二段階二肢CVMにおける選択回答の変化

農業農村整備事業を事例として

抄録

In the paper, firstly, we statistically investigate the willingness to pay (WTP) for multifunctionality sustained by improvement project of agriculture and agricultural village. We apply the double bounded version of contingent valuation method (CVM) to estimate the mean of WTP. This version is extended one of dichotomous choice CVM and generally thought to be an effective method relative to original dichotomous choice CVM. However, because, in double bounded version of CVM, respondents are continuously asked the question that you could pay a listed value or not twice, 1st WTP derived from 1st responses and 2nd WTP derived from 2nd responses might not follow same distribution.
So, We estimate the mean of 1st WTP estimated by using 1st responses only and the mean of 2nd WTP estimated by using 2nd responses only. In all survey area, the mean of 1st WTP is larger than that of 2nd WTP.
Secondly, we test the change of responses between 1st bound and 2nd bound by using nonparametric method. We decide to use Gehan's test statistics W to do hypothesis test. This statistics are modified by Mantel and good for testing the sameness of distribution between two samples. This statistics also can be used even in the case that samples are arbitrary left or right censoring.
We now suggest new method applying bootstrap method to nonparametric test and do hypothesis test by this new method. As a result, in 6 survey area, except one survey area, we reject the null hypothesis that the difference of responses between 1st bound and 2nd bound is not significant statistically at 5% significance level. So, we conclude that when estimating WTP by using double bounded version of CVM, change of responses between 1st bound and 2nd bound generates in these area.
Finally, from our results of tests, we think that we have to take the change of responses into account to build new model when using double bounded version of CVM. This is our rested problem.