JISCMail - RISK-AND-DECISION Archives

Below is the title and abstract for this week's LJDM talk.

LJDM seminars are held Wednesdays at 5:00 in Room 313 at the UCL
Psychology Department, 26 Bedford Way WC1H 0AP.

The full schedule is at http://ljdm.info

Wednesday, 13 May, 5 PM:

Allowing for promotion effects in judgmental forecasting: Effects of series type and provision of formal forecasts

Shari De Baets

Vlerick Business School

Abstract

Judgmental forecasting aims to combine the strengths of statistical forecasting and judgments made by the forecasters. While statistical forecasting methods can generate predictions based on the logical and systematic processing of information and can handle large amounts of data (Goodwin & Wright, 2010), they have difficulties with discontinuities, unexpected events and external influences (Armstrong & Collopy, 1998; Goodwin, 2002; Hughes, 2001; Taleb, 2007). One such example is the occurrence of promotional investments by the company and its effect on future sales numbers. Especially, the prediction of rare promotional events remains a difficult hurdle to take (Goodwin & Fildes, 1999). This study was conducted to examine how forecasters perform in predicting sales numbers when the time series are subject to occasional promotional events. This was investigated in a series of experiments (n1 = 39, n2 = 36, n3 = TBD, n4= TBD) by varying within experiments (a) the difficulty of the time series: independent versus autoregressive (AR(1)), (b) presence of a statistical forecast and forecast history, and across experiments: (c) the presence of feedback and (d) the relationship between the promotional expenditure and the effect on sales. Preliminary analysis of Experiment 1 (linear relationship, no feedback) indicates that forecasts are too low when promotions are present and too high without promotions, suggesting an anchoring effect on the mean of the series. This effect increases with the presence of a statistical forecast. Results of Experiment 2 (S-shaped relationship, no feedback) generally confirm those of Experiment 1. At the time of the presentation, at least two more experiments will have run: one investigating the effect on forecasting performance of providing immediate outcome feedback and one experiment on the effect on forecasting performance of providing the relationship between the promotional expenditure and sales increase. The provision of additional information is hypothesized to increase judgmental forecasting.