Behavioral differences across bet pairs.

In preference reversals, subjects express different rankings over a set of alternatives depending on how preferences are elicited. In classical reversal tasks, for instance, subjects often select a safe bet over a risky one when given a choice between the two in a pair, but then assign a higher monetary evaluation to the risky bet. Motivated by a rich literature on context-dependent preferences, we conjecture that comparisons across bets in a pair can influence both Choice and Evaluation. Yet deciders are less likely to mentally compare the bets in the latter case, as bets are typically evaluated in isolation. This asymmetry between Choice and Evaluation is, we surmise, one cause of the reversals. If we further assume that memory decay affects mental comparisons in Evaluation, the account predicts order and timing effects on the reversal probability. We run several treatments designed to facilitate or hinder the retrieval from memory of the alternative bet during evaluation of a bet. However, the reversal rate does not vary across treatments in the predicted direction, and we find no systematic order or timing effects. We conclude that reversals are not influenced by the ease with which subjects recall the alternative bet during the evaluations, which suggests in turn that a relatively smaller frequency of comparisons across bets during the (typically isolated) evaluations is not a significant cause of reversals.