Valuing forest recreation in a multidimensional environment

Transcription

1 Bordeaux Regional Centre Research unit ADER Valuing forest recreation in a multidimensional environment The contribution of the Multi-Program Contingent Valuation Method Bénédicte Rulleau, Jeoffrey Dehez & Patrick Point IUFRO Division VI Symposium: Integrative Science for integrative management August 2007, Saariselkä, Finland

2 OUTLINE OF THE PRESENTATION Context The Multi-Program Contingent Valuation Method Theoretical framework Protocol and data collection Some descriptive statistics Results Hanemann (1984) Cameron (1988, 1991) Comparison of the two regressions Concluding remarks

3 OUR RESEARCH AREA State-owned coastal forests in Gironde

4 OUTDOOR RECREATION IN GIRONDE Recreation supply 20% of the coastline are urbanized 87% of beaches are natural Diversity of sites Very attractive because remains natural Non-market service Recreation demand 34 millions visits, 13.6 of them for natural beaches (Dehez, 2003) Multidimensional activities Residents and tourists High seasonality Visitors expectations? Economic value?

5 RESEARCH QUESTIONS Since visitors have access on a single site to 3 wilderness areas which personal characteristics and which features of the site influence visit choice? do they get more satisfaction than if they could recreate just in one? how do they trade-off between the price of a recreation conservation policy and the natural setting(s) concerned? What are they willing to pay to preserve the recreation quality on the entire site? On each asset? OBJECTIVES Identification of explanatory variables of visit choice Determining of the nature of the relations between the 3 wilderness areas Important before using another method dedicated only to the forest Inference of visitors WTP

6 WHY USING THE MPCVM? RPM do not satisfy all stakes of environmental valuation, especially in multidimensional contexts SPM are an alternative but Traditional CVM nor adapted to the study of multidimensional changes Conjoint analysis more linked with marketing than with economics In CE The value of the good = sum of the values of its attributes studies the trade-offs between attributes i.e. considers them as substitutes The number of attributes must be limited does not always reflect the complexity of the sites We suggest to use a multi-attribute extension of CVM called The Multi- Program Contingent Valuation Method and to adapt its protocol Still explanatory 2 different procedures in the literature we first studied them separately and propose here a comparison

7 Protocol first proposed in Cameron (1987) and developed by Santos (1998) on 2 basis: 1. Lancaster multi-attribute approach (1966, 1971) The utility does not come from the consumption of a good but from the consumption of its various features Independent Valuation and Summation 2. Hoehn (1991) THEORETICAL FRAMEWORK ' 2 2 ' ' 2 s q e q e q = Ag. Ah Ah q A g q Ag. A h If SE > CE, the programs are complements in valuation valuation of one program increases with the valuation of the other If SE < CE, the programs are substitutes in valuation ( embedding effect )

8 STAGES IN A MPCVM STUDY Information collection and pre-survey (April 2006, 93 quest.) Population sample and survey strategy Identification of the most pertinent way to describe the policies Programs and schemes design 1 program/wilderness area and 4 criteria/program Schemes = combinations of programs 7 schemes Status-quo recreation quality is not maintained Price = extra-distance to cover (in km) 5 bid levels 3 amounts/scheme/bid level depending on the number of programs Price of status-quo = 0 Survey (July and August 2006, 389 quest.) Each scheme is opposed to the status-quo Respondents must choose the one they prefer repeated 7 times Bid levels are randomly affected

9

10

11 REFUSALS AND REAL PROTESTS ACCORDING TO BID LEVELS 100% 80% 60% 40% 20% 0% Min. Inter. 1 Inter. 2 Inter. 3 Max. No answers Real protests

12 ACCEPTANCE OF THE SCHEMES 80% 60% 40% 20% 0% Oceansandforest Sandforest Oceanforest Oceansand Forest Ocean Sand Tourists Residents

13 TWO METHODS (1) Responses give the following information: I = 1 (accepts to pay) if B C ip ip ip I = 0 (refuses to pay) if B < C ip ip ip 1. Difference in utility function (Hanemann, 1984): from utility function to demand one Respondent accepts to pay for a scheme only if ( ) ( ) ( ) Pr Iip = 1 = Pr Vip p, xi Vi 0 0, xi ε i 0 ε ip with random utility specifications ( ) V = α + β y B + λx ip p i ip i Vi0 = α0 + β yi + λxi Bid taken as an explanatory variable Regression on the schemes WTP for each scheme = its mathematical expectation ( ) EC p α p = β

14 2. Difference in expenditure function (Cameron, 1988): from demand function to utility one Respondent accepts to pay for a scheme only if Regression on the programs TWO METHODS (2) (,, ) η ip (,, ) e z p U + + B e z p U + η ( ) Pr Iip = 1 = Pr εip Bip xi ' λ Procedure to calculate the true variance-covariance matrix and confidence intervals (Cameron, 1991) based on the censored regression (1988) Gives the WTP without recalculation ( Bip xi λ ) = 1 Φ ' According to McConnell (1990), each method is dual to the each other concerning WTP estimates only results on the relations between the programs will be presented here σ

15 RELATIONS BETWEEN THE PROGRAMS (1) Estimated coefficients of the composed schemes are compared to the outcome of the IVS Estimations IVS Wald test O-S-F S-F O-F O-S (S-F) + O 3.20 *** (O-F) + S (O-S) + F F+ O+ S F+ S F+ O O+ S *** ** ** The programs are complements in valuation when they are combined in order to compose the complete scheme Otherwise, they are independent in valuation

16 RELATIONS BETWEEN THE PROGRAMS (2) Regression that includes the second and third-order interactions O S F ** O x S 35.15** O x F 47.64*** S x F 57.75*** O x S x F *** Dispersion parameter 60.60*** The programs are complements in valuation at the second-step They remain complements at the third-step The ocean and the sand are not valuated when they are presented alone The forest appears to be an economic bad

17 Hanemann COMPARISON OF THE TWO METHODS IN THE LITTERATURE WTP for the status-quo more adapted to dichotomous choice format Can be used for more than 2 levels of environmental quality - Does not valuate a demand function but log-odds probabilities - Mean WTP may be affected by unusual observations or mistakes in data collection Cameron No need to specify the form of the utility function not restricted to random utility specification Reparametrization of the standard probit/logit Links with OLS Coefficients are easily interpreted as the variation of the value of the good according due to variation in explanatory variables Better interpretation of the information collected - Only binomial probit/logit

18 SOME IMPORTANT QUESTIONS REMAIN Interesting result of independency since for Hoehn (1991) substitution occurs in all situations for Santos (1998) programs implemented on one single site can behave as complements in valuation But, results are identical when recreation quality is maintained on the entire site and different when it is maintained only on 2 assets WHY? LINK WITH THE REGRESSION CHOSEN? Comparison of visitors WTP in order to confirm empirically McConnell s findings HOW?

19 Bordeaux Regional Centre Research unit ADER THANK YOU FOR YOUR ATTENTION!