Statistical analysis of the performance of more than 3,000 hedge funds suggests that returns are predictable. Under changing market conditions, the predictability pattern observed is consistent with economic rationale and largely attributable to differences in key fund characteristics. A simple strategy of combining return forecasts based on individual predictors is recommended for achieving superior performance.
The authors develop a unified approach to comprehensive analysis of hedge fund return predictability. They determine the relationship between each selected macro variable and individual fund returns. To examine the exploitability of this strategy in hedge fund selection, they further examine its applicability out of sample. Finally, to gain a deeper understanding of the drivers of performance, they examine whether the predictability patterns documented in sample explain when and why the recommended strategies outperform out of sample.
How Is This Research Useful to Practitioners?
For the full sample, the authors find that the returns of 63% of hedge funds are predictable, with wide variation in predictability across investment styles. In addition, the time variation in performance is significant, ranging from 6.6% to 7.2% a year for 25% of sample funds. The predictability is primarily attributable to variation in alpha (i.e., managerial skill).
From their in-sample analysis covering the period of January 1994–December 2008, the authors identify the following relationships between hedge fund performance and macro variables, which are generally consistent with economic rationale:
- Net inflows into hedge funds and future performance are inversely related, which is in line with the presence of capacity constraints and is more pronounced for strategies that operate in crowded spaces (e.g., convertible bonds).
- Expansion in default spreads leads to superior future performance of emerging market funds and also—albeit to a lesser extent—of global macro funds using foreign exchange carry trades. This relationship is consistent with a flight to quality in bad markets.
- A high dividend yield predicts lower future returns, suggesting reduced access to leverage during downturns.
- Higher market uncertainty, captured by the Chicago Board Options Exchange Volatility Index (VIX), generally signals lower future performance, potentially reflecting the greater failure rate of deals in uncertain times.
- Managed future funds, however, perform better in uncertain times because of their lower dependence on prime broker funding (dividend yield) and their ability to exploit trend reversals through commodity trading advisers when volatility (i.e., the VIX) is high.
- The financial crisis of 2008 caused structural breaks that mostly affect the default spread and the dividend yield but do not affect the VIX, allowing VIX-based strategies to experience better performance.
By applying these results to an out-of-sample dataset, the authors find that using a simple combination strategy that averages the individual predictor-based return forecasts leads to superior performance. An unconditional portfolio that uses past returns to select funds is found to be the second-best strategy. Single-predictor strategies lack consistency, and the combination strategy performs better because it is less affected by the asymmetry in the forecasting ability of individual predictors, provides a hedge against structural breaks in data gathering, and benefits from the diversification of forecast errors.
How Did the Authors Conduct This Research?
The authors look at monthly returns of more than 8,000 funds, selected based on data availability, from five different data providers: BarclayHedge, Trading Advisor Selection System (TASS), Hedge Fund Research (HFR), Center for International Securities and Derivatives Markets (CISDM), and Morgan Stanley Capital International (MSCI).
Changes in economic conditions are tracked through four variables: (1) the default spread between Moody’s BAA and AAA rated bonds, (2) dividend yield for the broad stock market as a surrogate for leverage, (3) the implied volatility of the S&P 500 Index (the VIX), and (4) the monthly net aggregate flow into the selected funds. The period of study is from January 1994 to December 2008.
Predictability is analyzed based on a time-series predictive regression, which is run separately for the return and alpha of each fund against the four macro variables. The authors strive to account for spurious predictability and small sample biases often present in hedge fund data. For out-of-sample predictability, they factor in such real-world constraints faced by institutional investors as lockup periods, number of funds that can be held, and so forth.
At the end of 2013, hedge funds managed assets of more than US$2.6 trillion. After years of impressive returns, the performance of hedge funds has been modest in recent times. Predictability of hedge fund performance is a key area of current research interest, although it is fraught with such challenges as data availability and a wide variety of strategies.