Improving the assessment of subjective probabilities of joint and conditional events

Subjects estimated probabilities of real (Studies 1 - 3) and hypothetical (Study 4) events, using direct estimation and/or the pair-wise approach. The direct estimates were combined to derive indirect (Studies 1 - 4) and scaled estimates (Studies 2, 4). Probabilities were also extracted from ratios (Studies 3, 4). In general probabilities elicited via the traditional direct method were the worst performing, in terms of internal consistency and accuracy, among the methods compared. In Study 4, more subjects opted to review their responses during direct elicitations than in pair-wise comparisons, but the increased attention did not result in greater accuracy. Findings also indicated that, for low-numerate individuals, judgments elicited using the pair-wise approach were more accurate than the direct ones. The pair-wise comparison approach is shown to be a compelling and viable assessment method for eliciting subjective probabilities. The approach can also be further extended to (a) elicit parameters of continuous distributions, (b) elicit experts\u27 subjective estimates for expressing uncertainty, (c) aid individuals in making decisions and (d) the design of risk communication materials. Future studies should also investigate the advantages of the approach for different subgroups (e.g., older, lowly educated) in the population