empirical financial economics current approaches to performance measurement
TRANSCRIPT
Empirical Financial Economics
Current Approaches to Performance Measurement
Overview of lecture
Standard approachesTheoretical foundationPractical implementationRelation to style analysisGaming performance metrics
Performance measurement
Leeson InvestmentManagement
Market (S&P 500) Benchmark
Short-term Government Benchmark
Average Return
.0065 .0050 .0036
Std. Deviation
.0106 .0359 .0015
Beta .0640 1.0 .0
Alpha .0025(1.92)
.0 .0
Sharpe Ratio
.2484 .0318 .0
Style: Index Arbitrage, 100% in cash at close of trading
Frequency distribution of monthly returns
0
5
10
15
20
25
30
35
Universe Comparisons
5%
10%
15%
20%
25%
30%
35%
40%
Brownian ManagementS&P 500
One Quarter
1 Year 3 Years 5 Years
Periods ending Dec 31 2002
Average Return
Total Return comparison
A
BCD
rf = 1.08%
Average Return
RS&P = 13.68%
Total Return comparison
AS&P 500
BCD
Treasury Bills
Manager A best
Manager D worst
Average Return
Total Return comparison
A
BCD
Average Return
Standard Deviation
Sharpe ratio comparison
A
BC
D
rf = 1.08%
σS&P = 20.0%
Average Return
Standard Deviation
RS&P = 13.68%
Sharpe ratio comparison
^
AS&P 500
BC
D
Treasury Bills
rf = 1.08%
σS&P = 20.0%
Average Return
Standard Deviation
RS&P = 13.68%
Sharpe ratio comparison
^
AS&P 500
BC
D
Treasury Bills
Manager D bestManager C worstSharpe ratio =
Average return – rf
Standard Deviation
rf = 1.08%
σS&P = 20.0%
Average Return
Standard Deviation
RS&P = 13.68%
Sharpe ratio comparison
^
AS&P 500
BC
D
Treasury Bills
13
rf = 1.08%
βS&P = 1.0
Average Return
Beta
RS&P = 13.68%
Treynor Measure comparison
A
S&P 500
BCD
Treasury Bills
14
rf = 1.08%
Average ReturnRS&P =
13.68%
Treynor Measure comparison
A
S&P 500
BCD
Treasury Bills
Manager B worstManager C best
Treynor measure =Average return – rf
Beta
βS&P = 1.0Beta
rf = 1.08%
Average Return
RS&P = 13.68%
Jensen’s Alpha comparison
AS&P 500
BCD
Treasury Bills
Manager B worstJensen’s alpha = Average return
–
{rf + β (RS&P - rf )}
βS&P = 1.0Beta
Manager C best
Intertemporal equilibrium model
Multiperiod problem:
First order conditions:
Stochastic discount factor interpretation:
“stochastic discount factor”, “pricing kernel”
0
Max ( )jt t j
j
E U c
,( ) (1 ) ( )jt t i t j t jU c E r U c
, , ,
( )1 (1 ) ,
( )t jj
t i t j t j t jt
U cE r m m
U c
,t jm
Value of Private Information
Investor has access to information
Value of is given by where and are returns on optimal portfolios given and
Under CAPM (Chen & Knez 1996)
Jensen’s alpha measures value of private information
Other pricing kernels:
1 0I I
1 0I I 1 0[( ) ]t tE R R m 1R 0R1I 0I
1 0 1 1 1[( ) ] ( )t t t ft t mt ftE R R m r r
( )Mm a b r c
The geometry of mean variance
a
b
a
b
E
2 1a
1 1
2
1/1/
0
bx b
22
2
2a bE cE
ac b
Note: returns are in excess of the risk free rate
fr
Informed portfolio strategy
Excess return on informed strategy where is the return on an optimal orthogonal portfolio (MacKinlay 1995)
Sharpe ratio squared of informed strategy
Assumes well diversified portfolios
1 0f fR r R r
2 1 1 2 2 21 0 0 0 0( ) ( )f fr r
Informed portfolio strategy
Excess return on informed strategy where is the return on an optimal orthogonal portfolio (MacKinlay 1995)
Sharpe ratio squared of informed strategy
Assumes well diversified portfolios
1 0f fR r R r
2 1 1 2 2 21 0 0 0 0( ) ( )f fr r
Used in tests of mean variance efficiency of benchmark
Practical issues
Sharpe ratio sensitive to diversification, but invariant to leverage
Risk premium and standard deviation proportionate to fraction of investment financed by borrowing
Jensen’s alpha invariant to diversification, but sensitive to leverage
In a complete market implies through borrowing (Goetzmann et al 2002)
2 0
Changes in Information Set
How do we measure alpha when information set is not constant?
Rolling regression, use subperiods to estimate
(no t subscript) – Sharpe (1992)
Use macroeconomic variable controls – Ferson and Schadt(1996)
Use GSC procedure – Brown and Goetzmann (1997)
1 1 1 ( )t t ft t mt ftr r 1tI
1 1 1( )f m ftr r
Style management is crucial …
Economist, July 16, 1995
But who determines styles?
Characteristics-based Styles
Traditional approach …
are changing characteristics (PER, Price/Book)
are returns to characteristics Style benchmarks are given by
jt Jt Jt t jtr I j J
jt Jt jtr j J
JttI
Jt
Returns-based Styles
Sharpe (1992) approach …
are a dynamic portfolio strategy are benchmark portfolio returns Style benchmarks are given by
jt Jt Jt t jtr I j J
jt Jt jtr j J
JttI
Jt
Returns-based Styles
GSC (1997) approach …
vary through time but are fixed for style
Allocate funds to styles directly using
Style benchmarks are given by
jt Jt Jt t jtr I j J
jt Jt jtr j J
,jT Jt
Jt
J
Jt
Basis Assets
GSC (1997) approach …
vary through time but fixed for risk class
Allocate equities to risk classes directly using
Style benchmarks are given by
jt Jt Jt t jtr I j J
jt Jt jtr j J
,jT Jt
Jt
J
Jt
Brown, Stephen J. and William N. Goetzmann, 1997 Mutual Fund Styles. Journal ofFinancial Economics 43:3, 373-399.
Switching Regression
Quandt (1958)
If regimes not observed
1 1 1
2 2 2
3 3 3
1 0
0 0
t t t t
t t t t
t t t
tt
t
y X I
y X I
Z X
if ZI
if Z
= 1 (Regime 1)
= 0 (Regime2)
K means procedure
Hartigan (1975)
If regimes not observed, use iterative algorithm to determine regime membership
1 1
2 2
t t t
t t t
t Kt Kt
y
y
y K
(Regime 1)
(Regime2)
(Regime )
Switching Regression
Quandt and Ramsey (1978)
Method of moments ...
2 21 2
2 21 2
2 21 1 2 2
( ) ( )
2 21 2
1 2
2 2 2 21 2 1 2
( ) ( , ) (1 ) ( , )
1
2 2
(1 )
(1 ) 2 (1 )( )
.
y y
y
y
f y N N
e e
etc
Eight style decomposition
GSC1 GSC2 GSC3 GSC4 GSC5 GSC6 GSC7 GSC80%
20%
40%
60%
80%
100%
US Equity HedgeNon-US Equity HedgeEvent DrivenNon Directional/Relative ValueGlobal MacroPure Leveraged Currency
Five style decomposition
GSC1 GSC2 GSC3 GSC4 GSC50%
20%
40%
60%
80%
100%
US Equity HedgeNon-US Equity HedgeEvent DrivenNon Directional/Relative ValueGlobal MacroPure Leveraged Currency
Style classifications
GSC1 Event driven internationalGSC2 Property/Fixed IncomeGSC3 US Equity focusGSC4 Non-directional/relative valueGSC5 Event driven domesticGSC6 International focusGSC7 Emerging marketsGSC8 Global macro
Regressing returns on classifications: Adjusted R2
Year N GSC 8
classifications
GSC 5 classificati
ons
TASS 17 classificati
ons
1992 149 0.3827 0.1713 0.4441
1993 212 0.2224 0.132 0.1186
1994 288 0.1662 0.104 0.0986
1995 405 0.0576 0.0548 0.0446
1996 524 0.1554 0.0769 0.1523
1997 616 0.3066 0.1886 0.2538
1998 668 0.2813 0.2019 0.1998
Average 0.2246 0.1328 0.1874
“Informationless” investing
Analytic Optioned
Our fund purchases a stock and simultaneously sells a call option against the stock. By doing this, the fund receives both dividend income from the stock and a cash premium from the sale of the option. This strategy is designed for the longer term investor who wants to reduce risk. It is particularly suited for pension plans IRAs and Keoghs. Our defensive buy/write strategy is designed to put greater emphasis on risk reduction by focusing on “in-the-money” call options. The results speak for themselves. Over a twelve year period, of 153 institutional portfolios in the Frank Russell Co. universe, no other portfolio had a higher return with less risk than our All Buy/Write Accounts Index. In the terminology of modern portfolio theory, our clients’ portfolios dominated the market averages.
Modern Portfolio Theory
Covered Call Strategy
Stock
Value
Profit to Covered Call
Payoff to Covered Call
Unoptioned Portfolio Return
Portfolio Return
Expected
Return
Optioned Return
Portfolio Return
Expected
Return
Optioned Return (incl. premium)
Portfolio Return
Expected
Return
Concave payout strategies
Zero net investment overlay strategy (Weisman 2002)
Uses only public informationDesigned to yield Sharpe ratio greater than
benchmarkUsing strategies that are concave to
benchmark
Concave payout strategies
Zero net investment overlay strategy (Weisman 2002)
Uses only public informationDesigned to yield Sharpe ratio greater than
benchmarkUsing strategies that are concave to
benchmark
Why should we care?
Sharpe ratio obviously inappropriate hereBut is metric of choice of hedge funds and
derivatives tradersGoetzmann, William N., Ingersoll, Jonathan E., Spiegel, Matthew I. and Welch, Ivo, 2007 Portfolio Performance
Manipulation and Manipulation-proof Performance Measures, Review of Financial Studies 20(5) 1503-1546.
Sharpe Ratio of Benchmark
-50% 0% 50% 100%-2
-1.5
-1
-0.5
0
0.5
1
Benchmark
Sharpe ratio = .631
Maximum Sharpe Ratio
-50% 0% 50% 100%-2
-1.5
-1
-0.5
0
0.5
1
Benchmark
Maximum Sharpe Ra-tio Strategy
Sharpe ratio = .748
Short Volatility Strategy
-50% 0% 50% 100%-2
-1.5
-1
-0.5
0
0.5
1
Benchmark
Short volatility
Benchmark return
Port
foli
o r
etu
rn
Sharpe ratio = .743
Concave trading strategies
-50% 0% 50% 100%-2
-1.5
-1
-0.5
0
0.5
1
Benchmark
Loss Averse Trading (Median)
Maximum Sharpe Ra-tio Strategy
Examples of concave payout strategies
Long-term asset mix guidelines
Unhedged short volatilityWriting out of the money
calls and puts
Examples of concave payout strategies
Loss averse trading a.k.a. “Doubling”
Examples of concave payout strategies
Examples of concave payout strategies
Long-term asset mix guidelines
Unhedged short volatilityWriting out of the money
calls and puts
Loss averse trading a.k.a. “Doubling”
Forensic Finance
Implications of concave payoff strategies
Patterns of returnsare returns concave to benchmark?
Patterns of security holdingsdo security holdings produce
concave payouts?Patterns of trading
does pattern of trading lead to concave payouts?
Manipulation proof measure
Criteria:Ranks portfolios based on investor
preferencesCannot reward informationless tradingShould be scale invariantShould be consistent with market
equilibrium models
1
1
1 1ˆ n (1 ) (1 ) .(1 )
T
t ftt
r rt T
1
1
1 1ˆ n (1 ) (1 ) .(1 )
T
t ftt
r rt T
1
1
1 1ˆ n (1 ) (1 ) .(1 )
T
t ftt
r rt T
1
1
1 1ˆ n (1 ) (1 ) .(1 )
T
t ftt
r rt T
Goetzmann, William N., Ingersoll, Jonathan E., Spiegel, Matthew I. and Welch, Ivo, 2007 Portfolio Performance Manipulation and Manipulation-proof Performance Measures, Review of Financial Studies 20(5) 1503-1546.
Manipulation proof measure
1
1
1 1ˆ n (1 ) (1 ) .(1 )
ˆ
T
t ftt
r rt T
T
t
Certainty equivalent of portfolio return
Number of observations
Length of time between observationsChosen to make holding benchmark optimal for an uninformed investor
Implied risk aversion parameter
ln[ (1 )] ln[(1 )]
(1 )m f
m
E r r
Var r
What implied risk aversion parameter makes the
market participant indifferent to holding the
market portfolio?
Performance of beta ranked portfolios
Jan-00 Sep-02 Jun-05 Mar-08 Dec-100
0.5
1
1.5
2
2.5
3
lobethighbetmktCRSPvw
Performance of vol ranked portfolios
Jan-00 Sep-02 Jun-05 Mar-08 Dec-100
0.5
1
1.5
2
2.5
lowvolhivolCRSPvw
Descriptive statistics of beta ranked portfolios
Monthly returns: January 1980 - December 2011
lobeta hibeta mktbeta lovol hivol lobeta-hibeta lovol-hivol
Mean 1.00% 0.71% 0.95% 1.03% 0.32% 0.29% 0.71%
Std.Dev 2.78% 9.38% 3.71% 2.76% 8.31% 8.36% 7.10%
Skewness -1.741 -0.253 -1.146 -1.207 -0.239 0.036 -0.228
Kurtosis 14.619 4.199 9.208 7.933 5.601 4.564 7.393
Beta 0.339 1.791 0.667 0.464 1.450 -1.449 -0.983
Sharpe 0.2080 0.0310 0.1415 0.2223 -0.0123 -0.0159 0.0411
Alpha 0.391% -0.697% 0.159% 0.352% -0.903% 0.667% 0.833%
t-value 2.805 -3.022 1.331 3.556 -3.522 2.339 2.810
FF alpha 0.322% -0.489% 0.158% 0.218% -0.776% 0.391% 0.574%
t-value 2.483 -2.866 1.645 2.426 -4.628 1.507 2.668
MPPM -.00205 .00391 -.00148 -.00221 .00453
Descriptive statistics of beta ranked portfolios
Daily returns: January 1980 - December 2011
lobeta hibeta mktbeta lovol hivol lobeta-hibeta lovol-hivol
Mean 0.05% 0.03% 0.04% 0.05% 0.01% 0.01% 0.04%
Std.Dev 0.34% 1.96% 0.56% 0.49% 1.39% 1.92% 1.11%
Skewness -2.926 -0.133 -1.829 -1.171 -0.597 -0.172 0.385
Kurtosis 46.557 10.471 30.116 37.263 12.931 10.886 16.465
Beta 0.070 1.615 0.416 0.378 1.033 -1.545 -0.655
Sharpe 0.0764 0.0060 0.0411 0.0570 -0.0085 -0.0029 0.0182
Alpha 0.024% -0.032% 0.012% 0.018% -0.040% 0.037% 0.038%
t-value 5.907 -3.423 2.994 5.574 -4.180 3.564 3.707
FF alpha 0.025% -0.034% 0.013% 0.012% -0.041% 0.039% 0.034%
t-value 6.394 -4.984 4.018 4.637 -6.527 4.505 4.615
MPPM -.00009 .00012 -.00007 -.00009 .00013
Hedge funds follow concave strategies
R-rf = α + β (RS&P- rf) + γ (RS&P- rf)2
Hedge funds follow concave strategies
R-rf = α + β (RS&P- rf) + γ (RS&P- rf)2
Concave strategies: tβ > 1.96 & tγ < -1.96
Hedge funds follow concave strategies
Concave NeutralConve
x N
Convertible Arbitrage
Dedicated Short Bias
Emerging Markets
Equity Market Neutral
Event Driven Fixed Income
Arbitrage Fund of Funds Global Macro Long/Short Equity
HedgeManaged Futures
Other
5.38%0.00%
21.89%1.18%
27.03%2.38%
16.38%4.60%
11.19%2.80%5.00%
94.62%100.00
%77.25%97.06%72.64%95.24%82.06%91.38%86.62%94.17%91.67%
0.00%0.00%0.86%1.76%0.34%2.38%1.57%4.02%2.18%3.03%3.33%
13027
233170296126574174
109942960
Grand Total 11.54% 86.53% 1.93% 3318
R-rf = α + β (RS&P- rf) + γ (RS&P- rf)2
Source: TASS/Tremont
Standard deviation as a function of the number of funds in FoHFs
0.35 3.5 35 3500
0.01
0.02
0.03
0.04
0.05
Std. Deviation S&P500 Index
Std. Dev. (naïve 1/N)
Median FoHF Std. Deviation
Number of underlying funds
Std
.Dev
. o
f p
ort
foli
o r
etu
rn
Brown, Stephen J., Gregoriou, Greg N. and Pascalau, Razvan C., Diversification in Funds of Hedge Funds: Is it Possible to Overdiversify? Review of Asset Pricing Studies 2(1), 2012, pp.89-110 http://ssrn.com/abstract=1436468
0.35 3.5 35 350-2
-1.5
-1
-0.5
0
Skewness S&P500 Index
Skewness (naïve 1/N)
Median of FoHF Skewness
Number of underlying funds
Ske
wn
es o
f p
ort
foli
o r
etu
rn
Skewness as a function of the number of funds in FoHFs
0.35 3.5 35 3500
1
2
3
4
5
6
7
8
9
Kurtosis S&P500 Index
Kurtosis (naïve 1/N)
Median FoHF Kurtosis
Number of underlying funds
Ku
rto
sis
of
po
rtfo
lio
ret
urn
Kurtosis as a function of the number of funds in FoHFs
Conclusion
Value of information interpretation of standard performance measures
New procedures for style analysis
Return based performance measures only tell part of the story