Do hedge funds managers misreporting of returns follows monetary policies? Evidence from macro-funds during the great moderation

Abstract

This thesis investigates whether the misreporting behavior of hedge fund managers is influenced by the interest rate environment, contributing to the limited literature on the impact of macroeconomic variables on the misreporting of hedge funds. Using a comprehensive hedge fund dataset covering more than 4,800 funds and 560,000 monthly returns between 1995 and 2021, combined with U.S. Treasury rates at three different maturities (3-month, 2-year, and 10-year) and Nelson-Siegel yield curve coefficients, the analysis identifies potential misreporting via a discontinuity test developed by Bollen and Pool (2012). This test aims to identify a "kink," or a discontinuity in returns around zero, where the abnormal clustering of small positive returns and lack of small negative returns is interpreted as a deliberate inflation of small losses. Misreporting is first assessed over rolling windows of 3, 6, and 12 months, using the pooled distribution of hedge fund returns across all funds. This reveals that higher levels and volatility of interest rates are significantly associated with stronger signs of misreporting. The study is then extended by computing the kink statistic over 12-month rolling windows, separately for different fund subgroups based on their characteristics (e.g., management fees, size, country of registration), showing that the relationship between interest rates and misreporting varies across fund types. Notably, we find that the performance fee structure, the length of the redemption periods, and the liquidity of the strategy employed by the funds significantly affect their response to interest rate conditions. These results have both academic and practical implications. They suggest that misre- porting might not only be driven by internal fund incentives, but also by broader financial conditions. In that sense, further research using other interest rate maturities, other types of model specification (since we used linear regression), or other misreporting flags could further develop these findings.