Revealing Manager Skill Through Path-Dependent Risk Management

Active managers are consistently evaluated over finite horizons, yet standard risk models implicitly assume infinite time. This mismatch allows measurement error to dominate the signal, frequently leading to the termination of skilled managers. While standard practice targets a static level of active risk, this paper proposes a path-dependent framework where managers explicitly condition risk levels on the time remaining and accumulated excess returns. Using dynamic programming, the authors demonstrate that this approach significantly tightens the distribution of realized outcomes. The result is a more consistent information ratio that better aligns realized performance with true skill, substantially improving the long-term survival rate of skilled managers.