Article contents
Experimental study of cursed equilibrium in a signaling game
Published online by Cambridge University Press: 14 March 2025
Abstract
The laboratory experiment described in this paper provides evidence on play in signaling games in the context of electoral competition. In this game, voters must infer the preferred policy of each candidate from the candidate’s choice of whether to announce (truthfully) his preferred policy or to take no position. Bayesian voters would put high probability on a candidate having an extreme policy preference after observing him take no position, but cursed voters would not fully appreciate the informational content of the decision to take no position. Stated beliefs reveal substantial uncertainty about other players’ strategies. Based on estimates of a structural model of cursed equilibrium allowing for heterogeneity in the degree of cursedness, 32% of choices between candidates are consistent with Bayesian updating, 32% imply no inferences about others’ types after observing their actions, and the remainder indicate partial updating. Though the experiment also includes treatments with subjects in both roles, these estimates are based on interactions with programmed candidates, implying that uncertainty about others’ rationality and strategic sophistication is not driving the result. We also find that the quantal response error structure in which errors depend on payoff differences cannot explain the pattern of errors that subjects make.
- Type
- Original Paper
- Information
- Copyright
- Copyright © 2017 Economic Science Association
Footnotes
Electronic supplementary material The online version of this article (doi:https://doi.org/10.1007/s10683-017-9545-0) contains supplementary material, which is available to authorized users.
I thank Prachi Jain, Kaywana Raeburn, and an anonymous reviewer for helpful comments. I also thank Ted O’Donoghue, Steve Coate, Ori Heffetz, Dan Benjamin, and Alex Rees-Jones for their comments on an earlier version of the experiment. I thank Zhongyi Tang for research assistance. I would also like to thank participants in the Cornell Behavioral Economics Research Group and Cornell Behavioral/Experimental Economics lab meeting, and seminar/conference audiences at Cornell University, University of Virginia, Trinity College, Western Economic Association International Annual Conference (2013), Haverford Meeting on Behavioral and Experimental Economics (2013), and RSF Early Career Behavioral Economists Conference (2015) for their feedback.