Forecast Combination Modeling Using A Pooled Forecast Weights In A Linear And Homoskedastic Environment

ISBN: 978-981-08-8407-9

Authors: Yip Chee Yin and Lim Hock Eam

Abstract:

This paper proposes forecast combination model based on Weighted Averaging Forecast Weights (WAF) of Bates and Granger Model Averaging (BGMA) and Simple Model Averaging (SMA). We find that WAF is an asymptotically unbiased estimate of both the mean squared forecast error (MSFA) and one-step-ahead mean squared forecast error (MSFE). We show how to compute WAF weights in forecasting settings. In empirical experiment, we find that WAF forecasts have low MSFE. In addition, it has much lower maximum regret than others feasible forecasting models like equal weighting and BIC selection combination

Keywords: Forecast combination model; Weigthed Averaging Forecast Weights; Simple Model Averaging; Bates and Granger Model Averaging.