Broken promises and hidden partnerships: An experiment

Previous research indicates that cheap-talk promises promote cooperation. We extend the empirical study of promises to a three person environment. Doing this enabled us to study two features of promises in the naturally occurring world that have not been studied in the previous literature. For one, we are able to determine the impact of “contingent” promises, which refers to promises that need to be kept (or broken) only in the event that another event first occurs. Second, we were able to study “promise chains”, by which we mean a set of sequentially-made promises. In both cases, we are able to apply existing theories to these novel environments to predict promise keeping behavior. Consistent with theory predictions, we find that promises are not necessary, and do not change behavior, when incentives are aligned. When there is conflict of interest, and when a promise is definite (that is, not contingent), we find that people are indeed willing to forgo monetary payoff to keep their promises. However, when the promise is contingent, promises no longer promote cooperation: people are equally like to choose a selfish action regardless whether they made a promise. Finally, we find that promise chains do not promote trustworthiness, with people at the end of the chain no more likely to honor their promises than people at earlier positions, despite the additional indirect harm that people later in the chain cause by defecting. Our findings offer insights relevant for the behavioral economic theory of guilt aversion.