Referee Finance

Home of the Financial Intelligence Lab

  • Increase font size
  • Default font size
  • Decrease font size
Hedge funds
In this section, the reports deal with articles on hedge funds: their risk and return performance, their specificities,  whether they can be replicated etc...

Risk in Fixed-Income Hedge Fund Styles

E-mail Print PDF

In this article the authors identify several risk factors that underlie hedge fund styles. This work should help to better understand and manage the risk exposures of fixed-income hedge funds. 

They use data from HFR covering the period 1998-2000. HFR classifies fixed income hedge-funds in to five different styles:

-          Fixed-Income Convertible Bond

-          Fixed-Income High-Yield Bond

-          Fixed-Income Mortgage-Backed

-          Fixed-Income Arbitrage

-          Fixed-Income Diversified 

From the HFR classification, the authors extract return-based style factors using principal component analysis. For convertible bond, high-yield bond and mortgage-backed, the first component explains more than 50% of the variation in returns. The other styles (arbitrage and diversified) need a second component.

The authors then link the return-based style factors to asset-based style factors (ABS factors), which are related to observed asset prices.They define the ABS factors by their products (or location) and their strategies (long-only, passive spread trading, trend-following and convergence trading).

To proxy for passive spread trading, the authors use as ABS the spread against treasuries of the particular product. To proxy for trend-following and convergence (non trend-following) strategies, they use the return of look-back straddle strategies on the respective spreads. A look-back straddle is a combination of a look-back call and a look-back put option. A look-back call (put) option gives the buyer the right to purchase (sell) an underlying security at the minimum (maximum) price over a look-back period.

The authors find that the hedge fund convertible bond (high-yield bond) index is mostly exposed to the passive long-only convertible (high-yield) spread return.The hedge fund mortgage bond index is correlated with the change in mortgage rate, the 10-year swap rate and the 10-year Treasury rate. The out-of-sample performance (predicting the returns with the factors only) is however poor for mortgages.

High-yield spread returns and convertible spread returns are the most significant factors for the fixed income arbitrage index.The Lehman corporate bond index returns and the JP Morgan emerging market index spread returns are the most significant factors for the fixed income diversified index. The long horizon straddle is also significant.

The authors find a large exposure of hedge fund fixed-income returns to credit spreads (negative correlation).

They run the same analysis for mutual funds and confirm that the mutual funds have mostly passive exposure to the standard benchmarks.  

Hsieh, David A. and William Fung, September 2002, Risk in Fixed-Income Hedge Fund Styles, The Journal of Fixed Income 


Can Hedge-Fund Returns Be Replicated?: The Linear Case

E-mail Print PDF

In this paper, the authors Jasmina Hasanhodzic and Prof. Andrew Lo from MIT examine the possibility of reproducing hedge fund returns with liquid exchange-traded instruments (the “clones”).

They use hedge fund data from the TASS database (, which covers 1610 hedge funds over the period 1986 to 2005. The database classifies the funds into 11 different investment styles:

  • Long/short equity hedge
  • Funds of funds
  •  Event driven
  • Managed futures
  • Emerging markets
  • Fixed income arbitrage
  • Global macro
  • Convertible arbitrage
  • Dedicated short bias
  • Equity market neutral
  • Multistrategy

The authors try to replicate each series of fund returns with six factors:

  • USD dollar index return
  • The Lehman corporate AA index return
  • The spread between BBB credit and the treasury index
  • The S&P500 equity return
  • The Goldman Sachs Commodity index return
  •  The Monthly change in stock market volatility VIX

They use linear regressions of hedge fund returns by style on the factors to calculate the weights on each factor and build the “clones” as a portfolio of factors. All the factors can be traded except for the credit spread variable since it is not a return (it is always positive for instance). This can be problematic as the return of the “clone” will be biased upward and not achievable.

The authors run the regressions in sample (“fixed weights”) and out-of-sample (“rolling-windows”, they choose 24 months).  Only the latter should be relevant since they can be performed in real time. Overall, after considering the averages of returns of clones and funds, they find that they can produce similar returns.  They also consider the performance of equally weighted portfolio of clones and portfolio of funds and find that the rolling-window clones underperform.  

They repeat their analysis by style: the clones perform better for some strategies (dedicated short-bias 6.83% for the clones vs. 2.53% for the funds, global macro 12.97% vs. 9.01% and managed futures 19.24% vs. 11.84%). The clones underperform in emerging markets, event driven and fixed income arbitrage. The authors argue that illiquidity might play a role since the clones are liquid but not necessarily the funds and investors might earn an extra liquidity premium. They indeed document a gap between the clones and the funds using the autocorrelation of returns as a proxy for illiquidity.  

The risk exposures are similar between the portfolio of clones and the portfolio of funds therefore the clones should provide the same diversification benefits.  

Hasanhodzic, Jasmina and Lo, Andrew W., "Can Hedge-Fund Returns Be Replicated?: The Linear Case" (August 16, 2006). Available at SSRN:




Do Market Timing Hedge Funds Time the Market?

E-mail Print PDF

In this paper, the authors Bing Liang and Yong Chen test for the market timing capacities of market timing hedge funds.

This literature originates from the study of mutual funds market timing skills (Treynor and Mazuy 1966) and has documented limited evidence of such skills. 

Their hedge funds data come from the Center for International and Derivatives Markets, Hedge Fund Research and Lipper TASS. Their sample covers 221 hedge funds over the period from January 1994 to June 2005.  

They first correct for survivorship bias by adjusting the hedge fund returns by the ratio of the unconditional probability of fund survival to an estimated conditional probability of fund survival. 

To detect market-timing skills, they perform their analysis on the aggregate return of hedge funds and on individual hedge funds. They regress the hedge fund returns on a series of systematic factors expected to be priced (Fama and French factors, the Carhart Momentum factor) and on the square of the market Sharpe ratio. This allows them to combine return timing and volatility timing skills analysis. A positive coefficient on the Sharpe ratio will be evidence of such skills. 

Indeed, with such configuration, when the market outperforms (underperforms) or is less (more) volatile, the fund returns will be higher (lower). The authors also estimate their model with returns timing and volatility timing separately. 

In the rest of the paper, the authors go through many robustness checks such as controlling for the illiquidity of hedge fund investments and the presence of option strategies. They also perform a bootstrap analysis to show that their results are not because of chance alone. 

Overall, we found their analysis detailed and convincing. However, the usual caution is that they have to rely on self-reporting hedge funds returns data.    

Carhart, Mark M., 1997, On Persistence in Mutual fund Performance, Journal of Finance 52,57-82. 

Liang, Bing and Chen, Yong, "Do Market Timing Hedge Funds Time the Market?" (May 2006). EFA 2005 Moscow Meetings Available at SSRN: 

Treynor, Jack, and Kay Mazuy, 1966, Can mutual funds outguess the market? HarvardBusiness Review 44, 131-136.


How Smart are the Smart Guys? A Unique View from Hedge Fund Stock (..)

E-mail Print PDF

How Smart are the Smart Guys? A Unique View from Hedge Fund Stock Holdings 

In this paper, the authors use stock holdings information from 13F filing forms to assess the performance of hedge funds. 

Hedge funds with assets of more than $100 million of 13F securities need to fill out the 13F on a quarterly basis for all long US equity positions of more than $200,000 or of more than 10,000 shares.  Some exemptions apply but are limited. Short positions are not reported in the 13F. 

They collect hedge fund data from Nelson’s Directory, Altvest, TASS and the MAR graveyard and manage to cover around 306 firms representing over 1,000 funds (a firm can have several hedge funds but the 13F reporting is done at the firm level). 

To address the problem of reporting long-only positions, the authors regress the returns of the hedge funds (from hedge fund indices) on the returns on their holdings and find a positive correlation except for a few funds (8.5%). Most funds therefore seem to be long-biased. 

They collect similar holding data from mutual funds, which report with the N-30D file (data are from Thomson Financial CDA/Spectrum S12) and compare the hedge funds with the mutual funds. 

They find that compared to the mutual funds: 

-          Turnover is higher for hedge funds than for mutual funds (median quarterly 102% vs. 63%).

-          Hedge funds deviate more from the value-weighted index holdings

-          Hedge funds hold smaller stocks, riskier stocks, value stocks, low momentum stocks, fewer past winners

-          Hedge funds hold less liquid, cheaper and less covered (by analysts) stocks 

There is no predictability from hedge fund stockholdings to future stock performance except for the number of hedge funds holding the stock (defined as the breadth; see Chen, Hong and Stein 2002). 

The authors break up the hedge fund performance into stock picking, characteristic timing and average style, following the approach of Daniel, Grinblatt, Titman and Wermers (1997). Hedge funds have slightly better stock picking abilities than mutual funds. Both fare poorly at characteristic timing and average style. The relative outperfomance of hedge funds is concentrated in 1999 and 2000 during the Internet boom period.   

Daniel, K., M. Grinblatt, S. Titman, and R. Wermers, 1997, “Measuring Mutual Fund Performance withCharacteristic-based Benchmarks,” Journal of Finance 52, 1035-1058. 

Griffin, John M. and Xu, Jin, "How Smart are the Smart Guys? A Unique View from Hedge Fund Stock Holdings" (March 2007). Available at SSRN:   

Chen, J., H. Hong, and J. Stein, 2002, “Breadth of Ownership and Stock Returns,” Journal of FinancialEconomics 66, 171-205.


Hedge Funds: Performance, Risk and Capital Formation

E-mail Print PDF

In this paper Fung, Hsieh and Naik explore the performance of funds of hedge funds.  They also examine the risk and the capital flows in and out of these funds.

They use a database built from CSFB/Tremont TASS, HFR (Hedge Fund Research, Inc) and CISDM (Center for International Securities and Derivatives Markets) that contains 1603 funds from January 1995 to December 2004.

To identify alphas, they use a seven-factor model introduced by Fung and Hsieh (2004). The factors are:

-          Excess return of the S&P500

-          Wilshire small minus big cap index

-          A portfolio of look back straddle options on currencies

-          A portfolio of look back straddle options on commodities

-          A portfolio of look back straddle options on bonds

-          10-year Treasury yield duration-adjusted spread over the 3-month t-bill

-          BAA bond duration-adjusted spread over the 10-year Treasury bond

The look back straddle options capture the performance of trend-following strategies.

The residual of this model is the alpha of the fund since these factors represent systematic risk exposures (beta exposures).

In time series analysis, they find the alpha is significant only during the October 1998-March 2000 period.

In cross-section analysis, they identify have alpha and have beta funds depending on the statistical significance of the intercepts in the rolling regression of fund returns on the seven factors (have alphas=significant positive intercepts, have betas=the rest).

They find that on average 22.5% of funds are of have alpha type, that have alphas have a greater likelihood of remaining have alphas in the next two years vs. have betas to become have alphas. There is therefore some persistence in alpha performance. The have alphas have also a greater chance of surviving than the have betas.

With capital flows, the have alphas experienced of average positive inflows vs. zero inflows for the have betas. In the most recent period, both types of funds receive positive inflows. Finally, they document that funds that receive above median inflows of capital have a lower likelihood of becoming a have alpha.

Their study confirms the model of Berk and Green (2004) on active portfolio management.

We found this paper insightful on the dynamic of alphas, the transition between have alphas and have betas, and capital flows. The study is however focused on funds of funds for which the alpha is an average of alpha produced by the fund of funds manager and alphas produced by each individual hedge fund manager. We do not know at which level the alpha is created and how it disappears. The alpha seems also to follow some systematic pattern (present in bull period). Could it be that it is a proxy for a missing risk factor?  

Berk, Jonathan and Richard C. Green, “Mutual Fund Flows and Performance in Rational Markets”, Journal of Political Economy, Vol. 112, pp. 1269-1295, December 2004

Fung, William, Hsieh, David A., Naik, Narayan Y. and Ramadorai, Tarun, "Hedge Funds: Performance, Risk and Capital Formation" (July 19, 2006). AFA 2007 Chicago Meetings Paper Available at SSRN:

Fung, William, Hsieh, David A., “Hedge Fund Benchmarks: A Risk-Based Approach”, Financial Analysts Journal, Vol. 60, No. 5, pp. 65-80, September/October 2004  


Page 2 of 2


You can follow us on Twitter  @fintlab and facebook