The evolution of conventions under condition-dependent mistakes

We study the long-run conventions emerging in a stag-hunt game when agents aremyopic best responders. Our main novel assumption is that errors converge to zero ata rate that is positively related to the payoff earned in the past. To fully explore theimplications of this error model, we introduce a further novelty in the way we model theinteraction structure, assuming that with positive probability agents remain matchedtogether in the next period. We find that, if interactions are sufficiently persistentover time, then the payoff-dominant convention emerges in the long run, while ifinteractions are quite volatile, then the maximin convention can emerge even if it isnot risk-dominant. We contrast these results with those obtained under two alternativeerror models: uniform mistakes and payoff-dependent mistakes.