This paper studies the choice between centralization and decentralization of fiscal policy in a political economy setting. With centralization, regional delegates vote over agendas comprising sets of region-specific projects. The outcome is inefficient because the choice of projects is insufficiently sensitive to within-region benefits. The number of projects funded may be non-monotonic in the strength of project externalities. The efficiency gains from decentralization, and the performance of “constitutional rules” (such as majority voting) which may be used to choose between decentralization and centralization, are then discussed in this framework. Weaker externalities and more heterogeneity between regions need not increase the efficiency gain from decentralization.

A matching model is analyzed in which firms imperfectly test workers prior to hiring them. If (some) firms hire only workers who pass the test, there is an informational externality; unemployment duration is a signal of productivity. In equilibrium, if it is profitable for a firm to test, it is also profitable for it to condition its hiring decision on duration, hiring those whose duration is less a than critical value. Sensitivity analysis of the latter suggests explanations for the dependence of reemployment probabilities on duration and the instability of the U-V curve.

This paper considers a class of two-player dynamic games in which each player controls a one-dimensional action variable, interpreted as a level of cooperation. The dynamics are due to an irreversibility constraint: neither player can ever reduce his cooperation level. Payoffs are decreasing in one's own action, increasing in one's opponent's action. We characterize efficient symmetric equilibrium action paths; actions rise gradually over time and converge, when payoffs are smooth, to a level strictly below the one-shot efficient level, no matter how little discounting takes place. The analysis is extended to incorporate sequential moves and asymmetric equilibria.