Machine Learning and Forecast Combination in Incomplete Panels
55 Pages Posted: 5 Dec 2013
Date Written: December 2, 2013
Abstract
This paper focuses on a number of newly proposed on-line forecast combination algorithms in Sancetta (2010), Yang (2004), and Wei and Yang (2012). We first establish certain asymptotic properties of these algorithms and compare them with the Bates and Granger (1969) method. We then show that when implemented on unbalanced panels, different combination algorithms implicitly impute missing data differently, so that the performance of the resulting combined forecasts are not comparable. Using forecasts of several important macroeconomic variables from the U.S. Survey of Professional Forecasters, we evaluate the performance of the combination methods, after explicitly accounting for the missing data. We find that even though equally weighted average is hard to beat, the new algorithms deliver superior performance especially during periods of volatility clustering and structural breaks.
Keywords: On-line learning, Recursive algorithms, Unbalanced panel, SPF forecasts
JEL Classification: C22; C53; C14
Suggested Citation: Suggested Citation