Environmental Valuation through Stated Preference Techniques: Contingent Valuation
Prepared for Module CB9005 “Environmental Valuation” by Carlos Ferreira
Submitted on the 8th April, 2009 1
Environmental goods have a value for society. However, this value is usually not explicit, because of the nonexistence of markets where they can be traded. As a result, when confronted with the choice between conservation of an environmental good or development, policy makers can seldom reach a decision based on society's preferences, since usually only the estimates of the value gained by society from the development are readily available. It is from the need to estimate society's real preferences that environmental valuation arises: to indicate how much society stands to lose from foregoing conservation. The problem of providing economic valuation of a nonmarket good has been tackled using two distinct types of methods: revealed preference techniques and expressed (or stated) preference techniques. The former infers the value society attributes to an environmental good from surrogate markets or indirect approaches, and as a result is limited to use values; the later resorts to a direct approach, explicitly questioning individuals on how they value the environmental good in question. As a result, stated preference techniques can be used to determine both use and nonuse values (such as existence value, option value or bequest value). One of the most widely used stated preference techniques of environmental valuation is Contingent Valuation. The method works by questioning participants what their actions would be in a hypothetical situation – specifically, how much would they be willing to pay for the conservation of a certain environmental good, or what is the minimum compensation they would be willing to accept to foregone the utility derived from an environmental good. Choosing between one of these situations is a matter of whom property rights are assigned to: if the participant does not own a good, it is relevant to ask what is the maximum he or she will be willing to pay to acquire it; on the contrary, if he or she does own the good, it makes sense to ask the individual the minimum amount of compensation he or she would be willing to accept in order to return to his or her initial utility level when deprived of that good. Contingent Valuation studies are undertaken by following a series of steps. The first among these steps is to create a hypothetical market for the environmental good, setting up the reason for the payment of services. During this stage, the bid vehicle – the way the proposed payment will be collected – is also presented. It is paramount that this vehicle is
2
credible, taxes and utility bills being commonly used. It is also important that the vehicle has some connection to the amenity being valued, and that it is perceived as being as equitable and fair as possible, The second step is to obtain willingness to pay (or to accept) estimates. These estimates (bids) are usually obtained through survey methods, using one among several elicitation methods: openended questions, closeended questions, dichotomous choice questions, iterative bidding or payment cards. The third step concerns estimating the willingness to pay (or accept) amounts, and consists of a process of statistical analysis and econometrics, different for each case. The fourth step is to aggregate the data, converting the findings to the total relevant population for the case at hand. Finally, the last step is to carry out validity checks, to estimate how trustworthy the results obtained are. Appropriate validity tests include content validity (is the study actually measuring what it is suppose to value?), criterion validity (how do the results compare with actual or simulated market experiments?) and construct validity (do results converge towards values obtained by other techniques, where these are available, and are they consistent with theoretical expectation?). The potential of Contingent Valuation studies to obtain estimates of the value of environmental good can be enhanced by the usage of more than one elicitation method, in order to overcome some of the limitations of each one method. The example we have chosen (Garcia et al., 2009) does exactly this. The study in question is a survey intended to value biodiversity in French forests. Since biodiversity can have both a use value (when participants visit forests) and a non use value (nonvisitors could consider biodiversity possesses option, bequest or existence value), usage of revealed preferences was precluded. However, this fact also means there's a selectivity bias to the sample: respondents can either be visitors or nonvisitors to forests; the former have both use and nonuse value to forests, while the later are limited to nonuse value. This is not the only potential source of selectivity bias: protest votes, usually excluded from the analysis, also result in estimate bias. The authors tackle both these problems, using two different econometric models (probit and tobit). The study consisted of a telephone survey to a sample of 4504 French households,
3
randomly chosen by different départements, in 2002. The survey was directed at investigating the maximum contribution respondents would accept to pay for forest conservation. The design of the questionnaire is discussed in a previous article (Garcia et al., 2007), but the authors do not explicit how they defined the bidding vehicle – whether it was a tax or utility fee, for example. Before the willingness to pay questions, participants were also asked whether or not they were users of forests (by reporting if they visited forests for recreational activities the previous year). After this, participants were asked to report on a series of socioeconomic characteristics that were thought to potentially impact on their willingness to pay for biodiversity conservation, such as composition of the household, whether they lived in a rural or urban area, and type of housing, plus a series of dummy variables on respondents' opinions about forest exploitation. The questions that elicited participants willingness to pay for biodiversity conservation in forests were in two formats, with all participants answering both formats: dichotomous choice followed by openended. After answering how much they would be willing to pay in the dichotomous format question (where options ranged from €6 to €90 per year, in €6 intervals), participants were asked what was the maximum amount they were willing to pay, per year, for conservation of biodiversity in forests. Of course, it's expected there will be anchoring from the values accepted in the dichotomous choice. In case participants answered €0 to the open ended question, configuring what could be thought of as a protest answer, there was a followup question, intended to differentiate true €0 answers (no value attributed to the environmental good) from protest answers, where the subject values the environmental good, but doesn't think he should have to pay for its conservation. Asking this question is one of the strongest aspects of this study's methodology: eliminating all €0 answers, regardless of whether they were protest answers or real novalue answers results in over estimation of the true value of the environmental good when the mean value is estimated. The estimation of the willingness to pay values was obtained by a probit function (in the dichotomous choice case) and a tobit function (in the open ended question). In the first case, this procedure allows for the possibility of estimating the willingness to pay accounting for whether they are users or nonusers, therefore countering the selectivity bias; not using this method to account for selection control means considering that a
4
participant being either a user or a nonuser is a random variable. This is not the case: users and nonusers self select, because they are different. In the case of the open ended question, the objective was to account for the effect of protest answers. The econometric procedure used (tobit regression) allows for the estimation of the willingness to pay maintaining the protestanswering participants in the sample, and infering what the mean willingness to pay would be if these participants disclosed their “real” WTP, reducing the bias in the results. The results for the dichotomous choice question were €65 per year in the case of users and €11.59 per year in the case of nonusers, after sample correction. Without sample correction, the values were respectively €63.23 and €30.35 in the case of non users, therefore showing the potential overestimation of nonuser willingness to pay without control of sample selection. The results of the open ended question were somewhat different: the authors estimated a mean willingness to pay of €39.84 for users and €33.42 for nonusers. Different as these values are compared to the dichotomous choice results, they are less biased because of the treatment of the €0 values, and so are the best estimates of the respondents' willingness to pay.
5
References ●
Garcia, S., Harou, P., Montagne, C., Stenger, A. (2007) Valuing forest biodiversity from a national survey in France: a dichotomous choice contingent valuation. Cahier du LEF no 200708, [Internet] . [Accessed 01 April 2009]
●
Garcia, S., Harou, P., Montagne, C., Stenger, A. (2009) Models for sample selection bias in contingent valuation: Application to forest biodiversity. Journal of Forest Economics, 15, pp.5978.
6