The Predictive-Ability Criterion and User Prediction Models: A Comment.

Abstract The article presents comments of the author on the article "The Predictive Ability Criterion and User Prediction Models," by Robert H. Ashton published in the October 1974 issue of the periodical "The Accounting Review." Ashton cites a number of potential problems associated with an application of the Brunswik Lens Model to an accounting context. Among other things, he specifically emphasizes infeasibility of prediction model alteration without criterion value availability, desirability of focusing on linear rather than nonlinear prediction models and probable periodic reassessment of prediction models. This comment has two purposes. First, I suggest that criterion values are available and model alteration is, therefore, feasible. Second, aspects of issues desirability of focusing on linear rather than nonlinear prediction models and probable periodic reassessment of prediction models not considered in the original article will be discussed. Ashton notes three characteristics of criterion values. They are typically either separated from the cues by a period of time, imprecise, because of measurement difficulties, or affected by events which are uncontrollable by the decision-maker.