How does pay-for-performance (P4P) impact productivity and the composition of workers in mission-oriented jobs when output has multiple dimensions? This is a central issue in the public sector, particularly in areas such as education and health care. We conduct an experiment, manipulating compensation and prosocial elements of the job, to answer these questions. We find that P4P has significantly smaller positive effects on productivity on the incentivized dimension in the prosocial setting relative to the non-prosocial setting. On the other hand, P4P generates no loss in performance on the non-incentivized dimension of effort in the prosocial setting, whereas it does so in the non-prosocial setting. In both settings, P4P attracts higher ability workers, but it does so at the expense of attracting prosocially-motivated workers in the prosocial setting.