A Structural Estimation Algorithm for Heterogeneous Agent Macro Models with Micro Data

I propose a computational algorithm for estimating heterogeneous agent macro models with micro data. The main challenge is that the state vector is infinite-dimensional and the likelihood of the stationary distribution is intractable. The key feature of the framework is that it minimizes the density difference between the observed distribution and stationary distribution to estimate the parameter. As an illustration, I implement the algorithm to conduct a likelihood-free structural estimation of a household's portfolio choice model with imperfect insurance. The results show that the assumption of imperfect insurance and limited market participation is not sufficient to solve the equity premium puzzle with the estimated model that produces reasonable equilibrium heterogeneity