Comparing Forecasting Methods: Why Do Traditional Macroeconometric Models Remain Popular?

Abstract

No macroeconometeric forecasting technique anticipated the financial market meltdown and subsequent Great Recession in 2008-2009. This failure is not surprising because it is hard to forecast unprecedented events. The experience explains, however, why many forecasters, particular those supplying information directly to decision makers, continue to rely on forecasting systems based on large scale econometric models. "Purist" forecasting methods include dynamic stochastic general equilibrium models (DSGE) and vector autoregressions (VAR), but each of these techniques has severe limitations for practical forecasting. By contrast, the traditional forecasting system using a structural model combined with the forecasters judgment allows the forecaster to react to unprecedented events and sudden regime changes. Combining judgmental and model-based forecasting is common in other fields, and there is likely greater marginal value in developing this approach further than in refining "purist" forecasting techniques such as DSGE models and VARs.

Keywords: forecasting

JEL Classification: C53, E17