Joerg Oechssler

Learning in Cournot Oligopoly: An Experiment

This experiment was designed to test various learning theories in the context of a Cournot oligopoly. We derive theoretical predictions for the learning theories and test these predictions by varying the information given to subjects. The results show that some subjects imitate successful behavior if they have the necessary information; and if they imitate, markets are more competitive. Other subjects follow a best reply process. On the aggregate level we find that more information about demand and cost conditions yields less competitive behavior, while more information about the quantities and profits of other firms yields more competitive behavior.

Stability of the Cournot Process: Experimental Evidence

We report results of experiments designed to test the predictions of the best-reply process. In a Cournot oligopoly with four firms, the best-reply process should theoretically explode if demand and cost functions are linear. We find, however, no experimental evidence of such instability. Moreover, we find no differences between a market which theoretically should not converge to Nash equilibrium and one which should converge because of inertia. We investigate the power of several learning dynamics to explain this unpredicted stability.

Herding With and Without Payoff Externalities - An Internet Experiment

Most real world situations that are susceptible to herding are also characterized by direct payoff externalities. Yet, the bulk of the theoretical and experimental literature on herding has focused on pure informational externalities. In this paper we experimentally investigate the effects of several different forms of payoff externalities (e.g., network effects, first-mover advantage, etc.) in a standard information-based herding model. Our results are based on an internet experiment with more than 6000 subjects, including a subsample of 267 consultants from an international consulting firm. We also replicate and review earlier cascade experiments. Finally, we study reputation effects (i.e., the influence of success models) in the context of herding.

Rage against the machines: how subjects learn to play against computers

We use an experiment to explore how subjects learn to play against computers which are programmed to follow one of a number of standard learning algorithms. The learning theories are (unbeknown to subjects) a best response process, fictitious play, imitation, reinforcement learning, and a trial & error process. We test whether subjects try to influence those algorithms to their advantage in a forward-looking way (strategic teaching). We find that strategic teaching occurs frequently and that all learning algorithms are subject to exploitation with the notable exception of imitation. The experiment was conducted, both, on the internet and in the usual laboratory setting. We find some systematic differences, which however can be traced to the different incentives structures rather than the experimental environment.

When is Tit-for-Tat Unbeatable?

We characterize the class of symmetric two-player games in which tit-for-tat cannot be beaten even by very sophisticated opponents in a repeated game. It turns out to be the class of exact potential games. More generally, there is a class of simple imitation rules that includes tit-for-tat but also imitate-the-best and imitate-if-better. Every decision rule in this class is essentially unbeatable in exact potential games. Our results apply to many interesting games including all symmetric 2

Does information about competitors’ actions increase or decrease competition in experimental oligopoly markets?

\times

An Evolutionary Interpretation Of Mixed-Strategy Equilibria

2 games, and standard examples of Cournot duopoly, price competition, public goods games, common pool resource games, and minimum effort coordination games.