Book Description

Abstract: This dissertation aims to provide answers to some of issues in dichotomous choice contingent valuation: the temporal structure of willingness to pay, practical guideline for survey design and generalized estimation method. The first essay proposes the temporal willingness to pay (TWTP) as an alternative definition of the present value of willingness to pay. In the survey of contingent valuation, a respondent compares TWTP with the present value of randomly assigned cost. TWTP enables the test for consistency of respondent's valuation with respect to payment schemes. Using a sequential test, the insensitivity of TWTP is tested on the data of oyster reef restoration programs in the Chesapeake Bay. The test result shows that TWTP is insensitive to the offered payment schedule or on the length of the stream of benefits of the project, which implies consistent willingness to pay for the environmental project. However, discount rates estimated from the data vary significantly across project lengths and time span between offered payment schedules. The second essay suggests a practical alternative design named a uniform design, to existing optimal or robust bid designs in contingent valuation. The uniform design draws cost assigned to respondent from a predetermined uniform distribution. Analytics and simulations show that the uniform design has lower bound of efficiency at 84 percent of D-optimum. Simulations demonstrate that the uniform design outperforms optimal designs when initial information is poor and outperforms robust designs when true values of parameters are known. The third essay challenges the theoretical and technical background of the simple logit model. Standard logit model in contingent valuation assumes i.i.d error distribution between initial and proposed states. Relaxing the restrictive assumption in the simple logit model requires a generalized estimation technique that utilizes a Gumbel mixed model. Estimation results show that correlation between two states is usually minimal, but homoskedastic errors are rejected in many cases. Heteroskedasticity or correlation provides willingness to pay estimate different from estimate of the simple logit, thus different policy implication in benefit-cost analysis.