On the distribution of information in the moment structure of DSGE models

There is a long tradition in macroeconomics of using selected moments of the data to determine empirically relevant values of structural parameters. This paper presents a formal approach for evaluating the implications of DSGE models for the distribution of information in the moment structure of their variables. Specifically, it shows how to address the following questions: (1) what are the efficiency gains from using more instead of fewer moments; (2) what is the efficiency loss from assigning suboptimal weights on the used moments; and (3) which particular dimensions of the data - first and second order moments in the time domain, and sets of frequencies in the fre quency domain - are most informative about individual structural parameters. The analysis is based on the asymptotic properties of maximum likelihood and moment matching estimators and is simple to perform for general linearized models. A standard real business cycle model is used as an illustration.