The Forecast Combination Puzzle: A Simple Theoretical Explanation

This paper offers a theoretical explanation for the stylized fact that forecast combinations with estimated optimal weights often perform poorly in applications. The properties of the forecast combination are typically derived under the assumption that the weights are fixed, while in practice they need to be estimated. If the fact that the weights are random rather than fixed is taken into account during the optimality derivation, then the forecast combination will be biased (even when the original forecasts are unbiased), and its variance will be larger than in the fixed-weight case. In particular, there is no guarantee that the ‘optimal’ forecast combination will be better than the equal-weight case, or even improve on the original forecasts. We provide the underlying theory, some special cases, and a numerical illustration.

[1]  B. M. Pötscher.  Effects of Model Selection on Inference , 1991, Econometric Theory.

[2]  Giuseppe De Luca,et al.  Weighted‐Average Least Squares (WALS): A Survey , 2016 .

[3]  H. Akaike,et al.  Information Theory and an Extension of the Maximum Likelihood Principle , 1973 .

[4]  N. Hjort,et al.  Frequentist Model Average Estimators , 2003 .

[5]  C. H. Oh,et al.  Some comments on , 1998 .

[6]  T. W. Anderson,et al.  An Introduction to Multivariate Statistical Analysis , 1959 .

[7]  David R. Anderson,et al.  Model selection and multimodel inference : a practical information-theoretic approach , 2003 .

[8]  J. Stock,et al.  How Did Leading Indicator Forecasts Perform during the 2001 Recession? , 2003 .

[9]  Andreas Graefe,et al.  Combining Forecasts: An Application to Elections , 2013 .

[10]  Anja Vogler,et al.  An Introduction to Multivariate Statistical Analysis , 2004 .

[11]  K. Wallis,et al.  A Simple Explanation of the Forecast Combination Puzzle , 2009 .

[12]  Graham Elliott,et al.  Averaging and the Optimal Combination of Forecasts , 2011 .

[13]  J. M. Bates,et al.  The Combination of Forecasts , 1969 .

[14]  T. W. Anderson.  An Introduction to Multivariate Statistical Analysis , 1959 .

[15]  Colin L. Mallows,et al.  Some Comments on Cp , 2000, Technometrics.

[16]  C. Mallows.  More comments on C p , 1995 .

[17]  G. Schwarz.  Estimating the Dimension of a Model , 1978 .

[18]  Bruce E. Hansen,et al.  Least-squares forecast averaging , 2008 .

[19]  Alan T. K. Wan,et al.  Optimal Weight Choice for Frequentist Model Average Estimators , 2011 .

[20]  Shui Ki Wan,et al.  Is there an optimal forecast combination , 2014 .

[21]  Paul Kabaila,et al.  The Effect of Model Selection on Confidence Regions and Prediction Regions , 1995, Econometric Theory.

[22]  Jan R. Magnus,et al.  On the harm that ignoring pretesting can cause , 2004 .