Consistent Bayesians Are No More Accurate Than Non-Bayesians. Economists Surveyed About PSA

Abstract

This paper looks for, but cannot find, evidence that links belief inconsistency to belief inaccuracy or economic loss. Economists with consistent (i.e., Bayesian) conditional beliefs about the sensitivity and positive predictive value of the Prostate Specific Antigen (PSA) test have unconditional beliefs about the risk of prostate cancer that are, if anything, less accurate than the beliefs of inconsistent non-Bayesians. PSA decisions depend more on the advice of doctors and family members than on beliefs about cancer risks. Men’s beliefs about the pros and cons of PSA testing do not explain self-reported PSA decisions. This absence of evidence that non-Bayesian beliefs lead to economic loss suggests that belief consistency may be relatively unimportant as a normative criterion in high-stakes decision tasks that reward accuracy instead of consistency. A technique is introduced for eliciting measures of both consistency and accuracy of an individual’s probabilistic beliefs.