We depart from the usual assumption in noncooperative game theory that games arise exogenously. We assume instead that games between agents are formed endogenously in the sense that agents choose their opponents through a costly search process. Since agents are aware in this situation that both parties have the option of rejecting a match, they have an incentive to make themselves as attractive a partner as they can. This is accomplished in a pregame in which agents consider all the potential strategies in a game but choose to learn only a subset. We assume that this choice is observable by their potential partners in the matching game. We motivate this as a proxy for subscribing to a code of ethics, accepting a set of social norms, or being a member of a religious, philosophical, or political group. We show that agents will sometimes choose to constrain their actions sets in the pregame in order to achieve better matches and higher payoffs. We suggest that this might provide at least a partial explanation for experimental observations that agents apparently choose strategies that do not maximize their payoffs.

We depart from the usual assumption in noncooperative game theory that games arise exogenously. We assume instead that games between agents are formed endogenously in the sense that agents choose their opponents through a costly search process. Since agents are aware in this situation that both parties have the option of rejecting a match, they have an incentive to make themselves as attractive a partner as they can. This is accomplished in a pregame in which agents consider all the potential strategies in a game but choose to learn only a subset. We assume that this choice is observable by their potential partners in the matching game. We motivate this as a proxy for subscribing to a code of ethics, accepting a set of social norms, or being a member of a religious, philosophical, or political group. We show that agents will sometimes choose to constrain their actions sets in the pregame in order to achieve better matches and higher payoffs. We suggest that this might provide at least a partial explanation for experimental observations that agents apparently choose strategies that do not maximize their payoffs.