A Computational Approach to Dynamically Inconsistent Decision Models

Abstract
I develop a numerical approach to computing Markov Equilibria in non-recursive decision problems and use it to study the savings behavior of households with dynamically inconsistent preferences. It is found that for relatively wealthy households, time-consistent strategies typically embed sufficient sophistication to approximately attain the first-best. Households with relatively little wealth typically deviate significantly from the first best. The implication for the distribution of wealth (in the Harris-Laibson-Aiyagari type settings) is that dynamic inconsistency can lead to bimodal wealth and income distributions. In "resource allocation" problems, the paper's numerical algorithm is guaranteed to find an (epsilon-) Markov equilibrium, and it also makes the analysis of steady states straightforward.
This paper is based on the paper "The Ego Loss Approach to Dynamic Inconsistency" and is also related to the paper "Equilibrium Analysis in Behavioral One-Sector Growth Models" (both available at Martin's homepage).