Jinzhe Wang (王今者)

This paper studies the joint design of portable diagnostic expertise and performance pay. A firm designs a worker's expertise, the ability to conduct a statistical experiment, through training. The worker decides whether to leave with her expertise before diagnosing and solving an unknown problem. Expertise improves diagnoses and incentivizes hidden effort. Expertise profiles can vary in productivity while keeping average diagnostic accuracy. We show that the optimal expertise profile admits a Threshold Diagnostic Structure. It preserves exact diagnosis of common types while pooling rarer types with more common ones in nested categories. In response to higher retention pressure, the firm alternates between preserving expertise with a retention premium and coarsening it with a lower bonus, generating a non-monotone bonus response. Productive training persists even under extreme retention pressure with zero average diagnostic accuracy gain. The model helps account for how firms adjust training composition and make non-monotonic adjustments to bonuses.

This paper studies preference aggregation in sequential majority voting with two candidates. Voters have heterogeneous preferences, are uncertain about the distribution of preferences, and incur a social-pressure cost from voting for the losing candidate. We characterize all symmetric Markov equilibria exhibiting a two-stage structure: a learning stage followed by herding once either candidate’s lead reaches a threshold. We characterize the conditions for equilibria existence and show that stronger social pressure and greater distributional uncertainty hinder preference aggregation. We also identify a class of resistant equilibria, in which voters become more willing to act sincerely as their preferred candidate falls behind, until herding begins. Finally, we study welfare across voting positions. An influence–risk trade-off implies that early voters bear more risk but exert greater influence, while later voters are more protected but less influential. Consequently, although stronger social pressure reduces welfare for all voters, it makes later voters relatively better off than early voters within an equilibrium. Moreover, when comparing coexisting equilibria under the same parameters, improvements in preference aggregation need not be Pareto improving.