An Experimental Investigation of Strategies in the Dynamic Common Pool Game

This paper studies experimentally the selection of strategies in a canonical dynamic environment, the dynamic common pool game. The main goal is to evaluate if behavior can be rationalized with strategies that condition on past play that can support efficient outcomes, or with stationary Markov strategies that only condition on the “state of the system.” The main finding is that modal behavior is Markovian even when the gains from cooperation are large. However, the popularity of Markov strategies is decreasing in the incentives to cooperate. The evidence also suggests that early coordination failures when trying to implement the efficient outcome can contribute to coordination on Markov play