An Empirical Study of Subjective Probability Revision in a Financial Context

By William F. Wright
1978| Working Paper No. 414

Bayesian consistency and accuracy of subjective probability assessments for a financial parameter are the major research hypotheses addressed here. Individuals familiar with the information and task environment provided probability assessments for the systematic risk of securities, conditional on alternative profiles of accounting information. A computer-oriented, graphical display procedure was used to elicit the subjective probabilities. The reported posterior probabilities were usually slightly less in magnitude than the probabilities implied by conditional data likelihoods also reported by the subjects. A few individuals displayed considerable insight into environmental data relationships: however a range of accuracy statistics across subjects is reported. The subjects were more internally consistent in their probability assessments than they were accurate in estimating the environmental probabilities. A tendency for individuals to revise probabilities in the direction suggested by environmental evidence, but to an inadequate extent, has been noted frequently in simplified situations using abstract tasks. Given a more realistic task, this group of sophisticated subjects also displayed a tendency for “conservative” probability revision. Discussion of possible cognitive heuristics used by the subjects to generate the probability distributions is included.