The Winner's Curse: Conditional Reasoning & Belief Formation

Abstract

We investigate the relevance of conditional reasoning and belief formation for the occurrence of the winner's curse with the help of two experimental manipulations.First, we compare results from a very simple common-value auction game with results from a transformed version of this game that does not require any conditioning on future events. In human opponent settings, we observe significant differences in behavior across the two games. Second, we investigate subjects' behavior when they face naïve computerized opponents and after they have faced them. We find that both strong and weak assistance in belief formation changes subjects' play significantly in the auction game. Overall, the results suggest that the difficulty of conditioning on future events is at least as important in explaining frequent occurrences of the winner's curse as is the challenge to form beliefs.JEL classification: D44, D81, D82

Hypothetical Thinking and the Winner's Curse: An Experimental Investigation

“…Ivanov, Levin, and Niederle (2010) showed that belief-based models, like cursed equilibrium, might be not that powerful in explaining the winner's curse, but they provided no alternative explanation. The paper of Koch and Penczynski (2014) is also closely related to my work. So far they are the only ones who combined contingent reasoning and belief-based models in their experiment.…”

“…Whereas this is well documented in the psychological literature (see, for example, Evans, 2007;Nickerson, 2015;Singmann, Klauer, and Beller, 2016), economists devoted little attention to this issue for a long time. However, in the more recent economic literature this topic appears more and more frequently (see, for example, Charness and Levin, 2009;Louis, 2013;Esponda and Vespa, 2014;Koch and Penczynski, 2014;Ngangoue and Weizsäcker, 1 Both models fall into the category of belief-based models, since the cause manifesting in the winner's curse is seen in the belief formation of individuals. The general assumptions of Bayesian Nash Equilibrium are still fulfilled, in the sense that subjects best-respond to beliefs, but the assumption about the consistency of beliefs is relaxed.…”

“…Additionally I only changed one parameter in stage II, whereas Koch and Penczynski (2014) used two different games, but with the same best-response functions and equilibria. Levin, Peck, and Ivanov (2016) used an experiment to investigate the impact of Bayesian updating and non-probabilistic reasoning (referred to as contingent reasoning in this paper) on avoiding the winner's curse.…”

Separating Bayesian Updating from Non-Probabilistic Reasoning: An Experimental Investigation

“…There was a total of 30 periods. 18 At the end of each common-value auction, each subject was provided feedback on the market price, the resale value, the winning bidder's profit or loss, her own profit or loss, her cash balance entering the period, and her current cash balance. She could also see her own signal, the winning bidder's signal, and the signals of all buyers in her market.…”

“…If a subject was randomly selected to sit out during the first three dry runs, we ran a fourth dry run. 18 Due to time constraints, session 12 had to be stopped after period 27.…”