Testing for misspecification in the referendum contingent valuation approach

Binary choice models are frequently used in the analysis of referendum contingent valuation data. Because omitted regressors, heteroskedasticity, and distribution asymmetry in these models result in inconsistent parameter estimates, it is important that misspecification tests be undertaken. In this article, conditional moment tests for these problems are presented and applied to data from two referendum contingent valuation studies. Results show that some models have misspecification problems and that these problems affect the estimation of welfare measures.