Estimation on stated-preference experiments constructed from revealed-preference choices

Constructing stated-preference (sp) experiments from a choice that the respondent made in a revealed-preference setting can enhance the realism of the sp task and the efficacy of preference revelation. However, the practice creates dependence between the sp attributes and unobserved factors, contrary to the independence assumption that is maintained for standard estimation procedures. We describe a general estimation method that accounts for this non-independence and give specific examples based on standard and mixed logit specifications of utility. We show conditions under which standard estimation methods are consistent despite the non-independence. We illustrate the general methodology through an application to shippers’ choice of route and mode along the Columbia/Snake River system.