Evolutionary Learning in the Ultimatum Game

The ultimatum game is (in)famous for its `anomalies': The outcomes of laboratory experiments are very different from the results generated by traditional game theory. This paper aims to find to what extent these discrepancies between theory and experiments can be explained by the effects of bounded rationality and learning dynamics. These are modeled by several agent based models and computer simulations of evolutionary learning by pure imitation as well as imitation and experiments. The main result of the analysis is surprisingly clear and robust: Proposers do not play a subgame perfect strategy but instead `learn' to make offers of about 20 to 25 % of the total amount to their opponents.