Representative Sampling Equilibrium


We present an equilibrium concept based on the idea that agents evaluate actions using sample data drawn from the equilibrium distribution, where the number of observations about an alternative is proportional to its usage in a relevant population. Agents naively extrapolate from their data, using the sample mean payoff from each alternative as a predictor of their payoff from choosing it. The endogeneity of sample sizes gives rise to a novel equilibrium effect: Players’ assessment of less frequently played actions is noisier. We study the implications of this effect in a single-agent, binary-choice model, as well as in various examples of games.


If you have any questions regarding the seminar, please contact the seminar organizers Heidi Thysen or Linda Hundsnes.