Comparing Payment-Vehicle Effects in Contingent Valuation Studies for Recreational Use in Two Protected Spanish Forests

By Campos, Pablo; Caparrs, Alejandro; Oviedo, Jos L

We compare two alternative payment-vehicles for contingent valuation studies to estimate economic recreational values. We analyze the potential effects of each wording to determine the most appropriate vehicle. Four contingent valuation surveys, carried out in two different protected Spanish forests, are compared. In each forest, one contingent valuation survey used entrance-fees and the other used an increase in trip-expenditures as payment-vehicles. As expected, results show statistically significant differences and great divergences between estimations, some three times higher in the second type. These differences remain, regardless of the format and the estimation technique used. The tests suggest that the second type approximates better welfare values for recreation.

KEYWORDS: Contingent valuation, recreational use, payment- vehicle, trip-expenditures, entrance-fee.

(ProQuest-CSA LLC: … denotes formulae omitted.)

There are many studies (Roach, Boyle, & Welsh, 2002) comparing the impact of different question formats (open-ended, simple- dichotomous, double-dichotomous and so forth) in contingent valuation (CV) but only a few papers have compared different payment- vehicles (see below). However, one of the crucial elements for the validity of CV studies is the paymentvehicle since it provides the context for payment, implying that differences in cultural and institutional factors can affect results (Morrison, Blarney, & Bennett, 2000).

The good valued in this study is a one-day recreational visit to two Mediterranean forests. This value is estimated using two different paymentvehicles: (a) an entrance-fee to access the area for one day (entrance-fee CV survey), and (b) an increase in trip- expenditures, specified as an increase in gasoline prices (trip- expenditures CVsurvey).

Establishing entrance-fees to protected forests is not a common practice in Spain and people do perceive free access as a right, although this is not always legally so (private owners have the right to exclude free access although they do not generally use this right). Despite this institutional setting, virtually all CV studies that have sought recreational use values of Spanish forests have used entrance-fees as the payment-vehicle. Therefore, this paper analyzes the use of increases in trip-expenditures as an alternative paymentvehicle to avoid the possible culturally induced context effects associated with entrance-fees.

If it is assumed that an individual ought to be indifferent to the payment-mechanism, a question positing an entrance-fee or its alternative, based on a hypothetical increase in trip-expenditures, should bring forth similar results. However, there are reasons that can explain divergence between these payment-vehicles, as will be discussed below.

To analyze these issues, this paper compares four CV surveys conducted for two different protected forests in Spain: the Scotch Pine Forests of Sierra de Guadarrama (Madrid) and the Cork Oak Forests of Alcornocales Natural Park (Cdiz-Mlaga). For each forest, one survey framed the valuation question in terms of an entrance- fee whilst the other survey posed the valuation question in terms of an increase in trip-expenditures. As expected, the results show relevant differences between the estimations obtained with the two payment-vehicles and the different tests made suggest that the trip- expenditures CV survey approximates better welfare values for recreation.

The rest of the article is structured as follows. The next section reviews related literature. The Methodology section describes the survey and estimation methods used, payment-vehicle effects and the tests used to estimate their significance. The subsequent section sets forth the results and the conclusions are given in the final section.

Literature Review

Although CV has often been used to value the recreational benefits of natural areas in monetary units, the payment-vehicles used for this purpose have received little attention, being chiefly entrance-fees (Bateman, Brainard, & Lovett, 1995a; Richer & Christensen, 1999) and increased tripexpenditures (Bishop & Heberlein, 1979; Boyle, Welsh, & Bishop 1993; Cooper & Loomis, 1992; Teisl, Boyle, McCollum, & Reiling, 1995). In Spain, researchers have used the entrance-fee option almost exclusively (e.g., Campos, de Andrs, Urzainqui, & Riera, 1996; Gonzlez, Gonzlez, Polom, & Prada, 2001).

Table 1 shows the different CV studies that have compared paymentvehicles. As can be seen, payment vehicles for recreation have not yet been compared. Other studies have analyzed the different effects caused by payment-vehicles. Blarney (1995) suggests that unfamiliarity with institutional procedures taken from CV studies applied in USA can lead to plausibility problems if used in Australia. Likewise, Morrison et al. (2000) conclude that, in countries other than the USA, the use of tax levies and referenda as payment-vehicles in CV studies reduces their plausibility. Kontoleon, Yabe & Darby (2005) argue that the differences with Bergstrom, Boyle and Yabe’s (2004) results (Table 1) are related to different institutional and cultural factors between the UK and the USA.

Jorgensen, Syme, Bishop, and Nancarrow (1999) and Strazzera, Genius, Scarpa, and Hutchinson (2003) discuss the problems associated with the presence of protest-response, concluding that respondents who have stated a positive WTP may hold beliefs that can be considered as some kind of protest (implying a deviation from their maximum WTP).

Chase, Lee, Schulze, and Anderson (1998) and Richer and Christensen (1999) have analyzed the differences between the maximum WTP and the price considered by respondents as fair or appropriate, when using entrancefees for recreation. Bateman et al. (1995a) conducted a meta-analysis of extant studies in the UK on the valuation of the recreational use of forests and found that the intercept was the most powerful explanatory variable. The authors’ explanation of this fact is that individuals may have stated what they thought was a fair value, and not their maximum WTP. The authors added that the use of entrance-fees as payment-vehicles may provide respondents with a pointer about what a fair value is (for example, the price of nearby parking facilities). Similarly, Hanley and Ruffell (1993, p. 226) found that the price paid for parking at over 60 recreational areas studied was a good predictor of stated WTP. As suggested by Santos (1997, p. 294), this may be caused by the use of the parking price paid as a pointer when stating their WTP.

Methodology

Contingent Valuation Surveys Performed and Data Treatment

The Scotch Pine Forests of Sierra de Guadarrama (PSG) cover about 100,000 hectares (ha) and are in a high mountain area with a MediterraneanContinental climate. Visiting rates are high, at 15 visits per ha per annum, due to the forests’ proximity to Madrid (60 to 100 km). This area is currently protected as a ZEPA (Zona de Especial Protection para las Avey, a protection category for birds, since the area has one of the largest black vulture colonies in the world) and partially as a Natural Park.1 The Cork Oak Forests of Akornocales Natural Park (ANP) are in a typical Mediterranean area in the south of Spain. ANP covers about 170,000 ha, it is in the humid part of the Andalusia region and its mountains are not very high. Although it has significant environmental assets, visiting rates are relatively low, about 0.5 visits per ha per annum.

As stated in the introduction, the good valued in this study is a one-day recreational visit. The wording used in the case of each forest and for each payment-vehicle can be found in Appendix I.2 The Spanish word for entrance-fee (entrada) originally referred to a ticket to a show and people undoubtedly understand that it refers to a one-day visit (and not to a payment to grant access for a full year, for example, which would be called abono). Furthermore, respondents were being asked for the WTP for that day’s visit.3

The entrance-fee CV survey at PSG (not in the case of ANP) started with an introductory paragraph which explained that the funds obtained would be used (together with current public funds) to ensure the area’s conservation. This probably means that some of the respondents included in their response a value for conservation and not only the intended recreational value. Nevertheless, this strategy has been used in previous studies in Spain to reduce the rejection of the entrance-fee and the study’s focus group also showed that this can reduce rejection of this payment-vehicle (if differences are found with this formulation, they should appear even clearer without this introduction).

Respondents’ awareness of their expenditures during the day was guaranteed by asking them, in a question immediately prior to the valuation one, to specify their per person expenditures during the day (Table 8 footnote ‘c’). In the PSG survey, respondents were asked to give separate values for transport, food, and others and were then asked to add them up. In the case of ANP they were merely asked to give an overall value, although they were explicitly requested to take into acco\unt the three kinds of expenditure described above. Interviewers were requested to remind respondents that per person values should be given.

At PSG, a double-dichotomous question was posed, without giving a prior indication of the second question, followed by an open-ended question. At ANP, a simple-dichotomous question was used followed by an open-ended question. For PSG, the vector of values was designed to be offered independently for each survey type,4 whilst for ANP the same vector of values for both main surveys5 (both based on two pre-test surveys) was used.

In the case of PSG, the trip-expenditures CV surveys were conducted from October 1998 to September 1999, whilst the entrance- fee CV surveys were conducted from September to November 1999.6 At ANP, surveys were performed simultaneously from June 2002 to May 2003 and the survey-type to be applied to each respondent was decided randomly. All questions were put to visitors by an interviewer. At PSG, different numbers of tripexpenditures and entrance-fee CV surveys were conducted. Nevertheless, relative errors obtained were 8% for the trip-expenditures survey and 10% for the entrance-fee survey. This similarity in the surveys’ precision allows for comparison. At ANP, the same number of entrance-fee and trip-expenditure interviews was conducted. Relative errors were 6% for the trip-expenditures survey and 9% for the entrance-fee survey. Table 2 shows the main characteristic of the four surveys made.

Statistical Treatment

Simple-dichotomous modeling. Of the various suggested ways to calculate estimators based on data from simple-dichotomous surveys, the methods proposed by Cameron (1988, 1991) were used. The following model was assumed:

WTP^sub i^ = x’^sub i^β + u^sub i^ (1)

Where u^sub i^ is the stochastic component which follows logistic distribution with zero mean and dispersion parameter k; x’^sub i^ is a vector of explanatory variables for which observations are available; and β is a vector of parameters to be estimated. The response given by each individual i to the offered value t^sub i^ enabled the following variable to be constructed: I^sub i^ = 1 if WTP^sub i^ ≥ t^sub i^; otherwise I^sub i^ = O. Thus, the probability of an affirmative response to the question can be expressed as: Pr(I^sub i^ = 1) = Pr(WTP^sub i^ ≥ t^sub i^) = PT(x’^sub i^β + u^sub i^ ≥ t^sub i^) = Pr(u^sub i^ ≥ t^sub i^, – x’^sub i^β) = Pr[u^sub i^/k ≥ (t^sub i^ - x'^sub i^β]. The associated log-likelihood can be written and optimized directly using a non-linear iterative optimization algorithm (Cameron, 1988).

However, Cameron (1991) proposes an alternative that allows programming and computation time to be reduced by making it possible, through a change of variable, to use conventional logit procedures such as those provided in most statistical packages. The procedure consists of carrying out a logit that includes among the explanatory variables die value offered to respondents. In the new logit, γ* = (β’/k, -I/A) is die vector of increased parameters and x* = (x’^sub i^, t^sub i^) is the new vector of explanatory variables. After estimating this conventional logit model, the censored logit’s values are recovered by undoing the reparametrization. The confidence intervals were also estimated following Cameron (1991). A conventional logit was estimated to obtain point estimates for γ* and the associated variance- covariance matrix (Σ^sub γ*^). This information yields the parameter variance-covariance matrix (Σ^sub β^) after some manipulations of Σ^sub γ*^ and its related information and transformation matrix.

Double-dichotomous modeling. The approach developed by Cameron and Quiggin (1994), which allows for the first and second question to stem from different valuation functions, was used. Cameron and Quiggin”s (1994) model assumes the existence of two unobservable values (WTP^sub 1^ and WTP^sub 2^), one for the first valuation question and the other for the second. They are the function of a vector (x^sub 1^ and x^sub 2^) of observable respondent attributes plus an unobservable random component (E^sub 1^ and E^sub 2^) distributed in accordance with two related normals (N(0, σ^sup 2^^sub 1^) and N(O, σ^sup 2^^sup 2^)). Unlike in the case of other models, it is not necessary for WTP^sub 1^ to be equal to WTP^sub 2^. They are allowed to differ and both values are estimated jointly using a bivariate normal.

Following Cameron and Quiggin (1994), a censored regression was implemented, defining the following variables (for j = 1, 2): I^sub ji^ = 1 if WTP^sub ji^ ≥ t^sub ji^; I^sub ji^ = O if WTP^sub ji^ < t^sub ji^, Possible combinations of these values for a given individual are (1,1), (1,0), (0,0) and (0,1). Since only models without additional explanatory variables were estimated, the underlying model is WTP^sub i^ = β^sub i^ + u^sub i^, with β^sub i^ being a scalar (one for each valuation function). Assuming a bivariate normal distribution BVN (β^sub 1^, β^sub 2^, σ^sup 2^^sub 1^, σ^sup 2^^sup 2^, ρ) and omitting subscript i, which indicates the individual in question only, the likelihood function logarithm is as shown in equation (2) (to simplify notation the density function of the standard bivariate normal is expressed as g(z^sub 1^, Z^sub 2^) where z^sub 1^ = (t^sub 1^ β^sub 1^)/σ^sub 1^ and z^sub 2^ = (t^sub 2^ - β^sub 2^)/σ^sub 2^). The Newton-Rapson optimization method was used to estimate β^sub 1^, β^sub 2^, σ^sub 1^, σ^sub 2^, and ρ.

… (2)

Payment-Vehicle Effects

Table 3 shows the different reasons why WTP with the entrance- fee (WTPE) and with trip-expenditures (WTPT) surveys may differ. Although the analysis focussed on the reject, fair, substitutes, round-trip and other areas effects, which imply WTP^sub T^ > WTP^sub E^, Table 3 also reflects effects that can cause WTP^sub T^ < WTP^sub E^. However, if H^sub 0^: WTP^sub T^ = WTP^sub E^ is rejected, and the direction of the inequality is WTP^sub E^ > WTP^sub E^, it can be assumed that the effects that imply WTP^sub T^ > WTP^sub E^ are large enough to compensate the other effects.

The reject effect can appear due to the perception that expenditures associated with protected natural areas should be funded through taxes. In the study-cases in question, the current situation is free access even though a significant part of the two forest areas is privately owned. Thus, respondents may have an incentive to defend the current situation and to strategically bias their answer so as to prevent private owners from establishing an entrance-fee. Both if respondents are reluctant to there being an entrancefee and if they have a strategic behavior, they can either refuse to pay or deviate from their maximum WTP (Jorgensen et al., 1999; Strazzera et al., 2003). The first strategy makes for a smaller and less representative sample (after eliminating protest- responses as usual). The second strategy would modify the stated WTP in an amount that the researcher would be unable to determine.

The fair effect (stating a value considered fair and not the maximum WTP) is more likely to appear with the entrance-fee essentially for the same reasons discussed as for strategic behavior. That is, because people feel that they can have more influence on the price set for the entrance-fee than on gasoline prices.

The substitute effect arises if respondents know close substitutes and feel that the entrance-fee will only be implemented in one particular area whereas they will probably assume that the increase in trip-expenditures will apply to all areas visited.

If the respondent values the round-trip positively (round-trip effect) or if he or she visits several areas on the same day (other areas effect), stated WTP in the trip-expenditures survey will be higher than corresponds to the welfare reported for the visit to the site itself, since the wording includes the whole excursion in the valuation.

It is assumed, a priori, that the reject, fair and substitutes effects reduce the value obtained with the entrance-fee survey and that the round-trip and other areas effects tend to increase the value obtained with the trip-expenditures survey when compared with the true value of the good to be valued (the one-day visit to a particular forest).

Testing Divergence and Effects

Divergence between the results can be tested directly through a Convergent Validity Test, checking correspondence between WTP values estimated using the same theoretical model but with a different payment-vehicle (Morrison et al., 2000).

As pointed out by McFadden and Leonard (1993), the mean without explanatory variables is an unbiased estimator of the mean with explanatory variables, and it has the additional advantage that it permits to use the highest number of observations. In addition, some explanatory variables were significant in one study area and not in the other, making it difficult to perform a homogeneous comparison. Thus, the comparison is based primarily on the models without explanatory variables.

Three simple-dichotomous models without explanatory variables were estimated for each survey7: (a) one model with all valid responses (DSP model), (b) one eliminating protest-responses (DS model), and (c) a logarithmic version of the DS model (DSLN model). In all these models x’^sub i^ = (1) and β = (β^sub 0^) in equation (1). Open-ended versions of the first two models were also calculated using the information of the closing open-ended question (OP model and O model). A double-dichotomous model without explanatory variables (DD model) and with the same treatment of the protest-response as the DS model was also estimated, and the logarithmic versions of this model (DDLN).8 The central model for the discussion is DS, by virtue of it bei\ng dichotomous, eliminating protest-responses and amenable to estimation for both forests studied. Further models that include other explanatory variables were based on the DS model.

The treatment of protest-response aims to identify, through a follow-up question (Appendix 1), the reasons that lead respondents to state a zero amount to the open-ended question. Respondents who gave reasons that implied a rejection of the scenario (and not a real zero) were eliminated in the O and DS models. The main reasons identified as protest-response were Natural areas should have neither boundaries nor limitations, It is a public area and we do not have to pay, We should not have to pay to enjoy natural areas and We already pay enough taxes (Table 2).

The means of the described models were compared and checked for overlapping confidence intervals. This comparison was carried out directly using the reported values but also using the reported values for the entrancefee survey (since it was not possible to give precise estimates of the impact of the different effects considered) and adjusted values for the tripexpenditures survey to take into account the round-trip and other places effects.

The following procedure was adopted to deduct the round-trip and other places effects. Valuation of the journey itself assumed that respondents, who said they were valuing the entire journey or a part of it, placed an equal value on the time spent on the round-trip and on the time spent in the forest. The reduction in respect of the influence of visits to other locations was implemented by whichever of the following two methods brought about the largest reduction: (a) estimating the value only with respondents who had not visited any other site, and (b) reducing the value from the total sample by the percentage resulting from respondents’ subjective valuation of the importance of the survey site as against the group of sites visited in their decision to go on the excursion.9 However, all the tests below were done using reported values, without adjusting the values for the trip-expenditures CV survey.

Additional explanatory variables described in Table 4 were used to identify the studied effects. Variables Efee, Other, Travel and Substitutes were used, respectively, to analyze the reject, other areas, round-tup and substitutes effects. One model was estimated by pooling the data obtained with both payment-vehicles and these explanatory variables (following Carson et al. (2003) this model has been called the Valuation Function Approach). Furthermore, Expectations and Income were included to increase the model’s explanatory power, since they are significant and not correlated with other variables. This model can be written substituting in equation (1) the following vectors:

x’ = (1, Efee, Other, Travel, Substitutes, Expectations, Income)

β’ = (β^sub 0^, β^sub Efee^,β^sub Other^, β^sub Travel^, β^sub Substitutes^, β^sub Expectations, β^sub Income^)

A model was also estimated for each study area including only Efee as the explanatory variable and pooling the data obtained with both paymentvehicles. Finally, models were prepared for both areas using the data of one of the payment-vehicles in each case and only one of the following explanatory variables: Other, Travel or Substitutes. All these one-variable models can be written explicitly substituting x’ = (1, Variable) and β’ = (β^sub 0^, β^sub Variable^) in equation (1), where Variable stands for the variable under consideration.

Through a Protest-Response Rates Test (Morrison et al., 2000), another possible divergence between entrance-fee and trip- expenditures CV surveys was analyzed (testing the reject effect). This test compares the percentage of responses that has been qualified as protest. A larger ratio of protestresponses obtained with a concrete payment-vehicle is a factor that can make it inadequate to capture adequate welfare values, mainly due to possible selectivity bias (Strazzera et al., 2003). This is really just an indicator of the impact on the WTP of the individuals who entirely reject the paymentvehicle, but may also be an indirect indicator of the potential reduction of the valuation stated by individuals who would be willing to pay some price (Jorgensen et al., 1999; Strazzera et al.).

Measuring whether rejection to entrance-fees is maintained in the event of the hypothetical restriction of free access, respondents were asked to choose between there being an entrance-fee to regulate congestion and the introduction of a system of daily visitor quotas (on a first come, first served basis). A logit analysis (based onjakus and Shaw (1997)) using Protestas the dependent and Quota as the explanatory variable (Table 4) makes it possible to identify whether the choice of a quota plays a significant role in the appearance of protest-response (it is assumed that the variable Quota is a proxy of rejection of the entrance-fee, that is, a preference for quotas even in a congestion context).

On the other hand, if the number of individuals preferring a quota to an entrance-fee is significantly higher than the number of people whose refusal to pay is identified as a protest-response, it can be said that some respondents not eliminated by protest- response treatment are influenced by their opposition to the introduction of an entrance-fee.

This assertion can be tested directly through a Procedural Variance Test10 (Holmes and Kramer, 1995). Since this test focusses only on respondents who do not raise a protest, it is possible to analyze whether the rejection of an entrance-fee (in a congestion context) is one of the factors that leads individuals to state a different WTP for each payment-vehicle, even when they have not explicitly rejected the payment.

This test compares actual responses to the simple-dichotomous question of the entrance-fee survey with counterfactual responses. To do this, the WTP of the entrance-fee survey is first predicted by multiplying the entrance-fee survey data by the behavioral parameters of the trip-expenditures survey, plus an equation error randomly drawn from the error distribution. Once the simulated WTP for each respondent has been obtained, a counterfactual response to the simple-dichotomous question for the entrance-fee survey is created by comparing the simulated WTP with the bid offered to each respondent of the entrance-fee survey (1 if WTP ≥ bid; 0 if not). Then a Maximum Likelihood Test for Binomial Probabilities is done by comparing actual and counterfactual responses. The χ^sup 2^ statistic indicates if there is a statistically significant difference between them. This procedure was applied to a model without explanatory variables (similar to DS) and to a model using the explanatory variables of the Valuation Function Approach (except Efee, which is meaningless in this sub-sample based only on entrance-fee questions) . This procedure was repeated but predicting the responses to the tripexpenditures survey with the parameters of the entrance-fee survey.

In the event that these differences are statistically significant (that is, procedural variance exists), a new variable is created for each respondent that takes value 1 if the actual and counterfactual responses to the entrancefee survey are different (that is, if procedural variance is present) and value 0 if not (for the trip- expenditures sub-sample this test is not relevant). With this binary variable as dependent, two logit regressions were made including the explanatory variables of each model used to obtain the counterfactual responses, plus Quota (Table 4) as an additional explanatory variable.11 To obtain an empirical distribution of the regression coefficients, bootstrapping was carried out with 500 replacements for each model. A significant t-ratio indicates the origin of the procedural variance.

Following Bateman et al. (1995a) it was also possible to analyze the fair effect, at least to some limited extent, comparing the explanatory power of the intercept and trying to use the price of nearby parking facilities to predict stated WTP for each question type.

Finally, an analysis was made of whether the different samples compared differ in relevant socio-economic factors, since this can explain part of the divergence obtained.

Results

Figures 1 and 2 show the proportion of respondents who answered yes for each payment-vehicle, for the whole sample and for the sample obtained after eliminating protest-response. As can be seen, for both areas positive answers for entrance-fees are systematically lower (in percentage) than for trip-expenditure increases. Moreover, the exclusion of protest-response has a higher impact on the entrance-fee surveys in both areas.

Table 5 provides relevant information for comparing the results obtained with each payment-vehicle, reporting mean values, with their respective confidence intervals, for models without explanatory variables and without logarithmic transformation. Appendix 2 shows the statistics associated with all these models. For PSG, the results of the entrance-fee survey amounted to 30% of the values estimated with the trip-expenditures survey for the mean of the preferred dichotomous models (DS and DD). For ANP, this percentage (for the DS model) was 50% (Table 5). A comparison of models OP, O and DSP brings out even greater differences, as do models with logarithmic transformation (Appendix 2). In none of the cases was there overlapping between confidence intervals (at 95%).

Taking into account the valuation of the round-trip and of other places visited made for a 13% reduction12 of the estimated values at PSG (for the DS^sub T^ model the final value for the mean was euro12.24) and 3.1% at ANP13 (for the DS^sub T^ model the final value for the mean was euro21.52). Moreover, confidence intervals do not overlap when applying the reductions descri\bed above in order to take into account the valuation of the round-trip and of the other places visited. For PSG, the lower limit of the trip- expenditures survey model that gives the lowest result (the OP^sub T^ model) was euro8.80 for the minimum reduced value, whilst the upper end of the model comprising the highest values of the entrance- fee survey (the DSE model) was euro5.11. For ANP, the lower limit of the trip-expenditures survey model giving the lowest result (the OP^sub T^ model) was euro13.78 for the reduced value, whilst the upper limit of the model comprising the highest values obtained with the entrance-fee survey (the DSE model) was euro13.02. The differences are still greater if models estimated in a similar manner are compared. That is, the Convergent Validity Test shows that the differences are statistically significant.

Another result shown in Table 5 is that WTP values are higher in the ANP than in the PSG case-studies. This may be surprising, since PSG is very close to Madrid, a high income area (individual income was also higher in our sample for PSG than in the sample for ANP, see Table 8 and note (b) within this table). However, this difference may be explained by the type of recreational use prevailing in both areas. In PSG, the typical visitor makes frequent visits (7.11 times per year) and stays for a relatively short period of time (5.11 hours per day on average). On the contrary, in the case of ANP, the typical visitor makes 1.65 visits, on average, and stays for 7.52 hours. In addition, and as reported in the Methodology section, the PSG survey was performed in 1998-99 and the ANP in 2002-03 (the Euro was introduced14 in 2002). Compared with other studies, it can be said that entrance-fee values in PSG are amongst the lowest in Spain whilst the ANP values are amongst the highest. Entrance-fee values in other studies range from 4.96 to 12.24 euro per visit (Campos, Caparros & Sanjurjo, 2005). In any case, the main objective of this paper is not to compare these two forests but to compare the results within each forest.

Another way to test the differences between WTP^sub T^ and WTP^sub E^ within each forest is through the dummy variable Efee. In the PSG study, the regression coefficient of variable Efee in the DSP model is significant at the 99% level (t-ratio = -12.8551). With the DS model, the value of the t-ratio is -9.6747 (99%). In the case of ANP, the DSP model has a t-ratio of -1.8044 (90%). For the DS model, the t-ratio is -1.3488 (thus, only statistically significant at the 80% level). Hence, in all cases signs are negative (higher values with the trip-expenditures wording) and in most cases these results are clearly statistically significant.

Concerning the Valuation Function Approach, the first step is to test the correlations among variables through a χ^sup 2^- Pearson Test. No significant correlation was found in the ANP selected variables but, in the case of PSG, Efee was strongly correlated with all the others. This is probably due to the different bids used for the different payment-vehicles in PSG. Since it was of interest to keep Efee in the analysis, it was decided to focus on the ANP case. As Table 6 shows, all the variables considered are significant, except Travel Thus, the round-trip does not seem to have a strong impact on the valuation function. Visiting other places (Other) has a significant impact, although the regression coefficient is much lower than that of the Efee variable (also significant). Income is significant and has a positive impact, as expected, and Expectations and Substitutes also play a major role. The positive effect of the Substitutes variable is contrary to what intuitively may have been expected. However, this may have been caused simply by the fact that people who know substitutes have more knowledge and awareness of natural areas and consequently place more value on them.

The one-explanatory-variable models showed that Other and Travel are not significant in the trip-expenditures survey, either in PSG or ANP (variables Other and Travel are not relevant with the entrance-fee data). In other words, at least in the case-studies in question, the effects of the tripexpenditures survey (other places and round-trip effects) seem to be less important than those observed for the entrance-fee survey (see below), so that it is to be expected that better welfare measures will be obtained using the former survey. In addition, the substitutes effect is not relevant in the cases in question (probably due to the uniqueness of the Natural Parks analyzed in their respective areas), since the Substitutes variable is only significant in the case of ANP in the trip-expenditures survey (90% level). That is, in the one- explanatory-variable models it is never significant in PSG and it is not significant in ANP using (i) entrance-fee data only or (ii) entrance-fee data and trip-expenditures data together.

Regarding the Protest Response Rates Test and the reject effect, a highlight in Table 5 is the scant effect of protest-response treatment on WTP for the trip-expenditures surveys (2% for PSG and 4% for ANP on the DS^sub T^ model) due to the small number of responses deemed as protest (3% in both cases; see Table 2). For entrance-fee surveys, however, protest-response treatment does have a marked influence on the final outcome, which increased by 94% on eliminating protest-responses for PSG and by 71% for ANP in the DSE model. This is due to the higher protest-response rates elicited with this survey-form (35% at PSG and 23% at ANP; see Table 2). That is, protest-responses are significantly higher with the entrance- fee surveys in both areas and the effect of the treatment of this kind of response has a greater impact on the result of entrance-fee survey. This fact indicates that the entrance-fee payment-vehicle raises more problems when obtaining adequate WTP values, whilst the scant effect of protest-response in the trip-expenditures survey makes it more suitable for measuring individual welfare values due to the absence of distorting factors.

The logit analysis of the protest-response shows that Quota is significant at the 99% level (t-ratio = 3.6444). The positive sign indicates that the election of the quota increases the probability of the appearance of protestresponse, supporting the idea of a relationship between the rejection of the entrance-fee and the protest response.

Continuing with the reject effect, it can be seen that the confidence intervals for the proportion of protest-responses in the entrance-fee survey is 29-41% at PSG and 19-27% at ANP, whereas the confidence intervals for the proportion of respondents preferring an admission quota is 77-85% and 7583%, respectively. This significant difference indicates that some of the individuals whose response was not eliminated as a protest-response are against the introduction of entrance-fees, even if the alternative was to restrict free access.

Respondents probably oppose the idea of entrance-fees due to the perceived right to free access. This notion is supported by the fact that 79% of reasons identified as protest-response in ANP and 67% in PSG referred to people’s right of free access to natural areas (the remaining percentage of protest-response was mainly related to the We already pay enough taxes reason). In addition, they may believe that their answer can influence future policies in the direction of establishing payments to enter protected natural areas.

The Procedural Variance Test was conducted only for ANP due to the correlation problems at PSG, mentioned previously. The χ^sup 2^-statistic shown by the Maximum Likelihood Test for Binomial Probabilities indicates that the differences between the actual and the counterfactual responses are statistically significant in the two models and for both payment-vehicles (99% significance level).15 Table 7 shows the results for the procedural variance regression based on the DS model and Valuation Function Approach. In the DS model, Quota is not a significant variable and the intercept has all the explanatory power (Table 7). The reason is probably the high percentage of people choosing the quota (in the pooled sample but also in the entrancefee and trip-expenditures sub- samples). In the case of the Valuation Function Approach, the results of Quota are similar to the previous one. In this model, only intercept (90%) and Expectations (95%) are significant explanatory variables (Table 7). Thus, when analyzing the reasons that lead to respondents that accept the payment to diverge the values elicited with each wording, the dissatisfaction of visitors’ expectations is the factor that provokes the appearance of procedural variance.

To sum up, it has been shown that protest-responses are more frequent with the entrance-fee survey than with the trip- expenditures one. It has been shown that choosing quota (assumed to be an indicator of a rejection to the entrance-fee) increases the probability of a protest-response being given. It has also been shown that amongst respondents who did not give a protestresponse, the two payment vehicles produced different results. It has also been shown that, amongst those not eliminated as protest-response, a significant portion chooses quota. However, it was not possible to show that Quota is a good variable to explain the difference observed between the results obtained for those not eliminated as protest-responses (one reason may be that ANP does not have a congestion problem at present, as is the case at PSG; unfortunately, the Procedural Variance Test could only be applied to ANP).

As regards the identification of the possible fair effect through comparison of the intercept’s explanatory power, nothing can be ascertained, because the intercept exerted the greatest explanatory power in all cases. Nonetheless, it may be pointed out that for the entrance-fee surveys the \intercept is very similar to the value paid at fee-paying parking facilities in some areas of the forests, which was not the case for the trip-expenditures surveys. As suggested earlier, this may be due to the fact that respondents used that datum as a pointer for their answer. If so, this would compromise the validity of their response as a maximum WTP in the entrance-fee survey.

Table 8 shows the main socio-economic indicators for the two surveys, since a significant difference between these variables may explain the divergences observed. At PSG, the differences between Age and Level of studies are not statistically significant, unlike Monthly net income and Total day-trip expenditure. However, these two variables are higher in the entrance-fee subsample; hence, on the face of it, this divergence should tend to increase the value obtained with the entrance-fee survey. For ANP, there are no statistically significant differences for any of these socio- economic variables, except for Age at 95% of significance (however, Age is not significant if included as an additional explanatory variable in the models tested above). These variables were not included in the models described above due to correlation problems and because the aim was to focus on the effects studied (only income was included in the main analysis).

Conclusions

The clearest conclusion to be drawn is that the values obtained for maximum WTP for recreational use of forests differ significantly depending on how the contingent valuation question is worded. The results obtained from a CV survey using entrance-fees as the payment- vehicle and another CV survey using trip-expenditures increases were compared and the means estimated on the preferred models are two to three times higher with the latter option than with the former (confidence intervals at no point overlap). These differences do not appear to be accounted for by any of the elements with a suspected influence on the result (socio-economic differences, valuation of the round-trip, influence of other places visited, substitute options).

The evidence assessed suggests that the increased trip- expenditures CV survey better approximates maximum WTP for, rather than forgo, recreational use. This is mainly due to the high number of protest-responses obtained with the entrance-fee CV survey, and the high impact on the final results of the treatment of this protest-response.

Although these results are not definitive, they may allow the use of tripexpenditures increases in CV surveys designed to value the welfare reported for the recreational use of forests to be recommended, at least in countries like Spain and most European countries where payment of entrance-fees to recreational areas is not common.

1 Similar to the International Union for the Conservation of Nature and Natural Resources (IUCN) V protection category of protected natural areas (International Union for the Conservation of Nature and Natural Resources [IUCN], 1994).

2 The full questionnaire is available from researchers upon request.

1 This is probably clearer in both trip-expenditures CV surveys and in the ANP entrance-fee CV survey, since the PSG entrance-fee CV survey did not refer to ‘your visit today’ but to ‘your visit’ (see Appendix 1).

4 We used the operational design proposed by Alberini (1995) to situate the values offered in dichotomous questions in the four quintiles of estimated a priori log-normal distribution (with the open-ended pre-test survey data) and the values of the second question at the median of truncated log-normal distribution. Use of this method for the entrance-fee survey produced a group of very low vectors (the median was 400 pesetas (euro2.40) and percentiles 25 and 75, respectively, were 200 and 500 pesetas (euro1.20 and euro3.00)). As it is one of the requirements for double-dichotomous surveys to be meaningful that values differ sufficiently for the respondent to appreciate a real difference, we chose to increase the values to be offered (following the criterion suggested by Dufneld and Patterson (1991) of separating values so that the logarithmic differences are approximately equal). This probably entailed increasing the estimates obtained with the entrance-fee survey due to the yea-saying effect (acceptance of any value, and even all the more ready acceptance of greater values). As the bias tends towards increasing estimates produced by the entrance-fee survey, accepting the hypotheses described with this criterion assures that they would also have been accepted if lower values had been posited.

5 We followed the criterion suggested by Duffield and Patterson (1991) of separating values so that logarithmic differences are approximately equal and tried to obtain a vector of values to be offered that would cover the responses obtained from the trip- expenditures and from the entrance-fee question in the pre-test surveys.

6To study the influence of different time intervals, we created the dummy variable Time (O for the six summer months and 1 for the six winter months) for the trip-expenditures survey. The results show that Time is not statistically significant (t-ratio = – 0.1979).

7 Trip-expenditures models are denoted by a T subscript; entrance- fee models are denoted by a ^sub E^ subscript.

8 see Appendix 2 for the statistical information associated with these models.

9 Scored from one to five, with five being the maximum.

10 We are indebted to one of our anonymous referees for proposing this test.

11 For the logit regression corresponding to DS model we used only Bid and Quota as explanatory variables and in the case of the model related with the Valuation Function Approach we used the variables Bid, Other, Travel, Substitutes, Income, Expectations and Quota.

12 THe average obtained for the subjective valuation question was 4.82, which means that 96% of total value can be ascribed to the survey site. The reduction in respect of the value attributable to the approach trip involved (a) not reducing the value for the 3% who said that they did not place a value on the trip, (b) reducing by 15% the value obtained for 45% of respondents who said that they valued the entire approach trip; and (c) reducing by 7% the value obtained for the 52% of respondents who said that they valued only that part of the journey travelled within the Siena de Guadarrama area.

13 The average obtained for the subjective valuation was 4.92 (98% of the value can be ascribed to the survey site). The reduction in respect of the value attributable to the approach trip involved (a) not reducing the value for the 11% who said that they did not place a value on the trip, (b) reducing by 4% the value obtained for 74% of respondents who said that they valued the entire approach trip; and (c) reducing by 1% the value obtained for the 15% of respondents who said that they valued only a part of the journey within Akornocala Natural Park.

14 The exchange rate is 166.386 pesetas/euro. The new currency brought a relatively important inflationary process, at least for low prices, since approximations were made systematically upwards.

15 For the entrance-fee sub-sample the value of the χ^sup 2^- statistic is 43.9505 for DS^sub E^ model and 44.8997 for the Valuation Function Approach. These values are 109.5666 and 83.2501 respectively in the case of the trip-expenditures sub-sample.

References

Alberini, A. (1995). Optimal design for discrete choice contingent valuation surveys: singlebound, double-bound and bivariate models. Journal of Environmental Economics and Management, 28(3), 287-306.

Bateman, I. J., Brainard J., & Lovett A. (1995a). Modeling woodland recreation demand using geographical information systems: A benefit transfer study. CSERGE Working Paper GEC 95-06. Norwich, University of East Anglia, UK.

Bateman, I. J., Langford I. H., Turner R. K., Willis K. G., & Garrod G. D. (1995b). Elicitation and truncation effects in contingent valuation studies. Ecological Economics, 12(2), 161-179.

Bergstrom, J. C., Boyle, K. J. & Yabe, M. (2004). Trading taxes vs. paying taxes to value and finance public environmental goods. Environmental and Resources Economics, 28, 533-549.

Bishop, R., & Heberlein T. (1979). Measuring values of extramarket goods: are indirect measures biased? American Journal of Agricultural Economics, 61, 926-930.

Blarney, R. K. (1995). Consumers and contingent valuation: An investigation into respondent behaviour. Unpublished PhD Thesis. Canberra, Australian National University.

Boyle, K. J., Welsh, M. P., & Bishop R. C. (1993). The role of question order and respondent experience in contingent valuation studies. Journal of Environmental Economics and Management, 25, S- 80-S-99.

Brookshire, D. S., Randall, A., & Stoll, J. R. (1980). Valuing increments and decrements in natural resource service flows. American Journal of Agricultural Economics, 62, 478-488.

Cameron, T. A. (1988). A new paradigm for valuing non-market goods using referendum data: maximum likelihood estimation by censored logistic regression. Journal of Environmental Economics and Management, Ii, 355-379.

Cameron,T. A. (1991). Interval estimates for non-market resource values from referendum contingent valuation surveys. Land Economics, 67, 413-421.

Cameron, T. A., & Quiggin, J. (1994). Estimation using contingent valuation data from a “Dichotomous choice with follow-up” questionnaire. Journal of Environmental Economics and Management, 27, 218-234.

Campos, P., de Andrs, R., Urzainqui, E., & Riera, P. (1996). Valor econmico total de un espacio de inters natural. In D. Azqueta & L. Prez (Eds.): Gestin de espacios naturales: la demanda de servicios recreativos. Madrid, Spain. McGraw-Hill, 193-215.

Campos, P., Caparrs, A., & Sanjurjo, E. (2005). Spain. In M. Merlo & L. Croitoru (Eds.): Valuing Mediterranean forests: Towards total economic value. Cambridge, UK. CABI, 319-330.

Carson, R. T., Mitchell, R. C., Hanemann, W. M\., Kopp, R. J., Presser, S., & Rudd, P. A. (2003). A contingent valuation study of lost passive use values resulting from the Exxon Valdez Oil Spill. Environmental and Resources Economics, 25, 257-286.

Champ, P. A., Flores, N., Brown, T. C., & Olivers, J. (2002). Contingent valuation and incentives. Land Economics, 78, 591-604.

Chase, L. C., Lee, D. R., Schulze, W. D., & Anderson, D. J. (1998). Ecoturism demand and differential pricing of national park access in Costa Rica. Land Economics, 74, 466-482.

Cooper, J., & Loomis, J. (1992). Sensitivity of willingness-to- pay estimates to bid design in dichotomous choice contingent valuation models. Land Economics, 68, 211-224.

Duffield, J., & Patterson, D. (1991). Inference and optimal design for a welfare measure in dichotomous choice contingent valuation. Land Economics, 67, 225-239.

Gonzlez, M., Gonzlez, X. M., Polom, P., & Prada, A. (2001). Galician natural heritage: Use values and prservation values. In A. Prada & M. X. Vzquez (Eds.): Environmental economics and society. Santiago de Compostela, Spain. Consello da Cultura Gallega, 103-123

Greenley, D. A., Walsh, R. G., & Young, R. A. (1981). Option value: Empirical evidence from a case study of recreation and water quality. Quarterly Journal of Economics, 96, 657-672.

Hanley, N., & Ruffell, R. (1993). The contingent valuation of forest characteristics: Two experiments. Journal of Agricultural Economics, 44, 218-229.

Holmes, T. P., & Kramer, R. A. (1995). An independent sample test of yea-saying and starting point bias in dicholomous-choice contingent valuation. Journal of Environmental Economics and Management, 29(1), 121-132.

International Union for the Conservation of Nature and Natural Resources (1994). Guidelines for protected anas management categories. UICN, Cambridge.

Jakobsson, K. M., & Dragun, A. K. (2001). The worth of a possum: Valuing species with the contingent valuation method. Environmental and Resources Economics, 19, 211-227.

Jakus, P., & Shaw, D. (1997). Congestion at recreation areas: Empirical evidences on perceptions, mitigating behaviour and management preferences. Journal of Environmental Management, 50, 389- 101.

Jorgensen, B. S., Syme, G. J., Bishop, B. J., & Nancarrow, B. E. (1999). Protest response in contingent valuation. Environmental and Resources Economics, 14, 131-150.

Kontoleon, A., Yabe, M., & Darby, L., (2005). Alternative payment vehicles in contingent valuation: The case of genetically modified foods. Paper presented at 14th EAERE Annual Conference. Bremen, Germany.

McFadden, D., & Leonard, G., (1993). Issues in the contingent valuation of environmental goods: Methodologies for data collection and analysis. In J. Hausman (Ed.): Contingent valuation: A critical assessment. Amsterdam, Netherlands. North Holland, 165-208.

Morrison, M. D., Blarney, R. K., & Bennett, J. W. (2000). Minimizing payment vehicle bias in contingent valuation studies. Environmental and Resources Economics, 16, 407-422.

Richer, J., & Christensen, N. (1999). Appropriate fees for wilderness day use: Pricing decisions for recreation on public land. Journal of Leisure Research, 31, 269-280.

Roach, B., Boyle, K., & Welsh, M. (2002). Testing bid design effects in multiple-bounded contingent valuation questions. Land Economics, 78, 121-131.

Santos, J. M. L. (1997). Valuation and cost-benefit analysis of multi-attribute environmental changes: Upland landscapes in England and Portugal. PhD Dissertation. Newcastle, University of Newcastle upon Tyne.

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 Resources Economics, 25, 461- 476.

Teisl, M. F., Boyle, K. J., McCollum, D. W., & Reiling, S. D. (1995). Test-retest reliability of contingent valuation with independent sample pretests and posttest control groups. American Journal of Agricultural Economics, 77, 613-619.

Pablo Campos

Institute of Economics and Geography

Spanish Council for Scientific Research

Alejandro Caparrs

Institute of Economics and Geography

Spanish Council for Scientific Research

Jos L. Oviedo

Institute of Economics and Geography

Spanish Council for Scientific Research

Address correspondence to: Alejandro Caparros, PhD, Department of Economics, Institute of Economics and Geography (IEG)-Spanish Council for Scientific Research (CSIC), Pinar 25 – 28006 Madrid. Phone: 914111098. Fax: 915625567. Email: acaparros@ieg.csic.es.

Author note: We would like to thank an associate editor, two anonymous referees and the participants to the AERNA-2004 Conference for their helpful suggestions. This paper has benefited from the CICYT-AMB99-1161 project, funded by the Comision Interministerial de Ciencia y Tecnologia, and from the MEDMONT project, funded by the European Union. It has also benefited from funding from the S.A. Belga de tos Pinares de El Paular and the Consejeria de Media Ambiente de la Junta de Andalucia.

APPENDIX 1

Contingent Valuation Questions Used

Scotch Pine Forests of Sierra de Guadarrama survey

* Entrance-fee question

[Imagine that, besides the funding already allocated by Government, conservation of nature in this area (PVL/PV^sub T^) partly depended on money paid by visitors.]

10. In accordance with the extent to which you enjoyed your visit, would you pay an entrance-fee (ticket) of . . . pesetas per person for this natural area? Bear in mind that we are asking you to imagine a real payment and that you would be unable to spend the entrance-fee on other things. [white square] yes (Q. 11) [white square] no (Q. 12) [white square] don’t know (Q. 15)

11. If yes: And would you be willing to pay . . . pesetas as an entrance-fee?

[white square] yes (Q. 13a) [white square] no (Q. 13b) [white square] don’t know

12. If no: And would you be willing to pay . . . pesetas as an entrance-fee?

[white square] yes (Q. 13c) [white square] no (Q. 13d) [white square] don’t know

* Trip-expenditures increase question

[As you know trip-costs have changed in the last decades (i.e. gas prices have gone up and down). Now we are going to ask you to imagine that total expenditures of your visit increase for this reason, even though you realize exactly the same activity you have done (same transport, same food . . .)]

10. If the per person total expenditures of your visit would have been . . . pesetas more than the quantity you have just calculated, would you still have come today? Bear in mind that we are asking you to imagine a real payment and that you would be unable to spend the entrance-fee on other tilings. [white square] yes (Q. 11) [white square] no (Q. 12) [white square] don’t know (Q. 75)

11. If yes: And if the increase in total personal expenditures would have been . . . pesetas, would you still have come today?

[white square] yes (Q. 13a) [white square] no (Q. 13b) [white square] don’t know

12. If no: And if the increase in total personal expenditures would have been . . . pesetas, would you still have come today?

[white square] yes (Q. 13c) [white square] no (Q. 13d) [white square] don’t know

Alcornocales Natural Park survey

* Entrance-fee question

[Imagine that, besides the funding already allocated by Government for the management of recreation areas, reception centers and paths (maintenance and surveillance), these infrastructures partly depended on money paid by visitors.]

10. Would you pay an entrance-fee (ticket) per person of … euros (. . . pesetas) to access the recreation areas and paths rather than forgo the enjoyment provided by today’s visit?

Bear in mind that we are asking you to imagine a real payment, and that whatever you spent you would then be unable to spend on other things, and if you didn’t wish to pay an entrance-fee to access the recreational areas and paths you could still use public roads. [white square] yes (Q. 11) [white square] no (Q. 11) [white square] don’t know (Q 12)

* Trip-expenditures increase question

[As you know trip-costs have changed in the last decades (i.e. gas prices have gone up and down relatively independently of generalized increases in prices and live costs). Now we are going to ask you to imagine that total expenditures of your visit increase for this reason, even though you realize exactly the same activity you have done (same transport, same food . . .)]

10. If the total expenditures of your visit today would have been . . . euros per person (. . . pesetas) more than the quantity you have just calculated, would you still have come today? Bear in mind that we are asking you to imagine a real payment and that you would be unable to spend the entrancefee on other things.

[white square] yes (Q. 11) [white square] no (Q. 11) [white square] don’t know (Q. 12)

In all questionnaires dichotomous questions were followed by an open-ended question. If the respondents replied ‘don’t know’ or a zero amount to the open-ended question, the following follow-up question was made to identify the protest-response:

(If you answered zero, a negative value or don’t know) Could you state the reasons why you don’t know/you are not willing to pay any additional amount as entrance-fee/current total expenditures per person?

Copyright National Recreation and Park Association First Quarter 2007

(c) 2007 Journal of Leisure Research. Provided by ProQuest Information and Learning. All rights Reserved.