For those condemned to study the past: Reflections on historical judgment

Abstract

Formal Modeling. While the past never repeats itself in detail, it is often viewed as having repetitive elements. People make the same kinds of decisions, face the same kinds of challenges, and suffer the same kinds of misfortune often enough for behavioral scientists to believe that they can detect recurrent patterns. Such faith prompts psychometricians to study the diagnostic secrets of ace clinicians, clinicians to look for correlates of aberrant behavior, brokers to hunt for harbingers of price increases, and dictators to ponder revolutionary situations. Their search usually has a logic paralleling that of multiple regression or correlation. A set of relevant cases is collected and each member is characterized on a variety of dimensions. The resulting matrix is scoured for significant relationships that might aid us in predicting the future. The Daily Racing Form, for example, offers the earnest handicapper some one hundred pieces of information on each horse in any given race. The handicapper with a flair for data processing might commit to some computer’s memory the contents of a bound volume of the Form and try to derive a formula predicting speed as a weighted sum of scores on various dimensions. For example: y = b 1 x 1 + b 2 x 2 + b 3 x 3 (1) where y is our best guess at a horse’s speed, x 1 is its percentage of victories in previous races, x 2 is its jockey’s percentage of winning races, and x 3 is the weight it will carry in the present race. When scores are standardized (by subtracting the mean and then dividing by the standard deviation), the b i reflect the importance of the different factors. If b 1 = 2b 2, then a given change in the horse’s percentage of wins affects our speed prediction twice as much as an equivalent change in jockey’s percentage, because past performances have proved twice as sensitive to x 1 as to x 2.