Averaging model forecasts and expert forecasts: Why does it work?
This paper addresses a situation in which a manager has forecasts of country-specific stock-keeping unit (SKU)-level sales data from both a statistical model and from a range of experts. Empirical evidence suggests that averaging model forecasts and expert forecasts could give more accuracy than the individual forecasts do. At the same time, empirical evidence demonstrates that expert forecasts tend to be biased. So, why is it that this average could work? This paper assumes that expert forecasts can be decomposed into a judgmental bootstrapping equation, to be created by the manager, and a part based on unobserved intuition. When the bootstrapping equation dominates the intuition, it is this underlying model for the expert that can make the combined average forecast more accurate. An illustration for no less than 1,221 cases supports this conjecture.
|Keywords||Combining forecasts, Expert forecasts, Model forecasts|
|Persistent URL||dx.doi.org/10.1287/inte.1100.0554, hdl.handle.net/1765/23930|
|Series||ERIM Top-Core Articles , Econometric Institute Reprint Series|
Franses, Ph.H.B.F. (2011). Averaging model forecasts and expert forecasts: Why does it work?. Interfaces (Hanover), 41(2), 177–181. doi:10.1287/inte.1100.0554