Incentive-Compatible Forecasting Competitions

We initiate the study of incentive-compatible forecasting competitions in which multiple forecasters make predictions about one or more events and compete for a single prize. have two objectives: (1) to incentivize report truthfully (2) award prize most accurate forecaster. Proper scoring rules truthful reporting if all are paid according their scores. However, incentives become distorted only best-scoring forecaster wins prize, since can often increase probability having highest score by extreme beliefs. In this paper, we introduce novel competition mechanisms. Our first mechanism is incentive compatible guaranteed select with higher than any other Moreover, show that standard single-event, two-forecaster setting under mild technical conditions, no selects probability. second when forecasters’ beliefs such information event does not lead belief updates on events, it best approaching as number grows. notion compatibility general previous definitions dominant strategy allows reports be correlated outcomes. our mechanisms easy implement generalized related problems outputting ranking over hiring high accuracy future events. This paper was accepted Yan Chen, behavioral economics decision analysis. Funding: work supported European Research Council [Grant ERC StG 307036] National Science Foundation CCF-1445755].