Complexity and Model Selection in Macroeconomic Forecasting with DSGE Models

• Activity Number: 298
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Daniel McDonald*+
• Companies: Carnegie Mellon University

Abstract

Economists argue that dynamic stochastic general equilibrium models forecast as well as or better than reduced form models while avoiding the Lucas critique and providing analysts with economic intuition that makes forecasts easier to interpret. To improve forecasts, analysts expand their models, adding structure to handle sticky prices, risk premia, investment-specific technology and other economic phenomena. I apply results from the statistical learning literature to control the cost of making large forecast errors. This method bounds for probability of making costly prediction errors contingent only on the observed data and the complexity of the model space under consideration but not on the unknown data generating process. They can allow researchers to judge the predictive utility of increased model complexity and to compare competing models on the basis of predictive risk.