Providing Incentives for Better Cost Forecasting

Abstract

The dissertation examines contractual schemes for eliciting information about an uncertain outcome. Such information is requested regularly in the course of central planning, government contracting, risk-analysis, exchange-rate prediction, weather forecasting, and in a variety of other contexts. In the basic framework, the agent, usually a manager, knows the probability distribution of a random variable (say "project cost", or "achievable output"), or perceives an unknown outcome as if it were a random variable with this distribution. The planner "principal" wants to obtain some summary measure of this information. After the reports are made, the outcome will be observed by both parties. The planner's problem is to devise a compensation scheme providing positive incentives to the manager to reveal his beliefs truthfully. Such a scheme is called a proper scoring rule. The dissertation provides a virtually complete characterization of proper scoring rules for a broad class of probability indicators, thereby unifying and extending previous work on scoring rules. Some practical complications confronting potential users of these schemes are explored, including conditional project acceptance, risk-aversion, cost padding dangers, multi-attribute evaluation, multi-agent bidding, and multi-period contracting. In addition, agent learning costs are incorporated into the model, allowing for the derivation of optimal contracts and explicit consideration of the costs of agency.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 84 citations exist upstream.