You are on page 1of 17


Evaluation of Investment Performance

This chapter, which is concerned with measuring portfolio performance, is a logical concluding chapter for an investments text. The bottom line of investing for investors is deciding how well their portfolio performed and if they need to make changes. Chapter 22 is a typical chapter in this area, covering the composite measures of portfolio performance as well as some of the newer concepts such as performance attribution. Although the composite measures have some problems, there is nothing better that is widely accepted, these measures are often seen and used in both textbooks and in the popular press, and they are easy to understand. The chapter begins with a discussion of what is involved in evaluating portfolio performance, including the need to adjust for differential risk and differential time periods, the need for a benchmark, and constraints on portfolio managers. t also considers the difference between the portfolio!s performance and the manager"s performance. ncluded in the discussion is an outline of A #$!s presentation standards, which likely will have a major impact in this area. These are minimum standards for presenting investment performance. These standards are being widely adopted by investment management firms. The dollar%weighted and time%weighted returns measures are described, and the issue of risk is considered. &or expositional and other purposes at this level, the discussion centers around the three well%known risk%adjusted 'composite( measures of portfolio performance) *harpe, Treynor and +ensen. ,ach of these is developed and illustrated separately. Actual mutual fund data is used to show how the measures are computed and evaluated. Additional discussion involves a comparison of the measures to each other, such as how the *harpe and Treynor measures compare. 63

Along with the discussion of the measures themselves is related discussion of such issues as how to measure diversification. Also, capital market issues related to +ensen"s measure are considered. The chapter next considers the problems with portfolio measurement. This remains an unsettled issue in investments, with people using different measures and techni-ues, different benchmarks, and so forth. *tudents should be made aware that there is no standardi.ed, widely accepted approach to this problem. The chapter concludes with other issues in performance evaluation. /erformance attribution is briefly considered. CHAPTER OB ECTIVE!

To outline the issues involved in measuring portfolio performance, including the problems that remain. To present A #$!s new presentation standards, and develop well%known measures of return consistent with these standards. To develop and illustrate the three composite measures of portfolio performance. To consider other important issues, such as performance attribution.


"A OR CHAPTER HEA#I$%! &Contents' (rame)or* for Evaluatin+ Portfolio Performance

*ome 0bvious &actors to Consider 1differential risk levels2 differential time periods2 appropriate benchmarks2 constraints on portfolio managers2 other considerations3 A #$!s /resentation *tandards 1outline of some of the minimum standards for presenting investment performance3

Return an, Ris* Consi,erations

#easures of $eturn 1Total $eturn2 dollar%weighted return2 time%weighted return3 $isk #easures 1total risk2 nondiversifiable risk3

Ris*-A,.uste, "easures of Performance

The *harpe /erformance #easure 1e-uation2 diagram2 example with mutual fund data3 The Treynor /erformance #easure 1e-uation2 diagram2 example with mutual fund data2 comparing the *harpe and Treynor #easures2 measuring diversification3 +ensen!s 4ifferential $eturn #easure 1development of e-uation2 diagram2 how to use2 significance2 a comparison of the three composite measures3

Pro/lems With Portfolio "easurement 1$oll!s arguments about beta2 benchmark error2 other problems3 Other Issues In Performance Evaluation 65

#onitoring /erformance /erformance Attribution 1definition2 issues2 bogey3 Can /erformance 5e /redicted6

POI$T! TO $OTE ABO0T CHAPTER 22 Ta/les an, (i+ures Table 22%7 is an outline of A #$!s /erformance /resentation *tandards. These standards are discussed at various points in the chapter. Table 22%2 contains the mutual fund data discussed in the chapter. As such, it is for informational purposes in illustrating the calculations of the composite measures, the measure of diversification, and so forth. Table 22%8 shows the risk%adjusted measures for the five e-uity mutual funds used in the analysis of the risk%adjusted measures. The figures in this chapter are standard figures showing the three measures of composite performance and characteristic lines. They are keyed to discussion in the text in terms of the mutual fund data used, but otherwise there is nothing uni-ue about them. nstructors can easily use comparable figures of their own to illustrate the major points about the composite measures of performance. &igure 22%7 shows *harpe!s measure of performance for five mutual fund portfolios, with three funds plotting above the C#9 and two below. &igure 22%2 shows Treynor!s measure of performance for three mutual fund portfolios, with three of these funds plotting above the *#9 and two below. &igure 22%8 illustrates +ensen!s measure of portfolio performance for three hypothetical funds. 66

Bo1 Inserts There are no box inserts for this chapter.


A$!WER! TO E$#-O(-CHAPTER 20E!TIO$! 22-34 The framework for evaluating portfolio performance primarily involves an analysis of both the returns and the risk of the portfolio being evaluated and a comparison to a proper benchmark. 0ther issues include the diversification of the portfolio and an evaluation of the manager as opposed to the portfolio itself. t is important to determine if a portfolio manager is following the stated objectives for the portfolio. t is also necessary to analy.e any constraints imposed on a portfolio manager. f a portfolio outperforms its expected benchmark, such results may be attributable to the manager. f the portfolio manager ceases to manage this portfolio, potential investors will want to be aware of this fact. The three composite measures use either standard deviation 'from the C#9( or beta 'from the *#9( as a risk measure. t is easy to see how +ensen!s measure is directly related to the CA/#. There is a derivable relationship between +ensen"s measure and Treynor!s measure, thereby connecting Treynor!s measure to the CA/#. &inally, *harpe!s measure can be related to capital market theory. The *harpe measure takes into account how well diversified the portfolio was during the measurement period. Any difference in rankings between the two measures should reflect the lack of diversification in the portfolio. :ith perfect diversification, the two measures produce identical rankings. $egressing a portfolio!s returns against the market!s returns for some period of time results in a characteristic line for the portfolio. *uch a line shows the linear relationship between the fund!s returns and the market!s returns2 that is, it shows how well the former is explained by the latter. /ortfolio diversification can be measured by the coefficient of ,etermination 9R2:, which is produced when a characteristic line is fitted. f the fund is totally diversified, the $2 will approach 7.;, indicating that






the fund!s returns are completely explained by the market!s returns. 0n average, mutual funds have high degrees of diversification 'i.e., an $2 of .<= or .>;, or higher(. 22-;4 22-<4 An index fund should show complete diversification. nvestors who have all 'or substantially all( of their assets in a portfolio of securities should rely more on the *harpe measure because total risk is important. &or investors whose portfolio constitutes only one part of their total assets, systematic risk is the relevant risk and the Treynor measure would be appropriate. 22-=4 *tart with the CA/# e-uation, using a portfolio subscript p. f investor!s expectations are, on average, fulfilled, the e-uation can be approximated ex post as) $pt ? $&t @ bp1$mt%$&t3 @ ,pt $earranging, $pt % $&t ? bp1$mt%$&t3 @ ,pt '22%A( '22%=(

&inally, an intercept term, alpha, can be added to 22%A to identify superior or inferior portfolio performance. 22-3>4 The +ensen measure is computationally efficient because the beta for the portfolio is estimated simultaneously with the alpha, or performance measure. The alpha must be statistically significant to be meaningful. f it is not significantly different from .ero, it doesn!t mean much, whether positive or negative. $oll has yet been measures theory(, argued that no unambiguous test of the CA/# has conducted. *ince the composite performance are based on the CA/# 'or capital market they are also called into -uestion.




The use of different market indices can result in different betas for a portfolio. This, in turn, could lead to different rankings for a portfolio2 that is, if

a BwrongC market index was used, a portfolio could rank lower than it otherwise would. 224364 n theory, the proper market index to use would be the BtrueC market portfolio%%the portfolio of all risky assets, both financial and real, in their proper proportions. *uch a portfolio is completely diversified. 22-374 The steeper the angle, the higher the slope of the line, and the better the performance. A portfolio with a line steeper than that of the market has outperformed the market. Do. *harpe and +ensen use different measures of risk, and different procedures2 therefore, different rankings of performance can be obtained. b b c


22-3;4 22 22-3<4 22-3=4

A$!WER! TO E$#-O(-CHAPTER PROB?E"! 22-34 'a( The $EA$ are) &und 7 2 8 F = #arket 'b( The $E09 are) &und 7 2 8 F = #arket 'c( $ank '7F%A(G7.7= '7A%A(G7.7; '2A%A(G7.8; '7H%A(G .>; '7;%A(G .F= '72%A(G7.;; ? A.>A ? >.;> ? 7=.8< ? 72.22 ? <.<> ? A.;; fifth third first second fourth $ank '7F%A(G27 ? '7A%A(G27 ? '2A%A(G8; ? '7H%A(G2= ? '7;%A(G7< ? '72%A(G2; ? .8< .F< .AH .FF .22 .8; fourth second first third fifth

Ies, there are differences. Dote that the numerator values for the $EA$ are exactly the same as for the $E09. The differences arise from the differences in risk as measured by the *4 and the beta. The difference in rankings is caused by the degree of diversification. &or the $EA$, only &und = failed to outperform the market. &or the $E09, all of the funds outperformed the market. &und 7. t has the highest $2.



'a( 'b( 'c( 'd(

*tandard deviations are needed to answer this -uestion. &und F has the lowest market risk, and fund = is the highest. &unds 2 and = are the only funds with alphas that are positive and statistically different from .ero. The $EA$ are)



&und A 5 C 4 , & J K #arket

$ank '7H.;%<.A(G2;.; '7>.;%<.A(G7H.< '72.8%<.A(G2=.; '2;.;%<.A(G2F.= '7=.;%<.A(G7H.F '7>.;%<.A(G7<.; ' <.A%<.A(G7>.; '2;.;%<.A(G27.= ? .F2 ? .=<F ? .7= ? .FH ? .8H ? .=H< ? ;.; ? .=8 fifth first tenth fourth sixth second eleventh third .77H '=( '7( '7;( 'F( 'A( '2( '77( '8(

'77.;%<.A(G2;.= ?

'&or some of the funds, the $EA$ are shown to an additional number of decimal places in order to eliminate ties(. 'b( The $E09 are) &und A 5 C 4 , & J K #arket '7H.;%<.A(G .<< '7>.;%<.A(G .A= '72.8%<.A(G .<8 '2;.;%<.A(G7.;; '7=.;%<.A(G .H> '7>.;%<.A(G .<8 ' <.A%<.A(G .>7 '2;.;%<.A(G .>8 $ank ? ? ? ? ? ? ? ? >.=F= 7A.;;; F.FH< 77.F;; <.7;7 72.=8; ;.; 72.2=< fifth first ninth fourth sixth second eleventh third 2.F; '=( '7( high '>(L 'F( 'A( '2( '77( low '8(

'77.;%<.A(G7.;; ?

'L ? ranking by $E09 different from that of $EA$.( 'c( The highest $2 'the greatest diversification( is for &und J, the lowest is for &und C. ,xcept for &unds 5 and C, it appears that the funds are highly diversified. &unds & and K have positive alphas 'and positive t values greater than 2.8A=(, and therefore have above average performance.



betas original excess % orig. %.;7 %.;F @.;8 @.;8 @.;2 ;.; @.;7 @.;2 excess

alphas original excess % orig. % .>A %2.<> %7.=F % .;2 %7.<7 %7.F2 % .<= % .AF


A 5 C 4 , & J K

.<H .A7 .<A 7.;8 .<7 .<8 .>2 .>=

.<< .A= .<8 7.;; .H> .<8 .>7 .>8

A.=H <.<7 7.=< <.>< F.8F <.A> %2.22 <.<<

H.=8 77.H; H.=8 8.72 A.7= 7;.77 %7.8H >.=2

Msing the 1excess%original3 beta differences, two betas are less, one unchanged, and five are larger. Do generali.ation about larger or smaller betas is possible because it depends on the relative covariance of $& with the market return and the covariance of $& with the portfolio return. Therefore, the betas using excess returns may be greater or smaller than the betas with the original returns. :ith the alpha differences 1excess%original3, all are negative, but the magnitudes vary widely. This is somewhat illusory, because no beta is greater than unity. f we had used a constant for $& 'the average $& for the period( then the betas above would have not been e-ual to the excess betas, and) excess alpha ? 1$&'N%7( @ original alpha3 *o, for portfolios with betas less than unity, the excess alpha is smaller2 and for portfolios with betas greater than unity, the excess alpha is larger.


Msing the data as given in the problem, the results are as follows) #arket $eturn #ean *4 5eta $2 $EA$ $E09 Alpha <.; 78.8 7.; 7.; ;.;8 ;.F; ;.;; $& H.A % % % % % % /ortfolio 7 7;.8 7=.< 7.7= ;.>8 ;.7H 2.8= 2.2F /ortfolio 2 >.> 7H.72 7.2F ;.>8 ;.78 7.<= 7.<;

5ased on these results, the answers are as follows) 'a( 'b( 'c( 'd( 'e( 'f( 'g( 'h( portfolio 7 ranks higher portfolio 7 ranks higher portfolio 7 has the higher alpha since the $2 is the same for each, unsystematic risk is a tie portfolio 2 has the larger beta portfolio 2 has the larger standard deviation portfolio 7 has the larger mean return portfolio 7 has a larger mean return but a smaller *4 and beta compared to portfolio 2. The result is a larger $EA$ and $E09. portfolio 8 has the widest range of returns and the largest beta portfolio 8 for the same reason portfolio 7!s returns are closest to the market, and therefore best explained by the market there is no way to be certain without doing the calculations


'a( 'b( 'c( 'd(


the results using the data as given are as follows) $& H.7H % % % % % % /ortfolio 7 7A.88 72.A8 7.;; ;.>A ;.H8 >.7H F.;; /ortfolio 2 7H.7= 78.== 7.;A ;.>F ;.HA >.H8 F.<F /ortfolio 8 7=.7H 7<.= 7.8= ;.<7 ;.F8 =.>8 7.;8

#arket $eturn #ean 72.88 *4 72.8A 5eta 7.; 2 $ 7.; $EA$ ;.F2 $E09 =.7H Alpha ;.;;

5ased on this, the portfolios rank as follows) on $EA$) on $E09) 'f( 'g( 'h( 22-84 'a( 'b( 'c( 2, 7, 8 2, 7, 8

on $2, the portfolios rank 7, 2, 8 portfolio 2 had the largest alpha portfolio 2, using composite measures portfolio 7 has returns identical to the market and a beta of 7.; portfolio 2 has returns exactly twice those of the market, and a beta of 2.; the $2 is going to be 7.; in each case because the market returns will explain each portfolio!s returns perfectly the market has an alpha of .ero, so portfolio 7 does also it would have to be identical since the returns are the same

'd( 'e( C(A 22-;4 C(A 22-<4

C(A 22-=4 C(A 22-3>4 C(A 22-334

a c A4 The Treynor measure 'T( relates the rate of return earned above the risk%free rate to the portfolio beta during the period under consideration. Therefore, the Treynor measure shows the risk premium 'excess return( earned per unit of systematic risk) $i%$f Ti ? OOOOO, i where $i ? average rate of return for portfolio i during the specified period, $f ? average rate of return on a risk%free investment during the specified period, and i ? beta of portfolio i during the specified period. /erformance $elative to the #arket '*P/ ==( 0utperformed

Treynor #easure 7;Q%AQ T ? OOOOOO ? A.HQ ;.A;

#arket '*P/ =;;( 72Q%AQ T# ? OOOOOO ? A.;Q

7.;; T examines portfolio performance in relation to the security market line '*#9(. 5ecause the portfolio would plot above the *#9, it outperformed the *P/ =;; ndex. 5ecause T was greater than T#, A.H percent versus A.; percent, respectively, the portfolio clearly outperformed the market index. The *harpe measure '*( relates the rate of return earned above the risk free rate to the total risk of a portfolio by including the standard deviation of returns. Therefore, the *harpe measure indicates the risk premium 'excess return( per unit of total risk) $i%$f *i ? OOOOO, i $i ? average rate of return for portfolio i during the specified period, $f ? average rate of return on a risk%free investment during the specified period, and i ? standard deviation of the rate of return for portfolio i during the period. *harpe #easure 7;Q%AQ * ? OOOOOO ? ;.222 7<Q /erformance $elative to the #arket '*P/ =;;( Mnderperformed

#arket '*P/ =;;( 72Q%AQ *# ? OOOOOO ? ;.FA2 78Q

The *harpe measure uses total risk to compare portfolios with the capital market line 'C#9(. The portfolio would plot below the C#9, indicating that it under%performed the market. 5ecause * was less than *#, ;.222 versus ;.FA2, respectively, the portfolio under%performed the market. B4 The Treynor measure assumes that the appropriate risk measure for a portfolio is its systematic risk, or beta. Kence, the Treynor measure implicitly assumes that the portfolio being measured is fully diversified. The *harpe measure is similar to the Treynor measure except that the excess return on a portfolio is divided by the standard deviation of the portfolio. &or perfectly diversified portfolios 'that is, those without any unsystematic or specific risk(, the Treynor and *harpe measures would give consistent results relative to the market index because the total variance of the portfolio would be the same as its systematic variance 'beta(. A poorly diversified portfolio could show better performance relative to the market if the Treynor measure is used but lower performance relative to the market if the *harpe measure is used. Any difference between the two measures relative to the markets would come directly from a difference in diversification. n particular, /ortfolio R outperformed the market if measured by the Treynor measure but did not perform as well as the market using the *harpe measure. The reason is that /ortfolio R has a large amount of unsystematic risk. *uch risk is not a factor in determining the value of the Treynor measure for the portfolio, because the Treynor measure considers only systematic risk. The *harpe measure, however, considers total risk 'that is, both systematic and unsystematic risk(. /ortfolio R, which has a low amount of systematic risk, could have a high amount of total risk, because of its lack of diversification. Kence, /ortfolio R would have a high Treynor measure 'because of low systematic risk( and a low *harpe measure 'because of high total risk(.