Files
Abstract
We present a demonstration of a Bayesian spatial probit model for a dichotomous choice
contingent valuation method willingness-to-pay (WTP) questions. If voting behavior is spatially
correlated, spatial interdependence exists within the data, and standard probit models will
result in biased and inconsistent estimated nonbid coefficients. Adjusting sample WTP to
population WTP requires unbiased estimates of the nonbid coefficients, and we find a $17
difference in populationWTP per household in a standard vs. spatial model. We conclude that
failure to correctly model spatial dependence can lead to differences in WTP estimates with
potentially important policy ramifications.