Download this Paper Open PDF in Browser

Groupthink and the Failure of Information Aggregation in Large Groups

33 Pages Posted: 24 Dec 2014 Last revised: 30 Nov 2017

Matan Harel

University of Geneva

Elchanan Mossel

University of Pennsylvania - Statistics Department

Philipp Strack

University of California, Berkeley - Department of Economics

Omer Tamuz

California Institute of Technology - Division of the Humanities and Social Sciences

Date Written: December 22, 2014

Abstract

We study how effectively long-lived rational agents learn from repeatedly observing each others’ actions. We find that in the long run, information aggregation fails, and the fraction of private information transmitted goes to zero as the number of agents gets large. With Normal signals, in the long-run, agents learn less from observing the actions of any number of other agents than they learn from seeing three other agents’ signals. We identify rational groupthink — in which agents ignore their private signals and choose the same action for long periods of time — as the cause of this failure of information aggregation.

Keywords: Bayesian learning

JEL Classification: C73, D82, D83

Suggested Citation

Harel, Matan and Mossel, Elchanan and Strack, Philipp and Tamuz, Omer, Groupthink and the Failure of Information Aggregation in Large Groups (December 22, 2014). Available at SSRN: https://ssrn.com/abstract=2541707 or http://dx.doi.org/10.2139/ssrn.2541707

Matan Harel

University of Geneva ( email )

102 Bd Carl-Vogt
Genève, GA CH - 1205
Switzerland

Elchanan Mossel

University of Pennsylvania - Statistics Department ( email )

Wharton School
Philadelphia, PA 19104
United States

Philipp Strack (Contact Author)

University of California, Berkeley - Department of Economics ( email )

549 Evans Hall #3880
Berkeley, CA 94720-3880
United States

HOME PAGE: http://philippstrack.com

Omer Tamuz

California Institute of Technology - Division of the Humanities and Social Sciences ( email )

1200 East California Blvd.
Pasadena, CA 91125
United States

Paper statistics

Downloads
381
rank
65,983
Abstract Views
1,560