Commentary on Fischhoff et al., “Construal Processes in Preference Assessment”

Abstract

Suppose a naive researcher has decided to study risk preferences. This researcher first approaches an economist and inquires about the best empirical methods for studying preferences. The economist replies that the best approach is to study existing data on individual choices (e.g., investment behavior) or perhaps to pose a series of choice problems (e.g., choice among lotteries) in an experimental setting. The economist argues that, after observing the choices individuals actually make, preferences may be inferred. In contrast, when the researcher questions a cognitive psychologist, the advice is to ask hypothetical choice questions or perhaps qualitative survey questions eliciting responses on a seven-point scale. What is the naive researcher to infer from the responses provided by the economist and the cognitive psychologist? He or she could infer, for example, that the economist has considered and rejected the verbal methods for eliciting preferences described by the psychologist. Similarly, the psychologist may have considered and rejected the revealed preference approach. However, what if the economist has assumed that the researcher is an economist? Then one may infer that the economist simply believed that any method other than revealed preference analysis is unlikely to be valued in the profession and therefore other approaches were not even considered. The cognitive psychologist, working under the assumption that the researcher is instead a cognitive psychologist, may have behaved similarly. Thus, if both respondents construed the question to be: "What is the best method for someone in your profession?" then the responses may be misinterpreted. How then should this researcher identify the best empirical methods for studying preferences? Some might argue that the correct approach is to forget about asking other researchers what they think is best to do and instead investigate what they actually do. Such an approach would probably involve an extensive literature search over many relevant disciplines. This study would be quite time intensive and would ultimately rely upon (1) a model in which the best methods are asserted to be those that have led to publication and (2) a judgment by the researcher as to which among the published methods is actually best. It would seem that much time and effort could be saved by simply devoting more attention to the interview