Individual risk preferences have largely been empirically analyzed through surveys and experiments. These studies have shed serious doubt on one of the bedrocks of applied economic modeling---the expected utility hypothesis. In this paper, we show how price variation in betting and prediction markets (e.g., the odds market at horse racetracks) can identify the structure of risk preferences, thus providing an opportunity to test the expected utility hypothesis using real world market data rather than experimental data. Since betting/prediction markets naturally present themselves as one period exchange economies, we setup a general equilibrium model to explain the pricing of bets. Under very mild regularity assumptions on the distribution of information and preferences in the population, we show that our model admits a unique fully revealing rational expectations equilibrium (REE), which very naturally generates price variation across markets. By restricting the model so that individual differences in risk preferences are one dimensional, we show how it is possible to nonparametrically recover the distribution of risk preferences from the observed pattern of price variation across markets. Using a data set consisting of the prices and winners from all North American horse races over a three year period (2001-2003), we estimate the distribution of preferences in our general equilibrium model. We find that there exists economically significant heterogeneity of preferences. Moreover, almost all of the variation in prices can be explained by simple CRRA preferences. Thus our structural analysis shows that market data paint a very different picture of the empirical validity of the expected utility hypothesis as compared to experimental data.
展开▼
