Hedge Funds: Performance, Risk and Capital Formation.

1
Hedge Funds Performance, Risk and Capital
Formation.

• Bill Fung (London Business School)
• David Hsieh (Duke University)
• Narayan Naik (London Business School)
• Tarun Ramadorai (Univ. of Oxford and CEPR)

2
Focus of this paper

• What is the relationship between risk-adjusted
  performance (alpha), alpha persistence, and
  capital formation in the hedge fund industry?
• Testing whether Berk and Greens (JPE 2004)
  rational model of active portfolio management is
  appropriate for the hedge fund industry.

3
The Berk and Green Model

• Berk and Green (JPE 2004) Rational model of
  active portfolio management has three features
• Investors competitively provide capital to funds.
  
• Managers have differential ability to generate
  high returns, but face decreasing returns to
  scale in deploying their ability.
• Investors learn about managerial ability from
  past performance and direct more capital towards
  funds with superior risk-adjusted performance.
• Results in zero alpha and zero alpha persistence
  in equilibrium.

4
Preview of our main conclusions

• Implication 1 of Berk and Green Funds with
  higher ability managers should experience greater
  capital inflows .
• We show that funds that generate statistically
  positive alpha experience far greater and
  steadier capital inflows than their less
  fortunate counterparts.

5
Preview of our main conclusions

• Implication 2 of Berk and Green Diminishing
  returns to scale combined with the inflow of new
  capital leads to erosion of superior performance
  over time, even for high ability funds.
• We show that capital inflows lead to lower alpha,
  and lower alpha persistence.
• We provide evidence that even alpha-producing
  funds have experienced a recent, dramatic decline
  in risk-adjusted performance.

6
A Few Facts About Hedge Fund Returns

• Individual hedge fund data is subject to biases
• Fung and Hsieh (2000) and Liang (2000)
• Hedge funds exhibit different styles/systematic
  risk.
• Fung and Hsieh (1997, 2001, 2002, 2004),
  Brown and Goetzmann (2003), Mitchell and Pulvino
  (2001), Agarwal and Naik (2004) etc
• Hedge fund performance measurement is hard
  because of illiquidity in their holdings.
• Getmansky, Lo, Makarov (2004), Asness, Krail
  and Lew (2001)
• The best (and worst) hedge funds exhibit
  statistically detectable performance persistence.
• Kosowski, Naik and Teo (2005), Jagannathan,
  Malakhov and Novikov (2006)
• And much more interesting research that I have
  little time to mention today!

7
The Data

• Consolidate data from CSFB/Tremont TASS, HFR and
  CISDM. The most comprehensive set of Funds of
  Hedge Funds that is publicly available - 1603 in
  total.
• Ten-year period (Jan 1995 to Dec 2004).
  Breakpoints in the data confirmed by the Chow
  test.
• Period I January 1995 to September 1998.
• Period II October 1998 to March 2000.
• Period III April 2000 to December 2004.
• We use the Fung-Hsieh seven-factor model to
  estimate alpha and find

8
For the Average FoF No Detectable Alpha Except
in Period II (October 1998-March 2000) see
Table II
9
Differentiating Funds on Alpha Production

• For each two-year period, do the Kosowski,
  Timmerman, White and Wermers (2006) bootstrap.
• Basic intuition trying to avoid Type I errors in
  detecting funds with alpha.
• OLS t-stat is misleading if errors are
  non-normally distributed and non i.i.d.
• This is very likely a problem for FoF returns.
• Robustness check the results in a variety of
  ways.
• Getmansky-Lo-Makarov (2004) correction.
• Politis-Romano (1994) stationary bootstrap.

10
Have Alphas and Have Betas see Table III
11
Have Alphas and Have Betas Transitions see
Table III
12
Have Alpha, Have Beta Liquidations Table IV
13
Do Investors Recognize Differences Between Have
Alphas and the Rest? Yes! See
Figure 1
14
Differences in Flows see Table V
15
Do The Flows Behave Differently?

• Are our results driven by total return-chasing
  rather than alpha-perception on the part of
  investors?
• Perhaps there are clientele differences between
  have alphas and the rest?
• Regress quarterly flows on past quarters returns
  and past quarters flows.

16
Do The Flows Behave Differently? see Table VI

• Perhaps there are clientele differences between
  have alphas and have betas?

17
Do the Flows Generate Declines in Alpha?

• First, condition two-year transition
  probabilities, t-stat of alpha and magnitude of
  alpha on level of flow (above median or below
  median flow).
• Second, identify time-variation in alpha for have
  alphas and have betas, to check if there are
  corresponding trends.

18
Do Flows Generate Declines in Alpha? Results
for Have Alphas see Table IX

• Same is true for the t-stat of alpha and
  magnitude of alpha, see Table XI

19
Do Flows Generate Declines in Alpha? Results
for Have Betas see Table X

• No detectable difference on t-stat and
  magnitude of alpha (see Table XII), but recall
  these are beta-only funds anyway.

20
Has the Alpha Generating Ability of Alpha
Producers Changed Over Time? See Table XIII
21
Conclusions

• Consistent with Berk and Green (JPE 2004), there
  are significant differences in the ability of
  managers to deliver alpha.
• Investors appear to perceive these ability
  differentials, and direct a steady stream of
  capital to the managers with higher ability.
• The persistent inflow of capital is associated
  with a decline in the alpha produced in the hedge
  fund industry.
• The decline is experienced by all managers
  regardless of ability.

22
Question?

• Recall that Berk and Greens model predicts zero
  alpha and zero alpha persistence in equilibrium
• Is the hedge fund industry heading that way?