On the Forecast Combination Puzzle

  title={On the Forecast Combination Puzzle},
  author={Wei Qian and Craig Rolling and Gang Cheng and Yuhong Yang},
It is often reported in the forecast combination literature that a simple average of candidate forecasts is more robust than sophisticated combining methods. This phenomenon is usually referred to as the “forecast combination puzzle”. Motivated by this puzzle, we explore its possible explanations, including high variance in estimating the target optimal weights (estimation error), invalid weighting formulas, and model/candidate screening before combination. We show that the existing… 

Figures and Tables from this paper

Forecast combinations: An over 50-year review

A Heuristic for Combining Correlated Experts When There is Little Data

It is intuitive and theoretically sound to combine experts’ forecasts based on their proven skills, while accounting for correlation among their submissions. Simpler combination methods, however,

Combining Predictions of Auto Insurance Claims

Overall, model combination methods improve the prediction accuracy for auto insurance claim costs and Adaptive Regression by Mixing (ARM), ARM-Tweedie, and constrained Linear Regression can improve forecast performance when there are only weak learners or when no dominant learner exists.

Forecast combinations for benchmarks of long-term stock returns using machine learning methods

Forecast combinations are a popular way of reducing the mean squared forecast error when multiple candidate models for a target variable are available. We apply different approaches to finding

Model averaging for asymptotically optimal combined forecasts

A High-Dimensional Focused Information Criterion

This work extends the focused selection process to the high-dimensional regression setting with potentially a larger number of parameters than the size of the sample and obtains an alternative expression of the low-dimensional focused information criterion that can directly be applied.


Estimating a treatment’s effect on an outcome conditional on covariates is a primary goal of many empirical investigations. Accurate estimation of the treatment effect given covariates can enable the

Optimal Designs for Model Averaging in non-nested Models

In this paper we construct optimal designs for frequentist model averaging estimation. We derive the asymptotic distribution of the model averaging estimate with fixed weights in the case where the

Composite versus model-averaged quantile regression



Chapter 4 Forecast Combinations

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

Another look at forecast selection and combination: Evidence from forecast pooling

To Combine Forecasts or to Combine Information?

When the objective is to forecast a variable of interest but with many explanatory variables available, one could possibly improve the forecast by carefully integrating them. There are generally two

Optimal Selection of Expert Forecasts with Integer Programming

Combinations of point forecasts from expert forecasters are known to frequently outperform individual forecasts. It is also well documented that combination by simple averaging very often has

A Simple Explanation of the Forecast Combination Puzzle

This article presents a formal explanation of the forecast combination puzzle, that simple combinations of point forecasts are repeatedly found to outperform sophisticated weighted combinations in

Online learning and forecast combination in unbalanced panels

It is found that the equally weighted average continues to be hard to beat, but the new algorithms can potentially deliver superior performance at shorter horizons, especially during periods of volatility clustering and structural breaks.