Incentive and information properties of preference questions
Richard T. Carson; Theodore Groves
This chapter is both a commentary on and extension of the Carson and Groves (2007) (hereafter CG) article reprinted in this volume. The substantial attention the paper has received has been enormously gratifying. Reception of CG has largely been positive with little if any substantive criticism directed toward it; and, there are many papers now being presented at conferences that are testing or relying on various aspects of it.Our remarks are organized into a series of short sections. The first points out that the main purpose of CG was to extend the revealed preference paradigm to cover some types of survey responses. The second notes that CG provides the theoretical foundation that some critics of contingent valuation (CV) had argued was missing. The third takes the concepts of “hypothetical” and “hypothetical bias” head on and argues that these concepts are, for the most part, ill-defined or simply wrong and have done enormous damage to clear and careful thinking about the nature of the response to stated preference questions. The fourth examines the properties of cheap talk which is often proposed as a way to reduced hypothetical bias. The fifth provides some elaboration on CG and the issue of how to interpret information extracted from preferences questions. The sixth poses an answer to the often asked question: Is a single binary discrete choice (SBC) question always the best elicitation format for a researcher to use? The seventh provides some elaboration on the payment card elicitation format, which in recent years has seen resurgence. The eighth turns to an examination of some of the properties of the now widely used discrete choice experiment. The ninth considers the usefulness of economic experiments to help determine the performance of preference elicitation formats. The last section addresses the relationship between CG and the behavioralist critique of neoclassical economics with a focus on the different-answersto-the-same-underlying-question issue.
