User Prediction Models in Accounting: An Alternative Use.

Abstract The article is concerned with improving the utilization of accounting information in prediction situations. In this context improvement is defined in terms of an increase in prediction accuracy. If adequate criterion information is available, increase in prediction accuracy may be "directly" assessed by measuring the correspondence between criterion values and predictions of those values. The article recognizes that there are at least two approaches to improving prediction accuracy. The first approach is to improve the data which are supplied to decision-makers and used as a basis for predictions. The second approach is to improve the way that predictions are made on the basis of whatever data are used. The author presents evidence from the psychology literature that individuals may usefully be replaced by their models in prediction tasks, and he also discusses the reasons for this situation. He presents evidence from the management literature which leads to the same conclusion and discusses implications for accounting of these findings.