Model Evaluation in Macroeconometrics: from early empirical macroeconomic models to DSGE models

This paper reconsiders the developments of model evaluation in macroeconometrics over the last forty years. Our analysis starts from the failure of early empirical macroeconomic models caused by stagflation in the seventies. The different diagnosis of this failure are then analyzed to classify them in two groups: explanations related to problems in the theoretical models that lead to problems in the identification of the relevant econometric model and explanations related to problems in the underlying statistical model that lead to misspecification of the relevant econometric model. Developments in macroeconometric model evaluation after the failure of the Cowles foundation models are then discussed to illustrate how the different critiques have initiated different approaches in macroeconometrics. The evolution of what has been considered the consensus approach to macroeconometric model evaluation over the last thirty years is then followed. The criticism moved to Cowles foundation models in the early seventies might apply almost exactly to DSGE-VAR model evaluation in the first decade of the new millenium. However, the combination of general statistical model, such as a Factor Augmented VAR, with a DSGE model seems to produce forecasts that perform better than those based exclusively on the theoretical and on the statistical model.