Abstract
We demonstrate that the so-called forecasting combination puzzle is a consequence of the methodologies commonly used to produce forecast combinations. By the combination puzzle, we mean that complex predictions formed by combination methods can often be outperformed by simpler, e.g. equally weighted, approaches. As a consequence, we show that tests of forecast superiority that seek to discriminate between the predictive accuracy of different forecast combination methods have low power, and can lack size control. Using these results, we show that the forecast combination puzzle can be completely avoided by using more effective inference strategies. We demonstrate these findings both in the context of forecasting a functional of interest and in terms of predictive densities.
About the presenter
Associate Professor David Frazier is an ARC DECRA and Associate Professor in Econometrics and Business Statistics at Monash University, working in the dual disciplines of econometrics and statistics. His research interests are broad, but his primary focus has been on the area of simulation-based interference.
He will be visiting the School of Economics on Friday 29th April and will be using room 520A Colin Clark Building (39). If you would like to meet with Associate Professor David Frazier or have lunch with him, please contact Professor Alicia Rambaldi (a.rambaldi@uq.edu.au), who will be his host while at The University of Queensland.
About School Seminar Series
The School of Economics General Seminar Series is held on Fridays. These are in-person and presented by a range of guest researchers from around Australia and internationally.