Dynamic Preference "Reversals" and Time Inconsistency

We study the possibility of identifying time-inconsistent preferences in empirical designs where preferences are elicited in advance at time 0 and then again at time 1, after the agent receives additional information. For single-peaked preferences, time consistency is rejected only when the time-1 ranking between a pair of alternatives is always the reverse of the time-0 ranking. We establish variations and generalizations of this result. Since such stark reversals are rarely observed, choice-revision designs require stronger identification assumptions than perhaps previously appreciated. But we show that time inconsistency is identifiable in environments where preferences over alternatives can be "priced out."