You are on page 1of 82

Topic 10 (Ch.

24)
Portfolio Performance Evaluation
Measuring investment returns
The conventional theory of performance
evaluation
Market timing
Performance attribution procedures

MeasuringInvestmentReturns
One period:
total proceeds
income capital gain
rate of return

intial investment
initial investment

Find the rate of return (r) that equates the present


value of all cash flows from the investment with the
initial outlay.

Example:
Consider a stock paying a dividend of $2 annually
that currently sells for $50.
You purchase the stock today and collect the $2
dividend, and then you sell the stock for $53 at yearend.

2 ( 53 50)
rate of return
10%
50
2 53
50
r 10%
1 r
3

Multiperiod:

Arithmetic versus geometric averages:

Arithmetic averages:

0.1189 0.221 0.2869 0.1088 0.0491


rA
0.021
5
4

Geometric averages:
The compound average growth rate, rG, is calculated
as the solution to the following equation:
(1 rG )5
(1 0.1189 )(1 0.221)(1 0.2869)(1 0.1088)(1 0.0491)

1.0275

rG 0.0054

In general:

(1 rG ) n (1 r1 )(1 r2 )...(1 rn )
where rt is the return in each time period.
5

Geometric averages never exceed arithmetic ones:


Consider a stock that doubles in price in period 1 (r1
= 100%) and halves in price in period 2 (r2 = -50%).
The arithmetic average is:
rA = [100 + (-50)]/2 = 25%
The geometric average is:
rG = [(1 + 1)(1 - 0.5)]1/2 1 = 0

The effect of the -50% return in period 2 fully offsets


the 100% return in period 1 in the calculation of the
geometric average, resulting in an average return of
zero.
This is not true of the arithmetic average.
In general, the bad returns have a greater influence
on the averaging process in the geometric technique.
Therefore, geometric averages are lower.

Generally, the geometric average is preferable for


calculation of historical returns (i.e. measure of past
performance), whereas the arithmetic average is
more appropriate for forecasting future returns:
Example 1:
Consider a stock that will either double in value (r =
100%) with probability of 0.5, or halve in value (r =
-50%) with probability 0.5.

InvestmentOutcome
Double
Halve

FinalValueofEach

DollarInvested
One-YearRateofReturn
$2.00
100%
$0.50
-50%
8

Suppose that the stocks performance over a 2-year


period is characteristic of the probability distribution,
doubling in one year and halving in the other.
The stocks price ends up exactly where it started, and
the geometric average annual return is zero:
2

(1 rG ) (1 100%)(1 50%) 1
rG 0
which confirms that a zero year-by-year return would
have replicated the total return earned on the stock.
9

However, the expected annual future rate of return on


the stock is not zero.
It is the arithmetic average of 100% and -50%:
(100 - 50)/2 = 25%.
There are two equally likely outcomes per dollar
invested: either a gain of $1 (when r = 100%) or a loss
of $0.50 (when r = -50%).
The expected profit is ($1 - $0.50)/2 = $0.25, for a 25%
expected rate of return.
The profit in the good year more than offsets the loss
in the bad year, despite the fact that the geometric
return is zero. The arithmetic average return thus
provides the best guide to expected future returns.
10

Example 2:
Consider all the possible outcomes over a two-year
period:

InvestmentOutcome
Double,double
Double,halve
Halve,double
Halve,halve

FinalValueofEach
DollarInvested
$4.00
$1.00
$1.00
$0.25

TotalReturn
overTwoYears
300%
0%
0%
-75%

11

The expected final value of each dollar invested is:


(4 + 1 + 1 + 0.25)/4 = $1.5625
for two years, again indicating an average rate of
return of 25% per year, equal to the arithmetic
average.
Note that an investment yielding 25% per year with
certainty will yield the same final compounded value
as the expected final value of this investment:
(1 + 0.25)2 = 1.5625.

12

The arithmetic average return on the stock is:


[300 + 0 + 0 + (-75)]/4 = 56.25%
per two years, for an effective annual return of 25%
since:
(1 + 25%)(1 + 25%) 1 = 56.25%.
In contrast, the geometric mean return is zero since:
[(1 + 3)(1 + 0)(1 + 0)(1 0.75)]1/4 = 1.0
Again, the arithmetic average is the better guide to
future performance.
13

Dollar-weighted
returns:

returns

versus

time-weighted

Example:

Time
0
1

1
2

Outlay
$50topurchasefirstshare
$53topurchasesecondshareayearlater

Proceeds
$2dividendfrominitiallypurchasedshare
$4dividendfromthe2sharesheldinthe
secondyear,plus$108receivedfromselling
bothsharesat$54each

14

Dollar-weighted returns:
Using the discounted cash flow (DCF) approach,
we can solve for the average return over the two
years by equating the present values of the cash
inflows and outflows:

53
2
112
50

1 r 1 r (1 r ) 2

r 7.117 %
15

This value is called the internal rate of return, or


the dollar-weighted rate of return on the
investment.
It is dollar weighted because the stocks
performance in the second year, when two shares
of stock are held, has a greater influence on the
average overall return than the first-year return,
when only one share is held.

16

Time-weighted returns:
Ignore the number of shares of stock held in each
period.
The stock return in the 1st year:

2 ( 53 50)
r1
10%
50
The stock return in the 2nd year:

2 ( 54 53)
r2
5.66%
53
17

The time-weighted (geometric average) return


is:

rG [(1 10%)(1 5.66%)]1 / 2 1 7.81%

This average return considers only the periodby-period returns without regard to the amounts
invested in the stock in each period.
Note that the dollar-weighted average is less than
the time-weighted average in this example
because the return in the second year, when
more money is invested, is lower.
18

Note:
For an investor that has control over
contributions to the investment portfolio, the
dollar-weighted return is more comprehensive
measure.
Time-weighted returns are more likely
appropriate to judge the performance of an
investor that does not control the timing or the
amount of contributions.

19

TheConventionalTheoryof
PerformanceEvaluation
Several risk-adjusted performance measures:
Sharpes measure:

rP r f
P

Sharpes measure divides average portfolio excess


return over the sample period by the standard
deviation of returns over that period.
It measures the reward to (total) volatility trade-off.
Note: The risk-free rate may not be constant over
the measurement period, so we are taking a sample
average, just as we do for rP.
20

Treynors measure:

rP r f

Like Sharpes, Treynors measure gives excess


return per unit of risk, but it uses systematic risk
instead of total risk.
Jensens measure:

P r P [r f P ( r M r f )]

Jensens measure is the average return on the


portfolio over and above that predicted by the
CAPM, given the portfolios beta and the average
market return.
Jensens measure is the portfolios alpha value.
21

Information ratio:

P
(eP )

The information ratio divides the alpha of the


portfolio by the nonsystematic risk of the portfolio.
It measures abnormal return per unit of risk that in
principle could be diversified away by holding a
market index portfolio.
Note:
Each measure has some appeal.
But each does not necessarily provide consistent
assessments of performance, since the risk
measures used to adjust returns differ substantially.
22

Example: Consider the following data for a particular


sample period:
Portfolio P

Market M

Average return

35%

28%

Beta

1.20

1.00

Standard deviation

42%

30%

Nonsystematic risk, (e)

18%

The T-bill rate during the period was 6%.

23

Sharpes measure:

SP

rP r f

35 6

0.69
42

28 6
SM
0.73
30
Treynors measure:

TP

rP r f

35 6

24.2
1 .2

28 6
TM
22
1
24

Jensens measure:

P r P [r f P ( r M r f )] 35 [6 1.2( 28 6)] 2.6%


M r M [ r f 1( r M r f )] 0
Information ratio:

P
2 .6
Portfolio P :

0.144
(e P ) 18

Market : 0
25

The M2 measure of performance


While the Sharpe ratio can be used to rank
portfolio performance, its numerical value is not
easy to interpret.
We have found that SP = 0.69 and SM = 0.73.
This suggests that portfolio P under-performed the
market index.
But is a difference of 0.04 in the Sharpe ratio
economically meaningful?
We often compare rates of return, but these ratios
are difficult to interpret.
26

To compute the M2 measure, we imagine that a


managed portfolio, P, is mixed with a position in Tbills so that the complete, or adjusted, portfolio
(P*) matches the volatility of a market index (such
as the S&P500).
Because the market index and portfolio P* have the
same standard deviation, we may compare their
performance simply by comparing returns.
This is the M2 measure:

M rP * rM

27

Example:
P has a standard deviation of 42% versus a market
standard deviation of 30%.
The adjusted portfolio P* would be formed by mixing
portfolio P and T-bills and :
weight in P: 30/42 = 0.714
weight in T-bills: (1 - 0.714) = 0.286.
The return on this portfolio P* would be:
(0.286 6%) + (0.714 35%) = 26.7%
Thus, portfolio P has an M2 measure:
26.7 28 = -1.3%.
28

29

We move down the capital allocation line


corresponding to portfolio P (by mixing P with Tbills) until we reduce the standard deviation of the
adjusted portfolio to match that of the market index.
The M2 measure is then the vertical distance (i.e., the
difference in expected returns) between portfolios P*
and M.
P will have a negative M2 measure when its capital
allocation line is less steep than the capital market
line (i.e., when its Sharpe ratio is less than that of the
market index).
30

Appropriate performance measures


in 3 scenarios
Suppose that Jane constructs a portfolio (P) and
holds it for a considerable period of time.
She makes no changes in portfolio composition
during the period.
In addition, suppose that the daily rates of return on
all securities have constant means, variances, and
covariances. This assures that the portfolio rate of
return also has a constant mean and variance.
We want to evaluate the performance of Janes
portfolio.
31

Jane's portfolio P represents her entire risky


investment fund:
We need to ascertain only whether Janes portfolio
has the highest Sharpe measure.
We can proceed in 3 steps:
Assume that past security performance is
representative of expected performance, meaning
that realized security returns over Janes holding
period exhibit averages and covariances similar to
those that Jane had anticipated.
32

Determine the benchmark (alternative) portfolio


that Jane would have held if she had chosen a
passive strategy, such as the S&P 500.
Compare Janes Sharpe measure to that of the
best portfolio.
In sum:
When Janes portfolio represents her entire
investment fund, the benchmark is the market
index or another specific portfolio.
The performance criterion is the Sharpe measure
of the actual portfolio versus the benchmark.
33

Janes portfolio P is an active portfolio and is mixed


with the market-index portfolio M:
When the two portfolios are mixed optimally, the
square of the Sharpe measure of the complete
portfolio, C, is given by:

SC2

2
SM

P 2
[
]
(eP )

where P is the abnormal return of the active


portfolio relative to the market-index, and (eP) is
the diversifiable risk.
34

The ratio P/ (eP) is thus the correct performance


measure for P in this case, since it gives the
improvement in the Sharpe measure of the overall
portfolio.
To see this result intuitively, recall the single-index
model:
rP r f P P ( rM r f ) e P
If P is fairly priced, then P = 0, and eP is just
diversifiable risk that can be avoided.

35

However, if P is mispriced, P no longer equals zero.


Instead, it represents the expected abnormal return.
Holding P in addition to the market portfolio thus
brings a reward of P against the nonsystematic risk
voluntarily incurred, (eP).
Therefore, the ratio of P / (eP) is the natural
benefit-to-cost ratio for portfolio P.
This performance measurement is the information
ratio.
36

Janes choice portfolio P is one of many portfolios


combined into a large investment fund:
The Treynor measure is the appropriate criterion.
E.g.:

Beta
Excess return

(r r f )

Alpha*

Portfolio P

Portfolio Q

Market

0.90

1.60

1.00

11%

19%

10%

2%

3%

* alpha excess return [beta market excess return]


( r r f ) ( r M r f ) r [ r f ( r M r f )]
37

38

Note:
We plot P and Q in the expected return-beta
(rather than the expected return-standard
deviation) plane, because we assume that P
and Q are two of many sub-portfolios in the
fund, and thus that nonsystematic risk will be
largely diversified away, leaving beta as the
appropriate risk measure.

39

Suppose portfolio Q can be mixed with T-bills.


Specifically, if we invest wQ in Q and wF = 1 - wQ in
T-bills, the resulting portfolio, Q*, will have alpha
and beta values proportional to Qs alpha and beta
scaled down by wQ:

Q* wQ Q

Q* wQ Q

Thus, all portfolios Q* generated from mixing Q


with T-bills plot on a straight line from the origin
through Q.
We call it the T-line for the Treynor measure,
which is the slope of this line.
40

P has a steeper T-line.


Despite its lower alpha, P is a better portfolio after
all.
For any given beta, a mixture of P with T-bills will
give a better alpha than a mixture of Q with Tbills.

41

Suppose that we choose to mix Q with T-bills to


create a portfolio Q* with a beta equal to that of P.
We find the necessary proportion by solving for wQ:

Q* wQ Q 1.6 wQ P 0.9
wQ 9 / 16
Portfolio Q* has an alpha of:

Q* wQ Q ( 9 / 16) 3 1.69%
which is less than that of P.
42

In other words, the slope of the T-line is the


appropriate performance criterion for this case.
The slope of the T-line for P, denoted by TP, is:

TP

rP r f

Treynors performance measure is appealing because


when an asset is part of a large investment portfolio,
one should weigh its mean excess return against its
systematic risk rather than against total risk to
evaluate contribution to performance.
43

An example:
Excess returns for portfolios P & Q and the
benchmark M over 12 months:

44

Performance statistics:

45

Portfolio Q is more aggressive than P, in the sense


that its beta is significantly higher (1.40 vs. 0.70).
On the other hand, from its residual standard
deviation P appears better diversified (2.02% vs.
9.81%).
Both portfolios outperformed the benchmark
market index, as is evident from their larger
Sharpe measures (and thus positive M2) and their
positive alphas.

46

Which portfolio is more attractive based on


reported performance?
If P or Q represents the entire investment fund, Q
would be preferable on the basis of its higher
Sharpe measure (0.49 vs. 0.43) and better M2
(2.66% vs. 2.16%).
As an active portfolio to be mixed with the market
index, P is preferable to Q, as is evident from its
information ratio (0.81 vs. 0.54).

47

When P and Q are competing for a role as one of a


number of subportfolios, Q dominates again
because its Treynor measure is higher (5.38 versus
3.97).
Thus, the example illustrates that the right way to
evaluate a portfolio depends in large part how the
portfolio fits into the investors overall wealth.

48

Relationships among the various


performance measures
The relation between Treynors measure and
Jensens :

TP

E ( rP ) r f

E ( rP ) [r f p ( E ( rM ) r f )] p ( E ( rM ) r f )

P
P
P

[ E ( rM ) r f ]
TM
P
P
49

The relation between Sharpes measure and Jensens


:

SP

E ( rP ) r f

E ( rP ) [r f p ( E ( rM ) r f )] p ( E ( rM ) r f )

P
2
P M

P M
cov( rP , rM )
PM

P M
P M
P

P PM

P
M
50

P p ( E ( rM ) r f )
SP

P
P
P PM

[ E ( rM ) r f ]
P M
P

PM S M
P

51

Performance measurement
with changing portfolio composition
Estimating various statistics from a sample period
assuming a constant mean and variance may lead to
substantial errors.
Example:
Suppose that the Sharpe measure of the market
index is 0.4.
Over an initial period of 52 weeks, the portfolio
manager executes a low-risk strategy with an
annualized mean excess return of 1% and standard
deviation of 2%.
52

This makes for a Sharpe measure of 0.5, which


beats the passive strategy.
Over the next 52-week period this manager finds
that a high-risk strategy is optimal, with an
annual mean excess return of 9% and standard
deviation of 18%.
Here, again, the Sharpe measure is 0.5.
Over the two-year period our manager maintains
a better-than-passive Sharpe measure.

53

Portfolio returns in last four quarters are more


variable than in the first four:

54

In the first 4 quarters, the excess returns are -1%,


3%, -1%, and 3%, making for an average of 1%
and standard deviation of 2%.
In the next 4 quarters the returns are -9%, 27%,
-9%, 27%, making for an average of 9% and
standard deviation of 18%.
Thus both years exhibit a Sharpe measure of 0.5.
However, over the 8-quarter sequence the mean and
standard deviation are 5% and 13.42%,
respectively, making for a Sharpe measure of only
0.37, apparently inferior to the passive strategy.
55

The shift of the mean from the first 4 quarters to


the next was not recognized as a shift in strategy.
Instead, the difference in mean returns in the two
years added to the appearance of volatility in
portfolio returns.
The active strategy with shifting means appears
riskier than it really is and biases the estimate of the
Sharpe measure downward.
We conclude that for actively managed portfolios it
is helpful to keep track of portfolio composition and
changes in portfolio mean and risk.
56

MarketTiming
Market timing involves shifting funds between a
market-index portfolio and a safe asset (such as
T-bills or a money market fund), depending on
whether the market as a whole is expected to
outperform the safe asset.
In practice, most managers do not shift fully, but
partially, between T-bills and the market.

57

Suppose that an investor holds only the marketindex portfolio and T-bills.
If the weight of the market were constant, say, 0.6,
then portfolio beta would also be constant, and the
security characteristic line (SCL) would plot as a
straight line with slope 0.6.

58

No market timing, beta is constant:

59

If the investor could correctly time the market and


shift funds into it in periods when the market does
well.
If bull and bear markets can be predicted, the
investor will shift more into the market when the
market is about to go up.
The portfolio beta and the slope of the SCL will be
higher when rM is higher, resulting in the curved
line.

60

Market timing, beta increases with expected


market excess return:

61

Such a line can be estimated by adding a squared


term to the usual linear index model:
rP r f a b( rM r f ) c ( rM r f ) 2 e P

where rP is the portfolio return, and a, b, and c are


estimated by regression analysis.
If c turns out to be positive, we have evidence of
timing ability, because this last term will make the
characteristic line steeper as (rM - rf) is larger.
62

A similar and simpler methodology suggests that the


beta of the portfolio take only two values: a large
value if the market is expected to do well and a
small value otherwise.

63

Such a line appears in regression form as:


rP r f a b( rM r f ) c ( rM r f ) D e P

where D is a dummy variable that equals 1 for rM > rf


and zero otherwise.
Hence, the beta of the portfolio is b in bear markets
and b + c in bull markets.
Again, a positive value of c implies market timing
ability.
64

Example:
Regressing the excess returns of portfolios P and Q on
the excess returns of M and the square of these
returns:

rP r f a P bP ( rM r f ) c P ( rM r f ) 2 e P
rQ r f aQ bQ ( rM r f ) cQ ( rM r f ) 2 eQ
we derive the following statistics:

65

The numbers in parentheses are the regression estimates


from the single variable regression (reported in Table
24.3).
66

Portfolio P shows no timing (cP = 0).


The results for portfolio Q reveal that timing has, in
all likelihood, successfully been attempted (cQ = 0.10).
The evidence thus suggests successful timing (positive
c) offset by unsuccessful stock selection (negative a).
Note that the alpha estimate, a, is now -2.29% as
opposed to the 5.26% estimate derived from the
regression equation that did not allow for the
possibility of timing activity.

67

PerformanceAttributionProcedures
Portfolio managers constantly make broad-brush
asset allocation decisions as well as more detailed
sector and security allocation decisions within asset
class.
Performance attribution studies attempt to decompose
overall performance into discrete components that
may be identified with a particular level of the
portfolio selection process.

68

The difference between a managed portfolios


performance and that of a benchmark portfolio
then may be expressed as the sum of the
contributions to performance of a series of
decisions made at the various levels of the portfolio
construction process.
For example, one common attribution system
decomposes performance into 3 components:
broad asset market allocation choices across
equity, fixed-income, and money markets.
industry (sector) choice within each market.
security choice within each sector.
69

The attribution method explains the difference in


returns between a managed portfolio, P, and a
selected benchmark portfolio, B (called the bogey).
Suppose that the universe of assets for P and B
includes n asset classes such as equities, bonds, and
bills.
For each asset class, a benchmark index portfolio is
determined.
For example, the S&P 500 may be chosen as
benchmark for equities.

70

The bogey portfolio is set to have fixed weights in


each asset class, and its rate of return is given by:
n

rB wBi rBi
i 1

where wBi: weight of the bogey in asset class i.


rBi: return on the benchmark portfolio of that
class over the evaluation period.

71

The portfolio managers choose weights in each


class (wPi) based on their capital market
expectations, and they choose a portfolio of the
securities within each class based on their security
analysis, which earns rPi over the evaluation
period.
Thus, the return of the managed portfolio will be:
n

rP wPi rPi
i 1

72

The difference between the two rates of return is:


n

i 1

i 1

i 1

rP rB wPi rPi wBi rBi ( wPi rPi wBi rBi )


We can decompose each term of the summation into
a sum of two terms as follows:
Contribution from asset allocation: ( wPi wBi ) rBi
+ Contribution from selection:

wPi ( rPi rBi )

= Total contribution from asset class i: wPi rPi wBi rBi


73

74

Example:
Consider the attribution results for a portfolio
which invests in stocks, bonds, and money market
securities.
The managed portfolio is invested in the equity,
fixed-income, and money markets with weights of
70%, 7%, and 23%, respectively.
The portfolio return over the month is 5.34%.

75

Bogey Performance and Excess Return

Component

Benchmark Weight

Return of Index
during Month (%)

Equity (S&P 500)

0.60

5.81

Bonds (Barclays
Aggregate Bond
Index)

0.30

1.45

Cash (money market )

0.10

0.48

Bogey = (0.60 5.81 ) + (0.30 1.45) + (0.10 0.48) = 3.97%

Return of managed portfolio

5.34%

- Return of bogey portfolio

3.97%

Excess return of managed portfolio

1.37%
76

Note:
The bogey portfolio is comprised of investments in
each index with the following weights:
60%: equity
30%: fixed income
10%: cash (money market securities).
These weights are designated as neutral or usual.
They depend on the risk tolerance of the investor and
must be determined in consultation with the client.
77

This would be considered a passive asset-market


allocation.
Any deviation from these weights must be justified
by a belief that one or another market will either
over- or underperform its usual risk-return
profile.

78

A. Contribution of Asset Allocation to


Performance

(1)

(2)

Actual

Benchmark

Weight in

Weight in

Excess

Market

Performance

Market

Market

Market

Weight

Return
(%)

(%)

Equity

0.70

0.60

0.10

5.81

0.5810

Fixed-income

0.07

0.30

-0.23

1.45

-0.3335

Cash

0.23

0.10

0.13

0.48

0.0624

(3)

(5) = (3) (4)

(4)

Contribution to

Contribution of asset allocation

0.3099

79

B. Contribution of Selection to Total Performance

(1)

(2)

(3)

Portfolio

Index

Excess

Performance Performance Performance

(4)

(5) = (3) (4)

Portfoli
o

Contribution

Market

(%)

(%)

(%)

Weight

(%)

Equity

7.28

5.81

1.47

0.70

1.03

0.44

0.07

0.03

1.06

Fixed1.89
1.45
income
Contribution of selection
within
markets

80

Sector selection within the equity market:

81

Portfolio attribution summary:

82