Research

This paper studies how a principal can influence an agent’s costly private learning about the state of the world through two instruments: (i) a public signal, and (ii) signal-realization-contingent monetary transfers. The agent gains from trade in the good state but loses in the bad state, while the principal benefits from trade in all states. Both parties share a common prior. Although the principal can, at no cost, disclose full information by providing a perfect signal, this never happens in equilibrium. Instead, the agent acquires costly information under some conditions. Specifically, when the prior is intermediate, the principal optimally designs a partially informative public signal that induces the agent’s private learning. Transfers are used only when the agent is sufficiently pessimistic to retain those at the participation threshold. Combining a public signal with monetary incentives yields strictly better outcomes for the principal than either instrument alone. Finally, when transfers may be negative—or in procurement environments where the agent must always be compensated—the principal optimally deters the agent’s private learning.

We study a multilateral bargaining game in which players can attempt short-lived commitments before bargaining. Existing work shows that simultaneous commitments under unanimity lead to inefficient delayed agreements due to players’ strategic over-commitments. We show that this inefficiency is not inherent in unanimity itself, but arises from a coordination failure caused by the timing of simultaneous commitments. We introduce a sequential commitment protocol: Players are randomly ordered to make their commitment attempts after observing prior stochastic commitments, modeling real-time credibility updating during negotiations. Sequential commitment yields efficient and immediate agreements in the unique stationary Markov-perfect equilibrium: The first successful committer effectively becomes the proposer, while subsequent players do not attempt to demand more than their continuation payoffs. The result is robust to heterogeneity in players’ time preferences, commitment powers, and their likelihood to be the proposer. Our findings highlight how inefficiency under unanimity arises from timing frictions—not the rule itself—and demonstrate that structured commitment can promote efficient cooperation.

This paper asks when organizations should use ex-ante screening rather than ex-post enforcement to mitigate hold-up when productivity-enhancing investment is unverifiable and sunk before contracting. The model features persistent type heterogeneity: nature-assigned high types and investment-upgraded low types coexist and are observationally identical at the contracting stage. Investment incentives must therefore be provided through high-type rents, which also accrue to nature-assigned high types and distort the continuation screening problem with hidden effort.

Two results follow. First, the optimal governance margin follows a three-regime pattern. In the region where enforcement remains viable after screening is no longer the preferred investment-inducing instrument, screening dominates at low investment costs because posterior redistribution sustains investment with less effort distortion. Enforcement dominates at intermediate costs because adjusting continuation rents after performance does not require the posterior redistribution that screening demands. At high costs, neither instrument induces investment.

Second, optimal enforcement need not be punitive. When investment incentives are difficult to sustain and low-productivity agents are sufficiently common, report-contingent adjustments take the form of continuation benefits—favorable renewal terms, reduced future obligations, or priority access—rather than penalties. Monitoring systems designed to sustain unverifiable investment may therefore optimally reward favorable reports rather than use audits solely to punish deviations.

This paper studies persuasive behavior when the sender can control both the information the receiver observes and the model through which it is interpreted (the narrative). We show that even when the receiver begins with a correctly specified model and understands the sender's strategic incentives, the sender can manipulate him and often secure her preferred action with probability one. The key mechanism highlights a strong complementarity between strategic communication of information and narratives, allowing the sender to strictly outperform a Bayesian persuasion even without commitment power herself. We fully characterize the sender-optimal equilibrium for a broad class of information technologies. Softer information lowers the bar for full manipulation, while harder information expands the set of environments where any manipulation is possible. The results provide a formal foundation for understanding the widespread success of disinformation.

We study the relationship between competitive division in economies of bads and the Nash rationing solution for utility-space rationing problems. In a negative division problem, feasible allocations generate a set of utility profiles below the no-allocation claim point. Bogomolnaia, Moulin, Sandomirskiy, and Yanovskaya (2017) characterize competitive utility profiles as critical points of the Nash product of utility losses. We first give an equivalent characterization of the Nash rationing solution in terms of supporting hyperplanes and Nash-product maximization, and use it to show that the set of competitive utility profiles in a negative division problem is identical to the set of Nash rationing solution points of the corresponding rationing problem. This gives a decentralized price-equilibrium foundation for the Nash rationing solution and, conversely, transfers the consistency and equity properties of Nash rationing to competitive division of bads. We also strengthen the minimal characterization of Mariotti and Villar (2005) into an exact axiomatic characterization by adding expansion consistency to positive affine invariance, weak symmetry, and lower contraction consistency. Finally, we show that each solution point is the centroid of its associated supporting simplex.