Improving the Simple Average Combined Forecast via Factor-Adjusted Regularization

This paper addresses the forecast combination puzzle—the empirical observation that the simple average forecast combination often outperforms complex weighting schemes—by employing a factor-adjusted regularization framework. In this framework, the simple average is treated as a common factor. We identify and incorporate idiosyncratic components, defined as the deviations of individual forecasts from the simple average. Our approach effectively manages the high correlation among forecasts by focusing on idiosyncratic components that enhance predictive content beyond what is captured by the simple average. Empirical applications in macroeconomic forecasting show that this factor-adjusted approach yields significant accuracy gains over the simple average.