valuation elicitation mechanisms. Environmental and Resource Economics 35, 137–168. Walpole, M.J., Goodwin, H.J., Ward, K.G.R., 2001. Pricing policy for tourism in protected areas: lessons from Komodo National Park, Indonesia. Conservation Biology 15, 218–227. Wang, H., Whittington, D., 2005. Measuring individuals' valuation distributions using a stochastic payment card approach. Ecological Economics 55, 143–154. Welsh, M.P., Poe, G.L., 1998. Elicitation effects in contingent valuation: comparisons to a multiple bounded discrete choice approach. Journal of Environmental Economics and Management 36, 170–185.