Reviews of Important Papers on Forecasting,
1985-1995 Reviews
Review of:

Ben-Shakhar, Gerson, M. Bar-Hillel, Y. Bliu, E. Ben-Abba, and A. Flug (1986), "Can graphology predict occupational success? Two empirical studies and some methodological ruminations," Journal of Applied Psychology, 71, 645-653.

[Review written with J. Scott Armstrong]

The analysis of handwriting (graphology) is used extensively for personnel selection in Europe. One survey indicated that 85% of European firms use it. Furthermore its use seems to be growing. Does this success in the market mean that the predictions are valid? No. Previous research has not provided evidence of predictive ability. Ben-Shakhar et al. add two carefully done studies: The first study deals with forecasting job success, and the second one asks subjects to predict the occupations of a group of highly successful people. In predicting job success, where the criterion was the supervisor's evaluation, three graphologists who were on the company's payroll did no better than a clinical psychologist with no knowledge of graphology who read the same materials. In the study for the prediction of occupations, the graphologists were given materials written by 40 highly successful people from various occupations mathematicians, psychologists, artists, physicians, executives, architects; and jurists. Each person wrote exactly the same words in their primary language, Hebrew. Five Israeli graphologists took part in the study, three of them being famous in this field. (A dozen other graphologists had refused to participate.) The graphologists predicted the occupation of the writers no more accurately than chance, even though some writing conventions are used in certain occupations. For example, architects use a "B"' made up of a small oval above a large oval. Combined with previous studies using different samples, in different countries, with different criteria, and different study designs, one might conclude with confidence that graphology has no predictive validity.

The authors then examine why graphology sells. It does so because it appears to work. First, it has high face validity. Second, the users do not systematically evaluate the predictive accuracy. Graphologists have certain tricks of the trade that make the predictions seem to be correct. [If we used our unaided observation of magicians to assess the world, it would lead to some strange conclusions about the behavior of objects.] Third, personnel predictions are important and difficult; executives are happy to pass along the responsibility for this task to someone who claims to have the professional ability to make such predictions. Fourth, while the graphologists do not add validity due to their handwriting analysis, they do read the application materials and provide predictions. While these predictive validities are a bit poorer than those derived from a battery of tests, handwriting analysis is less expensive than the tests. Given all these reasons, I predict that graphology will continue to enjoy success in large bureaucratic firms in Europe.