Endogenous Benchmarks

This paper develops a new approach to control for commonalities in actively managed investment fund returns when measuring their performance. Many investment fund managers systematically load on common priced factors that are omitted from popular models, exhibit similarities in their choices of specific stocks and industries, or vary their factor-loadings in a similar way over time. These commonalities create well-known problems in measuring the performance of groups of funds, since it is difficult to control for their correlated residuals in commonly used models. While, in principle, it is possible to add factors to control for these commonalities, the large number of potential strategies used by fund managers make this approach untenable. In this paper, we propose a simple approach to account for the commonalities in fund strategies that only uses information on fund returns and the investment objective of the fund. Our approach is to form an additional factor from the return on the group of funds to which a given fund belongs. We call this additional factor an quot;endogenous benchmark,quot; since each fund manager chooses the group within which it intends to compete. This choice of a group by the manager indicates the set of strategies from which the manager will choose in order to compete. Using only the returns and investment objectives of funds, our endogenous benchmarks substantially reduce the cross-sectional dependencies of residuals, across groups of funds and across individual funds within a group. Specifically, more than half of the cross-sectional correlations between individual funds is eliminated with our method. We also show that this improvement in estimation of alphas results in a better identification of U.S. equity and fixed-income funds with persistent performance. For instance, relative to a standard four-factor model, our model generates outperformance of more than 3% per year among U.S. growth equity funds