Working Paper: CEPR ID: DP4397
Authors: Hans Peter Grüner; Elisabeth Schulte
Abstract: This Paper provides a game theoretic extension of Radner's (1993) model of hierarchical information aggregation. It studies the role of the hierarchy design for the speed and quality of a collective decision process. The hierarchy is described as a programmed network of agents. The programme describes how information is processed within the network. The network of P identical managers has to aggregate information in the form of a set of n data items in order to make an informed decision. Each manager benefits from reaching an accurate decision but suffers from an individual cost of effort, which has to be provided in order to understand the information contained in a data item properly. We find that decentralized information processing increases incentives for information provision. There may be boundaries on the appropriate extend of decentralization, however. We also compare three different hierarchy designs: two balanced hierarchies and the fastest (skip-level) hierarchy, proposed by Radner. Skip-level reporting outperforms balanced hierarchies in terms of decision speed and in terms of decision quality.
Keywords: hierarchies; incentives for information provision; information processing
JEL Codes: D23; D70; D83; L22; P51
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| decentralized information processing (D87) | increased incentives for information provision (D82) |
| hierarchy structure (L22) | speed and quality of decisions (L15) |
| reduced tree hierarchy (Y10) | speed and quality of decisions (L15) |
| decentralized structure (L22) | speed and quality of decisions (L15) |