Total: 1
I model balanced task allocation where tasks stochastically arrive and must be matched to a fixed set of agents; the novel constraint is that agents must receive allocations that require the same level of average effort. Social work supervisors, call center managers, and courts all rotate allocation across workers to satisfy this constraint, but the Rotation mechanism is not Pareto efficient. I design the Dynamic Pseudomarket (DPM) mechanism, and it satisfies Pareto efficiency and asymptotic balance. I derive an explicit equation characterizing DPM's expected productivity gain over Rotation that can be estimated only from aggregate statistics in firm-level data. Simulation results indicate large average productivity gains. These results indicate that preference-based allocation can Pareto dominate the status quo.