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

Expertise, credibility of system forecasts and integration methods in judgmental demand forecasting

“…In an extensive demand forecasting study, Fildes et al (2009) showed that 50-50 combination (also known as the Blattberg-Hoch method) improves the accuracy by decreasing the harmful impact of unjustified large (and usually positive) adjustments. These results are also echoed using non-expert participants (e.g., in extrapolation tasks without contextual information, see Webby & O'Connor, 1996). The current study aims to fill this research gap through a formal comparison of these three integration methods via a demand forecasting task with real experts in their natural settings.…”

“…Judgmental adjustment may improve the accuracy, particularly when the expert judgment incorporates information about special events and contextual knowledge into unstable series (Fildes et al, 2009;Goodwin, 2002;Webby & O'Connor, 1996). However, such adjustments may be influenced by several biases, including overconfidence in the expert's own judgment (Friedman et al, 2001;Lawrence et al, 2006;Lim & O'Connor, 1996;Sanders, 1997); anchoring and adjustment (i.e., anchoring the forecast to a single cue like the last point or the system forecast, and then making insufficient adjustments to this cue; see Epley & Gilovich, 2006;Fildes et al, 2009;Goodwin, 2005;Lawrence & O'Connor, 1995); and a predisposition to adjust (forecasters making many small harmful adjustments to the system forecasts without any specific reason, leading to a deterioration in accuracy; see Fildes et al, 2009;Lawrence et al, 2006;Önkal, Gönül, & Lawrence, 2008;Sanders & Manrodt, 1994).…”

“…It is also necessary to have (i) the ability to integrate this information into the final forecast, and (ii) the motivation to do such integration (Gavrilova & Andreeva, 2012). The review by Webby and O'Connor (1996) showed that experiential knowledge of the cause-effect relationships encountered in the industry may not be a good predictor of superior accuracy. Another study, by Edmundson, Lawrence, and O'Connor (1988), showed that intimate domain knowledge elicited from experts was useful only for the most important products, not for the others.…”

“…Comparisons among such methods are quite uncommon, as the extant research has usually focused on each technique separately (Webby & O'Connor, 1996), leading Goodwin (2002) to call for more direct comparisons. Exploring the performance of judgmental integration methods is important both for the efficient design of forecast support systems (FSS) and for gaining an understanding of the conditions required for the effective elicitation and use of expert knowledge, in order to improve the functioning of these systems.…”

Biases of Professional Exchange Rate Forecasts: Psychological Explanations and an Experimentally Based Comparison to Novices

“…the output of time series analysis models, in order to observe behaviour in settings closer to reality. For a detailed discussion we refer to Webby and O'Connor, [1996], who reviewed the literature about judgmental and statistical time series forecasting intensively.…”