Posterior Probabilities of the Independence Axiom with Nonexperimental Data (or Buckle Up and Fan Out).

This article addresses the issues associated with the construction of posterior probabilities for violation of the independence axiom of expected utility from nonexperimental data. To illustrate the methodology of analyzing nonexperimental evidence, the authors consider seat-belt-usage-data. They find a posterior probability close to one of an Allais-type paradox in these data. In addition, the evidence is not inconsistent with Machina's Hypothesis II but is inconsistent with the "light" hypothesis of Chew and Waller.