Dynamic learning and context-dependence in sequential, attribute-based, stated-preference valuation questions

A hybrid stated-preference model is presented that combines the referendum contingent valuation response format with an experimentally designed set of attributes. A sequence of valuation questions is asked to a random sample in a mail-out mail-back format. Econometric analysis shows greater discrimination between alternatives in the final choice in the sequence, and the vector of preference parameters shifts. Lead and lag choice sets have a structural influence on current choices and unobserved factors induce positive correlation across the responses. These results indicate that people learn about their preferences for attribute-based environmental goods by comparing attribute levels across choice sets.