Three essays on the rationality requirements of expected utility theory

The dissertation is composed of three essays on the use of expected utility theory as a positive theory of behavior. The first essay examines new decision theories challenging expected utility theory with the claim of superior behavioral accuracy. Machina's Hypothesis II predicts that indifference curves fan out in the unit probability triangle. It is shown that expected utility with rank dependent probabilities (EURDP), dual theory, and prospect theory predict behavior contrary to fanning out. The results of a laboratory test of these alternative decision theories against each other and against expected utility theory is presented. Important elements of the experimental design are that subjects make choices about real lotteries and the lotteries are selected from inside the unit triangle to minimize framing effects and allow a more accurate test of prospect theory. The predictions of EURDP and dual theory with convex probability weighting functions fare poorly. The only theory that performs appreciably better than expected utility theory is EURDP with concave probability weighting function. The second essay reviews evidence of a disparity between the measures of value willingness to pay and compensation demanded. The expected utility theory prediction of equality in the two measures of value is tested in a laboratory experiment in which subjects participate in second price auctions to indicate willingness to pay and compensation demanded for lottery tickets with fixed odds of winning or losing money. When the auction institution used has truthful revelation of preferences as the dominant strategy, and that strategy is revealed to subjects, the difference between the measures of value decreases significantly. Also demonstrated is that the within subject comparison design provides a more discriminating measure of subjects' behavior than the between subject comparison design used in previous work. The third essay develops a simple dynamic programming model of the cigarette consumption decision. The model allows for the estimation of the marginal lung cancer mortality cost of smoking to the individual. The essay includes a discussion of whether individuals misperceive scientific estimates of the harm of smoking, and use of the marginal mortality cost of smoking as the basis for a misperception correction tax.