The persistence of specification problems in the distribution of combined forecast errors
Combinations of several forecasts are now quite commonly used as inputs into business planning models. In such cases, where uncertainty and risk are also being explicitly considered, the statistical specification of the distribution of the combined forecast errors becomes particularly important. In this paper, the shape of the combined forecast error distribution is investigated, non-normality is considered and skewness emerges as a relevant criterion for specifying the method of combination. Subsequently, the stochastic nature of the forecast errors is addressed, and an assessment of current empirical work is presented.