Sie sind auf Seite 1von 7

Diversification Returns and Asset Contributions

David G. Booth and Eugene F. Fama For a portfolio with a constant percentage invested in each asset, the compound return is the sum of the contributions of the individual assets in the portfolio. Theportfolio compound return is greater than the weighted average of the compound returns on the assets in the portfolio. The incremental return is due to diversification. The contribution of each asset exceeds its compound return by the amount it adds to the portfolio diversification return. The compound return on an asset is approximately the asset's average return minus one-half the asset's variance. A portfolio's average return is the weighted average of each asset's average return, but a portfolio's variance is the weighted average of each asset's covariance. Itfollows that the return contribution of an asset can be better approximated by subtracting one-half the covariance than one-half the variance.

ning and ending portfolio values. Knowing the beginning portfolio value, the compound return and the number of investment periods, we can easily compute the ending portfolio value.

LUI

-D

z
0
:D 1-

z
_i

The compound return on a portfolio is an important measure of performance because it provides 26 the connection between begin-

z z

4.44%. The return contributions are higher than the compound returns (11.31% and 4.36%) because of the gains from diversification. Portfolio diversification adds 48 basis points to the S&P 500 return and 8 basis points to The compound return on an asset the Treasury bond return. The is, however, a misleading mea- benefit of diversification for the sure of the contribution of the portfolio (28 basis points) is the asset to the compound return on average of the benefits for the a portfolio. We will show that, individual assets. with a constant percentage invested in each asset, the portfolio The increment of the return concompound return is greater than tribution of an asset over its comthe weighted average of the com- pound return depends on how pound returns on the assets in the much the asset's risk is reduced portfolio. This means that the through diversification. The incontribution of each asset to the crement is higher for the S&P500 portfolio compound return is because more of its risk is divergreater than the asset's com- sified away. Conversely, from a pound return. The difference is portfolio perspective, judging asan incremental return due to di- sets in terms of their compound versification. This incremental re- returns penalizes most those asturn is the subject of this article. sets that benefit most from diverTable I presents a simple exam- sification. ple using returns on the S&P500, Treasurybonds and a 50-50 port- The Compound Return folio of the two assets for the 50 The appendix shows that the years 1941-90. The annualized compound return (continuously compound return for the S&P500 compounded) on asset j is well is 11.31%,and the annualized re- approximated as: turn for Treasurybonds is 4.36%. The average of the two com- Eq. 1 pound returns is 7.84%. But a portfolio that maintains a 50% C, = ln[1 + E(R1)] weight in the S&P500 and a 50% weight in Treasury bonds has a Sj2 higher compound return8.11%. Diversification adds 28 ba2[1 + E(Rj)f sis points a year to the portfolio return. where Table I also shows the "return E(R.) = the expected contribution" of each asset. The (average) rereturn contribution-the central turn on asset j, concept of this article-is our esvariance S2 = the timate of the contribution of each (standard deviasset to a portfolio's compound ation squared) return. For the 50-50 portfolio, of the simple the return contribution of the returns on j S&P500 is 11.79%,and the return and contribution of Treasurybonds is

The CFA Institute is collaborating with JSTOR to digitize, preserve, and extend access to Financial Analysts Journal www.jstor.org

Table I Compound Returns, 1941-1990


S&P500 Index 50.00 1.03 4.14 20-Year Treasury Bonds 50.00 0.39 2.39 Returns (%) 4.36 0.08 4.44 8.11 7.84 0.28 8.11

Glossary *.Return Contribution:


Portfolio 100.00 0.71 2.62

Portfolio Weights (%) Average Monthly Returns (%) Standard Deviations (%)

The contribution of an asset to a portfolio's compound return.

* Variance:
The second moment of a distribution around its mean; the average of the squared deviations from the mean.

Annualized Continuously Compounded Compound Return Compound Return of Each Asset 11.31 and the Average Asset Compound Return Return Due to Diversification 0.48 Return Contribution 11.79

*'Asset-Class Investing:
Investing to capture the returns of an asset class without regard to stock selection. An index fund is one example.

ln[l + E(Rf)] = the average Eq. 3 return on j, excov(Rj, Rp) pressed as a = *.Diversification Returns: 2 continuously SP Sp For a portfolio, the difference compounded between its compound return return. Here cov(R., Rp)is the covariance and the weighted average of of asset j's returns with the porteach asset's compound return. The compound return on portfo- folio's returns and sp2 is the variFor an asset in a portfolio, the lio p can be expressed in terms of ance of the portfolio's returns. difference between the return the mean and variance of its contribution and the comRearrangingEquation (3), we can simple returns: pound return. express the risk of asset j in terms of its contribution to portfolio p's Eq. 2 risk, as measured by its variance: tween the compound return and the return contribution is thus due to the difference between Eq. 4 CP= ln[l + E(Rp)] b,pSp2and s.2. In a nutshell, the return contribution of asset j is cov (Rj,Rp) = bjpsp greater than its compound return 5p2 because the contribution of asset j 2[1 + E(Rp)] The covariance measure of each to the variance of the return on asset in the portfolio [cov(R,,Rp)], the portfolio, b. sp2, is less than weighted by the asset's represen- the variance of thereturn on j, s;. Equations (1) and (2) say that tation in the portfolio, will sum to This risk reduction, a result of compound returns are lower, the the portfolio's variance, diversification, enhances the consP2. higher the variance of returns. tribution of asset j to the comFor a portfolio, variance is an pound return on the portfolio. The Return Contribution appropriate measure of risk. But One estimate of the contribution the risk of an asset in a diversified of asset j to the compound return The weighted average of the return contributions estimated portfolio is less than the variance on portfolio p is: from Equation (5) is not exactly of the asset's returns. We show the portfolio compound return, below that the risk reduction due because the weighted average of to diversification is what causes Eq. 5 ln[1 + E(P)] is not exactly ln[l + an asset's compound return to E(Rp)J.We can arrange for the = + In[1 E(R1)] understate the contribution of the Di weighted average of the return asset to the compound return on contributions to equal the portfothe portfolio. lio compound return given by 2[1 + E(Rj)f Equation (2) if we change Equation (5) to: The Risk of an Asset in a Comparing Equations (1) and (5) Portfolio shows that the compound return Eq. 6 Asset j's beta relative to the port- on asset j and its contribution to folio is calculated by regressing the compound return on portfoln[1 + E(Rp)] its returns against the portfolio's lio p have the same first term=E(RJ) E(Rp) returns. It can be expressed as: ln[l + E(Rf)].The difference be-

z
-LJ D

27

Table II Portfolio with Small-CapStocks, 1941-1990


US. Small-Cap D6-10
10.00 1.24 5.51 Compounded 12.96 0.99

S&P500
Portfolio Weights (%) Average Monthly Returns (%) Standard Deviations (%) Annualized Compound Return Compound Return of Each Asset and the Average Asset Compound Return Return Due to Diversification 50.00 1.03 4.14 Continuously 11.31 0.41

20-Year Treasury Bonds


20.00 0.39 2.39 Returns

One-Month Treasury Bills


20.00 0.36 0.28

Portfolio
100.00 0.79 2.70 9.02 8.68 0.33

(%)
4.36 0.16 4.30 -0.00

Return Contribution

11.72

13.96

4.52

4.29

9.02

tas from quarterly data and average returns and standard devia2[1 + E(Rp)f tions from monthly data. The continuously compounded reThe difference between estimat- turns and return contributions ing the return contribution using are annualized from the monthly Equation (5) and using Equation data. (6) is minuscule for most periods. Results We will use Equation (6) because II Table III show reand Table of its additive property. The balanced portfor domestic sults weighted average of the first term for Table II uses folios 1941-90. on the right-hand side is exactly 6 10 through (D6-10) for deciles ln[l + E(Rp)I and the weighted and Table III small-cap stocks average of the second term is 10 (D9through uses deciles 9 sp2/2[1 + E(Rp)]2.Thus: 10). The results for the S&P 500, Treasurybonds and Treasurybills Eq. 7 are almost identical in the two tables. The annualized return contribution for the S&P 500 in Dp = ln[l + E(Rp)] both is 11.72%, an increase of S2 Sp -C 0.41% over the index's com2[1 + E(Rp)]2 P pound return. The incremental return for Treasury bonds is 0.16%. Examples We estimate below the return contributions of seven asset The incremental returns due to a.) classes for different time periods diversification are greater for and portfolios.' We estimated be- small-cap stocks than for the
b)jpsp

other assets. The incremental returns (the difference between the contribution of small stocks to the compound return on the portfolio and the compound return on small stocks) are 0.99%for D6-10 stocks and 1.36% for D9-10 stocks. Measuring the contributions of these assets to portfolio returns, rather than the compound returns on the assets, substantially increases the premium of small-cap stocks over the S&P 500. The difference between the return contributions of D9-10, 15.42%,and the S&P500, 11.72%, is 3.70%. The difference between the return contributions of D610, 13.96%, and the S&P 500 is 2.24%. The corresponding differences between the compound returns are only 2.75% and 1.65%. Why are the contributions of small stocks to the compound returns on portfolios so much larger than the compound returns on small stocks? Small stocks have high return variances,

Table III Portfolio with Very Small-CapStocks, 1941-1990


z
D

F-

Portfolio Weights (%) Average Monthly Returns (%) StandardDeviations (%)

S&P500 50.00 1.03 4.14

U.S.Small-Cap D9-10 10.00 1.37 6.20 Compounded 14.06 1.36 15.42

20-Year TreasuryBonds 20.00 0.39 2.39 Returns (%) 4.36 0.16 4.52

One-Month TreasuryBills 20.00 0.36 0.28

Portfolio 100.00 0.80 2.72 9.17 8.79 0.37 9.17

z z

28

Annualized Continuously Compound Return 11.31 Compound Return of Each Asset and the Average Asset Compound Return 0.41 Return Due to Diversification 11.72 Return Contribution

4.30 -0.00 4.29

Table IV

Domestic Portfolio, 1926-1940


S&P500 US. Small-Cap D6-10
10.00 0.99 13.62 Compounded 1.73 4.79 6.53

20-Year TreasuryBonds
20.00 0.42 1.45 Returns

One-Month TreasuryBills
20.00 0.11 0.14

Portfolio
100.00 0.59 6.09 4.94 3.39 1.55 4.94

Portfolio Weights (%) Average Monthly Returns (%) Standard Deviations (%) Annualized Compound Return Compound Return of Each Asset and the Average Asset Compound Return Return Due to Diversification Return Contribution

50.00 0.78 9.56 Continuously 3.96 2.09 6.05

(%)
1.31 0.03 1.34

4.87 0.10 4.96

which lower their compound returns (see Equation (1)). Because small stocks are not highly correlated with other assets, however, their risk in a diversified portfolio (the covariance of their returns with portfolio returns) is much less than their return variance. This diversification benefit substantially increases the contribution of small stocks to the compound returns on portfolios. (The story is similar for international stocks, large and small.) Tables II, IV and V show results for the same domestic portfolio over different time periods. The results for the last 20 years (Table V) are similar to the results for the last 50 years (Table II). The results for the Great Depression (Table IV) are quite different. During the Depression, it was important to be diversified. The return contribution for small-cap stocks in this period is greater than the return contribution for the S&P 500, but the compound return for small-cap stocks is

lower than the compound return understated by 0.75% (1.31-0.56). The international large-cap prefor the S&P 500. mium over the S&P is underTables VI and VIIshow results for stated by 0.74%. two internationally diversified portfolios. With two additional Table VII shows the results for a equity asset classes, the negative portfolio that equally weights all bias of the asset compound re- four equity asset classes. Its turn is more apparent. Table VI monthly standard deviation is shows that the compound return only slightly greater than that of understates the annualized return the portfolio in Table VI-2.99% contributions for stocks by the versus 2.94%. Even though the following percentages: S&P 500 standard deviation is much lower than the standard S&P500 0.56 deviations of the other equity asU.S. Small-CapStocks 1.31 set classes, the portfolio standard (D6-10) deviation does not increase sigInternational Large-Cap 1.30 nificantly because the S&P 500 Stocks commitment is reduced. The imInternational Small-Cap 1.23 provement in diversification Stocks about offsets the increase in average standard deviation. Because the S&P500 benefits the least from diversification among Finally, the difference between the four asset classes and it has a the contribution of Treasury large weight in the portfolio, its bonds to the compound returns return increment due to diversifi- on portfolios and the compound cation is the lowest of the four return on the bonds has widened asset classes. The U.S. small-cap in the last 20 years (see Tables I to premium over the S&P is thus VII). The reason is that the vari-

LU z

Table V Domestic Portfolio, 1971-1990


S&P 500 Portfolio Weights (%) Average Monthly Returns (%) Standard Deviations (%) Annualized Compound Return Compound Return of Each Asset and the Average Asset Compound Return Return Due to Diversification Return Contribution 50.00 0.99 4.65 Continuously 10.57 0.48 11.06 U.S. Small-Cap D6-10 10.00 1.11 6.10 Compounded 10.97 1.19 12.16 20-Year Treasury Bonds 20.00 0.75 3.30 Returns One-Month Treasury Bills 20.00 0.62 0.22 Portfolio
U

100.00 0.88 3.13 9.95 9.53 0.42 9.95

-J

z
:D

(%)
7.39 -0.00 7.39

8.36 0.28 8.64

z
z

29

Table VI 10% International Portfolio, 1971-1990


S&P 500 US. Small D6-10 Intl. Large Intl. Small 20-Year T-Bonds One-Month T-Bills Portfolio

Portfolio Weights (%) Average Monthly Returns (%) Standard Deviations (%)

45.00 0.99 4.65

5.00 1.11 6.10

5.00 1.63 5.86

5.00 2.02 5.50

20.00 0.75 3.30

20.00 0.62 0.22

100.00 0.96 2.94 10.95 10.44 0.50 10.95

Annualized Continuously Compounded Returns (%) Compound Return Compound Return of Each Asset 22.21 10.57 10.97 17.46 8.36 and the Average Asset Compound Return Return Due to Diversification 0.56 1.31 1.30 1.23 0.31 Return Contribution 12.28 18.76 11.13 23.45 8.67

7.39 -0.00 7.39

ance of long-bond returns has increased. In terms of variability, long bonds are now more like stocks. Thus it is now more important that long bonds be held in diversified portfolios that include stocks.

Diversification vs. Investment Uncertainty

Returns

Thus far we have shown that diversification improves portfolio compound returns. We will now show that the incremental returns may be lost by engaging in active management. In order to measure diversification returns, we engaged in asset-class investing, investing in asset classes in fixed proportions. This runs counter to active management. The essence of active management is to try to "beat the market"by shifting asset weights. Active management can either increase or decrease the portfolio compound return, depending on

both skill and luck. As we will see, were 14.33% or higher, while the luck component can be larger 10% were 10.52% or lower. than the skill component. The average asset mix is 50% S&P To see how shifting weights intro- 500 and 50% Treasury bonds. duces uncertainty, suppose we However, the average standard randomly invest in the S&P 500 deviation of returns is somewhat and Treasurybonds on a monthly higher for the timing portfolios basis. One-half of the months we than it is for the 50/50 portfolio, invest 65% in the S&P 500 and because of the shifting of asset 35% in Treasurybonds. The other weights. The average standarddehalf of the months we invest 35% viation for the random timing in the S&P 500 and 65%in Trea- portfolio is equal to that of a constant-mix portfolio invested sury bonds. 53% in the S&P 500 and 47% in The returns from this strategywill Treasury bonds. This 53/47 portdepend on the luck of the draw. folio serves as a suitable benchTo show how uncertain the out- mark for comparison with the comes can be, we simulated the random timing portfolios. The approach 1000 times for the five benchmark outperforms the ranyears 1986-90. It is as though we domly generated portfolios. Its had 1000 managers engaged in a 12.53% compound return is 14 market timing contest. The sum- basis points higher than the avermary of the results are displayed age compound return for the ranin Table VIII. domly generated portfolios, and has a 52-basis-point annual diverWe calculated the five-year com- sification return. pound return for each of the simulations, then ranked from high- Thus shifting weights lowers exest to lowest. As Table VIIIshows, pected compound return. But, 10% of the annualized returns perhaps more importantly, shift-

Table VII 30% International Portfolio, 1971-1990


z
S&P 500
H-

US. Small D6-10

Intl. Large

Intl. Small

20-Year T-Bonds

One-Month T-Bills

Portfolio

Portfolio Weights (%) Average Monthly Returns (%) Standard Deviations (%)

15.00 0.99 4.64

15.00 1.11 6.10

15.00 1.63 5.86

15.00 2.02 5.50

20.00 0.75 3.29

20.00 0.62 0.22

100.00 1.14 2.99 13.04 12.33 0.71 13.04

36

Annualized Continuously Compounded Returns (%) Compound Return 22.21 17.46 Compound Return of Each Asset 10.57 10.97 8.36 z and the Average Asset Compound Return zL Return Due to Diversification 1.18 1.08 0.66 1.35 0.36 18.64 8.72 Return Contribution 11.23 12.32 23.30
-J

7.39 -0.01 7.39

TableVIII Random Timing Strategy, Annualized Compound Returns

(%), 1986-1990
10th Percentile 50th Percentile 90th Percentile lOth-90th Benchmark Portfolio S&P 500 Treasury Bills 14.33 12.39 10.52 3.81 12.53 13.14 10.74

ing weights also introduces uncertainty. The median return is the most likely return, but achieving the median return is highly uncertain. About 10% of the random portfolios have returns 200 basis points a year higher than the benchmark return. About 10% of the portfolios have returns 200 basis points a year lower than the benchmark return. Most investors would take 12.53% for certain rather than invest in a strategy that has uncertain returns and an equal chance of a 14.33% return or a 10.52% return.

percentiles measures the uncertainty introduced by active management. The implication is clear: Investors should prefer the certain return from a portfolio that maintains fixed asset weights to a strategy whose outcome is uncertain and has an equal chance of underperforming or outperforming the benchmark by 200 basis points per year. Stated differently, investors need a large premium to be willing to incur the additional uncertainty of active management. Recent studies indicate that investors cannot expect such a premium.2 The premium return from active management is negative rather than positive. Every year, about half the funds outperform their benchmarks. Those in the upper half in one year have only about a 50% chance of repeating in any other year.

benefits these assets (reduces their risks) the most. For the S&P 500, the difference is smaller, but important. The increased variance of long-bond returns over the last 20 years has widened the spread between their return contributions and their compound returns. By shifting asset weights, active management introduces uncertainty about portfolio compound returns. Diversification returns are assured only for those portfolios that maintain relatively fixed asset weights.

Append'x

The continuously compounded return on an asset is In[1 + , where R.is the simple return. The expected value of the compound return of an asset, E[ln(l + R.)], can be derived from the Taylor Active management introduces so series expansion of the natural much uncertainty that we cannot logarithm about the mean return document a premium return over of the asset: benchmark returns. By contrast, fully diversified portfolios reli- Eq. Al The uncertainty of achieving ably increase portfolio combenchmark returns extends to pound returns through the diverreal portfolios. Table IX displays sification process and eliminate Ci = In[1 + E(R1)] the distribution of total plan re- "benchmark risk." M2 M3 turns for institutional funds in the 1 + SEICorporation database. The in2[1 + E(R)]2 3[ + E(Rj stitutional portfolios behave like Conclusion the randomly generated portfo- The portfolio compound return is M4 lios. About 10%of the funds have useful for describing portfolio re4+ _ returns 200 basis points a year sults. But the compound return of 4[1 + E(Rj)] above the median. About 10% of an asset in a portfolio understates the funds have returns 200 basis the contribution of the asset to where Mk = the kth moment of points a year below the median. the portfolio compound return the asset return about its mean, m because it ignores the benefit of or: z The 400-basis-point spread of re- diversification. turns between the 10th and 90th Eq. A2 The "return contribution" de0 fined by Equation (6) is an im- Mk = E[(R- E(R1))k]. z Table IX Total Plan Sponsor H proved measure of the contribu-D Annualized Com- tion of an asset to the portfolio Equation (1) is just Equation (A1) pound Returns (%), compound return. The portfolio with the higher moments 1986-1990* compound return is the average dropped. For the assets and portof the asset return contributions. folios in Tables I through VII, z 10th Percentile 12.38 these higher-order terms are al50th Percentile 10.47 The increments of return contri- ways minuscule, except during U 90th Percentile 8.35 bution over compound returns the Great Depression. Table AI z 4.03 are especially large for small-cap displays the accuracy of Equation z lOth-90th stocks and international stocks, (1) for the portfolios displayed in primarily because diversification Tables I through VII. 31 Source: SEI Corporation.
-J -c

LUJ

Table AI Annualized Portfolio Compound Returns Obtained by Linking Monthly Returns Compared with Compound Returns Derived from Equation (1)
Portfolio by Table Actual Compound Return % Derived Compound Return % Difference (Formula Error) %

Time Period

I II III LV V VI VII

1941-90 1941-90 1941-90 1926-40 1971-90 1971-90 1971-90

8.11 9.02 9.17 4.97 9.95 10.95 13.04

8.11 9.02 9.17 4.94 9.95 10.95 13.04

0.00 -0.00 -0.00 0.03 -0.00 -0.00 -0.00

Footnotes
1. S&P500 and long-term Treasury bond data from R. G. Ibbotson and R. A. Sinquefield, Stocks, Bonds, Bills and Inflation (Chicago: Ibbotson Associates). D9-10 is the Dimensional Fund AdvisorsInc. (DFA) US. 9-10 Small Company Portfolio, net of all fees, from 1982 on and the CRSP 9-10 index (courtesy of the Center for Research in Securities Prices, Universityof Chicago) for 1926-81. The D6-10 is the DFA US. 6-10 Small Company Portfolio, net of administrativefees, from June 1986 on and the CRSP6-10 index for 1926 through May 1986. The international small-cap portfolio uses the following data. For the UK: the Hoare Govett Smaller Co. Index (courtesy London Business School) for 1956 through March 1986 and the DFA United Kingdom Small Company Portfolio, net of allfees, for April 1986 on. For Japan: the smaller half of the first section of the Tokyo Stock Exchange (courtesy of Nomura Securities Investment Trust Management Co., Ltd., Tokyo)for 1970 through March 1986 and the DFAJapanese Small Company Portfolio, net of allfees, for April 1986 on. The international large-cap portfolio uses thefollowing data. For the UK: the Financial Times All Shares Index for 1956 on. For Japan: The larger half of the first section of the Tokyo Stock Exchange (courtesy Nomura Securities)for 1970 through June 1986 and the Japan National Index (courtesy of Morgan Stanley Capital International) for July 1986 on. Data on large-cap continental stocks (excluding the UK) and on Asia and Australia (excluding Japan) from Morgan Stanley. The international large and small-cap portfolios were weighted as follows. For 1970June 1988&50% Japan, 50% UK

For July 1988-September 1989:50% Japan, 30% Continent, 20% UK For October 1989-March 1990: 40% Japan, 30% Continent, 20% UK, 10% Asia-Australia.For April 1990 on: 40% Japan, 35% Continent, 15% UK, 10% Asia-Australia.Returns for various portfolios over various years are available from the authors. 2. See G. P. Brinson, L. R. Hood and G. L. Beebower, "Determinants of Portfolio Performance, Financial
Analysts Journal, July/August 1986,

and G. P. Brinson, B. D. Singer and G. L. Beebower, "Determinants of Portfolio Performance II: An Update," Financial Analysts Journal,

May/June 1991.

uJ

z
.D

32

Das könnte Ihnen auch gefallen