Does systemic risk of hedge funds predict future asset returns?
1University of Oulu, Oulu Business School, Department of Finance, Finance
|Online Access:||PDF Full Text (PDF, 5.1 MB)|
|Persistent link:|| http://urn.fi/URN:NBN:fi:oulu-201409251887
|Publish Date:|| 2014-09-29
|Thesis type:||Master's thesis
In this study, we derive a new measure of aggregate systemic risk, denoted ASR, from the Hedge Fund Research (HFR) database that complements systemic risk of hedge funds by forecasting future asset returns up to six months ahead. We examine the ability of ASR measure of hedge funds to predict fixed income and equity returns. We apply the methodology following Linda, Bali, and Tang (2012) to estimate ASR and construct the main framework of this study. Subsequently, the micro level systemic risk exposure of individual hedge funds, that is 5% value at risk (VaR) of each fund, is aggregated. The VaR of each fund is estimated using a non-parametric quantile approach. This method is appropriate because usually hedge funds have shorter return series. ASR is then constructed as cross sectional average of VaR estimates of funds over a fixed rolling window of thirty six months. We run multiple time series regression models to predict asset returns.
The regression results reveal mix evidence of predictability because ASR measure of hedge funds is able predict some asset returns while others remain unpredictable. Most of the U.S. constant maturity yields, especially five years and higher, are predictable along with regional ETFs. On the contrary, other assets show no strong evidence of predictable because only few countries’ bonds and or ETFs are predictable. Results indicate predictability improves slightly after adjusting hedge fund data for backfill bias. Based on the broad strategies of hedge funds, results suggest that systemic risk among hedge funds under multi-process and security selection have predictability stronger than other categories. Smaller funds seem to have no impact on predictability implying that predictability is mostly driven by larger funds. Predictability results across different time periods suggest that the ability of ASR to predict asset returns is strongest during 2003–2007 period followed by 1997–2002 period. In contrast to these findings, predictability drops down at longer horizons. The long term predictability i.e. two to six months ahead, tends to drop because the number of assets predictable by ASR decline for all assets except for constant maturity yields. These results indicate that micro level risk exposure of hedge funds when aggregated provides a tool to forecast asset returns. Although, not all assets are predictable by ASR, it has the potential to be used as a forecasting tool in conjunction with other forecasting tools to enhance the forecast accuracy.
© Shahbaz Qureshi, 2014. This publication is copyrighted. You may download, display and print it for your own personal use. Commercial use is prohibited.