Stand by me—Experiments on help and commitment in coordination games

We present experiments studying how high-ability individuals use help to foster efficient coordination. After an initial phase that traps groups in a low-productivity equilibrium, incentives to coordinate are increased, making it possible to escape this performance trap. The design varies whether high-ability individuals can offer help and, if so, whether they must commit to help for an extended period. If help is chosen on a round-by-round basis, the probability of escaping the performance trap is slightly reduced by allowing for help. The likelihood of success significantly improves if high-ability individuals must commit to help for an extended time. We develop and estimate a structural model of sophisticated learning that provides an explanation for why commitment is necessary. The key insight is that potential leaders who are overly optimistic about their ability to teach their followers are too fast to eliminate help in the absence of commitment.