A Survey of Estimations of Dynamic Discrete Choice Models

The DDCM is a dynamic program (DP) with discrete controls, which is very important in understanding agents' behavior in various settings. This paper begins with the essential features and hypothesis. On the one hand, to estimate the model with two sources of heterogeneity when the number of periods is small, it obtains non-linear bias corrected estimator (NBC) and minimized integrated mean square error (MIMSE). On the other hand, “Curse of Dimensionality” makes the estimation of the dynamic model infeasible even in a relatively simple setting, it proposes Bayesian estimation, Bayesian-MCMC-DP estimation and MCMC-ANN etc.