In this paper, we consider an environment in which agents’ skills are private information, are potentially multi-dimensional, and follow arbitrary stochastic processes. We allow for arbitrary incentive-compatible and physically feasible tax schemes. We prove that it is typically Pareto optimal...