Integrating judgment in statistical demand forecasting: An approach to confront uncertainty

This paper investigates the potential value of judgment in forecasting demand after sudden changes in the external environment and in the presence of a high level of uncertainty. We forecast the daily load demand in Japan after the country fs 2011 severe energy crisis. The study examines statistical and judgmental techniques as competing or as complementary approaches, in the light of the availability of contextual information and relevant time-series data. The result indicates that immediately after a special event, the availability and dominance of contextual information seem to be the determinants of judgmental superiority over statistical models. However, when relevant time-series data are observed, statistical forecasting outperforms judgmental forecasting. When neither contextual information nor relevant time-series data is dominant, a combination of both methods succeeds in generating accurate forecasts. In addition, judgment is better in a combination framework than in the adjustment of statistical outputs.