Allowing for Zeros in Dichotomous-Choice Contingent-Valuation Models.

Dichotomous-choice contingent-valuation data are modeled using a mixture distribution. The standard parametric survival model is modified such that respondents in the lowest willingness-to-pay category may have either zero willingness to pay or a small positive willingness to pay. In comparison to the standard model, the mixture model leads to a dramatic reduction in estimates of mean willingness to pay. Covariates such as income are found to be more significant in determining the positive portion of the distribution of willingness to pay.