An Experimental Analysis of Information Aggregation in Decision Markets

Abstract

Prediction markets, which allow experts to trade assets with payouts that depend on outcomes of future events, are well-established tools to aggregate distributed information. Decision markets are mechanisms derived from prediction markets to aggregate information for decision-making. They predict the consequences of possible actions, and then use a decision rule to select an action based on these predictions. Deterministic and stochastic (i.e. randomizing) decision rules have been suggested, with theory showing that only stochastic decision rules allow for an incentive-compatible elicitation of forecasts. We here present an experimental study of decision markets with participants trading assets in prediction markets, as well as decision markets with stochastic and deterministic decision rules. We find that decision markets with deterministic decision rules are statistically significantly less accurate than prediction markets of the same complexity. The accuracy of decision markets with stochastic decision rules falls between prediction markets and decision markets with deterministic decision rules.

Keywords: Decision Markets, Prediction Markets, Experimental Asset Markets, Collective Decision-making