Assessing Hedge Fund Performance with an Information-based Multiple Test

Abstract

We develop fwer+, an information-based multiple-testing procedure for selecting outperforming funds while controlling the family-wise error rate. The method conditions rejection thresholds on fund-level and macroeconomic covariates, improving power relative to unconditional procedures in hedge-fund simulations. Applied to U.S. hedge funds from 1997 to 2023, fwer+ selects portfolios that outperform passive benchmarks and generate significant out-of-sample alphas across targets and factor models. The information driving selection is concentrated in macroeconomic and risk-exposure variables, suggesting that skill is tied to navigating changing economic conditions.

Keywords: Data Snooping, Multiple Testing, Multivariate Family-Wise Error Rate, Informative Covariates, Hedge Fund