Applying the method of simulated moments to estimate a small agent-based asset pricing model

The paper takes a recent agent-based asset pricing model by Manzan and Westerhoff from the literature and applies the method of simulated moments to estimate its six parameters. In selecting the moments, the focus is on the fat tails and autocorrelation patterns of the daily returns of several stock market indices and foreign exchange rates. It is argued that it may be meaningful to abandon the econometrically optimal weighting matrix in the objective function and instead invoke the moments' t-statistics in an intuitively appealing way. This modification gives rise to estimations whose moment matching, given the model's parsimony, can be largely considered to be satisfactory. Also the parameter estimates across different markets make good economic sense.