Why Do Simple Logit CVM Models Often Work Well in Spite of Misspecification?1

Monte Carlo evidence has shown that simple, misspeciøed discrete choice models for referendum contingent valuation data can lead to good estimates of mean willingness to pay (WTP). Empirical studies have found that estimates of mean WTP derived from simple parametric models often diœer little from those derived from nonparametric methods. This seems to indicate that simple models can yield WTP estimators that are relatively unbiased. This note shows that it is possible to estimate mean WTP consistently using a very simple logit model, regardless of the true form of the bid acceptance probability, provided that the survey bids are drawn from a uniform distribution.