Estimation of forest values using choice modeling: An application to Spanish forests

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 07 ) 30 5 –3 12

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / e c o l e c o n

Estimation of forest values using choice modeling: An application to Spanish forests Raul Breya,⁎, Pere Rierab , Joan Mogasc a

Department of Economics, Pablo de Olavide University, Ctra. de Utrera km. 1, 41013, Sevilla, Spain Department of Applied Economics, Autonomous University of Barcelona, 08193 Bellaterra, Spain c Department of Economics, Rovira i Virgili University, E-43204 Reus, Spain b

AR TIC LE I N FO

ABS TR ACT

Article history:

This paper presents a Choice Modeling application of forest goods and services valuation for

Received 16 May 2006

an afforestation program in the Northeast of Spain. The results from a random parameter

Received in revised form

logit model reveal that, on average, individuals would annually pay an average of 11.79

29 June 2007

euros for the forests to sequester 68,000 tones of CO2, 0.12 euros for delaying the loss of land

Accepted 6 July 2007

productivity for ten years, and 6.33 euros from picnic users for being allowed to picnic in the

Available online 14 September 2007

new forests. On the other hand, individuals would experience a loss in welfare equivalent to − 9.67 euros if four-wheel driving is allowed in the new forests. Finally, being allowed to pick

Keywords:

mushrooms in the new forests is valued at an average of 12.82 euros by those who live in

Stated preference methods

rural areas.

Forest valuation

1.

Introduction

One of the most significant land-use changes over the last decades in Europe has been the afforestation of large areas of post-agricultural lands. Social and economic changes that occur within developed societies lead up to a situation in which large pieces of agricultural land are being left abandoned and many rural areas are becoming depopulated. Under such circumstances, afforestation may be an attractive way of managing fallow lands. Partly, afforestation also results from incentive measures taken by individual countries and regions as well as by the European Community. Numerous European countries implement subsidized afforestation programs, which provide land owners with financial support for afforestation and management of the planted forests. Within the European Union such funding was mainly setup under EC Regulation 2080/92, followed by EC Regulation 1257/99. Spain became a beneficiary of these programs, and until 1999 they had resulted in the afforestation of some 1 million hectares of agricultural lands of the lowest quality in the European Union.

© 2007 Elsevier B.V. All rights reserved.

The new forest ecosystems generate a variety of goods and services not only to the forest owners, but also to society at large. They provide enjoyment from recreational opportunities, subproducts like mushrooms, berries, or medicinal plants, carbon sequestration, erosion prevention and biodiversity preservation, among others. As many other countries, Spain has been reforming its forest policy to incorporate social concerns. However, the knowledge about the extent of the benefits that the new forests provide to society is very limited. Their estimation could constitute a significant source of information for furthering forest policy design and the development of financial instruments. The need for methodologies for assessing the value of the various forest goods and services has been recognized in the international arena, e.g. in the Pan-European Process (Ministerial Conference on the Protection of Forests in Europe — MCPFE), the Intergovernmental Panel on Forests (IPF) or in the Third Ministerial Conference on the Protection of Forests in Europe. However forest benefits valuation is not a straightforward task. Since there is no market from which to directly observe

⁎ Corresponding author. E-mail addresses: [email protected] (R. Brey), [email protected] (P. Riera), [email protected] ( J. Mogas). 0921-8009/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolecon.2007.07.006

306

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 7 ) 3 0 5 –3 12

the prices of most of the different forest goods and services, a researcher has to rely on special techniques to estimate their marginal values. The field of economics has developed several methods for estimating the monetary value of environmental changes. Most forest valuation exercises use stated preference methods. Those typically involve surveying people and asking them for choices or values that would reflect their preferences with respect to trade-offs between a payment and some good provision (Mitchell and Carson, 1989). An increasingly popular family of stated preference methods is Choice Modeling or Attribute-Based Valuation Methods, where several attributes of a good – or several goods – are valued simultaneously (Hensher et al., 2005). There are several variants in the Attribute-Based Valuation Methods, depending on the way people are requested to state their choices. One of these variants consists in comparing the business-as-usual (BAU) situation with an alternative involving a payment in exchange for getting a bundle of goods. Then the respondent has to state which of the two she prefers. Brey et al. (2005) provides a recent variant of this question format, where individuals were required to classify alternatives as “better than” or “worse than” a business-as-usual or reference situation. This approach was taken to estimate the marginal economic value of several forest functions in Catalonia, a region in the Northeast of Spain that was considering implementing an afforestation program. The structure of the paper is as follows. Section 2 introduces the valuation method. Section 3 provides some background on Catalan forests and describes the survey application. Section 4 reports the estimated model and the WTP estimates, while Section 5 discusses the results. Finally, the last section summarizes the main conclusions.

2.

Methodology

Stated preference methods use survey questionnaires to define hypothetical markets and ask individuals to express their preferences. These methods are usually classified in two groups: the Contingent Valuation Method (CVM) with its many variants, and the Attribute-Based Valuation Methods (Hanley et al., 2001). CVM and ABVM represent two different ways of dealing with forest ecosystems valuation. CVM is a stated preference method where respondents are asked to state their maximum willingness to pay (or minimum willingness to accept in compensation) for a predetermined environmental change. In the dichotomous choice version of CVM, respondents are offered a change in the quantity or quality of a good at a given cost, and the respondent either accepts or refuses the payment of the suggested cost. Applying a discrete choice model (Hanemann, 1984) to the data collected, the economic value of a specific change can be estimated. CVM has been used to estimate the value of a wide variety of environmental resources and a number of studies have used CVM in the study of forest ecosystem services (Kramer et al., 2003). For example, in Loomis and Gonzalez-Caban (1998), a contingent valuation survey was used to estimate the economic value to California and New England residents of implementing a fire management plan to reduce acres of old growth forests that burn in California and Oregon, and Scarpa et al. (2000) used data from

a large-scale contingent valuation study to investigate the effects of forest attributes on willingness to pay for forest recreation in Ireland. However, its use has also been subject to criticism in terms of its ability to deliver reliable and accurate estimates of the willingness to pay (Diamond and Hausman, 1994). In ABVM, different sets of alternatives (choice sets) defined by attributes with different levels (varying across the sample) are presented to individuals, who express their preferences for the alternatives. By making one of the attributes a price or cost term, marginal utility estimates can be converted into willingness-to-pay estimates for changes in attribute levels, and welfare estimates obtained for combinations of attribute changes. These methods tend to collect more information than CVM as respondents get multiple chances to express their preference for a valued good over a range of payment amounts, but at the expense of requiring respondents to evaluate larger and more complex sets (Louviere et al., 2000; Hanley et al., 1998a, 2001). The kind of choice tasks individuals have to perform depends on the variant or elicitation method used. A very common ABVM is Contingent Choice (CC). In a CC experiment the respondent selects the most preferred alternative from several differing in terms of attributes and levels. A status quo or BAU option is usually included in each choice set. This elicitation format is consistent with the Random Utility Maximization model (RUM) and is widely accepted. It is a relatively easy elicitation format for respondents since it resembles the kind of choices individuals face in actual markets, although from a statistical point of view this is the variant providing less information per choice set and individual (Hanley et al., 2001). Researchers have achieved positive results using ABVM for valuing the benefits of non-market environmental commodities or services (Adamowicz et al., 1994, 1998; Boxall et al., 1996; Hanley et al., 1998a,b, 2002; Morrison and Bennett, 1999; Carlsson et al., 2003). CC exercises have also been applied to forest ecosystems. Previous applications to forest related goods include: Adamowicz et al. (1994) on water-based recreation in Alberta; Boxall et al. (1996) and Adamowicz et al. (1998) on moose hunting in Alberta; Bullock et al. (1998) on deer hunting in Scotland; Hanley et al. (1998a) on visits to environmentally sensitive areas in Scotland; or Boxall and MacNab (2000), examining the preferences of wildlife recreationists among different boreal forest scenarios. Often, the choice sets in CC contain the BAU and three, four, or more alternatives. However, choice sets of two alternatives and the BAU option are also frequently used, forming what is usually labeled Pairwise Choice. A less used variant consists in comparing different alternatives to the BAU option, one by one. For instance, a choice set could consist of BAU and one alternative. Respondents would then typically see a number of successive different choice sets. Another possibility, as in Brey et al. (2005), is to face respondents with choice sets of three or four alternatives and BAU, asking for sorting the alternatives as better than or worse than BAU. In summary, each alternative is compared to BAU individually. This is consistent with RUM and provides welfare measures conforming to standard consumer theory. More formally, consider an individual i ∈ {1,…, N} facing a set of at least two alternatives, including a BAU alternative (q).

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 07 ) 30 5 –3 12

Assume that the utility provided by an alternative j can be expressed as Uij ¼ biVxij þ eij ;

ð1Þ

where xij is a vector of observable variables describing the j alternative for individual i, βi is a vector of unobserved coefficients representing the tastes of individual i, and εij is an unobserved stochastic component following an iid type I extreme value distribution, and independent of xij and βi. Vector βi varies from 1 to N, with density f(βi|θ) representing taste differences in the population, where θ denotes a vector of parameters characterizing the density function (for example, mean and standard deviation). The aim is to estimate θ. This is the basic formulation of random parameter logit models. They are very flexible models since they can approximate any pattern of substitution across alternatives through appropriate specification of random parameters and variables. In particular, McFadden and Train (2000) show that, under mild regularity conditions, this kind of model can approximate any discrete choice model derived from random utility maximization. Next, consider that each individual is requested to undertake a series of binary comparisons between the BAU option and each alternative. Conditional on βi, the likelihood function for individual i would consist of the product of the binary probabilities for their observed binary choices:

Pi ¼ j

expðbiVxij Þ expðbiVxiq Þ j þ expðbiVxiq Þ jϵB expðbiVxij Þ þ expðbiVxiq Þ

jϵW expðbiVxij Þ

ð2Þ

where B and W denote the sets of alternatives classified as “better” or “worse” than the BAU alternative, respectively. The selection of the BAU option might be influenced sometimes by whether the respondent likes or dislikes changes in general. A BAU dummy variable can be introduced to account for it. This variable takes a value of 1 when the alternative is the BAU option and 0 for the other alternatives, acting as an alternative specific constant. Since actual tastes are unknown to the researcher, expression (2) has to be integrated over all the possible values of βi weighted with its density, leading to expression (3). Z Pi ¼

j

expðbiVxij Þ expðbiVxiq Þ j f ðbi jhÞdbi : expðbiVxiq Þ jϵB expðbiVxij Þ þ expðbiVxiq Þ

jϵW expðbiVxij Þ þ

ð3Þ Thus, the model is formulated here as a random parameter logit model for panel data, where the coefficients entering the utility function vary over individuals, and are constant over the choice situation for each individual. The θ estimates are those values maximizing the logP likelihood function N i¼1 lnPi , where Pi denotes the probability of the binary choices observed from individual i. However, since the integral in Eq. (3) cannot be solved analytically, exact maximum likelihood estimation is not feasible. Instead, Pi is approximated by simulation, maximizing the simulated loglikelihood function. For a more detailed description of the simulation procedures and their properties, see for instance Train (2003), and Hensher and Greene (2003).

3.

Application

3.1.

Catalan forests

307

Catalonia is a region in the Northeast of Spain, with an area of 32,100 km2. The majority of the region is under Mediterranean climate, with winter precipitation and summer drought (Terrades and Piñol, 1996). Forests cover 1.3 million ha, or about 40% of its total area. The forest is composed of various Mediterranean species. 31 evergreen species are found in 86% of the total forested area (73% of which is occupied by conifers), while deciduous trees (58 species) cover the remaining 14%. The most abundant species are Aleppo pine (Pinus halepensis), which is the dominant coniferous species, and Holm oak (Quercus ilex), a deciduous species. Regarding forest ownership, slightly more than three-quarters of the Catalan forest area belongs to private owners, while the rest is publicly owned (DARP, 1994). Commercial forests represent only 2% of the agrarian production in Catalonia (García, 1997), due to the low profitability of the timber industry. However, Catalan forests provide society with multiple non-market goods and services. These have seen an increase in their demand due to the economic development of the last decades (Merlo and Rojas, 1999). Fire is the principal disturbance within the Catalan forest (Piñol et al., 1998; Díaz-Delgado and Pons, 2001). It plays a relevant role in determining landscape structure and plant community composition, and also in the amount of carbon sequestration and soil erosion.

3.2.

Design

A valuation survey was conducted in Catalonia to estimate the non-market values of forest attributes. The program proposed in the valuation exercise involved an increase in forest coverage from 40%, the current share, to 50%. The additional 10% of forest area would be created at the expense of marginal agricultural land. This does not represent an actual program implemented by the administration, but is consistent with the Farmland Foresting Program (EC Regulation 2080/92). The 10% was chosen following suggestions made by forest experts who took into account the typology of the Catalan forests and the agricultural land available. Moreover, focus group participants found this quantity to be credible and understandable. In the final survey, practically all of surveyed people perceived the afforestation program as an overall positive or very positive initiative. The afforested area would be spread around Catalonia. Fig. 1 reproduces the map included in the questionnaire showing the forested areas before and after the proposed afforestation. The new forests were characterized by six attributes with varying levels, as shown in Table 1. Four focus groups were conducted with different Catalan residents in order to gauge attitudes toward the forests and their problems and to identify the attributes by which the forest could best be characterized. The information obtained during the focus group helped in the design of the choice experiments, including which forest attributes were included in the experiments, how the attributes were described, and the levels that each attribute could

308

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 7 ) 3 0 5 –3 12

Fig. 1 – Forest areas in Catalonia before and proposed expansions. take. Then a pre-test study was conducted on 100 residents in Catalonia in order to uncover misinterpretations of the questions and the difficulty of the choice tasks. The attributes selected included recreational activities (such as picnicking, picking mushrooms, and driving motor vehicles on forest ways), environmental functions of forests (CO2 sequestration and erosion prevention), and the program cost. The payment vehicle was a mandatory yearly contribution that all Catalan citizens would make to a fund exclusively devoted to the reforestation and conservation of the new forest. This payment mechanism does not seem to have evolved into much protest from the general public, and it was Table 1 – Attributes and levels used in the choice sets Attribute Picnic

Description

Levels

adopted in the study. In the focus group sessions no concern was raised from the participants, and they generally accepted the idea that improvements had to be paid for. This may be due to the strong perception of Mediterranean forests as public or collective goods, in the economic sense of the concept. The price vector used was chosen based on available contingent valuation studies and the pre-test. From these attributes and levels, 512 different alternatives were obtained (23 × 43). Since this universe was very large, statistical design methods were used to construct choice sets with four alternatives (Louviere, 1988). A main effects fractional factorial design reduced the number of alternatives to 16, which were grouped in 4 choice sets composed of 4 alternatives, including the option of the status quo or business-as-usual situation. Fig. 2 reproduces an example of an alternative used in a choice set. The questionnaire was structured in three different sections. The first part described briefly the current situation of forests in Catalonia, the proposed increase in forest area, and some information on the likely effects of the afforestation program. The second section was devoted to the preference elicitation questions. One of the four aforementioned choice sets was randomly assigned to each individual. The respondent was requested to classify each of the four programs as better or as worse than the status quo or business-as-usual situation. The last section contained some debriefing and socioeconomic questions, to obtain information about the respondents.

3.3.

Data collection

The questionnaire was administered through in-person individual home interviews. A stratified random sample of 800 individuals was selected in Catalonia, obtaining 730 valid responses. The strata used were age, gender, and size of the town of residence, following the information from official statistics (IDESCAT, 1999). No significant problems were detected in the survey application. The interviews were undertaken in 25 randomly selected locations, and administered to the population of 18 years of

Yes No a Drive Automobiles allowed in new forests Yes No a Mushrooms Picking mushrooms allowed in the new Yes forests No a CO2 Annual level of CO2 sequestered by new 300,000 forests, measured in equivalence to annual 400,000 CO2 pollution generated annually by a city of 500,000 a given number of people 600,000 Erosion Number of years until non-afforested lands 100 a 300 suffer enough erosion to become 500 unproductive 700 Payment Annual household afforestation program 6 euros cost 12 euros 18 euros 24 euros a

Picnicking allowed in the new forests

BAU: levels used in the business-as-usual alternative. For the attributes CO2 and payment the level value used for BAU was 0.

Fig. 2 – Example of afforestation alternative used in the valuation exercise, translated into English.

309

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 07 ) 30 5 –3 12

age or older proportionally to the population of each location. In each location, the questionnaires were distributed using random survey routes.

4.

Results

This section reports the estimated model, the marginal WTP estimates and the 95% confidence intervals for the different attributes. Table 2 contains the estimations of the model from the series of binary choices provided by respondents. The first model shows the estimates obtained using a fixed parameter logit model and including exclusively the attributes and a BAU dummy as explanatory variables. All the attribute coefficients have the expected sign, but only BAU and Payment are significant at 5% significance level. The coefficient of the BAU variable is significant and negative indicating that people are prone to prefer changes. In the second model some socioeconomic variables are included. The variables Use and Nonuse are binary variables reflecting whether respondents went to the forest for recreational activities in the last year. Smalltown and Largetown are also binary variables, indicating whether the respondent resides in localities with less, or more, than 3000 inhabitants. Finally, another dummy variable Age1 was added interacting with the BAU variable. This variable takes a value of 1 when the individual is older than 60 years of age and 0 otherwise. In this model, the coefficients for the attributes CO2 and Erosion are not significant, whereas the parameters associated with

the payment and the option of driving in the new forest area are highly significant and with the expected sign. The parameters for the Picnic and Mushrooms attributes are significant and with the expected sign only when interacting with variables Nonuse and Smalltown, respectively. The model also suggests that individuals older than 60 years of age seem to be more prone than others to select the BAU alternative. The last model allows for some parameters of the previous model to vary over respondents. The random coefficients were determined combining two different approaches. One was the classical procedure based on estimating different models and using the likelihood ratio test to select from them. The other procedure was based on the Lagrange Multiplier test proposed by McFadden and Train (2000). This test works by constructing P artificial variables, ztj ¼ 12 ðxtj  xtC Þ, with xtC ¼ jaC xtj Pj , where t represents the parameters where heterogeneity is suspected, C the set of alternatives being offered, and Pj the conditional logit choice probability for alternative j. The conditional logit model is reestimated including these artificial variables. The null hypothesis of fixed parameters for t is rejected when the coefficients for the artificial variables are significantly different from zero. This can be tested using a Wald or Likelihood Ratio test (Brownstone, 2001; Hensher and Greene, 2003). The application of both approaches determined coefficients of variables CO2 and Payment as random. Since it is reasonable to think that respondents will not consider a negative reduction of CO2, or tend to select more expensive alternatives, everything else held constant, these variables were expected to have only positive and negative values, respectively. Several statistical

Table 2 – Model estimations Fixed parameter logit model

Fixed parameter logit model with socioeconomic characteristics

Random parameter logit model with socioeconomic characteristics

Log-likelihood = −1947.299

Log-likelihood = −1910.200

Log-likelihood = −1716.952

Log-likelihood at 0 = −2023.990

Log-likelihood at 0 = −2015.672

Log-likelihood at 0 = −2015.672

N = 730

N = 727

N = 727

BAU

Picnic

−1.030 (−5.225)

BAU ⁎ Age1

0.111

Picnic ⁎ use

(1.460) Mushrooms

Drive Erosion CO2 Payment

BAU

Picnic ⁎ nonuse

0.059

Mushrooms ⁎ smalltown

(0.773)

Mushrooms ⁎ largetown

−0.358 (−4.702) −0.030 (−1.754) 0.008 (0.227) −0.035 (−6.136)

Drive Erosion CO2 Payment

−1.164 (−5.774) 0.433 (4.900) 0.254 (2.857) −0.117 (−1.126) 1.045 (3.178) 0.018 (0.236) −0.366 (−4.747) −0.027 (−1.545) 0.007 (0.193) −0.036 (−6.211)

BAU BAU ⁎ Age1 Picnic ⁎ use Picnic ⁎ nonuse Mushrooms ⁎ smalltown Mushrooms ⁎ largetown Drive Erosion CO2 (mean and spread) Payment (mean and spread)

0.793 (3.398) 0.688 (3.989) 0.441 (3.648) 0.053 (0.364) 0.892 (1.947) −0.066 (−0.630) −0.673 (−6.434) −0.085 (−3.654) 0.821 (17.267) −0.087 (−9.645)

The model was estimated using Halton sequences with 2000 replications. In the results reported in this paper, variables CO2 and Erosion have been divided by 105 and 102, respectively, to facilitate estimation. Values in brackets are t-statistics. Point estimates and t-statistics for the spreads have to be considered in absolute values.

310

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 7 ) 3 0 5 –3 12

Table 3 – Point estimate and 95% confidence intervals of the marginal WTP estimates for the five attributes (in € of 2006) Picnic a Mushrooms a Drive Erosion CO2

6.332 (3.840, 8.452) 12.819 (0.232, 30.116) −9.671 (− 12.422, − 7.956) 1.216 (0.565, 1.802) 11.793 (9.741, 14.847)

The first and last values in brackets represent the 95% confidence interval limits, and the values on top represent the corresponding point estimates. a Marginal WTP for Picnic and Mushrooms have been calculated from individuals visiting a forest area for recreational purposes in the last year and people who reside in localities with less 3000 inhabitants, respectively.

distributions were tested for both variables, the triangular distribution being the one found to fit the best empirical data. A triangular density function has the shape of a triangle (Evans et al., 2000). It has the advantage of not having large tails that may strongly influence the estimation of the mean of the distribution or provide unbounded willingness to pay for an attribute, as it happens with the log-normal distribution (Hensher and Greene, 2003). The main disadvantage of the triangular distribution is that it may prove difficult to explain sign changes for some variables. To avoid this problem, Hensher and Greene propose to make the standard deviation or spread of this distribution a function of its mean. Constraining the spread to be equal to the mean ensures that the distribution provides only positive or negative values. This has been the approach followed in this paper. The estimated spreads and means for these parameters were significantly different from zero at 5% significance level, and with the expected sign for their means. The likelihood ratio index increased substantially with respect to the fixed parameter models (from 0.052 to 0.148), indicating that the explanatory power of the model improved as a result of allowing CO2 and Payment parameters to vary over respondents. All the attribute parameters are significant at 5% significance level and with the expected sign, except for Picnic and Mushrooms interacting with Nonuse and Largetown, respectively, and for Mushrooms with Smalltown, although this one is close to the 5% significance level. The alternative specific constant for the BAU alternative is significant and positive, supporting the idea that respondents are not favorable to changes, especially for respondents older than 60 years.

4.1.

Marginal WTP estimates

From the estimated random parameter logit model, the mean marginal WTP for the different forest attributes can be estimated. The simplest approach assumes coefficients as given points. Assuming a linear utility model, the WTP will be the negative of the ratio between the mean coefficients for each attribute and the mean coefficient of the Payment attribute.

However, this approach does not incorporate the sampling variance. To do that, the Krinsky–Robb procedure can be adopted (Krinsky and Robb, 1986). This procedure uses random draws from the estimated asymptotic normal distribution of parameter estimates to calculate numerous WTP estimates. From these estimates, any empirical statistic of the mean marginal WTP can be obtained. Table 3 shows WTP point estimates and the 95% confidence intervals calculated with 1000 replications.

5.

Discussion

The marginal WTP reported as point estimates for each attribute corresponds to the average of the maximum amount (in 2006 euros) that the surveyed population would be prepared to pay annually and indefinitely for a one unit improvement in the attribute level. A positive (negative) marginal value for an attribute denotes that the average person would be better off with an increase (decrease) in the level of the attribute. The WTP for allowing picnic activities in the new forests is of 6.33 euros per year for the population that has used the forest in the last twelve months for recreational purposes. For those individuals not using forest areas in the last year, the coefficient for this variable was insignificant and very close to zero. Therefore, the willingness to pay for this variable seems to encompass mainly used values. The estimated value of allowing picking mushrooms, for those living in rural areas, is of 12.82 euros per year. The non-significance of the Mushrooms coefficient for residents in urban areas is not counter-intuitive. In a market research study, Cervera (1997) finds that picking mushrooms is an important part of the Catalan rural economy. Furthermore, a recent survey (GESOP, 2005) about the opinion of Catalan people on rural areas, the idea of paying a price for using the forest to pick mushrooms has large support among residents of townships with less than 5000 inhabitants. On the other hand, allowing driving in the new forests has a negative effect on the overall welfare of the Catalan population, with an average of 9.67 euros per year. The negative marginal WTP for Drive suggests that people do not hold positive preferences for this attribute. Thus, four-wheel driving does not, on average, form part of the benefits individuals receive from forested areas. Allowing driving may be interpreted by individuals as a source of pollution, conflicting with some of the environmental values associated with forests. Four-wheel driving may bring to forests one thing many respondents might be trying to escape from. The marginal WTP of 11.79 euros for CO2 reflects the value that the new forests would provide to society by sequestering a quantity equivalent to the emissions of CO2 that, on average, a Catalan city of 100,000 inhabitants generates annually in production and consumption activities, i.e. approximately equivalent to 68,000 tons of CO2 per year (Departament de Medi Ambient, 1996). Finally, 1.22 euros is the marginal WTP for delaying the loss of land productivity for a hundred years. In other words, continuing erosion for a 100 year period would imply an annual land productivity cost of 1.22 euros per person and year. Overall, results suggest that the public holds positive preferences for forest attributes, reinforcing the idea that forests provide significant non-market values to the public, beyond those held by forest landowners.

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 07 ) 30 5 –3 12

6.

Conclusions

In taking into account the multidimensional nature of forests from the viewpoint of society, researchers might be interested in the estimation of the value of some of their attributes. This paper has used a variant of the AttributeBased Valuation Methods (ABVM) to analyze the way in which individuals regard some attributes representing both recreational and environmental functions of forests. The variant consists of showing to respondents a choice set with the business-as-usual (BAU) situation and several alternatives. The task for respondents is then to indicate which alternatives they prefer to BAU. This is equivalent to having individual comparisons between each alternative and the BAU situation. The econometric model is formulated as a random parameter logit model for panel data, where each individual makes a series of binary comparisons to the BAU option. The results obtained the expected sign for all the significant attribute coefficients. The random parameter logit model estimates suggest the presence of heterogeneous preferences with respect to recreational forest functions. Allow picnicking in the new forest areas seems to be appreciated only by those individuals who went to a forest area for recreational purposes in the last year. These respondents are willing to pay an average of 6.33 euros per year for the chance of picnicking in the new forest areas. For picking mushrooms, people residing in small localities are willing to pay 12.82 euros per year on average, whereas people from larger cities seem to be indifferent. On the other hand, respondents value negatively the possibility of allowing four-wheel driving in the proposed forests, with an average annual cost of 9.67 euros. Finally, respondents show positive preferences regarding the two environmental forest functions considered, CO2 sequestration and erosion prevention. On average, they would be willing to pay 1.22 euros per year to delay for a period of a hundred years the loss of land productivity caused by erosion, and 11.79 euros per year for the CO2 sequestered annually by the new forests, when this is equivalent to the emissions from a city of 100,000 inhabitants. The results of this study may be used by forest planners when social preferences are to be taken into account and by decision makers when trying to evaluate different policy options.

Acknowledgements The authors acknowledge the partial financial support from the Ministry of the Environment of the Spanish and Catalan governments, and from the Forest Technological Center of Catalonia through its MEDFOREX program. The comments of three anonymous referees are also greatly acknowledged.

REFERENCES Adamowicz, W., Louviere, J., Williams, M., 1994. Combining revealed and stated preference methods for valuing

311

environmental amenities. Journal of Environmental Economics and Management 26 (3), 271–292. Adamowicz, W., Boxall, P., Williams, M., Louviere, J., 1998. Stated preference approaches for measuring passive use values: choice experiments and contingent valuation. American Journal of Agricultural Economics 80 (1), 65–75. Boxall, P.C., MacNab, B., 2000. Exploring the preferences of wildlife recreationists for features of boreal forest management: a choice experiment approach. Canadian Journal of Forest Research 30 (12), 1931–1941. Boxall, P., Adamowicz, W., Williams, M., Swait, J., Louviere, J., 1996. A Comparison of stated preference approaches to the measurement of environmental values. Ecological Economics 18 (3), 243–253. Brey, R., Bergland, O., Riera, P., 2005. A contingent grouping approach for stated preferences. Working Paper, vol. 22/2005. Department of Economics and Resource Management, Norwegian University of Life Sciences. Brownstone, D., 2001. Discrete choice modelling for transportation. In: Hensher, A. (Ed.), Travel Behaviour Research: the Leading Edge. Pergamon Press, Oxford, pp. 97–124. Bullock, C.H., Elston, D.A., Chalmers, N.A., 1998. An application of economic choice experiments to a traditional land use—deer hunting and landscape change in the Scottish Highlands. Journal of Environmental Management 52 (4), 335–351. Carlsson, F., Frykblom, P., Liljenstolpe, C., 2003. Valuing wetland attributes: an application of choice experiments. Ecological Economics 47 (1), 95–103. Cervera, M., 1997. Análisis comercial del sector de la seta silvestre en Cataluña. Distribución detallista en la ciudad de Lleida. Proyecto Final de Carrera. Escola Tècnica Superior d’Enginyeria Agrària, University of Lleida, Spain. 120 pp. Departament d’Agricultura, Ramaderia i Pesca, DARP, 1994. Pla General de Política Forestal. Generalitat de Catalunya, Barcelona. 156 pp. Department de Medi Ambient, 1996. Les emissions a l’atmosfera a Catalunya, 1996. Una aproximació quantitativa. Quaderns de Medi Ambient, vol. 5. Generalitat de Catalunya, Barcelona. Diamond, P., Hausman, J., 1994. Contingent valuation: is some number better than no number? Journal of Economic Perspectives 8 (4), 45–64. Díaz-Delgado, R., Pons, X., 2001. Spatial patterns of forest fires in Catalonia (NE España) along the period 1975–1995. Analysis of vegetation recovery after fire. Forest Ecology and Management 147 (1), 67–74. European Council Regulation No 2080/92. Community aid scheme for forestry measures in agriculture, 30 June 1992. European Council Regulation No 1257/99. Support for rural development from the European Agricultural Guidance and Guarantee Fund (EAGGF) and amending and repealing certain Regulations, 17 May 1999. Evans, M., Hastings, N., Peacock, B., 2000. Statistical Distributions. John Wiley and Sons, New York. 248 pp. García, F., 1997. Estimació de les macromaginituds agràries de les comarques de Catalunya, 1993. Serveis de Publicacions Universitat de Lleida, LLeida. 125 pp. GESOP, 2005. La imatge del món rural. Informe de resultats. Desembre 2005. http://www.congresmonrural.com/secc6.html. Hanemann, W.M., 1984. Welfare evaluations in contingent valuation experiments with discrete responses. American Journal of Agricultural Economics 66 (3), 332–341. Hanley, N., MacMillan, D., Wright, R., Bullock, C., Simpson, I., Parsisson, D., Crabtree, B., 1998a. Contingent valuation versus choice experiments: estimating the benefits of environmentally sensitive areas in Scotland. Journal of Agricultural Economics 49 (1), 1–15. Hanley, N., Wright, R., Adamowicz, W., 1998b. Using choice experiments to value the environment. Environmental and Resource Economics 11 (3–4), 413–428.

312

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 7 ) 3 0 5 –3 12

Hanley, N., Mourato, S., Wright, R.E., 2001. Choice modelling approaches: a superior alternative for environmental valuation? Journal of Economic Surveys 15 (3), 435–462. Hanley, N., Robert, E.W., Koop, G., 2002. Modelling recreation demand using choice experiments: climbing in Scotland. Environmental and Resource Economics 22 (3), 449–466. Hensher, D.A., Greene, W.H., 2003. The mixed logit model: the state of practice. Transportation 30 (2), 133–176. Hensher, D.A., Rose, J.M., Rose, W.H., 2005. Applied Choice Analysis. A Primer. Cambridge University Press, New York. 742 pp. Institut d’Estadística de Cataluya (IDESCAT). Population statistics 1999.http://www.idescat.net. Kramer, R., Holmes, T., Haefele, M., 2003. Contingent valuation estimation of forest ecosystem protection. In: Sills, E., Abt, K. (Eds.), Forests in a Market Economy. Kluwer Academic Publishers, Dordrecht, Netherlands, pp. 303–320. Krinsky, I., Robb, A., 1986. On approximating the statistical properties of elasticities. The Review of Economics and Statistics 68 (4), 715–719. Loomis, J., Gonzalez-Caban, A., 1998. A willingness-to-pay function for protecting acres of spotted owl habitat from fire. Ecological Economics 25 (3), 315–322. Louviere, J., 1988. Analysing individual decision making: metric conjoint analysis. Sage University Series on Quantitative Applications in the Social Sciences, vol. 67. Sage Publications, Inc., Newbury Park, CA. 95 pp.

Louviere, J.J., Swait, D.A., Swait, J.D., 2000. Stated Choice Methods. Analysis and Applications. Cambridge University Press, New York. 402 pp. McFadden, D.L., Train, K., 2000. Mixed MNL models for discrete response. Journal of Applied Econometrics 15 (5), 447–470. Merlo, M., Rojas, E., 1999. Policy instruments for promoting positive externalities of Mediterranean forests. European Forest Institute, Annual Conference, Chartreuse, Ittingen. Mitchell, R.C., Carson, R.T., 1989. Using Surveys to Value Public Goods: the Contingent Valuation Method. Resource for the Future, Washington, D.C. 463 pp. Morrison, M., Bennett, J.W., Blamey, R.K., 1999. Valuing improved wetland quality using choice modeling. Water Resources Research 35 (9), 2805–2814. Piñol, J., Terradas, J., Lloret, F., 1998. Climate warming, wildfire hazard, and wildfire occurrence in coastal eastern Spain. Climatic Change 38 (3), 345–357. Scarpa, R., Chilton, S.M., Hutchinson, W.G., Buongiorno, J., 2000. Valuing the recreational benefits from the creation of nature reserves in Irish forests. Ecological Economics 33 (2), 237–250. Terrades, J., Piñol, J., 1996. Els grans incendis: condicions meteorológiques i de vegetació per al seu desenvolupament. In: Terrados, J. (Ed.), Ecologia del foc, Proa editions. Barcelona, pp. 63–75. Train, K., 2003. Discrete Choice Methods with Simulation. Cambridge University Press, New York. 334 pp.