Modeling Hidden Alternatives in Random Utility Models: An Application to “Don’t Know” Responses in Contingent Valuation

Steven B. Caudill and Peter A. Groothuis


We develop a multinomial logit model with missing information that can be used to test for hidden alternatives in random utility models and contingent valuation analysis. To illustrate our latent-choice technique, we focus on the “don’t know” responses in dichotomous choice contingent valuation questions. In this version of the latent-choice model, we probabilistically reassign “don’t know” responses to either the “yes” or “no” categories (or leave them in the “don’t know” category), if indicated by the data. We find that in our two studies a statistically significant number of the “don’t know” responses are actually “no” responses. (JEL C25, Q26)