Andy B. Choi

A principal chooses a policy at a future date, and wishes to match the policy to an uncertain state. An agent chooses when to make an irreversible investment, and wishes to invest only if he expects the policy will be favorable. Information about the state is publicly and gradually revealed over time. Moral hazard arises because the agent wishes to wait for more information. To incentivize the agent to invest early, it is optimal for the principal to provide certainty about her future policy. This is inefficient -- the agent's benefit from certainty is always outweighed by the principal's cost from reduced policy flexibility. We characterize conditions under which the agent receives positive rent, and provide sufficient conditions for moral hazard to delay investment. Our results apply to environmental subsidy programs and the incentivization of R&D.

Motivated by applications such as cloud computing, gig platforms, and blockchain auctions, we study optimal selling mechanisms for dynamic markets with stochastic supply and demand. In our model, buyers with private valuations and homogeneous goods arrive stochastically and can be held in queues at a cost. The optimal mechanism pairs allocative efficiency with dynamic admission control: goods are assigned to the highest-value buyer, while entry is restricted by value thresholds that strictly increase with the queue length and decrease with available inventory. This policy smooths competitive pressure across time and is implemented in dominant strategies via auctions with dynamic reserve prices.

How should an employer design its retention and promotion policy when personnel decisions reveal private information about its workers to the labour market? We show it is always optimal to over-retain workers---inefficiently retaining workers who would be more productive elsewhere---to dampen the signalling effect of retention. In contrast, the optimal policy may over- or under-promote workers depending on the production technologies of competing firms. Our results shed light on the role of job titles and the Peter Principle. To solve for the optimal policy, we develop an approach to characterise the sender's highest ex ante payoff when she can commit to signalling actions. Our approach applies to general signalling games.

A principal wishes to promote an agent only if the state is good, and gradually receives private information about the state. The agent wants promotion but would rather leave than stay and fail promotion. To induce the agent to stay, the principal commits to commit; that is, she commits today to tell the agent tomorrow about his chances of promotion the day after. Because the principal cannot contract on her information, she may ignore it. We provide conditions under which the principal does not lead the agent on. Our results apply to worker retention, relationship-specific investment, and forward guidance.

This paper introduces search frictions into the two-sided matching model with transfers (the assignment game). A worker-firm pair can match only if they are acquainted. Workers may search at a cost to become acquainted with new firms. A matching is stable if no acquainted pair wishes to block, and no individual worker wishes to search. A matching is efficient if it maximizes total surplus taking acquaintance as given, and no search -- individual or coordinated -- can increase total surplus net of search costs in expectation. We show that all stable matchings are efficient. However, not all efficient assignments can be made stable.