Separating Bayesian Updating from Non-Probabilistic Reasoning: An Experimental Investigation

Through a series of decision tasks involving colored cards, we provide separate measures of Bayesian updating and non-probabilistic reasoning skills. We apply these measures to (and are the first to study) a common-value Dutch auction. This format is more salient than the strategically equivalent first-price auction and silent Dutch formats in hinting that one should condition one's estimate of the value on having the highest bid. Both Bayesian updating skills and non-probabilistic reasoning skills are shown to help subjects correct for the winner's curse, as does the saliency of the active-clock Dutch format.