Speed and quality of collective decision making: Incentives for information provision

We study a one-shot information aggregation problem in which agents have to provide effort in order to understand the information they are supposed to process. Agents have a common interest in reaching a good decision but suffer from an individual cost of providing effort. Showing that any problem which is incentive compatible for a single information processor is incentive compatible for a decentralized organization, but not vice versa, we derive a new rationale for decentralized information processing. For a class of problems, the fastest organization – the reduced tree proposed by Radner (1993) – yields also the best incentives for information processing.