You are on page 1of 80

0

Portfolio Management
3-228-07
Albert Lee Chun
Evaluation of Portfolio
Performance
Lecture 11
2 Dec 2008
Albert Lee Chun Portfolio Management
1
Introduction
 As portfolio managers, how can we evaluate the
performance of our portfolio?
 We know that there are 2 major requirements of a
portfolio manager‟s performance:
1. The ability to derive above-average returns
conditioned on risk taken, either through superior
market timing or superior security selection.
2. The ability to diversify the portfolio and eliminate
non-systematic risk, relative to a benchmark
portfolio.
Albert Lee Chun Portfolio Management
2
Today
 Performance Measurement
Risk Adjusted Performance Measures
Measures of Sharpe, Treynor and Jensen
Measures of Skill and Timing
 Attribution de performance
Concept de mesures ajustées pour le risque
Mesures de Sharpe, Treynor et Jensen
Mesure des habilités de timing
Albert Lee Chun Portfolio Management
3
Averaging Returns
Arithmetic Mean:
¿
=
=
n
t
t
n
r
r
1
Geometric Mean:
1 ) 1 (
/ 1
1
÷
(
¸
(

¸

+ =
[
=
n
n
t
t
r r
Example:
(.10 + .0566) / 2 = 7.83%
[ (1.1) (1.0566) ]
1/2
- 1
= 7.808%
Example:
17-3
Albert Lee Chun Portfolio Management
4
The arithmetic average provides unbiased estimates of the
expected return of the stock. Use this to forecast returns in
the next period.

The fixed rate of return over the sample period that would
yield the terminal value is know as the geometric average.

The geometric average is less than the arithmetic average and
this difference increases with the volatility of returns.

The geometric average is also called the time-weighted
average (as opposed to the dollar weighted average),
because it puts equal weights on each return.
Geometric Average
17-4
Albert Lee Chun Portfolio Management
5
Dollar-weighted returns
 Internal rate of return.
 Returns are weighted by the amount invested in each stock.


Time-weighted returns
 Not weighted by investment amount.
 Equal weighting
 Geometric average
Dollar- and Time-Weighted Returns
17-5
Albert Lee Chun Portfolio Management
6
Example: Multiperiod Returns
Period Action
0 Purchase 1 share of Eggbert‟s Egg Co. at $50
1 Purchase 1 share of Eggbert‟s Egg Co. at $53
Eggbert pays a dividend of $2 per share
2 Eggbert pays a dividend of $2 per share
Sell both shares for $108

17-6
Albert Lee Chun Portfolio Management
7
Period Cash Flow
0 -50 share purchase
1 +2 dividend -53 share purchase
2 +4 dividend + 108 shares sold
% 117 . 7
) 1 (
112
) 1 (
51
50
2 1
=
+
+
+
÷
= ÷
r
r r
Internal Rate of Return:
Dollar-Weighted Return
Dollar Weighted: The stocks performance in the second year,
when we own 2 shares, has a greater influence on the overall return.
Albert Lee Chun Portfolio Management
8
Time-Weighted Return
% 66 . 5
53
2 53 54
% 10
50
2 50 53
2
1
=
+ ÷
=
=
+ ÷
=
r
r
[ (1.1) (1.0566) ]
1/2
- 1
= 7.808%
17-8
Time Weighted: Each return has equal weight in the geometric
average.
Geometric Mean:
Albert Lee Chun Portfolio Management
9
Performance Measurement
Albert Lee Chun Portfolio Management
10
Early Performance Measure Techniques
 Portfolio evaluation before 1960
 Once upon a time, investors evaluated a portfolio‟s
performance based purely on the basis of the rate of
return.
 Research in the 1960‟s showed investors how to
quantify and measure risk.
 Grouped portfolios into similar risk classes and
compared rates of return within risk classes.
Albert Lee Chun Portfolio Management
11
Peer Group Comparisons

 This is the most common manner of evaluating
portfolio managers.
 Collects returns of a representative universe of
investors over a period of time and displays them in a
box plot format.
 Example: “US Equity with Cash” relative to peer
universe of US domestic equity managers.
 Issue: There is no explicit adjustment for risk. Risk is
only considered implicitly.

Albert Lee Chun Portfolio Management
12
Albert Lee Chun Portfolio Management
13
Treynor Portfolio Performance Measure
Albert Lee Chun Portfolio Management
14
Treynor (1965)
 Treynor (1965) developed the first composite measure of
portfolio performance that included risk.
 He introduced the portfolio characteristic line, which
defines a relation between the rate of return on a specific
portfolio and the rate of return on the market portfolio.


 The beta is the slope that measures the volatility of the
portfolio‟s returns relative to the market.
 Alpha represents unique returns for the portfolio.
 As the portfolio becomes diversified, unique risk
diminishes.

( )
t p p p , t M, t p,
R R c | o + + =
Albert Lee Chun Portfolio Management
15
A risk-adjusted measure of return that divides a portfolio's
excess return by its beta.

The Treynor Measure is given by





Treynor Measure
p
f p
p
r R
= T
|
÷
The Treynor Measure is defined using the average rate of return
for portfolio p and the risk-free asset.

Albert Lee Chun Portfolio Management
16
Treynor Measure
p
f p
p
r R
= T
|
÷
A larger Tp is better for all investors, regardless of their risk
preferences.

Because it adjusts returns based on systematic risk, it is the
relevant performance measure when evaluating diversified
portfolios held in separately or in combination with other
portfolios.
Albert Lee Chun Portfolio Management
17
Treynor Measure
 Beta measures systematic risk, yet if the portfolio is not fully
diversified then this measure is not a complete characterization
of the portfolio risk.
 Hence, it implicitly assumes a completely diversified portfolio.
 Portfolios with identical systematic risk, but different total
risk, will have the same Treynor ratio!
 Higher idiosyncratic risk should not matter in a diversified
portfolio and hence is not reflected in the Treynor measure.
 A portfolio negative Beta will have a negative Treynor
measure.
 Also known as the Treynor Ratio.
Albert Lee Chun Portfolio Management
18
T-Lines
17-18
Q has higher
alpha, but P has
steeper T-line.
P is the better
portfolio.
Albert Lee Chun Portfolio Management
19
Sharpe Portfolio Performance Measure
Albert Lee Chun Portfolio Management
20
 Similar to the Treynor measure, but uses the total risk of the
portfolio, not just the systematic risk.
 The Sharpe Ratio is given by







 The larger the measure the better, as the portfolio earned a
higher excess return per unit of total risk.

Sharpe Measure
p
f p
p
r R
= S
o
÷
Albert Lee Chun Portfolio Management
21
Sharpe Measure
It adjusts returns for total portfolio risk, as opposed to only
systematic risk as in the Treynor Measure.
Thus, an implicit assumption of the Sharpe ratio is that the
portfolio is not fully diversified, nor will it be combined with
other diversified portfolios.
It is relevant for performance evaluation when comparing
mutually exclusive portfolios.
Sharpe originally called it the "reward-to-variability" ratio, before
others started calling it the Sharpe Ratio.

Albert Lee Chun Portfolio Management
22
SML vs. CML
 Treynor‟s measure uses Beta and hence examines portfolio
return performance in relation to the SML.
 Sharpe‟s measure uses total risk and hence examines portfolio
return performance in relation to the CML.
 For a totally diversified portfolio, both measures give equal
rankings.
 If it is not a diversified portfolio, the Sharpe measure could
give lower rankings than the Treynor measure.
 Thus, the Sharpe measure evaluates the portfolio manager in
terms of both return performance and diversification.
Albert Lee Chun Portfolio Management
23
Price of Risk
 Both the Treynor and Sharp measures, indicate the risk
premium per unit of risk, either systematic risk (Treynor) or
total risk (Sharpe).





 They measure the price of risk in units of excess returns per
each unit of risk (measured either by beta or the standard
deviation of the portfolio).


T = r R
p p f p
| ÷
p p f p
S r R o = ÷
Albert Lee Chun Portfolio Management
24
Jensen Portfolio Performance Measure
Albert Lee Chun Portfolio Management
25
 Alpha is a risk-adjusted measure of superior performance





 This measure adjusts for the systematic risk of the portfolio.
 Positive alpha signals superior risk-adjusted returns, and that
the manager is good at selecting stocks or predicting market
turning points.
 Unlike the Sharpe Ratio, Jensen‟s method does not consider the
ability of the manager to diversify, as it is only accounts for
systematic risk.


Jensen’s Alpha
( )
t p t f p p
r
, , t M, t f, t p,
R r R c | o + ÷ + = ÷
Albert Lee Chun Portfolio Management
26
Multifactor Jensen’s Measure
Measure can be extended to a multi-factor setting, for example:

( )
t p
p p
t f
p
p
HML SML r
,
3 2
, t M,
1
t f, t p,
R r R c | | | o + + + ÷ + = ÷
Albert Lee Chun Portfolio Management
27
Information Ratio
Albert Lee Chun Portfolio Management
28
Information Ratio 1
 Using a historical regression, the IR takes on the form



where the numerator is Jensen‟s alpha and the denominator is the
standard error of the regression. Recalling that
c
o o
p p
IR =
( )
t p t f p p
r
, , t M, t f, t p,
R r R c | o + ÷ + = ÷
Note that the risk here is nonsystematic risk, that could, in theory,
be eliminated by diversification.
Albert Lee Chun Portfolio Management
29
Information Ratio 2
ER
b
p
p
R R
= IR
o
÷
Measures excess returns relative to a benchmark portfolio.
Sharpe Ratio is the special case where the benchmark equals
the risk-free asset.
Risk is measured as the standard deviation of the excess return
(Recall that this is the Tracking Error)
For an actively managed portfolio, we may want to maximize
the excess return per unit of nonsystematic risk we are bearing.

Albert Lee Chun Portfolio Management
30
Portfolio Tracking Error
Excess Return relative
to benchmark portfolio b

Average Excess Return

Variance in Excess Difference

Tracking Error




t b t p t
R R ER
, ,
÷ =
¿
=
=
T
t
t
ER
T
ER
1
1
( )
2
1
2
1
1
¿
=
÷
÷
=
T
t
t ER
ER ER
T
o
2
ER ER
o o =
Albert Lee Chun Portfolio Management
31
Information Ratio
 Excess return represents manager‟s ability to use information
and talent to generate excess returns.
 Fluctuations in excess returns represent random noise that is
interpreted as unsystematic risk.





Information to noise ratio.
 Annualized IR
ER
b
p
p
R R
= IR
o
÷
p p
IR T = IR
Albert Lee Chun Portfolio Management
32
Information Ratios
Albert Lee Chun Portfolio Management
33
M
2
Measure
Albert Lee Chun Portfolio Management
34
M
2
Measure

Developed by Leah and her grandfather Franco Modigliani.

M
2
= r
p*
- r
m

r
p*
is return of the adjusted portfolio that matches the volatility of the
market index r
m
. It is mixed with a position in T-bills.
If the risk of the portfolio is lower than that of the market, one has to
increase the volatility by using leverage.
Because the market index and the adjusted portfolio have the same
standard deviation, we may compare their performances by
comparing returns.












Albert Lee Chun Portfolio Management
35
M
2
Measure: Example
Managed Portfolio: return = 35% st dev = 42%
Market Portfolio: return = 28% st dev = 30%
T-bill return = 6%
Hypothetical Portfolio:
30/42 = .714 in P (1-.714) or .286 in T-bills
Return = (.714) (.35) + (.286) (.06) = 26.7%
Since the return of the portfolio is less than the market, M
2
is
negative, and the managed portfolio underperformed the market.

17-35
Albert Lee Chun Portfolio Management
36
M
2
of Portfolio P
17-36
17-36
Albert Lee Chun Portfolio Management
37
Excess Returns for Portfolios P and Q and
the Benchmark M
Albert Lee Chun Portfolio Management
38
Performance Statistics
17-38
Albert Lee Chun Portfolio Management
39
Which Portfolio is Best?
 It depends.
 If P or Q represent the entire portfolio, Q would be
preferable based on having higher sharp ratio and a
better M
2
.
 If P or Q represents a sub-portfolio, the Q would be
preferable because it has a higher Treynor ratio.
 For an actively managed portfolio, P may be
preferred because it‟s information ratio is larger (that
is it maximizes return relative to nonsystematic risk,
or the tracking error).
Albert Lee Chun Portfolio Management
40
Style Analysis
Albert Lee Chun Portfolio Management
41
Style Analysis
Introduced by William Sharpe
1992 study of mutual fund performance
 91.5% of variation in return could be
explained by the funds‟ allocations to bills,
bonds and stocks
Later studies show that 97% of the variation in
return could be explained by the funds‟ allocation
to set of different asset classes.
17-41
Albert Lee Chun Portfolio Management
42
Sharpe’s Style Portfolios for the Magellan Fund
17-42
Monthly returns on Magellan
Fund over five year period.
Regression coefficient only
positive for 3.
They explain 97.5% of
Magellan‟s returns.
2.5 percent attributed to
security selection within asset
classes.

Albert Lee Chun Portfolio Management
43
Fidelity Magellan Fund Returns vs Benchmarks
17-43
Fund vs Style and Fund vs SML
Impact of positive
alpha on abnormal
returns.
19.19%
Albert Lee Chun Portfolio Management
44
Average Tracking Error for 636 Mutual Funds
17-44
Bell shaped
Albert Lee Chun Portfolio Management
45
Market Timing
Albert Lee Chun Portfolio Management
46
Perfect Market Timing
 A manager with perfect market timing, that shifts assets
efficiently across stocks, bonds and cash would have a return
equal to
| | 0 , , max
t bt t st t pt
RFR R RFR R RFR R ÷ ÷ + =
Albert Lee Chun Portfolio Management
47
Returns from 1990 - 1999
Year Lg Stocks T-Bills
1990 -3.20 7.86
1991 30.66 5.65
1992 7.71 3.54
1993 9.87 2.97
1994 1.29 3.91
1995 37.71 5.58
1996 23.07 5.58
1998 28.58 5.11
1999 21.04 4.80
17-47
Albert Lee Chun Portfolio Management
48
 Switch to T-Bills in 90 and 94
 Mean = 18.94%,
 Standard Deviation = 12.04%
 Invested in large stocks for the entire period:
 Mean = 17.41%
 Standard Deviation = 14.11
With Perfect Forecasting Ability
17-48
Albert Lee Chun Portfolio Management
49
Performance of Bills, Equities and Timers
Beginning with $1 dollar in 1926, and ending in 2005....
Albert Lee Chun Portfolio Management
50
Value of Imperfect Forecasting
 Suppose you are forecasting rain in Seattle. If you predict rain,
you would be correct most of the time.
 Does this make you a good forecaster? Certainly not.
 We need to examine the proportion of correct forecasts for rain
(P1) and the proportion of correct forecasts for sun (P2).
 The correct measure of timing ability is
P = P1 + P2 – 1
An forecaster who always guesses correctly will show P1 = P2 =
P =1, whereas on who always predicts rain will have P1 = 1,
P2 = P = 0.
Albert Lee Chun Portfolio Management
51
If an investor holds only the market and the risk free security, and
the weights remained constant, the portfolio characteristic line
would be a straight line.

Adjusting portfolio weights for up and down movements in market
returns, we would have:

 Low Market Return - low weight on the market - low ßeta
 High Market Return – high weight on the market - high ßeta
Identifying Market Timing
17-51
Henriksson (1984) showed little evidence of market timing.
Evidence of market efficiency.
Albert Lee Chun Portfolio Management
52
Characteristic Lines: Market Timing
17-52
No Market Timing
Beta Increases with Return
Two Values of Beta
Albert Lee Chun Portfolio Management
53
Testing Market Timing
 The following regression equation, controls for the movements
in bond and stock markets, and captures the superior market
timing of managers






 Gamma was found to be equal to .3 and statistically significant,
suggesting that TAA managers were able to time the markets.
 However, the study also found a negative alpha of -.5.

( ) ( ) ( )
| | { }
t t bt t st
st s t bt b t pt
RFR R RFR R
RFR R RFR R RFR R
c ¸
| | o
+ ÷ ÷ +
÷ + ÷ + = ÷
0 , , max
Albert Lee Chun Portfolio Management
54
Performance Attribution Analysis
Albert Lee Chun Portfolio Management
55
Selectivity
 The basic premise of the Fama method is that overall
performance of a portfolio can be decomposed into a
portfolio risk premium component and a selectivity
component.
 Selectivity is the portion of excess returns that
exceeds that which can be attained by an unmanaged
benchmark portfolio.

 Overall performance = Portfolio Risk Premium +
Selectivity

Albert Lee Chun Portfolio Management
56
Overall performance = Portfolio Risk
Premium+ Selectivity
| | | | | | ) ( R - R R ) ( R = R
x p f x f p p p
R | | + ÷ ÷
Overall
Performance
Portfolio Risk
Premium
Selectivity
Selectivity measures the distance between the
return on portfolio p and the return on a
benchmark portfolio with beta equal to the beta
of portfolio p.
Albert Lee Chun Portfolio Management
57
Attribution Analysis
Portfolio managers add value to their investors by
1) selecting superior securities
2) demonstrating superior market timing skills by
allocating funds to different asset classes or market
segments.

Attribution analysis attempts to distinguish is the source
of the portfolio‟s overall performance.
Total value added performance is the sum of selection
and allocation effects.

Albert Lee Chun Portfolio Management
58
) (
&
1
1 1
1 1
Bi Bi
n
i
pi pi
n
i
Bi Bi
n
i
pi pi B p
n
i
pi pi p
n
i
Bi Bi B
r w r w
r w r w r r
r w r r w r
÷
= ÷ = ÷
= =
¿
¿ ¿
¿ ¿
=
= =
= =
Where B is the bogey portfolio and p is the managed portfolio.
Formula for Attribution
17-58
Set up a „Benchmark‟ or „Bogey‟ portfolio
Albert Lee Chun Portfolio Management
59
Allocation Effect
 Asset Allocation Effect




 Captures the manager‟s decision to over or
underweight a particular market segment i.
 Overweighting a segment i when the benchmark yield
is high is rewarded.




( ) ( ) | |
Bi Bi Pi i
r w w × ÷ E =
Albert Lee Chun Portfolio Management
60
Selection Effect
 Security Selection Effect



 Captures the stock picking ability of the manager, and
rewards the ability to form specific market segment
portfolios. Rewards the manger for placing larger weights
on those segments where his portfolio outperforms the
benchmark portfolio in that particular segment.









( ) ( ) | |
Bi Pi Pi i
r r w ÷ × E =
Albert Lee Chun Portfolio Management
61
Performance of the Managed Portfolio
17-61
Albert Lee Chun Portfolio Management
62
Performance Attribution
17-62
Albert Lee Chun Portfolio Management
63
Sector Selection within the Equity Market


17-63
Albert Lee Chun Portfolio Management
64
Portfolio Attribution: Summary


17-64
Albert Lee Chun Portfolio Management
65
Global Benchmark Problem
(Optional)
Albert Lee Chun Portfolio Management
66
Benchmark Error
 Market portfolio is difficult to approximate
 Benchmark error
 can effect slope of SML
 can effect calculation of Beta
 greater concern with global investing
 problem is one of measurement
 Note: Sharpe measure not as dependent on market
portfolio as the Treynor measure and others relying
on Beta.

Albert Lee Chun Portfolio Management
67
Differences in Betas
 Two major differences in the various beta statistics:
 For any particular stock, the beta estimates change
a great deal over time.
 There are substantial differences in betas estimated
for the same stock over the same time period when
two different definitions of the benchmark
portfolio are employed.

Albert Lee Chun Portfolio Management
68
Albert Lee Chun Portfolio Management
69
Global Benchmark Problem - SML
Albert Lee Chun Portfolio Management
70
Global Benchmark Problem - SML
Albert Lee Chun Portfolio Management
71
Global Benchmark Problem - SML
Albert Lee Chun Portfolio Management
72
Bond Portfolio Performance Measures
(Optional)
Albert Lee Chun Portfolio Management
73
Bond Portfolio Measures
 Returns-Based Bond Performance Measurement
 Early attempts to analyze fixed-income performance
involved peer group comparisons
 Peer group comparisons are potentially flawed because
they do not account for investment risk directly.
 How did the performance levels of portfolio managers
compare to the overall bond market?
 What factors lead to superior or inferior bond-portfolio
performance?


Albert Lee Chun Portfolio Management
74
Fama-French Measure
 Fama and French extended their 3-factor equity pricing
model with 2 additional factors to account for the return
characteristics of bonds




 TERM – captures the term premium in the slope of the yield
curve.
 DEF – captures the default premium in the credit spread
between corporate bonds and treasuries.
 These two bond factors are the dominate drivers of bond
portfolio returns.

( )
jt t j j1 mt t j2 t j3 t j4 t j4 t jt
R - RFR = α + b R - RFR + b SMB + b HML + b TERM + b DEF + e ( (
¸ ¸ ¸ ¸
Albert Lee Chun Portfolio Management
75
Seven Bond Portfolios
Albert Lee Chun Portfolio Management
76
Bond Performance Attribution
 A Bond Market Line
 Need a measure of risk such as beta coefficient for
equities
 Difficult to achieve due to bond maturity and
coupon effect on volatility of prices
 Composite risk measure is the bond‟s duration
 Duration replaces beta as risk measure in a bond
market line

Albert Lee Chun Portfolio Management
77
Bond Market Line
Albert Lee Chun Portfolio Management
78
Albert Lee Chun Portfolio Management
79
That’s all for today!