On the Estimation of Behavioral Macroeconomic Models via Simulated Maximum Likelihood

Abstract

In this paper, we introduce the simulated maximum likelihood method for identifying behavioral heuristics of heterogeneous agents in the baseline three-equation New Keynesian model. The method is extended to multivariate macroeconomic optimization problems, and the estimation pro-cedure is applied to empirical data sets. This approach considerably relaxes restrictive theoretical assumptions and enables a novel estimation of the intensity of choice parameter in discrete choice. In Monte Carlo simulations, we analyze the properties and behavior of the estimation method, which provides important information on the behavioral parameters of the New Keynesian model. However, the curse of dimensionality arises via a consistent downward bias for idiosyncratic shocks. Our empirical results show that the forward-looking version of both the behavioral and the rational model specifications exhibits good performance. We identify potential sources of misspecification for the hybrid version. A novel feature of our analysis is that we pin down the switching parameter for the intensity of choice for the Euro Area and US economy.