How much time-inconsistency is there and does it matter? Evidence on self-awareness, size, and effects

Abstract

Using students’ predicted and unpredicted delays in midterm preparation as measures of time-inconsistency and self-awareness, this paper shows that time-inconsistent behavior is associated with inferior class performance even after controlling for the actual time spent in preparation. Most students showed some time-inconsistency. Most were at least partially aware of their time-inconsistency. Despite full awareness, the sophisticates still under-performed in class relative to the time-consistent. Furthermore, they might still show over-confidence in their self-evaluation of future class performance. Those naive about future time-inconsistency also under-performed relative to the time-consistent.

Introduction

Using the class behaviors and class performances of two different groups of undergraduates, this paper empirically examines three questions. First, are people time-consistent? Most are not; in both samples, only 6–7% of the students behaved time-consistently in their preparation for midterm examination. Second, to what extent are people aware of their future time-inconsistency? Most are at least partially aware. Many cited reasons such as procrastination and laziness to explain why they predicted deviations from their own optimal plans. Third, does time-inconsistency matter? It does. The time-inconsistent under-performed relative to the time-consistent. Moreover, in the absence of pre-commitment devices, being aware of future time-inconsistency did not prevent under-performance in class. Compared to the time-consistent, being naive about future time-inconsistency may also hurt class performance.

The two samples consist of two semesters of students in my introductory macroeconomics class at the National University of Singapore (NUS). To measure class performance, I use scores in tutorial participation, midterm, final examination, their sum and the final letter grade awarded for overall performance. To identify time-inconsistency and self-awareness, I use survey responses to questions regarding their optimal, predicted, and actual study plans for the midterm examination. As the benchmark, the time-consistent students behave according to plan, and they plan according to what is optimal from the long-run perspective. Thus, they predict and experience no delay in implementing their ideal study plan (i.e., ideal from the prior perspective).

Time-inconsistency, on the other hand, can take three forms. First, the naifs, who are completely unaware of their future time-inconsistency, predict no delay in initiating their study plan but end up deviating from it. Second, the sophisticates, who are fully aware of their future time-inconsistency, incorporate their time-inconsistency into their plan. In the absence of pre-commitment devices, they choose the best plan among those they will actually follow when the moment of action arrives, given their true future time-inconsistency. This plan may not be the best plan from the long-run perspective if they can pre-commit themselves. Thus, the sophisticates predict deviations from their ideal study plan but experience no unpredicted delay in implementing their chosen plan. Finally, the partial naifs, who are aware of their future time-inconsistency but mis-estimate its severity, predict delay from the prior perspective but end up deviating further from their chosen plan. Thus, they show both predicted and unpredicted delays.

In summary, I use the students’ predicted delay in initiating their own ideal study plan from the prior perspective as a measure of sophistication about future time-inconsistency, and unpredicted delay in implementing their chosen study plan as a measure of naiveté about future time-inconsistency. While I expect predicted delay from optimal plan to be a good measure of sophistication, the interpretation of unpredicted delay needs to be qualified: in a world with external shocks, unpredicted delay is likely to be a noisy signal of naiveté because there may be other shocks that cause the students to deviate from their predicted plans for reasons unrelated to time-inconsistent preferences. However, if the other shocks are exogenous and random, unpredicted delay will still be an unbiased estimator of naiveté.

It turns out that both predicted and unpredicted delays have negative effects on all measures of class performance in both samples. Furthermore, the effect of unpredicted delay is comparable in magnitude across samples. The negative effects of predicted delay are comparable in magnitude to the effects of unpredicted delay in one sample. However, somewhat surprisingly they turn out to be larger in size and more statistically significant in the other sample. A priori, delays may be due to factors unrelated to time-inconsistency, such as an unusually busy schedule. However, it turns out that the negative effects of predicted and unpredicted delays remain even after controlling for the actual days or hours spent in midterm preparation although more hours spent did raise midterm score with diminishing returns. An interpretation of this finding is that the negative effects of delays are mainly due to sub-optimal intertemporal allocation of time rather than less time spent per se.

The failure to anticipate preference reversal also shows up as prediction error in class performance, defined as the difference between the students’ actual and predicted grade performance, where the prediction was made by the students themselves before the midterm and after they have had significant exposure to course materials. This prediction error in class performance is always significantly related to unpredicted delay. Furthermore, it is significantly related to predicted delay in one of the two samples. Thus, the evidence suggests that self-awareness of future time-inconsistency, as measured by predicted delay, is sometimes not fully incorporated into the self-evaluation of its consequence.