Working Paper: NBER ID: w28961
Authors: Philipp Strack; Dmitry Taubinsky
Abstract: Time inconsistency leads people to revise earlier plans, which has motivated empirical designs attempting to document such choice revisions. We study identification of time inconsistency in designs where an agent’s preferences are elicited in advance at time 0, and then again later at time 1, after they might have received additional decision-relevant information. We show that for single-peaked preferences, the only data that rejects time-consistent expected utility maximization is when an agent’s time-1 ranking between a pair of alternatives is the reverse of their time-0 ranking with probability one. We establish variations of this result under a variety of other assumptions. However, such patterns of choice are rarely observed in practice. To facilitate more robust identification, we present results about special conditions under which the degree of time inconsistency can be estimated.
Keywords: time inconsistency; preference reversals; expected utility maximization
JEL Codes: C90; D90
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Reversal of preferences over time (D11) | Failure of time-consistent expected utility maximization (TCEU) (D81) |
| Preference of x1 over x2 at time 0 (C69) | x2 not preferred to x1 at time 1 with certainty (D81) |
| Existence of at least one dimension of consumption where beliefs do not change between time 0 and time 1 (D11) | Degree of time inconsistency quantified (D15) |
| Assumptions about information revelation (D83) | Conclusions about time inconsistency (D15) |
| Existing data sets (Y10) | Hypothesis of time-consistent preferences not rejected (D15) |