The standard dual-self model of self-control, with a shorter-run self who cares only about the current period, is excessively sensitive to the timing of decisions and to the interpolation of additional "no-action" time periods in between the dates when decisions are made.We show that when the shorter-run self is not completely myopic, this excess sensitivity goes away.To accommodate the combination of short time periods and convex costs of self-control, we introduce a cognitive resource variable that tracks how the control cost depends on the self-control that has been used in the recent past.We consider models with both linear and convex control costs, illustrating the theory through a series of examples.We examine when opportunities to consume will be avoided or delayed, and we consider the way in which the marginal interest declines with delay.

We study the experimental play of the repeated prisoner’s dilemma when intended actions are implemented with noise. In treatments where cooperation is an equilibrium, subjects cooperate substantially more than in treatments without cooperative equilibria. In all settings there was considerable strategic diversity, indicating that subjects had not fully learned the distribution of play. Furthermore, cooperative strategies yielded higher payoffs than uncooperative strategies in the treatments with cooperative equilibria. In these treatments successful strategies were “lenient” in not retaliating for the first defection, and many were “forgiving” in trying to return to cooperation after inflicting a punishment.

We consider the robustness of extensive form mechanisms to deviations from common knowledge about the state of nature, which we refer to as information perturbations . First, we show that even under arbitrarily small information perturbations the Moore-Repullo mechanism does not yield (even approximately) truthful revelation and that in addition the mechanism has sequential equilibria with undesirable outcomes. More generally, we prove that any extensive form mechanism is fragile in the sense that if a non-Maskin monotonic social objective can be implemented with this mechanism, then there are arbitrarily small information perturbations under which an undesirable sequential equilibrium also exists. Finally, we argue that outside options can help improve efficiency in asymmetric information environments, and that these options can be thought of as reflecting ownership of an asset.