Estimation of Financial Agent-Based Models with Simulated Maximum Likelihood

Abstract

This paper proposes a general computational framework for empirical estimation of financial agent based models, for which criterion functions do not have known analytical form. For this purpose, we adapt a nonparametric simulated maximum likelihood estimation based on kernel methods. Employing one of the most widely analysed heterogeneous agent models in the literature developed by Brock and Hommes (1998), we extensively test properties of the proposed estimator and its ability to recover parameters consistently and efficiently using simulations. Key empirical findings point us to the statistical insignificance of the switching coefficient but markedly significant belief parameters defining heterogeneous trading regimes with superiority of trend-following over contrarian strategies. In addition, we document slight proportional dominance of fundamentalists over trend following chartists in main world markets.