An Experimental Analysis of Time-Inconsistency in Long-Run Projects

Abstract

In the first part of this paper, we elicit time preferences by using the experimental and econometric methods of Benhabib-Bisin-Schotter (2006). We follow the matching task procedure on money-time pairs with real rewards. Among the model specifications we use, the one with exponential discounting and quasi hyperbolic component of present bias appears to be the best model fitting the data. Unlike Benhabib et al., the present bias in the form of a fixed cost is not supported strongly by the data. In the second part of the paper, we test the theory of behavior of time-inconsistent agents in a long run project based on the quasi hyperbolic specification of O'donoghue and Rabin (2005) and Akin (2004). The preferences elicited in the first part are used to predict the behavior of agents in the long run project and categorize them based on their types. We find that the theory captures most of the subjects' observed behavior and helps understanding their types. We also find that some of the observations are compatible with alternative models, including sign effect, the preference for improving sequences, and anticipatory utility models.