Chapter 24 - Portfolio Performance Evaluation

CHAPTER 24: PORTFOLIO PERFORMANCE EVALUATION

PROBLEM SETS
1.

As established in the following result from the text, the Sharpe ratio depends on both
alpha for the portfolio ( P) and the correlation between the portfolio and the market
index ():
E (rP r f )

P
+ S M
P

Specifically, this result demonstrates that a lower correlation with the market index
reduces the Sharpe ratio. Hence, if alpha is not sufficiently large, the portfolio is inferior
to the index. Another way to think about this conclusion is to note that, even for a
portfolio with a positive alpha, if its diversifiable risk is sufficiently large, thereby
reducing the correlation with the market index, this can result in a lower Sharpe ratio.
2.

The IRR (i.e., the dollar-weighted return) can not be ranked relative to either the
geometric average return (i.e., the time-weighted return) or the arithmetic average
return. Under some conditions, the IRR is greater than each of the other two averages,
and similarly, under other conditions, the IRR can also be less than each of the other
averages. A number of scenarios can be developed to illustrate this conclusion. For
example, consider a scenario where the rate of return each period consistently increases
over several time periods. If the amount invested also increases each period, and then
all of the proceeds are withdrawn at the end of several periods, the IRR is greater than
either the geometric or the arithmetic average because more money is invested at the
higher rates than at the lower rates. On the other hand, if withdrawals gradually reduce
the amount invested as the rate of return increases, then the IRR is less than each of the
other averages. (Similar scenarios are illustrated with numerical examples in the text,
on page 824, where the IRR is shown to be less than the geometric average, and in
Concept Check 1, where the IRR is greater than the geometric average.)

3.

It is not necessarily wise to shift resources to timing at the expense of security selection.
There is also tremendous potential value in security analysis. The decision as to whether
to shift resources has to be made on the basis of the macro, compared to the micro,
forecasting ability of the portfolio management team.

24-1

Chapter 24 - Portfolio Performance Evaluation

4.

a.

Arithmetic average: r ABC = 10% ; r XYZ = 10%

b.

Dispersion: ABC = 7.07%; XYZ = 13.91%

Stock XYZ has greater dispersion.
(Note: We used 5 degrees of freedom in calculating standard deviations.)

c.

Geometric average:
rABC = (1.20 1.12 1.14 1.03 1.01)1/5 1 = 0.0977 = 9.77%
rXYZ = (1.30 1.12 1.18 1.00 0.90)1/5 1 = 0.0911 = 9.11%
Despite the fact that the two stocks have the same arithmetic average, the
geometric average for XYZ is less than the geometric average for ABC. The
reason for this result is the fact that the greater variance of XYZ drives the
geometric average further below the arithmetic average.

5.

d.

In terms of forward looking statistics, the arithmetic average is the

better estimate of expected rate of return. Therefore, if the data reflect the
probabilities of future returns, 10% is the expected rate of return for both
stocks.

a.

Time-weighted average returns are based on year-by-year rates of return:

Year

Return = (capital gains + dividend)/price

2005 2006
2006 2007
2007 2008

[($120 $100) + $4]/$100 = 24.00%

[($90 $120) + $4]/$120 = 21.67%
[($100 $90) + $4]/$90 = 15.56%

Arithmetic mean: (24% 21.67% + 15.56%)/3 = 5.96%

Geometric mean: (1.24 0.7833 1.1556)1/3 1 = 0.0392 = 3.92%

24-2

Chapter 24 - Portfolio Performance Evaluation

b.
Cash
Date
Flow
1/1/05 $300
1/1/06 $228
1/1/07 $110
1/1/08 $416

Explanation
Purchase of three shares at $100 each
Purchase of two shares at $120 less dividend income on three shares held
Dividends on five shares plus sale of one share at $90
Dividends on four shares plus sale of four shares at $100 each
416

110

Date: 1/1/05

1/1/06

1/1/07

1/1/08

228
300
Dollar-weighted return = Internal rate of return = 0.1607%
6.
Time
0
1
2
3
a.

Cash flow
3($90) = $270
$100
$100
$100

Holding period return

(10090)/90 = 11.11%
0%
0%

Time-weighted geometric average rate of return =

(1.1111 1.0 1.0)1/3 1 = 0.0357 = 3.57%

b.

Time-weighted arithmetic average rate of return = (11.11% + 0 + 0)/3 = 3.70%

The arithmetic average is always greater than or equal to the geometric average;
the greater the dispersion, the greater the difference.

c.

Dollar-weighted average rate of return = IRR = 5.46%

[Using a financial calculator, enter: n = 3, PV = 270, FV = 0, PMT = 100. Then
compute the interest rate.] The IRR exceeds the other averages because the
investment fund was the largest when the highest return occurred.

24-3

Chapter 24 - Portfolio Performance Evaluation

7.

a.

The alphas for the two portfolios are:

A = 12% [5% + 0.7(13% 5%)] = 1.4%
B = 16% [5% + 1.4(13% 5%)] = 0.2%
Ideally, you would want to take a long position in Portfolio A and a short position
in Portfolio B.

b.

If you will hold only one of the two portfolios, then the Sharpe measure is the
appropriate criterion:

SA =

12 5
= 0.583
12

SB =

16 5
= 0.355
31

Using the Sharpe criterion, Portfolio A is the preferred portfolio.

8.
a.
(i)
(ii)
(iii
)

Alpha = regression intercept

Information ratio = P /(eP)
*Sharpe measure = (rP rf)/P

(iv) **Treynor measure = (rP rf )/P

Stock A
1.0%
0.0971

Stock B
2.0%
0.1047

0.4907
8.833

0.3373
10.500

* To compute the Sharpe measure, note that for each stock, (rP rf ) can be
computed from the right-hand side of the regression equation, using the assumed
parameters rM = 14% and rf = 6%. The standard deviation of each stocks returns is
given in the problem.
** The beta to use for the Treynor measure is the slope coefficient of the
regression equation presented in the problem.
b.

(i) If this is the only risky asset held by the investor, then Sharpes measure is the
appropriate measure. Since the Sharpe measure is higher for Stock A, then A is the
best choice.
(ii) If the stock is mixed with the market index fund, then the contribution to the
overall Sharpe measure is determined by the appraisal ratio; therefore, Stock B is
preferred.
(iii) If the stock is one of many stocks, then Treynors measure is the appropriate
measure, and Stock B is preferred.

24-4

Chapter 24 - Portfolio Performance Evaluation

9.

We need to distinguish between market timing and security selection abilities. The
intercept of the scatter diagram is a measure of stock selection ability. If the
manager tends to have a positive excess return even when the markets
performance is merely neutral (i.e., has zero excess return), then we conclude that
the manager has on average made good stock picks. Stock selection must be the
source of the positive excess returns.
Timing ability is indicated by the curvature of the plotted line. Lines that become
steeper as you move to the right along the horizontal axis show good timing ability.
The steeper slope shows that the manager maintained higher portfolio sensitivity to
market swings (i.e., a higher beta) in periods when the market performed well. This
ability to choose more market-sensitive securities in anticipation of market upturns is
the essence of good timing. In contrast, a declining slope as you move to the right
means that the portfolio was more sensitive to the market when the market did poorly
and less sensitive when the market did well. This indicates poor timing.
We can therefore classify performance for the four managers as follows:

A.
B.
C.
D.
10. a.

b.

Selection
Ability
Bad
Good
Good
Bad

Timing Ability
Good
Good
Bad
Bad

Bogey: (0.60 2.5%) + (0.30 1.2%) + (0.10 0.5%) = 1.91%

Actual: (0.70 2.0%) + (0.20 1.0%) + (0.10 0.5%) = 1.65%
Underperformance:
0.26%
Security Selection:

Market
Equity
Bonds
Cash

(1)
Differential return
within market
(Manager index)

(2)

(3) = (1) (2)

Manager's
Contribution to
portfolio weight
performance

0.5%
0.70
0.2%
0.20
0.0%
0.10
Contribution of security selection:

24-5

0.35%
0.04%
0.00%
0.39%

Chapter 24 - Portfolio Performance Evaluation

c.

Asset Allocation:
Market
Equity
Bonds
Cash

(1)
Excess weight
(Manager benchmark)

(2)
Index
Return

(3) = (1) (2)

Contribution to
performance

0.10%
2.5%
0.10%
1.2%
0.00%
0.5%
Contribution of asset allocation:

0.25%
0.12%
0.00%
0.13%

Summary:
Security selection 0.39%
Asset allocation
0.13%
Excess performance 0.26%
11.

a.

Manager return:
(0.30 20) + (0.10 15) + (0.40 10) + (0.20 5) = 12.50%
Benchmark (bogey): (0.15 12) + (0.30 15) + (0.45 14) + (0.10 12) = 13.80%
Added value:
1.30%

b.

Added value from country allocation:

Country
U.K.
Japan
U.S.
Germany

c.

(1)
Excess weight
(Manager benchmark)

(2)
Index Return
minus bogey

0.15%
1.8%
0.20%
1.2%
0.05%
0.2%
0.10%
1.8%
Contribution of country allocation:

(3) = (1) (2)

Contribution to
performance
0.27%
0.24%
0.01%
0.18%
0.70%

Added value from stock selection:

Country
U.K.
Japan
U.S.
Germany

(1)
Differential return
within country
(Manager Index)

(2)

(3) = (1) (2)

Managers
country weight

Contribution to
performance

8%
0.30%
0%
0.10%
4%
0.40%
7%
0.20%
Contribution of stock selection:

Summary:
Country allocation 0.70%
Stock selection
0.60%
Excess performance 1.30%

24-6

2.4%
0.0%
1.6%
1.4%
0.6%

Chapter 24 - Portfolio Performance Evaluation

12. Support: A manager could be a better performer in one type of circumstance than in
another. For example, a manager who does no timing, but simply maintains a high beta,
will do better in up markets and worse in down markets. Therefore, we should observe
performance over an entire cycle. Also, to the extent that observing a manager over an
entire cycle increases the number of observations, it would improve the reliability of
the measurement.
Contradict: If we adequately control for exposure to the market (i.e., adjust for beta),
then market performance should not affect the relative performance of individual
managers. It is therefore not necessary to wait for an entire market cycle to pass before
evaluating a manager.
13. The use of universes of managers to evaluate relative investment performance does, to
some extent, overcome statistical problems, as long as those manager groups can be
made sufficiently homogeneous with respect to style.
14. a.
b.

15.

The managers alpha is: 10% [6% + 0.5(14% 6%)] = 0

From Black-Jensen-Scholes and others, we know that, on average, portfolios with
low beta have historically had positive alphas. (The slope of the empirical security
market line is shallower than predicted by the CAPM.) Therefore, given the
managers low beta, performance might actually be sub-par despite the estimated
alpha of zero.

This exercise is left to the student; answers will vary.

CFA PROBLEMS

1.

a.

Manager A
Strength. Although Manager As one-year total return was somewhat below the
international index return (6.0 percent versus 5.0 percent), this manager
apparently has some country/security return expertise. This large local market
return advantage of 2.0 percent exceeds the 0.2 percent return for the international
index.
Weakness. Manager A has an obvious weakness in the currency management area.
This manager experienced a marked currency return shortfall, with a return of 8.0
percent versus 5.2 percent for the index.

24-7

Chapter 24 - Portfolio Performance Evaluation

Manager B
Strength. Manager Bs total return exceeded that of the index, with a marked
positive increment apparent in the currency return. Manager B had a 1.0 percent
currency return compared to a 5.2 percent currency return on the international
index. Based on this outcome, Manager Bs strength appears to be expertise in the
currency selection area.
Weakness. Manager B had a marked shortfall in local market return. Therefore,
Manager B appears to be weak in security/market selection ability.
b.

The following strategies would enable the fund to take advantage of the
strengths of each of the two managers while minimizing their weaknesses.
1. Recommendation: One strategy would be to direct Manager A to make no
currency bets relative to the international index and to direct Manager B to
make only currency decisions, and no active country or security selection
bets.
Justification: This strategy would mitigate Manager As weakness by
hedging all currency exposures into index-like weights. This would allow
capture of Manager As country and stock selection skills while avoiding
losses from poor currency management. This strategy would also mitigate
Manager Bs weakness, leaving an index-like portfolio construct and
capitalizing on the apparent skill in currency management.
2. Recommendation: Another strategy would be to combine the portfolios of
Manager A and Manager B, with Manager A making country exposure and
security selection decisions and Manager B managing the currency exposures
created by Manager As decisions (providing a currency overlay).
Justification: This recommendation would capture the strengths of both
Manager A and Manager B and would minimize their collective weaknesses.

2.

a.

Indeed, the one year results were terrible, but one year is a poor statistical base
from which to draw inferences. Moreover, the board of trustees had directed Karl
to adopt a long-term horizon. The Board specifically instructed the investment
manager to give priority to long term results.

b.

The sample of pension funds had a much larger share invested in equities than did
Alpine. Equities performed much better than bonds. Yet the trustees told Alpine to
hold down risk, investing not more than 25% of the plans assets in common
stocks. (Alpines beta was also somewhat defensive.) Alpine should not be held
responsible for an asset allocation policy dictated by the client.

c.

Alpines alpha measures its risk-adjusted performance compared to the market:

= 13.3% [7.5% + 0.90(13.8% 7.5%)] = 0.13% (actually above zero)

24-8

Chapter 24 - Portfolio Performance Evaluation

3.

d.

Note that the last 5 years, and particularly the most recent year, have been bad
for bonds, the asset class that Alpine had been encouraged to hold. Within this
asset class, however, Alpine did much better than the index fund. Moreover,
despite the fact that the bond index underperformed both the actuarial return and
T-bills, Alpine outperformed both. Alpines performance within each asset class
has been superior on a risk-adjusted basis. Its overall disappointing returns were
due to a heavy asset allocation weighting towards bonds, which was the Boards,
not Alpines, choice.

e.

A trustee may not care about the time-weighted return, but that return is more
indicative of the managers performance. After all, the manager has no control
over the cash inflows and outflows of the fund.

a.

Method I does nothing to separately identify the effects of market timing and
security selection decisions. It also uses a questionable neutral position, the
composition of the portfolio at the beginning of the year.

b.

Method II is not perfect, but is the best of the three techniques. It at least attempts
to focus on market timing by examining the returns for portfolios constructed from
bond market indexes using actual weights in various indexes versus year-average
weights. The problem with this method is that the year-average weights need not
correspond to a clients neutral weights. For example, what if the manager were
optimistic over the entire year regarding long-term bonds? Her average weighting
could reflect her optimism, and not a neutral position.

c.

Method III uses net purchases of bonds as a signal of bond manager optimism.
But such net purchases can be motivated by withdrawals from or contributions to
the fund rather than the managers decisions. (Note that this is an open-ended
mutual fund.) Therefore, it is inappropriate to evaluate the manager based on
whether net purchases turn out to be reliable bullish or bearish signals.

4.

Treynor measure = (17 8)/1.1 = 8.182

5.

Sharpe measure = (24 8)/18 = 0.888

24-9

Chapter 24 - Portfolio Performance Evaluation

6.

a.

Treynor measures
Portfolio X: (10 6)/0.6 = 6.67
S&P 500:
(12 6)/1.0 = 6.00
Sharpe measures
Portfolio X: (10 6)/18 = 0.222
S&P 500:
(12 6)/13 = 0.462
Portfolio X outperforms the market based on the Treynor measure, but
underperforms based on the Sharpe measure.

b.

The two measures of performance are in conflict because they use different
measures of risk. Portfolio X has less systematic risk than the market, as measured
by its lower beta, but more total risk (volatility), as measured by its higher
standard deviation. Therefore, the portfolio outperforms the market based on the
Treynor measure but underperforms based on the Sharpe measure.

7.

Geometric average = (1.15 0.90)1/2 1 = 0.0173 = 1.73%

8.

Geometric average = (0.91 1.23 1.17)1/3 1 = 0.0941 = 9.41%

9.

Internal rate of return = 7.5%

10.

d.

11. Time-weighted average return = (15% + 10%)/2 = 12.5%

To compute dollar-weighted rate of return, cash flows are:
CF0 = $500,000
CF1 = $500,000
CF2 = ($500,000 1.15 1.10) + ($500,000 1.10) = $1,182,500
Dollar-weighted rate of return = 11.71%
12. b.
13. a.

24-10

Chapter 24 - Portfolio Performance Evaluation

14. a.

Each of these benchmarks has several deficiencies, as described below.

Market index:
A market index may exhibit survivorship bias. Firms that have gone out of
business are removed from the index, resulting in a performance measure that
overstates actual performance had the failed firms been included.
A market index may exhibit double counting that arises because of companies
owning other companies and both being represented in the index.
It is often difficult to exactly and continually replicate the holdings in the market
index without incurring substantial trading costs.
The chosen index may not be an appropriate proxy for the management style of
the managers.
The chosen index may not represent the entire universe of securities. For
example, the S&P 500 Index represents 65% to 70% of U.S. equity market
capitalization.
The chosen index (e.g., the S&P 500) may have a large capitalization bias.
The chosen index may not be investable. There may be securities in the index
that cannot be held in the portfolio.
Benchmark normal portfolio:
This is the most difficult performance measurement method to develop and
calculate.
The normal portfolio must be continually updated, requiring substantial resources.
Consultants and clients are concerned that managers who are involved in
developing and calculating their benchmark portfolio may produce an easilybeaten normal portfolio, making their performance appear better than it
actually is.
Median of the manager universe:
It can be difficult to identify a universe of managers appropriate for the
investment style of the plans managers.
Selection of a manager universe for comparison involves some, perhaps much,
subjective judgement.
Comparison with a manager universe does not take into account the risk taken in
the portfolio.
The median of a manager universe does not represent an investable portfolio;
that is, a portfolio manager may not be able to invest in the median manager
portfolio.
Such a benchmark may be ambiguous. The names and weights of the securities
constituting the benchmark are not clearly delineated.
The benchmark is not constructed prior to the start of an evaluation period; it is
not specified in advance.
A manager universe may exhibit survivorship bias; managers who have gone out of
business are removed from the universe, resulting in a performance measure that
overstates the actual performance had those managers been included.

24-11

Chapter 24 - Portfolio Performance Evaluation

b.

i. The Sharpe ratio is calculated by dividing the portfolio risk premium (i.e.,
actual portfolio return minus the risk-free return) by the portfolio standard
deviation:
Sharpe ratio = (rP rf)/P
The Treynor measure is calculated by dividing the portfolio risk premium (i.e.,
actual portfolio return minus the risk-free return) by the portfolio beta:
Treynor measure = (rP rf )/P
Jensens alpha is calculated by subtracting the market risk premium, adjusted
for risk by the portfolios beta, from the actual portfolio excess return (risk
premium). It can be described as the difference in return earned by the
portfolio compared to the return implied by the Capital Asset Pricing Model or
Security Market Line:
P = rP [rf + P (rM rf )]
ii. The Sharpe ratio assumes that the relevant risk is total risk, and it measures
excess return per unit of total risk. The Treynor measure assumes that the
relevant risk is systematic risk, and it measures excess return per unit of
systematic risk. Jensens alpha assumes that the relevant risk is systematic risk,
and it measures excess return at a given level of systematic risk.

15.

i.

The statement is incorrect. Valid benchmarks are unbiased. Median manager

benchmarks, however, are subject to significant survivorship bias, which results in
several drawbacks, including the following:
The performance of median manager benchmarks is biased upwards.
The upward bias increases with time.
Survivor bias introduces uncertainty with regard to manager rankings.
Survivor bias skews the shape of the distribution curve.

ii.

The statement is incorrect. Valid benchmarks are unambiguous and able to be

replicated. The median manager benchmark, however, is ambiguous because the
weights of the individual securities in the benchmark are not known. The
portfolios composition cannot be known before the conclusion of a measurement
period because identification as a median manager can occur only after
performance is measured.
Valid benchmarks are also investable. The median manager benchmark is not
investable. That is, a manager using a median manager benchmark cannot forego
active management and, taking a passive/indexed approach, simply hold the
benchmark. This is a result of the fact that the weights of individual securities in
the benchmark are not known.

iii.

The statement is correct. The median manager benchmark may be inappropriate

because the median manager universe encompasses many investment styles and,
therefore, may not be consistent with a given managers style.

24-12

Chapter 24 - Portfolio Performance Evaluation

16.

a.

Sharpe ratio = (rP rf)/P

Williamson Capital: Sharpe ratio = (22.1% 5.0%)/16.8% = 1.02
Joyner Asset Management: Sharpe ratio = (24.2% 5.0%)/20.2% = 0.95
Treynor measure = (rP rf )/P
Williamson Capital: Treynor measure = (22.1% 5.0%)/1.2 = 14.25
Joyner Asset Management: Treynor measure = (24.2% 5.0%)/0.8 = 24.00

b.

The difference in the rankings of Williamson and Joyner results directly from the
difference in diversification of the portfolios. Joyner has a higher Treynor measure
(24.00) and a lower Sharpe ratio (0.95) than does Williamson (14.25 and 1.202,
respectively), so Joyner must be less diversified than Williamson. The Treynor
measure indicates that Joyner has a higher return per unit of systematic risk than
does Williamson, while the Sharpe ratio indicates that Joyner has a lower return
per unit of total risk than does Williamson.

24-13