Evolution of Risk Preferences.pdfVIP

Paolo Pin pin@unive.it http://venus.unive.it/pin Evolution of Risk Preferences June 2005 Abstract This paper presents a simulation, made with a genetic algorithm (GA), in order to shape risk attitudes in a very simple environment of lotteries such as the Machina triangle. GA models the process of learning and imitation from successful traders in real world economics in a satisfactory way. An overlapping gener- ations case is used where bankrupted agents exit the market and are replaced by new-entrants that select their models with probabilities that can be proportional not only to instantaneous economical success, but also to longevity in the market or to financial stability (different kinds of fitness in the GA). The genetic code of the agents is nothing else but a permutation (the order of preference) on a finite number of lotteries forming a grid on the Machina triangle, with the possibility of winning or loosing fixed amounts. The resulting indifference curves are shaped graphically, moreover measuring how far they are from the ideal risk neutral case. The main result is that imposing the simple condition of the bank- ruptcy benchmark, with instantaneous wealth as fitness, is enough to induce stochastic dominance. JEL Classification Numbers: C61, D81, D83. Keywords: Risk Preferences, Genetic Algorithm. 1 Introduction After the contribution of Expected Utility Theory by von Neumann and Morgen- stern (1944) attitudes towards risk have almost always been directly related to utilities. In that framework, if an agent can measure her utilities and believes, and if she can compute weighted averages, she is able to give an utility to random 1 events. The assumptions has been criticized by some paradoxes, as Allais’ (1953) famous one, by empirical results from financial and insurance markets, and by the outcomes of experimental economics. The model remains nevertheless the most robust and widely used one to extend utilities to the world of uncertainty. We will not enter