Principles of Finance Grzegorz Trojanowski Lecture 4: Calculating efficient portfolios October 26, 2004 Principles of Finance - Lecture 4 1 Lecture 4 material • Required reading: 9 Elton et al., Chapters 5, 6 • Supplementary reading: 9 Luenberger, Chapter 6 9 Sharpe et al., Chapters 7, 8 9 Alexander et al., Chapters 7-9 October 26, 2004 Principles of Finance - Lecture 4 2 1 Lecture 4: Checklist • By the end of this lecture you should: 9 Understand the concepts of an efficient portfolio 9 Be able to compute two efficient portfolios when short sales are allowed 9 Be able to compute the efficient frontier when short sales are allowed 9 Be able to compute the capital market line when there is a risk free asset 9 Be able to compute the efficient frontier when short sales are not allowed October 26, 2004 Principles of Finance - Lecture 4 3 Last lectures recap (1) • Consider N assets with the expected return and variance-covariance matrix given by: ⎡ σ 12 σ 12 ⎡ Ε(r1 ) ⎤ ⎢ ⎢ Ε( r ) ⎥ σ σ 22 Ε( R ) = ⎢ 2 ⎥ and Ω = ⎢ 21 ⎢ M ⎢ M ⎥ M ⎢ ⎢Ε ( r ) ⎥ ⎣ N ⎦ ⎣σ N 1 σ N 2 • L σ 1N ⎤ ⎥ L σ 2N ⎥ O M ⎥ ⎥ L σ N2 ⎦ The expected return of the portfolio described by the (column) vector of weights W equals to E(rP) = W T E(R) • The portfolio variance is given by σ P2 = W Τ ΩW • Finally, the covariance between two portfolios whose portfolio weight vectors are W1 and W2 is given by σ = W Τ ΩW 12 1 2 October 26, 2004 Principles of Finance - Lecture 4 4 2 Last lectures recap (2) • The feasible set is the set that contains all possible portfolios made up of different combinations of the N assets • The minimum variance set, or the envelope, comprises the portfolios that have the lowest standard deviation (or variance) for any given expected return • Today: how to construct optimal portfolios given investor’s preferences? October 26, 2004 Principles of Finance - Lecture 4 5 Efficient portfolios (1) • Portfolios that have the lowest standard deviation given their expected return, and the highest expected return given their standard deviation, are known as efficient portfolios • The problem faced by the investor is to combine individual assets into a portfolio that is efficient, and given his or her risk preferences, to choose the efficient portfolio that is optimal October 26, 2004 Principles of Finance - Lecture 4 6 3 Efficient portfolios (2) • Formally, the problem is to choose portfolio weights {wi: i = 1, …, N} to solve the following minimisation problem N N N i =1 i =1 j =1 j ≠i min σ P = ∑ wi2σ i2 + ∑ ∑ wi w jσ ij N N i =1 i =1 subject to ∑ wi = 1 and ∑ wi Ε(ri ) = Ε(rP ) • Solving the minimisation problem for each possible expected return E(rP) yields the minimum variance frontier or the envelope October 26, 2004 Principles of Finance - Lecture 4 7 Efficient portfolios (3a) • The efficient frontier is the segment of the minimum variance frontier that lies above the GMV portfolio GMV Portfolio October 26, 2004 Principles of Finance - Lecture 4 8 4 Efficient portfolios (3b) • The efficient frontier is the segment of the minimum variance frontier that lies above the GMV portfolio GMV Portfolio October 26, 2004 Principles of Finance - Lecture 4 9 Efficient portfolios: Example (1) • Consider the four assets and two portfolios used in the example last week A 1 B C D E Variance-covariance matrix F Mean return vector 0.10 0.01 0.03 0.05 H I Portfolio weights W1 W2 6% 0.2 0.2 2 3 G 4 0.01 0.30 0.06 -0.04 8% 0.3 0.1 5 0.03 0.06 0.40 0.02 10% 0.4 0.1 6 0.05 -0.04 0.02 0.50 15% 0.1 0.6 • The four assets and all the possible combinations of the two portfolios can be plotted in mean-standard deviation space October 26, 2004 Principles of Finance - Lecture 4 10 5 Efficient portfolios: Example (2) 16.00% 15.00% Stock D 14.00% Expected return 13.00% 12.00% 11.00% 10.00% Stock C 9.00% 8.00% Stock B 7.00% 6.00% 5.00% 0.00% Stock A 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% Standard deviation October 26, 2004 Principles of Finance - Lecture 4 11 Efficient portfolios: Example (3) • It can be seen that stocks B and C are inefficient: there exist combinations of the two portfolios that have the same standard deviation but offer higher expected return • Moreover, not all the combinations of the two portfolios are minimum variance portfolios since if they were, then all other assets would lie within the graph (note that now stocks A and D lie outside the graph October 26, 2004 Principles of Finance - Lecture 4 12 6 Identifying efficient portfolios (1) • Consider a rate of return δ and the line originating from the point (0, δ) and tangent to the efficient set • The tangency point is known as an envelope portfolio E(rP) P δ October 26, 2004 σP Principles of Finance - Lecture 4 13 Identifying efficient portfolios (2) • In practice, the efficient frontier can be found by solving the following optimisation problem for every value of δ max Θ = Ε(rP − δ ) σP N subject to ∑ wi = 1 i =1 N N N N i =1 i =1 j =1 j ≠i i =1 where σ P = ∑ wi2σ i2 + ∑ ∑ wi w jσ ij and ∑ wi Ε(ri ) = Ε(rP ) October 26, 2004 Principles of Finance - Lecture 4 14 7 Identifying efficient portfolios (3) • Proposition 1: The weights of the envelope portfolio P are proportional to the vector z that solves the set of simultaneous equations E(R) - δ = Ω z where E(R) - δ is the ‘excess’ return vector given by ⎡ Ε(r1 ) − δ ⎤ ⎢ Ε( r ) − δ ⎥ ⎥ Ε( R ) − δ = ⎢ 2 M ⎥ ⎢ ⎢Ε ( r ) − δ ⎥ ⎣ N ⎦ and Ω is the variance-covariance matrix October 26, 2004 Principles of Finance - Lecture 4 15 Identifying efficient portfolios (4) • This system of equations is solved by z = Ω-1 [E(R) - δ] • Once we have solved this system for z, we can recover the actual weights of the envelope portfolio by re-scaling the elements of z so that they sum up to unity ∀i = 1,K, N : wi = zi N ∑zj j =1 October 26, 2004 Principles of Finance - Lecture 4 16 8 Identifying efficient portfolios (5) • Proposition 2: Given any two envelope portfolios, x and y, any combination of these two portfolios α x + (1 - α) y will also be an envelope portfolio • Proposition 1 gives a method of computing an envelope portfolio given some arbitrarily chosen constant δ • Proposition 2 says that once we have computed two such envelope portfolios, we can trace out the entire envelope simply by combining the two portfolios that we have computed in varying proportions October 26, 2004 Principles of Finance - Lecture 4 17 2 envelope portfolios: Example (1) • Recall the four-assets example discussed so far • For δ = 0, we can use Proposition 1 to solve for the first envelope portfolio A 1 B C D E F Variance-covariance matrix E[R] 2 0.10 0.01 0.03 0.05 3 0.01 0.30 0.06 -0.04 8.0% 4 0.03 0.06 0.40 0.02 10.0% 5 0.05 -0.04 0.02 0.50 15.0% 6.0% 6 =MMULT(MINVERSE(A2:D5),F2:F5) 7 8 z_x x 9 0.3861 0.3553 10 0.2567 0.2362 11 0.1688 0.1553 12 0.2752 0.2532 October 26, 2004 Cell C9 contains the function =B9/SUM(B$9:B$12) This function is copied down to cells C10:C12 Principles of Finance - Lecture 4 18 9 2 envelope portfolios: Example (2) • Arbitrarily choosing the value of δ (e.g. δ = 0.04), we can solve for the second envelope portfolio Compute ‘excess’ 1 Variance-covariance matrix E[R] E[R]- δ 2 0.10 0.01 0.03 0.05 6.0% 2.0% returns for δ chosen, e.g. cell G2 contains formula 3 0.01 0.30 0.06 -0.04 8.0% 4.0% =F2-F$7 0.03 0.06 0.40 0.02 10.0% 6.0% 0.02 0.50 15.0% 11.0% A 4 5 6 7 B C D Cell G9 contains the function =F9/SUM(F$9:F$12) 0.05 -0.04 This function is copied down to cells G10:G12 E δ F G 4.0% 8 z_x x z_y y 9 0.3861 0.3553 0.0404 0.0782 10 0.2567 0.2362 0.1386 0.2684 11 0.1688 0.1553 0.1151 0.2227 12 0.2752 0.2532 0.2224 0.4307 Enter here an arbitrarily chosen value of δ =MMULT(MINVERSE(A2:D5),G2:G5) October 26, 2004 Principles of Finance - Lecture 4 19 2 envelope portfolios: Example (3) • We can then compute the expected returns, variances and standard deviations as well as the covariance for the two envelope portfolios =MMULT(I4:L4,F2:F5) I 1 2 3 4 J K L 0.1553 0.2532 Transpose x 0.3553 0.2362 =MMULT(I4:L4,MMULT (A2:D5,G9:G12)) Transpose y 0.0782 0.2684 0.2227 0.4307 6 Mean(x) 9.37% Mean(y) 11.30% 7 Var(x) 0.0862 Var(y) 0.1414 8 Sigma(x) 29.37% Sigma(y) 37.60% 5 =SQRT(L7) =MMULT(I2:L2,MMULT (A2:D5,G9:G12)) 9 10 Cov(x,y) 0.1040 11 Corr(x,y) 0.9419 October 26, 2004 =J10/(J8*L8) Principles of Finance - Lecture 4 20 10 2 envelope portfolios: Example (4) • From Proposition 2 we can now trace the entire envelope by combining the two envelope portfolios that we have computed in varying proportions I J K L 14 W_x W_y Std. dev. Return 15 -2.00 3.00 60.78% 15.17% 16 -1.75 2.75 57.65% 14.68% : : : : : 23 0.00 1.00 37.60% 11.30% +J16^2*$L$7 24 0.25 0.75 35.20% 10.82% +2*I16*J16*$J$10) 25 0.50 0.50 33.00% 10.34% 26 0.75 0.25 31.04% 9.86% 27 1.00 0.00 29.37% 9.37% : : : : : 34 2.75 -1.75 29.00% 5.99% 35 3.00 -2.00 30.60% 5.51% October 26, 2004 =1-I15 =I15*$J$6+J15*$L$6 =SQRT(I16^2*$J$7 Portfolio y Portfolio x Principles of Finance - Lecture 4 21 2 envelope portfolios: Example (5) • Plotting the portfolios from the data table yields the envelope and the efficient frontier 16.00% Asset D Portfolio y 14.00% Portfolio x 12.00% Return 10.00% Asset C 8.00% Asset B 6.00% Asset A 4.00% 2.00% 0.00% 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% Std. deviation October 26, 2004 Principles of Finance - Lecture 4 22 11 Capital market line (1) • Suppose there exist a risk-free asset that offers the return rf • Consider a solution, z, to the simultaneous equations E(R) - rf = Ω z • This system is solved by z = Ω-1 [E(R) - rf], and the weights, z, re-scaled so that sum to unity • This yields the portfolio, M • If all the investors face the same optimisation problem, then M is the market portfolio October 26, 2004 Principles of Finance - Lecture 4 23 Capital market line (2) • Now, consider combinations of the market portfolio M and the risk-free asset • The expected return and standard deviation of such combinations are given by Ε(rP ) = αr f + (1 − α )Ε(rM ) σ P = α 2σ 2f + (1 − α ) 2 σ M2 + 2α (1 − α )σ fM = (1 − α )σ M • The locus of all such combinations is the capital market line October 26, 2004 Principles of Finance - Lecture 4 24 12 Capital market line (3) Portfolio expected return Capital market line (CML) Market portfolio (M) Risk-free rate (rf) Portfolio standard deviation October 26, 2004 Principles of Finance - Lecture 4 25 Capital market line: Example • For the four-asset case discussed so far, suppose that the risk-free asset earns a rate of return rf = 3% • We can then compute the market portfolio =MMULT(TRANSPOSE(C41:44),F2:F5) 38 A B rf 3.00% C D 39 =F2-$B$38 40 E[R]- rf z M 41 3.0% 0.1268 0.1924 42 5.0% 0.1682 0.2551 43 7.0% 0.1285 0.1949 44 12.0% 0.2356 0.3575 =MMULT(MINVERSE (A2:D5),G2:G5) October 26, 2004 E F rM 10.51% σM 33.75% =MMULT(TRANSPOSE (C41:44),MMULT (A2:D5,C41:C44)) Cell C41 contains the function =B41/SUM(B$41:B$44) This function is copied down to cells C42:C44 Principles of Finance - Lecture 4 26 13 Efficient frontier – recap (1) • The efficient frontier is found by solving the following optimisation problem for every value of δ max Θ = Ε(rP − δ ) σP N subject to ∑ wi = 1 i =1 N N N N i =1 i =1 j =1 j ≠i i =1 where σ P = ∑ wi2σ i2 + ∑ ∑ wi w jσ ij and ∑ wi Ε(ri ) = Ε(rP ) • The weights of tangency portfolio, P, are proportional to the vector z that solves the set of simultaneous equations E(R) - δ = Ω z October 26, 2004 Principles of Finance - Lecture 4 27 Efficient frontier – recap (2) • This system can be solved by z = Ω-1 [E(R) - δ], and the weights of the portfolio can be recovered by xi = zi N ∑zj j =1 • However, the solution may involve negative weights • A negative weight implies that the asset is short sold by the investor, and that the proceeds from the short sale can be used immediately to purchase another asset • In practice, many investors are prohibited from selling an asset short, and those are not likely to be able to use all of the short sale proceeds October 26, 2004 Principles of Finance - Lecture 4 28 14 Efficient frontier when short sales are not allowed • We can recast the portfolio choice problem with the added restriction of no short sales by including the following constraint in the previous optimisation problem xi ≥ 0 ∀ i = 1, …, N • This considerably complicates the problem, since the tangency portfolio weights can no longer be found as the solution to a set of simultaneous equations • Instead, we can solve the constrained optimisation problem numerically (e.g. using Excel’s Solver tool) October 26, 2004 Principles of Finance - Lecture 4 29 Efficient frontier with no short sales: Example (1) • Let’s come back to the four-asset example discussed before A B C D E F 1 Variance-covariance matrix E[R] 2 0.10 0.01 0.03 0.05 6.0% 3 0.01 0.30 0.06 -0.04 8.0% 4 0.03 0.06 0.40 0.02 10.0% 5 0.05 -0.04 0.02 0.50 15.0% • We can then arbitrarily choose a value for a constant, say δ = 5%, and set up the optimisation problem in Excel in the following way October 26, 2004 Principles of Finance - Lecture 4 30 15 Efficient frontier with no short sales: Example (2) A 15 B C Delta 5.00% Enter here an arbitrarily chosen constant, e.g. 5% 16 17 Optimal portfolio weights 18 x_1 0.2500 19 x_2 0.2500 20 x_3 0.2500 21 x_4 0.2500 22 Total 1.0000 =MMULT(TRANSPOSE(C18:C21), F2:F5) 24 Portfolio mean 9.75% =SQRT(MMULT(TRANSPOSE (C18:C21),MMULT(A2:D5,C18:C21))) 25 Portfolio sigma 31.22% 26 Theta 0.1521 Enter here some starting values for the portfolio weights =SUM(C18:C21) 23 October 26, 2004 =(C24-C15)/C25 Principles of Finance - Lecture 4 31 Efficient frontier with no short sales: Example (3) • The optimal portfolio weights can be computed using the Excel’s Solver tool • Go to the Tools/Solver option which will generate the following dialogue box (see last lecture) Note that this time we choose maximisation option Note that this time we change a number of cells (four of them) rather than just one Slide 33 shows how to add new constraints October 26, 2004 Principles of Finance - Lecture 4 32 16 Efficient frontier with no short sales: Example (4) • The user-defined entries in the dialogue box tell Solver to change the portfolio weights in a way that maximises the value of Ε ( RP − δ ) N and ∑ xi = 1 σP subject to constraints ∀i : xi ≥ 0 i =1 October 26, 2004 Principles of Finance - Lecture 4 33 Efficient frontier with no short sales: Example (5) • The constraints can be added by clicking on the Add button (in the Solver dialogue box), which generates the following dialogue box (completed here for the first constraint) Enter here the addresses of the cells on which you want to impose a constraint Select here the type of constraint Enter here the value corresponding to the constraint imposed Confirm here clicking ‘OK’ October 26, 2004 Principles of Finance - Lecture 4 by 34 17 Efficient frontier with no short sales: Example (6) A 15 B C Delta 5.00% 16 In the example discussed, the constrained optimisation procedure (explained on Slides 30-33) yields the following result 17 Optimal portfolio weights 18 x_1 0.0000 19 x_2 0.2618 20 x_3 0.2500 21 x_4 0.4981 22 Total 1.0000 23 October 26, 2004 24 Portfolio mean 11.97% 25 Portfolio sigma 41.18% 26 Theta 0.1692 Principles of Finance - Lecture 4 35 Efficient frontier with no short sales: Example (7) • By changing the value of δ, we can trace out the entire efficient set without the short sales 16.00% Asset D Expected return 14.00% 12.00% 10.00% Asset C 8.00% Asset B 6.00% 4.00% 0.00% Asset A 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% Standard deviation • Note that now having two portfolios is not enough to derive the whole set! October 26, 2004 Principles of Finance - Lecture 4 36 18