We develop an approach to jointly study four dimensions of hedge fund value creation—its drivers, split, dynamics, and optimality. This approach captures the large fund heterogeneity and controls for hedge fund complexities. We find that most funds add value via their unique skills but face strong scalability constraints—a feature that prevents them from systematically dominating mutual funds. Hedge fund investors slowly improve their fund capital allocation over time, which suggests an impactful but noisy learning process. Despite these efforts, they extract a modest fraction of the total value. These findings fit reasonably well with an equilibrium model featuring funds with heterogeneous skill and scalability and investors with limited bargaining power.

I study analyst reports for worldwide equity mutual funds to examine two key features of active management models: investor learning and decreasing returns to scale. Flows respond to report tone over and above analyst ratings, especially for direct-sold and young funds, indicating that some investors use qualitative research when allocating capital. Report tone also predicts future abnormal returns. Using dictionary and machine learning methods, I show that analysts emphasize fund size and capacity when assets deviate from model-implied optimal levels, providing signals about under- or overcapitalization and thus positive-NPV opportunities. A composite of tone and capacity language strongly predicts returns.