Judgemental and statistical time series forecasting: a review of the literature

Abstract

This paper reviews the literature on the contributions of judgemental methods to the forecasting process. Using a contingent approach, it first reviews the empirical studies comparing the performance of judgemental and statistical methods and finds emphasis for the importance of judgement in providing contextual information for the final forecasts. It then examines four methods of integrating contextual information with the output of statistical models. Although judgemental adjustment of statistical forecasts is a viable alternative, simple combination of forecasts may offer superior benefits. Promising developments can also be gained from the use of decomposition principles in the integration process.