Pages from f metrics hf_1

Financial Econometrics In Action:
Analyzing Hedge Funds and
Systemic Risk

Andrew W. Lo, MIT and NBER
Econometric Methods Lecture
July 16, 2010

© 2010 by Andrew W. Lo
All Rights Reserved

Outline of Presentation NBER

 Basic Facts About Hedge Funds
 Hedge-Fund Returns
 Hedge-Fund Risks and Liquidity
 Hedge Funds and Systemic Risk
 August 1998, August 2007,
May 2010,…

7/16/2010 © 2010 by Andrew W. Lo, All Rights Reserved Page 2

Basic Facts About
Hedge Funds
References
 Fung, W. and D. Hsieh, 1999, “A Primer on Hedge Funds”, Journal of Empirical
Finance 6, 309–331.
 Lo, A., 2002, “The Statistics of Sharpe Ratios”, Financial Analysts Journal 58, 36–52.
 Lo, A., 2008, Hedge Funds: An Analytic Perspective, Chapters 1–2. Princeton, NJ:
Princeton University Press.

7/16/2010 © 2010 by Andrew W. Lo, All Rights Reserved Slide 3

What Are Hedge Funds? NBER

Unregulated Investment Companies
 For “qualified” (sophisticated) investors
 Need not satisfy regulatory requirements (few investment constraints)
 High fees, high performance (historically), and high attrition

Alfred Winslow Jones
 First “hedge fund” in 1949 (market exposure vs. stock selection):

 Magnify stock selection (leverage). reduce market exposure (short positions)
 Hence the term “hedge”
 Charged 20% incentive fee
 Eventually included several managers (fund of funds)



An Example
 Suppose a mutual fund has $10MM of capital
– It can purchase $10MM of IBM
– If IBM goes up by 10%, mutual fund makes $1MM
– If IBM goes down by 10%, mutual fund loses $1MM
 Suppose a hedge fund has $10MM
– It can purchase $30MM of IBM (like a margin account)
– It can simultaneously sell $30MM of GE
– If IBM goes up by 10%, hedge fund makes $3MM
– If GE goes down by 10%, hedge fund makes $3MM
– What if IBM goes down and GE goes up by 10%?


Today:
 Over 10,000 hedge funds worldwide, controlling $1.7 trillion in assets
 Typical fee structure: 2% fixed, 20% incentive, high-water mark
 Most hedge funds are Registered Investment Advisors
 Most common legal form: Limited Liability Corporation (LLC)
 Most common corporate structure:
– Management company (the business entity)
– Fund (the assets), master-feeder structure

Domestic Feeder
(Delaware L.P.)
Master Fund Brokerage
(Delaware L.P.) Account
Offshore Feeder
(Cayman Ltd.)



The Hedge-Fund Industry Is Heterogeneous:
 Closer to 20–30 cottage industries
 CSFB/Tremont Index Categories:
Hedge Funds Event-Driven Multi-Strategy
Convertible Arbitrage Fixed Income Arbitrage
Dedicated Shortseller Global Macro
Distressed Long/Short Equity
Emerging Markets Managed Futures
Equity Market Neutral Multi-Strategy
Event-Driven Risk Arbitrage
 Does not include private equity, venture capital, real estate, natural
resources, insurance



Why Should We Care About Hedge Funds?
 Hedge funds play a key role in the financial industry
– During normal times, hedge funds are the “tip of the spear”
– During bad times, hedge funds are the “canary in the cage”
 As unregulated entities, hedge funds innovate rapidly
 Due to leverage, hedge funds have disproportionate impact on markets
 Investors in hedge funds include:
– Private investors
– Fund of funds
– Central banks and sovereign wealth funds
– Insurance companies
– Pension funds



Assets In The Hedge Fund Industry, 1990–2010
$ trillions
6.0
$6,000

5.0
$5,000

4.0
$4,000
Assets (U

3.0
$3,000

2.0
$2,000

1.0
$1,000

$0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Estimated assets 38 58 95 167 167 185 256 367 374 456 490 539 625 820 972 1,105 1,464 1,910 1,440 1,600 1,770
Market positions 95 145 238 418 418 463 640 918 935 1,140 1,225 1,887 1,877 1,968 2,620 2,674 4,247 5,231 2,730 3,670 4,200

Sources: through Q4 2009 – HFR industry report. 2010 projections – based on CS analysis



Hedge Funds Are The “Galapagos Islands” of Finance
 Relatively low barriers to entry and exit
 High levels of compensation
 Competition and adaptation are extreme
 New “species” are coming and going constantly
 Strategies wax and wane over time:
– Risk arbitrage waxing
– Statistical arbitrage waning
 Supports Adaptive Markets Hypothesis (Lo, 2004, 2005)


Basic Empirical Facts NBER

Aggregate and Individual Data:
 CSFB/Tremont hedge-fund indexes
– 14 indexes (13 style categories)
– Jan 1994 to present
 TASS individual hedge-fund returns
– Feb 1977 to present
– Monthly returns, AUM, style, fund info
 Other sources for individual hedge-fund data
– HFR, CISDM, MSCI, Eureka
– Some overlap, but many unique funds in each dataset
– See Fung and Hsieh (2007)



Intersections of Hedge-Fund Databases as of 2005

Source: Fung and Hsieh (2007)



Number of Funds in TASS, HFR, and CISDM Databases




Change in Style Composition of TASS Hedge Funds
(Percentage of Total AUM)




Summary Statistics, Jan 1994 to Feb 2009



Sharpe Ratio Measures Reward/Risk Trade-Off
 For monthly returns:

 Sometimes, Rf is assumed to be 0
 SR measures reward per unit risk
 Typical SR for long-only equities is 0.25 to 0.50
 Typical SR for liquid hedge funds is 0.25 to 1.00
 Typical SR for illiquid hedge funds is 0.50 to 1.50



High Sharpe Ratios May Be Due To:
 High-frequency trading (market-making)
 Illiquid securities (low volatility from smooth returns)
 Good information (too good?)
 Fraud?

Investors Are Drawn To
High Sharpe Ratios



Consider A Simple Example of Market-Making:
 Each day, you can make or lose 0.5%
 Probability of winning is 58%, probability of losing is 42%
 Expected return is 8 basis points a day (0.08%)
 Suppose risk-free rate is 1% per year
 Assume 250 days of independent bets
 Sharpe ratio is:



An Aside On Estimating Sharpe Ratios (Lo, 2002):
 For IID returns with finite fourth moments:



 For stationary and ergodic returns, use GMM:



 Time aggregation:



Summary Statistics for Individual Mutual Funds and Hedge Funds
Start Sample
Fund Date End Date Size Mean SD ρ1 ρ2 ρ3 p(Q 11)
(%) (%) (%) (%) (%) (%)

Mutual Funds
Vanguard 500 Index Oct-76 Jun-00 286 1.30 4.27 -4.0 -6.6 -4.9 64.5
Fidelity Magellan Jan-67 Jun-00 402 1.73 6.23 12.4 -2.3 -0.4 28.6
Investment Company of America Jan-63 Jun-00 450 1.17 4.01 1.8 -3.2 -4.5 80.2
Janus Mar-70 Jun-00 364 1.52 4.75 10.5 0.0 -3.7 58.1
Fidelity Contrafund May-67 Jun-00 397 1.29 4.97 7.4 -2.5 -6.8 58.2
Washington Mutual Investors Jan-63 Jun-00 450 1.13 4.09 -0.1 -7.2 -2.6 22.8
Janus Worldwide Jan-92 Jun-00 102 1.81 4.36 11.4 3.4 -3.8 13.2
Fidelity Growth and Income Jan-86 Jun-00 174 1.54 4.13 5.1 -1.6 -8.2 60.9
American Century Ultra Dec-81 Jun-00 223 1.72 7.11 2.3 3.4 1.4 54.5
Growth Fund of America Jul-64 Jun-00 431 1.18 5.35 8.5 -2.7 -4.1 45.4

Hedge Funds
Convertible/Option Arbitrage May-92 Dec-00 104 1.63 0.97 42.7 29.0 21.4 0.0
Relative Value Dec-92 Dec-00 97 0.66 0.21 25.9 19.2 -2.1 4.5
Mortgage-Backed Securities Jan-93 Dec-00 96 1.33 0.79 42.0 22.1 16.7 0.1
High Yield Debt Jun-94 Dec-00 79 1.30 0.87 33.7 21.8 13.1 5.2
Risk Arbitrage A Jul-93 Dec-00 90 1.06 0.69 -4.9 -10.8 6.9 30.6
Long/Short Equities Jul-89 Dec-00 138 1.18 0.83 -20.2 24.6 8.7 0.1
Multi-Strategy A Jan-95 Dec-00 72 1.08 0.75 48.9 23.4 3.3 0.3
Risk Arbitrage B Nov-94 Dec-00 74 0.90 0.77 -4.9 2.5 -8.3 96.1
Convertible Arbitrage A Sep-92 Dec-00 100 1.38 1.60 33.8 30.8 7.9 0.8
Convertible Arbitrage B Jul-94 Dec-00 78 0.78 0.62 32.4 9.7 -4.5 23.4
Multi-Strategy B Jun-89 Dec-00 139 1.34 1.63 49.0 24.6 10.6 0.0
Fund of Funds Oct-94 Dec-00 75 1.68 2.29 29.7 21.1 0.9 23.4



Correlations Between CS/Tremont Indexes and Other Indexes
January 1994 to February 2009


Hedge-Fund Returns
References
 Ackermann, C., McEnally, R. and D. Ravenscraft, 1999, “The Performance of Hedge
Funds: Risk, Return, and Incentives'', Journal of Finance 54, 833–874.
 Lo, A., 2008, “Where Do Alphas Come From?: A New Measure of the Value of
Active Investment Management”, Journal of Investment Management 6, 1–29.


Motivation NBER

Total Return Comparison, Jan 1994 to Feb 2009


Motivation NBER

What Do We Mean By “Active” Management?
 Excess expected return above some benchmark
 “Active” bets means departures from the benchmark Ft
 Benchmarks are considered “passive”
– Easy to achieve, very little skill involved, low fees
 Traditional measure of active management—IR:

 But what if there are other “systematic” factors?
 Such factors can yield positive IR, but require no skill


Motivation NBER

 For example:

αi ??
 In this case, excess returns may be passively achieved
 One natural implication: “hedge-fund beta” products
 So what do we really mean by “active management” in a multi-
factor world?
⇒ Where Do Alphas Come From?


A New Measure NBER

A New Definition of Active Management:
 Active management consists of time-varying portfolio weights
based on information
 Passive management consists of either non-stochastic
portfolios, or where the time-variation is not information-
based, e.g., value-weighted portfolios, random portfolios
 To measure the value of active management, ask whether
time-variation in weights increases expected return E[Rt]
 We don’t need to know Xt-1 to estimate the value of active
management


A New Measure NBER

A New Measure of the Value of Active Management:
 Recall the definition of covariance:

 True for sample estimators as well:


A New Measure NBER


Active Passive

A New Measure NBER


 Two sources of expected returns:
1. Average exposures to risk premia µi
2. Correlation between weights and returns (asset-timing ability)
 Note: weights are determined prior to returns (otherwise,
arbitrage is possible: buy low/sell high), hence they capture
predictability of current asset returns by past information


A New Measure NBER


 Active component can be estimated without covariances
 Active component can be estimated without weights ωi
 In fact, only average weights and returns are needed
 Ideal for institutional investors of active managers and hedge
funds (position transparency unnecessary)


A New Measure NBER

Implicit Assumptions:
 Returns are stationary and parameters are known
 Xt is information (need not be private)

Properties of Determine Active Value-Added
 If weights are non-stochastic, covariances are 0
 If weights are uncorrelated with returns, covariances are 0
 Passive portfolio earns risk premia on average
 Unlike IR, θ is a dynamic measure of predictability


Example 1 NBER

Weight 1 Asset Return 1

100% 2.50%

Portfolio Weight 1
80% 2.00%

Asset Return 1
60% 1.50%

40% 1.00%

20% 0.50%

0% 0.00%
1 2 3 4 5 6 7 8 9 10 11 12

 Constant weights (no timing)
 No active management
 Active component: 0%
 Passive component: 100%


Example 1 NBER


100% 2.50%

Portfolio Weight 1
80% 2.00%

Asset Return 1
60% 1.50%

40% 1.00%

20% 0.50%

0% 0.00%
1 2 3 4 5 6 7 8 9 10 11 12

 Time-varying weights
 Adds 0.13% per month
 Active component: 9.7%
 Passive component: 90.3%


Example 1 NBER


100% 2.50%

Portfolio Weight 1
80% 2.00%

Asset Return 1
60% 1.50%

40% 1.00%

20% 0.50%

0% 0.00%
1 2 3 4 5 6 7 8 9 10 11 12

 Subtracts 0.13% per month
 Active component: -12.05%


Example 2 NBER

Market Timing
 Equities vs. cash, where equity returns follow AR(1):

 Consider 0/1 investment policy based on lagged return


Example 2 NBER

Market Timing
 Then expected returns can be computed explicitly:


Example 2 NBER


The Sampling Interval NBER

Consider Warren Buffett vs. Steven Cohen
 Buffett makes a few decisions per year
 Cohen makes a few decisions per minute
 One symptom: very different Sharpe ratios!
 How frequently do we need to sample their weights?

Do these terms capture the
variation in the manager’s
weights, i.e., decisions?



Case 1: Decision Interval Coincides with Sampling Interval



Case 2: Decision Interval Longer Than Sampling Interval


A New Measure NBER

 Recall the definition of covariance:

 True for sample estimators as well:


Example 1 NBER


100% 2.50%

Portfolio Weight 1
80% 2.00%

Asset Return 1
60% 1.50%

40% 1.00%

20% 0.50%

0% 0.00%
1 2 3 4 5 6 7 8 9 10 11 12

 Constant weights (no timing)
 No active management
 Active component: 0%
 Passive component: 100%


Example 1 NBER


100% 2.50%

Portfolio Weight 1
80% 2.00%

Asset Return 1
60% 1.50%

40% 1.00%

20% 0.50%

0% 0.00%
1 2 3 4 5 6 7 8 9 10 11 12

 Adds 0.13% per month
 Active component: 9.7%


Example 1 NBER


100% 2.50%

Portfolio Weight 1
80% 2.00%

Asset Return 1
60% 1.50%

40% 1.00%

20% 0.50%

0% 0.00%
1 2 3 4 5 6 7 8 9 10 11 12

 Subtracts 0.13% per month
 Active component: -12.05%


Example 2 NBER

Market Timing
 Equities vs. cash, where equity returns follow AR(1):

 Consider 0/1 investment policy based on lagged return


Example 2 NBER

Market Timing
 Then expected returns can be computed explicitly:


Example 2 NBER



Consider Warren Buffett vs. Steven Cohen
 Buffett makes a few decisions per year
 Cohen makes a few decisions per minute
 One symptom: very different Sharpe ratios!
 How frequently do we need to sample their weights?

Do these terms capture the
variation in the manager’s
weights, i.e., decisions?



Case 1: Decision Interval Coincides with Sampling Interval


Empirical Analysis NBER

Consider Contrarian Trading Strategy:
 Lo and MacKinlay (1990)



 Profitability can be quantified:



Apply This To NASDAQ Size-Decile Returns
 January 2, 1990 to December 29, 1995
 Use only 5 smallest deciles to maximize returns
 May not be realistic, but is illustrative

Lead/Lag Effect Among NASDAQ Five Smallest Size-Decile Portfolios

R1t+1 R2t+1 R3t+1 R4t+1 R5t+1
R1t 10.0% 21.5% 15.8% 18.1% 16.7%
R2t 23.4% 15.4% 20.2% 19.7% 15.8%
R3t 26.2% 25.0% 15.2% 23.9% 21.6%
R4t 25.4% 27.0% 24.3% 18.2% 18.7%
R5t 25.4% 26.6% 26.5% 26.2% 19.4%



Cumulative Return of Contrarian Strategy on NASDAQ Deciles 1–5
January 2, 1990 to December 29, 1995



Summary Statistics for NASDAQ Deciles
and Contrarian Strategy, Daily
January 2, 1990 to December 29, 1995

Statistic Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Rpt

Mean*250 27.4% 17.5% 14.0% 13.7% 12.8% 31.4%
SD*sqrt(250) 12.2% 9.8% 8.9% 9.1% 9.5% 7.9%
SR*sqrt(250) 2.25 1.78 1.58 1.50 1.35 3.95
Min -2.9% -2.7% -2.7% -3.3% -3.5% -2.2%
Median 0.1% 0.1% 0.1% 0.1% 0.1% 0.1%
Max 6.7% 3.6% 2.0% 2.1% 2.3% 2.4%
Skew 0.6 0.0 -0.5 -0.7 -0.9 -0.1
XSKurt 5.1 2.4 2.1 3.1 3.9 1.7
ρ1 10.0% 15.4% 15.2% 18.2% 19.4% 4.7%
ρ2 10.3% 7.7% 10.1% 13.9% 10.5% 0.9%
ρ3 5.7% 4.2% 7.5% 9.2% 11.0% 7.5%



Measure of Active Management (Daily):

Estimate SE t-stat
Portfolio Mean * 250 31.4% 0.3% 91.00
Risk Premia * 250 -0.6% 3.5% -0.17
Active Component * 250 32.0% 3.5% 9.24
Active Ratio 101.9% 0.3% 354.40

 More than 100% of returns are active!
 Why is passive component negative? Mean reversion!
– On average, holding losers and shorting winners is costly



Now Consider Monthly Transparency
 Manager provides month-end weights and monthly returns
(summed) for all positions
 How informative is this information for the strategy?
 Note:



Summary Statistics for NASDAQ Deciles
and Contrarian Strategy, Monthly
January 1990 to December 1995

Statistic Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Rpt

Mean*12 27.5% 17.4% 13.9% 13.7% 12.8% -4.0%
SD*sqrt(12) 20.6% 17.7% 15.6% 15.0% 15.9% 8.8%
SR*sqrt(12) 1.34 0.98 0.89 0.91 0.80 -0.45
Min -8.0% -11.3% -9.0% -9.7% -11.4% -6.6%
Median 1.6% 1.0% 1.2% 1.0% 1.5% -0.5%
Max 26.4% 21.5% 18.1% 16.8% 16.2% 9.7%
Skew 1.2 0.9 0.7 0.5 0.1 1.2
XSKurt 3.3 3.0 2.1 2.0 1.2 4.7
ρ1 36.4% 43.7% 43.2% 41.4% 45.0% 8.2%
ρ2 17.3% 16.6% 18.9% 10.1% 13.7% 15.7%
ρ3 -5.6% -2.7% -3.1% -7.8% -7.0% -3.2%



Measure of Active Management (Monthly):

Estimate SE t-stat
Portfolio Mean * 12 -4.0% 1.0% -3.98
Risk Premia * 12 0.1% 4.0% 0.03
Active Component * 12 -4.1% 4.1% -1.01
Active Ratio 102.6% 11.8% 8.66

 Total return is negative based on month-end weights!
 Active component is responsible for this negative return
 Results bear no relation to the daily strategy!


Extension To Multi-Factor Models NBER

Consider Multi-Factor Model for Individual Securities:

Possible Factors:
 S&P 500  MSCI  Option-Based Strategies
 Lehman Bond  U.S. Growth/Value  Property & Casualty
 US Dollar  U.S. Large/Small Cap  Real Estate
 GSCI  Foreign Growth/Value  Timber
 VIX  Foreign Large/Small Cap  Hurricane
 Credit  Liquidity  Earthquakes



Now Consider An Active Manager:

Security Selection Factor Exposures

Factor Timing



Expected Returns Are the Sum of Three Components:
1. Risk Premia (Average Betas)
2. Security Selection (Alpha)
3. Factor Timing (Time-Varying Portfolio Betas)

Active Passive



Expected Returns Are the Sum of Three Components:
1. Risk Premia (Average Betas)
2. Security Selection (Alpha)
3. Factor Timing (Time-Varying Portfolio Betas)


Summary NBER

“Active” Management Involves Changing Weights
 Measure of active management must involve weights
 Correlation between weights and returns is key
 Good managers have large positive correlations
 Implies new definitions for “active” and “passive”
 Constant- or undirected-weight component is “passive”
 “Passive” is not always simple, but is often easier to implement
 Investors should pay less for passive management
 Investors should pay more for active management
 Implies new methods for performance attribution
 Requires new data: average weights, average returns
 Sampling interval must be at least as fine as decision interval
 Factor models provide additional transparency


Hedge-Fund Risks and
Liquidity
References
 Fung, W. and D. Hsieh, 2001, “The Risk in Hedge Fund Strategies: Theory and
Evidence from Trend Followers'', Review of Financial Studies 14, 313–341.
 Getmansky, M., Lo, A. and I. Makarov, 2004, “An Econometric Analysis of Serial
Correlation and Illiquidity in Hedge-Fund Returns”, Journal of Financial Economics
74, 529–609.
 Lo, A., 2001, “Risk Management For Hedge Funds: Introduction and Overview”,
Financial Analysts Journal 57, 16–33.


Risk Models NBER

Hedge Fund Risk Exposures are Different:
 Heterogeneity of Hedge-Fund Styles
Long/Short Equity Fund Fixed-Income Arb Fund
– Market Beta – Yield Curve Model
– Industry/Sector Exposure – Credit Exposure
– Value/Growth – Liquidity Exposure
– Stock-Loan Constraints – Leverage Constraints
– Execution Costs – Macroeconomic Factors

 Phase-Locking Regime Shifts
 Nonlinearities


Risk Models NBER

Correlation Analysis Can Be Misleading:
 Correlation Is A Linear Measure
 Hedge Fund Returns May Be Nonlinear
 One Example: “Phase-Locking”


Risk Models NBER

 When no event occurs (It = 0):

 When an event occurs (It = 1):


Risk Models NBER

 Unconditional Correlation:

 Let p = 0.001, σz = 10σε :


Risk Models NBER

Two-State Markov Regime-Switching Model:

State 1 State 0


Risk Models NBER

Regime-Switching Probability Estimates for
CSFB/Tremont Hedge-Fund Indexes
January 1994 to September 2006


Risk Models NBER
Regime-Switching Probability Estimates for CSFB/Tremont
Fixed-Income Arbitrage Index


Risk Models NBER
Convertible Arbitrage Index


Risk Models NBER
Managed Futures Index


Risk Models NBER

Aggregate Hedge-Fund Indexes:
 Estimate Parameters Via MLE
 Estimate Unconditional State Probabilities
 Aggregate Probabilities Across Indexes


Risk Models NBER


Liquidity NBER

Definition of Liquidity:
 Price, Time, and Size
 Many Hedge Funds Are Not “Liquid”
 Liquidity Risk Has Multiple Dimensions
– Valuation, Correlation, Timing
 Nothing Wrong With Illiquidity, If Fully Disclosed
 Problems When Illiquidity Is Not Recognized
 Simple Indicator of Illiquidity:


Pages from f metrics hf_1

More Related Content

Similar to Pages from f metrics hf_1 (20)

More from NBER (20)

Pages from f metrics hf_1