Supplanting the 'Minnesota' prior. Forecasting macroeconomic time series using real business cycle model priors

Although general equilibrium models are in wide use in the theoretical macroeconomic literature, their empirical relevance is uncertain. We develop procedures for using dynamic general equilibrium models to aid in analyzing the observed time series relationships among macroeconomic variables. Our strategy is based on that developed by Doan, Litterman, and Sims (1984), who constructed a procedure for improving time series forecasts by shrinking vector autoregression coefficient estimates toward a prior view that vector time series are well-described as collections of independent random walks. In our case, the prior is derived from a fully-specified general equilibrium model. We demonstrate that, like the atheoretical random-walk priors, real business cycle model priors can aid in forecasting.