We provide a novel methodology for estimating time-varying weights in linear prediction pools, which we call dynamic pools, and use it to investigate the relative forecasting performance of dynamic stochastic general equilibrium (DSGE) models, with and without financial frictions, for output...