Discounting behavior : a reconsideration

Abstract

We re-evaluate the theory, experimental design and econometrics behind claims that individualsexhibit non-constant discounting behavior. Theory points to the importance of controlling for thenon-linearity of the utility function of individuals, since the discount rate is defined over time-datedutility flows and not flows of money. It also points to a menagerie of functional forms tocharacterize different types of non-constant discounting behavior. The implied experimental designcalls for individuals to undertake several tasks to allow us to identify these models, and to severaltreatments such as multiple horizons and the effect of allowing for a front end delay on earlierpayments. The implied econometrics calls for structural estimation of the theoretical models,allowing for joint estimation of utility functions and discounting functions. Using data collectedfrom a representative sample of 413 adult Danes in 2009, we draw striking conclusions. Assumingan exponential discounting model we estimate discount rates to be 5.6% on average: this issignificantly lower than all previous estimates using controlled experiments. We also find noevidence to support quasi-hyperbolic discounting or “fixed cost” discounting, and only modestevidence to support other specifications of non-constant discounting. Furthermore, the evidence fornon-constant discounting, while statistically significant, is not economically significant in terms ofthe size of the estimated discount rates. We undertake extensive robustness checks on thesefindings, including a detailed review of the previous, comparable literature.