Incorporating theoretical restrictions into forecasting by projection methods

This work seeks to bridge the gap between two conflicting approaches in empirical macroeconomics, with ‘atheoretical’ econometric models on one side and “theoretical” models such as dynamic stochastic general equilibrium (DSGE) models on the opposite side. The former are often criticized for their lack of theoretical consistency and the latter are most likely misspecified, difficult to estimate and are typically outperformed by atheoretical models in empirical applications. We propose a method for modifying an atheoretical forecast and obtain a new forecast that satisfies the theoretical restrictions embedded in a DSGE model. We show that even imposing a simple Euler equation can improve the accuracy of the currently best performing atheoretical models, for example Bayesian VARs.