Brian P. Mathews and A. Diamantopoulos (1986), "Managerial
intervention in forecasting: An empirical investigation of forecast manipulation,"
International Journal of Research in Marketing, 3.
This study examined short-range forecasts for 281 repeat purchase products for a single
company operating in the U.K. health care industry. These products represent
all those for which management thought it worthwhile to make a subjective revision to the
forecast provided by Holts two-parameter exponential smoothing. The forecasts were
prepared six weeks prior to the start of the quarter. The revisions were made by the
product managers. (Each revision was based on only one subjective revision, but a total of
four product managers were involved.) Eight error measures were used; of these, three
showed no significant difference, four (absolute error, R2, percentage
better, and size of the improvement) showed that the subjective revisions were
significantly better, and the last (percentage error) showed that the subjective revisions
were significantly worse. The authors concluded that the revisions helped in this case.
While this study represents a useful first step on an important problem, these results
provide only weak support for subjective revisions because they were drawn from only one
company, one starting point, limited information on why the subjective revisions should be
better in this situation, and mixed results.