Estimating DSGE Models under Partial Information

Most DSGE models and methods make inappropriate asymmetric information assumptions. They assume that all economic agents have full access to measurement of all variables and past shocks, whereas the econometricians have no access to this. An alternative assumption is that there is symmetry, in that the information set available to both agents and econometricians is incomplete. The reality lies somewhere between the two, because agents are likely to be subject to idiosyncratic shocks which they can observe, but are unable to observe other agents’ idiosyncratic shocks, as well as being unable to observe certain economy-wide shocks; however such assumptions generally lead to models that have no closed-form solution. This research aims to compare the two alternatives - the asymmetric case, as commonly used in the literature, and the symmetric case, which uses the partial information solution of Pearlman et al. (1986) using standard EU datasets. We use Bayesian MCMC methods, with log-likelihoods accounting for partial information. The work then extends the data to allow for a greater variety of measurements, and evaluates the effect on estimates, along the lines of work by Boivin and Giannoni (2005).