The Use of Knowledge in Prediction Markets: How Much of Them Need He Know

Abstract

In this article, we extend an early agent-based spatial model of the prediction market by taking into account the heterogeneities of agents in their tolerance capacity (tolerance to neighbors with different political identities) and in their exploration capacity (exploration of the political identities of other agents). We then study the effects of these heterogeneities on the behavior of the prediction market, including prediction accuracy, determinants of earnings, and income distribution. First, in terms of prediction accuracy, we find that, compared to the homogeneous case, bringing heterogeneity into the model can generally improve the prediction accuracy, although its statistical significance is limited. In particular, the well-known empirical regularity known as the favorite-longshot bias remains almost unchanged with this extension. Second, through the heterogeneous-agent design, we find that both capacities (personality traits) of agents have a significant positive effect on earnings, and the effect of the exploration capacity is even more dramatic. Third, through their effects on earnings, both capacities also contribute to income inequality, but only to a mild degree with a Gini coefficient of 0.20.