Comparison of Mean-Variance Theory and Expected-Utility Theory through a Laboratory Experiment

In the 40's and early 50' two decision theories were proposed and have since dominated the scene of the fascinating field of decision-making. In 1944 - when von Neumann and Morgenstern showed that if preferences are consistent with a set of axioms then it is possible to represent these preferences by the expectation of some utility function - Expected Utility theory provides a natural way to establish "measurable utility". In the early 50's Markowitz introduced the Mean-Variance theory that is the basis of modern portfolio selection theory. Even if both models were analyzed from virtually all possible points of view; although they were tested against several generalizations; even though they seem to be the most attractive theories of decision making, they were never tested against each other. This paper will try to fill this gap. It investigates, using experimental data, which of these two models represent a better approximation of subjects' preferences.