Chapter Five The Mathematics of Diversification KEY POINTS The mathematics of diversification can be intimidating to students, largely because the nature of the equations is unfamiliar. The summation sign, especially double summation signs like in equation 5-2, may be imposing. Two principal points in this chapter are: 1) the expected return of a linear combination is a weighted average of the component expected returns, and 2) the variance of a linear combination is not a simple weighted average of the component variances. Usually, the variance of a combination of assets is lower than a simple weighted average (i.e., a lower risk). Other key topics include the idea of a minimum variance portfolio, the important role of return correlation, and the concept of covariance. Finally, the chapter addresses the utility of the single index model, particularly how it facilitates the computation of portfolio statistics. TEACHING CONSIDERATIONS Chapters 5 and 6 are probably the most quantitative chapters in the text. However, it is just as important to understand the concepts behind the math as it is to use the various formulae. I find it useful to focus on the reduction of risk for combinations of stocks in a portfolio and use the formulae to reinforce that concept. Be sure to incorporate the COV and MINVAR Excel files from the Strong Software into your discussion of this material. The use of these spreadsheet templates may reduce the tendency of students to be overwhelmed by the calculations. A good homework assignment is to provide students with a set of returns on two stocks and ask them to figure out the minimum variance portfolio. This requires using the COV file to get the two variances and the correlation of returns, and then using the MINVAR file to incorporate these statistics into determination of the minimum variance portfolio. I do not require students to memorize any formulae from this chapter except equations 51 and 5-8. I know some instructors will feel differently about this, and depending on teaching style, there is some value in the other approach. It is a matter of personal preference, I think. It may help relieve some anxiety to show that chapter six builds upon chapter five, and that additional clarity comes from this next chapter. It is better to give an exam after chapter six rather than after chapter five. 37 Chapter Five The Mathematics of Diversification ANSWERS TO QUESTIONS 1. Selling stock short brings cash in rather than requiring a cash outflow. In the absence of margin requirements (which is arguably true for large institutional investors), this means there is no initial investment, and any gain on no investment is an infinite return. 2. Student response. 3. The two-security portfolio is preferable, as it has higher expected return per unit of risk. 4. Covariance is the expected value of the product of two numbers. Each of the two numbers is a value minus its mean. Some values lie below the mean, some above. Consequently, each number can be positive or negative, and the product can therefore be positive or negative. Depending on the nature of the dispersions around the means, the expected value of the product can be positive or negative. ~ ~ ~ 5. E[( a~ a )( b b )] cov( a~, b ) and E[( b b )( a~ a )] cov( b , a ) . By the commutative law for multiplication, ab = ba. This means the order of the two products inside the ~ ~ expected value operator can be reversed and cov( a~, b ) cov( b , a~ ). 6. a am a m Where am = correlation between security a and the market a = standard deviation of security a m = standard deviation of the market 7. 0.25 x (4).5 x (6).5= 1.225 8. The size of the error term approaches zero as the number of portfolio components increases. 9. Standard deviations can only be positive, so a negative correlation means the covariance is also negative. 10. In a prediction model, R squared can only increase if additional explanatory variables are added. You cannot lose predictive ability by including additional data. 38 Chapter Five The Mathematics of Diversification ANSWERS TO PROBLEMS 1. (n 2 n) 1700 2 1700 1,444,150 2 2 Note: The first printing of this book had errors in Table 5-5. The covariance values should be as shown in the table below. A .280 .215 .136 .249 A B C D xa 2. B .215 .360 .170 .185 C .136 .170 .203 .114 D .149 .185 .114 .226 B2 A B AB A2 B2 2 A B AB .36 (.28) .5 (.36) .5 AB .28 .36 2(.28) .5 (.36) .5 AB AB ~ ~ cov( A, B ) A B .215 0.677 (.28) .5 (.36) .5 .36 (.28) .5 (.36) .5 (.677) xA .28 .36 2(.28) .5 (.36) .5 (.677) = 4 0.1450 = 69.2% 0.2101 1 4 3. p x1 i (1.05 + 1.20 + 0.90 + 0.95) = 1.025 i 1 4. 2p p2 m 2 ep2 . The error term approaches zero, so p2 (1.025) 2 (.25) 0.263. 5. xA = -.30 xB = .50 xC = .80 N ~ ~ E ( R p ) x i E ( Ri ) i 1 = (-.30)(.14) + (.50)(.16) + (.80)(.12) = .1340 = 13.4% 39 Chapter Five The Mathematics of Diversification N N x i x j ij i j 2 p i 1 j 1 x A2 A2 x B2 B2 xC2 C2 x A xC AC a C x A x B AB A B x B xC BC B C = (-.3)2 (.28) + (.5)2 (.36) + (.8)2 (.203) + (-.3)(.8) AC (.28).5(.203).5 + (-.3)(.5) AB (.28).5(.36).5 + (.5)(.8) BC (.36).5(.203).5 = .0252 + .09 + .1299 - .0572 AC - .0476 AB + .1081 BC AB AC BC ~ ~ cov( A, B ) A B .215 0.677 (.28) .5 (.36) .5 .136 0.570 (.28) .5 (.203) .5 .170 0.629 (.36) .5 (.203) .5 ~ ~ cov( A, C ) A C ~ ~ cov( B , C ) B C 2p .0252 .09 .1299 .0572(.570) .0476(.677) .1081(.629) 0.2483 ~ ~ 6. cov(C, D) C D m2 = (0.90)(0.95)(.32) = 0.274 7. 2 25% .25 =(.25).5 = .5 = 50% 8. BC 9. ~ ~ cov( B , C ) B C .170 = .629 (.36) .5 (.203) .5 cov(1,2) 1 2 m2 m2 40 cov(1,2) 1 2 1.55 1.127 (1.10)(1.25) Chapter Five The Mathematics of Diversification 10. See the computer printouts below. ENTER UP TO 100 RETURNS FOR UP TO FIVE SECURITIES IN LOTUS COLUMNS B, C, D, E, AND F HIT ESC to enter data or ALT S for menu. Return # EXAMPLE 1 2 3 4 5 6 7 8 Sec 1 10.00% 0.0270 0.0120 -0.0220 0.0130 -0.0110 -0.0330 0.0290 0.0550 Sec 2 -8.00% -0.0230 0.0000 -0.0100 0.0340 -0.0230 -0.0610 0.0260 0.0450 Sec 3 20.00% 0.0560 0.0130 -0.0150 0.0150 0.0120 -0.0350 0.0020 0.0470 Sec 4 10.00% 0.0020 0.0040 0.0020 0.0100 -0.0290 -0.0220 0.0000 0.0200 Sec 5 0.00% 0.0330 0.0170 -0.0450 0.0080 -0.0190 -0.0240 -0.0010 0.0560 Security Statistics Mean std dev variance Sec 1 0.88% 0.02736 7.49E-04 Sec Sec Sec Sec Sec 1 2 3 4 5 Sec 1 7.49E-04 7.05E-04 6.28E-04 3.06E-04 7.53E-04 1 2 3 4 5 Sec 1 1.000 0.781 0.824 0.739 0.901 Sec Sec Sec Sec Sec Sec 2 -0.15% 0.03301 1.09E-03 Sec 3 1.19% 0.02786 7.76E-04 COVARIANCE MATRIX Sec 2 Sec 3 1.09E-03 4.43E-04 3.95E-04 5.49E-04 7.76E-04 2.25E-04 7.26E-04 CORRELATION MATRIX Sec 2 Sec 3 1.000 0.481 0.792 0.545 Sec 4 -0.16% 0.01512 2.28E-04 Sec 5 0.31% 0.03054 9.33E-04 Sec 4 Sec 5 2.28E-04 2.95E-04 9.33E-04 Sec 4 1.000 0.534 0.853 1.000 0.640 Sec 5 1.000 ~ E ( R p ) .30(-.15%) + .70(1.19%) = 0.788% 2p x22 22 x32 32 x2 x3 23 2 3 = (.3)2 (.0011) + (.7)2 (.0008) + (.3)(.7)(.481)(.03301)( .02786) = 0.0006 41 Chapter Five The Mathematics of Diversification 11. Because we have the entire set of data, we can simply compute each periodic portfolio return and then determine the mean and variance of this series. Security 1 Security 2 Security 3 Security 4 Security 5 0.0270 -0.0230 0.0560 0.0020 0.0330 0.0120 0.0000 0.0130 0.0040 0.0170 -0.0220 -0.0100 -0.0150 0.0020 -0.0450 0.0130 0.0340 0.0150 0.0100 0.0080 -0.0110 -0.0230 0.0120 -0.0290 -0.0190 -0.0330 -0.0610 -0.0350 -0.0220 -0.0240 0.0290 0.0260 0.0020 0.0000 -0.0010 0.0550 0.0450 0.0470 0.0200 0.0560 Portfolio mean Portfolio variance ~ E ( R p ) 0.0041 Average 0.0190 0.0092 -0.0180 0.0160 -0.0140 -0.0350 0.0112 0.0446 0.0041 0.0005 p2 = 0.0005 12. The minimum variance combinations are as follows: Pair 1,2 1,3 1,4 1,5 2,3 2,4 2,5 3,4 3,5 4,5 42 Proportion in the First 100% 98.72% 100% 100% 34% 100% 41.5% 0.61% 80.5% 100% Proportion in the Second 0% 1.28% 0% 0% 66% 0% 58.5% 99.39% 19.5% 0% Chapter Five The Mathematics of Diversification 13. See the spreadsheet extracts below from the COV and MINVAR3 templates. The minimum variance portfolio involves going short security 5 and buying securities 3 and 4. ENTER UP TO 100 RETURNS FOR UP TO FIVE SECURITIES IN LOTUS COLUMNS B, C, D, E, AND F Return # EXAMPLE 1 2 3 4 5 6 7 8 Run the \S macro for the main menu. 4 5 -8.00% 20.00% 10.00% 0.0020 0.0330 0.0040 0.0170 0.0020 -0.0450 0.0100 0.0080 -0.0290 -0.0190 -0.0220 -0.0240 0.0000 -0.0010 0.0200 0.0560 3 10.00% 0.0560 0.0130 -0.0150 0.0150 0.0120 -0.0350 0.0020 0.0470 Security Statistics Mean std dev Variance 3 4 5 3 1.19% 0.02786 7.76E-04 3 1.000 0.534 0.853 4 -0.16% 0.01512 2.28E-04 5 0.31% 0.03054 9.33E-04 CORRELATION MATRIX 4 5 1.000 0.640 1.000 DETERMINING THE MINIMUM VARIANCE THREE-SECURITY PORTFOLIO Enter the INPUT in the boxes provided below:Variance of Stock A: 0.0008 (Standard deviation = 0.0279 Variance of Stock B: 0.0002 (Standard deviation = 0.0151 Variance of Stock C: 0.0009 (Standard deviation = 0.0305 Correlation between Stocks A and B: 0.5340 COV(A,B)= 0.0002 Correlation between Stocks B and C: 0.8530 COV(B,C)= 0.0004 Correlation between Stocks A and C: 0.6400 COV(A,C)= 0.0005 The Minimum Variance Combination is: 43 Chapter Five The Mathematics of Diversification 14.65% <-Stock A 135.65% <-Stock B -50.30% <-Stock C With these proportions, Portfolio variance is :-> Portfolio standard deviation is :-> 0.0001 0.0120 14. CFA Guideline Answer (reprinted with permission from the CFA Study Guide, CFA Institute, Charlottesville, VA. All Rights Reserved). A. Subscript OP refers to the original portfolio, ABC to the new stock, and NP to the new portfolio. i. The expected return is 0.728 percent. E(rNP) = wOPE(rOP) + wABCE(rABC) E(rNP) = 09 (.67) + 0.1 (1.25) E(rNP) = 0.603 + 0.125 = 0.728% ii. The expected covariance is 2.80. COV = r (σOP) (σABC) COV = 0.40 (2.37) (2.95) = 2.7966 or 2.80 iii. The expected standard deviation is 2.27 percent. σNP = [ wOP2 σNP2 + wABC2 σABC2 + wOP wABC COV ]1/2 σNP = [ 0.92 (2.372) + 0.12 (2.952) + 2 (0.9) (0.1) (2.80) ]1/2 σNP = [4.5497 + 0.0870 + 0.5040]1/2 = 2.2673 or 2.27% B. Subscript OP refers to the original portfolio, GS to government securities, and NP to the new portfolio. i. The expected return is 0.645 percent. E(rNP) = wOPE(rOP) + wGSE(rGS) E(rNP) = 09 (.67) + 0.1 (0.42) E(rNP) = 0.603 + 0.042 = 0.645% ii. The expected covariance is 0. 44 Chapter Five The Mathematics of Diversification COV = r (σOP) (σGS) COV = 0 (2.37) (0) = 0 iii. The expected standard deviation is 2.13 percent. σNP = [ wOP2 σNP2 + wGS2 σGS2 + wOP wGS COV ]1/2 σNP = [ 0.92 (2.372) + 0.12 (0) + 2 (0.9) (0.1) (0) ]1/2 σNP = [4.5497 + 0 + 0]1/2 = 2.133 or 2.13% C. Adding the risk-free government securities would cause the beta of the new portfolio to be lower. The new portfolio beta will be a weighted average of the individual security betas in the portfolio; the presence of the risk-free securities would lower that weighted average. D. The comment is correct. Although the standard deviations and expected returns of the two securities under consideration are the same, the covariances between each security and the original portfolio are unknown, making it impossible to draw the conclusion stated. For instance, if the covariances are different, selecting one security over the other may result in a lower standard deviation for the portfolio as a whole. In such a case, that security would be the preferred investment if all other factors are equal. E. i. Grace clearly expressed the sentiment that the risk of loss was more important to her than the opportunity for return. Using variance (or standard deviation) as a measure of risk in her case has a serious limitation because it does not distinguish between positive and negative price movements. ii. Two alternative risk measures that could be used instead of variance are: Range of Returns, which considers the highest and lowest expected returns in the future period, with a larger range being a sign of greater variability and therefore of greater risk; Semivariance, which can be used to measure expected deviations of returns below the mean or some other benchmark, e.g., zero. 45 Chapter Five The Mathematics of Diversification Either measure would potentially be superior to variance for Grace. Range of returns would help to highlight the full spectrum of risk she is assuming, especially the downside portion of the range about which she is so concerned. Semivariance would also be effective, because it implicitly assumes that the investor wants to minimize the likelihood of returns falling below some target rate; in Grace’s case, the target would be set at zero (to protect against negative returns). 46