Influence of bid and subsample vectors on the welfare measure estimate in dichotomous choice contingent valuation:: Evidence from a case-study

After briefly outlining recent studies on the formulation of dichotomous choice contingent valuation surveys, the authors describe the sampling design procedures followed to assess the recreational value of the 'Tablas de Daimiel' National Park, in which the truncated mean was used as the welfare measure. With the data obtained from 167 interviews, a posteriori Monte Carlo simulations were conducted to estimates bias and Variance of the welfare measure estimator, varying the following parameters: (1) the number and magnitude of the bids; (2) allocation of the sample among the various bid levels; and (3) the value used to truncate the willingness-to-pay distribution. Welfare measure estimator bias was found to be negligible in all the experiments conducted-each one with 1000 samples containing 400 observations - whereas sample allocation proved to have a greater effect on the accuracy of the welfare estimator than the number or magnitude of the bids. Here, the accuracy of the truncated mean estimates obtained with designs using the Cooper method is compared, for different truncation points, with the accuracy of the estimates obtained using an empirically constructed bid vector and a subsample size vector determined as per the Duffield and Patterson approach. The latter were found to be comparable to the Cooper optimal designs. (C) 2000 Academic Press.