Optimal Forecasting Incentives
研究如何设计合约，让一个能力未知的预测者既诚实报告又投入适当的学习，以最小化预期损失和支付，并比较了自我筛选和竞争机制的效果。
An agent of unknown expertise is requested to forecast the mean of an uncertain outcome. The agent can refine forecasts at a constant marginal cost per unit precision, but neither cost nor precision can be verified by the planner. The problem is to induce both truthful revelation and an appropriate degree of learning so as to minimize the expected sum of direct planning losses and agent payments. Optimal contracts are derived with and without self-screening of expertise and with and without competition between agents. Self-screening tends to be much less valuable than competition. Copyright 1989 by University of Chicago Press.