An Experimental Analysis of Nash Refinements in Signaling Games
- Creators
- Banks, Jeffrey
-
Camerer, Colin F.
- Porter, David
Abstract
Refinements of Nash equilibrium are investigated in two-person signaling game experiments. The experiments cover several nested refinements: Bayes–Nash, sequential, intuitive, divine, universally divine, NWBR, and stable. The experimental data suggest that subjects select the more refined equilibria up to divinity. However, an anomaly occurs in one game in which the stable equilibrium is preferred to an NWBR equilibrium. Since the refinements are nested this anomaly implies that outcomes are game specific. Deviations from Nash behavior do not seem to follow any specific decision rule (e.g., Nash, minimax, principle of insufficient reason, etc.). Most choices by both players are part of some Nash equilibrium, but deviations from equilibrium behavior occur when the choices are part of different equilibria.
Additional Information
© 1994 Academic Press. Received June 21, 1990. Available online 24 April 2002. We are indebted to Mark Olson for help in statistical analysis and programming. We also thank Richard McKelvey whose GAMBIT program was used to check the equilibria of our games. Audiences at Washington University, the University of Arizona, the University of Houston, and the University of Pennsylvania, and several referees provided helpful comments. This research was partially funded by the Wharton Risk and decision Processes Center. Support for the second author was provided by the National Science Foundation (SES 8708566).Additional details
- Eprint ID
- 22129
- Resolver ID
- CaltechAUTHORS:20110211-093516207
- Wharton Risk and Decision Processes Center
- SES-8708566
- NSF
- Created
-
2011-02-11Created from EPrint's datestamp field
- Updated
-
2021-11-09Created from EPrint's last_modified field