Risk, Time, Altruism, and Emotions

University Of California-San Diego, La Jolla CA

Investigators

Abstract

Since the introduction of Expected Utility (EU) the theory of preferences over risk has been an unsettled matter. At the same meeting in 1952 at which Savage presented his proof of EU, introducing the Independence Axiom (IA), Allais was gathering data from the audience on his famous counter-examples to IA. Since then, countless more rejections of IA have appeared, and many alternative models have gained attention. Using two contributions from my prior NSF support, Convex Risk Budgets and Uncertainty Equivalents, I show that EU actually performs better than popular alternatives when all outcomes are sufficiently risky, and that the independence axiom is most at peril when some outcomes involve certainty. We are, of course, not the first to notice this, but both of the experiments we have conducted demonstrate this in non-parametric tests, and are not confounded by other model or measurement restrictions. These findings suggest researchers may have "under-sampled" certainty, that is, we have spent too little effort exploring how IA breaks down at the edges of the probability space. This proposal describes a series of experiments designed to measure how preferences shift as the environment becomes less risky. The objective is to collect data necessary to characterize the restrictions on preferences throughout the choice space, and to do so in a completely non-parametric way such that the data are not confounded with framing or modelling assumptions. I will use these results to fit a new model of preferences which I call a model of Expecting the Unexpected.