Econometric Methods for Agent-Based Models

In this Chapter I review the recent developments of the Agent Based literature with respect to em- pirical estimation. The methods employed in the literature include Bayesian estimation, simulated minimum distance, simulated maximum likelihood. In the second part I focus on two distinct prob- lems. The first one is parameter calibration. This approach is indeed useful since Agent Based Models (ABMs) have typically a large parameter space. The second one regards the possibility of replacing ABMs with a metamodel, i.e. a statistical model linking the value of parameters to a set of model outputs, i.e. of moments of the simulated data. The metamodels provide the conditional expectation of the moments, which might be used for a variety of purposes, including estimation. In particular, we focus on sensitivity analysis and on the problem of parameter identification.