Inconsistent Incomplete Information: A Betting Experiment

We study two person-betting games with inconsistent commonly know beliefs, using an experimentalapproach. In our experimental games, participants bet against one another, each bettorchoosing one of two possible outcomes, and payoff odds are know at the time bets are placed.Bettors’ beliefs are always commonly known. Participants play a series of betting games, in someof which the occurrence probabilities of the two outcomes differ between bettors (inconsistent beliefs)while in others the same occurrence probabilities prevail for both bettors (consistent beliefs).In the betting games with consistent commonly know beliefs, we observe that participants refrainfrom betting. In the betting games with inconsistent commonly know beliefs, we observe significantbetting rates and the larger the discrepancy between the two bettors’ subjective expectations thelarger the volume of bets. Our experimental results contrast with the existing evidence on zero-sumbetting games according to which participants’ irrational inclination to bet is difficult to eliminate.