Dynamic Decisions Under Subjective Expectations: A Structural Analysis

Abstract

We study dynamic discrete choice models without assuming rational expectations. Agents' beliefs about state transitions are subjective, unknown, and may differ from their objective counterparts. We show that agents' preferences and subjective beliefs are identified in both finite and infinite horizon models. We estimate the model primitives via maximum likelihood estimation and demonstrate the good performance of the estimator by Monte Carlo experiments. Using the Panel Study of Income Dynamics (PSID) data, we illustrate our method in an analysis of women's labor participation. We find that workers do not hold rational expectations about income transitions.

Keywords: Dynamic discrete choice models, subjective expectations, rational expectations, nonparametric identification, estimation.