Implications of Partial Information for Applied Macroeconomic Modelling

Abstract

Implications of partial information for applied macroeconomic modelling along four dimensions are shown, and analysis provided on how they can be addressed. First, when permanent shocks are present a Vector Error-Correction Model including latent, as well as observed, variables is required to capture macroeconomic dynamics. Second, the assumption in Dynamic Stochastic General Equilibrium models that shocks are autocorrelated provides identifying information usable in Structural Vector AutoRe-gressions. Third, estimating models with more shocks than observed variables must yield correlated estimated structural shocks. Fourth, including measurement error, as commonly specified, implies a lack of co-integration between variables, even when actually present