We analyze a situation where common noise makes compensation based on relative performance evaluation (RPE) desirable, but where the agents' ability to hold-up values ex post obstruct the implementation of optimal RPE ...