Self-resolving Information Markets: A Comparative Study

Abstract

Traditional information markets (TIMs) are resolved with reference to events external to the markets, such as some particular candidate winning an election. However, when making long-term forecasts or evaluating counterfactuals, such resolution is not an option. Hence, the need for self-resolving information markets (SRIMs), resolved with reference to features internal to the markets themselves. The present paper demonstrates experimentally that the market profiles of otherwise identical TIMs and SRIMs show significantly higher degrees of correlation than do randomly paired markets, and that the average accuracies of TIMs and SRIMs are practically equivalent. This supports the so-called face-value hypothesis, on which a convention will arise on SRIMs of taking the question under consideration at face value and betting accordingly, in the same way as on TIMs—in which case SRIMs have the potential of matching TIMs in accuracy while shedding their limitations in relation to long-term predictions and counterfactuals.