OPTIMAL PORTFOLIO SELECTION: A NOTE
(Draft)
Ignacio Vélez-Pareja
ivelez@javeriana.edu.co
Department of Management
Universidad Javeriana
Bogotá, Colombia
Working Paper N 8
First version: August 23, 2000
This version: August 8, 2001
OPTIMAL PORTFOLIO SELECTION: A NOTE
(Draft)
Abstract
Usually in financial textbooks and courses the theory of portfolio selection
is taught in a strictly theoretical way. There is a model (Markowitz) that
stipulates that an investor has preferences and that she will choose the
best portfolio, given her preference curves and an efficient frontier. On the
other hand, the Capital Asset Pricing Model (CAPM) is presented as it is: a
genial idea that served to simplify and to make operative the Markowitz
setup.
Most students and practitioners conclude that those models are just
inapplicable theory. This is the most rational behavior one can expect.
What can an investor do with the textbook recipes to configure an optimal
portfolio? Very little.
My purpose with this note is to rescue a simple procedure presented by
Black (1972), Merton (1973) and later by Levy and Sarnat (1982), Elton
and Gruber (1995) and Benninga (1997). They just propose that the
optimal portfolio can be found maximizing the slope of the line that joins
the point of risk-free return and the efficient frontier. When this maximum
tangent is reached, that line is the capital market line (CML) (it is tangent
to the efficient frontier). This is a simple procedure that does not require
one to calculate the efficient frontier and is an easy task with Excel Solver.
It is just one point of the efficient frontier. An example is presented.
Keywords
CAPM, efficient frontier, porfolio selection, capital market line, optimal
portfolio
JEL Classification: G11,G12
OPTIMAL PORTFOLIO SELECTION: A NOTE
(Draft)
Usually in financial textbooks and courses, portfolio selection is taught in
a strictly theoretical way. There is a model (Markowitz) that stipulates that
an investor has preferences and that she will choose the best portfolio,
given her preference curves and an efficient frontier. On the other hand,
the Capital Asset Pricing Model (CAPM) is presented as it is: an ingenious
idea that served to simplify and to make operative the Markowitz setup.
Most students and practitioners conclude that those models are just
inapplicable theory. This is the most rational behavior one can expect.
What can an investor do with the textbook recipes to configure an optimal
portfolio? Very little.
My purpose with this note is to rescue a simple procedure presented by
Black (1972), Merton (1973) and later by Levy and Sarnat (1982), Elton
and Gruber (1995) and Benninga (1997). They just propose that the
optimal portfolio can be found maximizing the slope of the line that joins
the point of risk-free return and the efficient frontier. When this maximum
tangent is reached, that line is the capital market line (CML) (it is tangent
to the efficient frontier). This is a simple procedure that does not require
one to calculate the efficient frontier and is an easy task with Excel Solver.
It is just one point of the efficient frontier. An example is presented.
How to calculate the optimal portfolio
What is an optimal portfolio? It is, according to the CAPM theory, a
portfolio that lies in the efficient frontier and combined with certain
proportion of risk free investment, and given a desired risk level,
maximizes the return of the combined portfolio. This definition is valid
even if the desired risk level is less than the minimum defined by the
efficient frontier. Given the risk free rate of return, how to determine the
optimal risky portfolio?
That optimal risky portfolio is just the point of tangency between the
Capital Market Line and the efficient frontier. As this optimal portfolio has
to lie along the efficient frontier, then the point of tangency is located at
the line with the maximum tangent between that line and the horizontal
line. This solution is very good because it is not easy to determine the
indifference curves for each decision maker. However, as it was said
above, it is not necessary to generate the indifference curves nor even the
efficient frontier given the Separation Theorem posited by Tobin.
According to the CAPM theory, the investor will prefer a position in the
“market portfolio” either levered or unlevered. Then, the optimal portfolio
is given by this optimization problem.
3
Max tnθ =
Rm − r
m
m
∑ ∑α α σ
k
j
kj
k=1 i=1
s.t.
m
∑α
i
=1
i=1
where αi is the is weight of stock i in the portfolio, σkj is the covariance
between stocks k and j, Rm is the portfolio return, r is the risk-free rate of
return and m is the number of stocks in the analysis. The restriction that
the α’s be positive might be included. In that case there is no short
position.
This idea can be seen in the next figure
Figure 1. Capital Market Line, Efficient frontier and optimal portfolio
θ
Return
Capital Market Line
Optimal portfolio m
Standard Deviation
The solution to his optimization problem produces the αj‘s and hence the
optimal portfolio return. Done this, the investor will select the desired risk
level (for instance, she will reduce the risk beyond the minimum defined by
the efficient frontier) combining the optimal portfolio with an appropriate
portion of risk free investment.
I have been examining the solution for historical data from the Bolsa de
Bogota (Bogota Stock Exchange), and the resulting optimal portfolio has
been composed for a few stocks (in some cases the optimal solution is
comprised of only one stock). This apparently contradicts the basic idea
behind portfolio selection: a high degree of diversification. However, when
contrasted with the real practice performed by stock traders, they
intuitively form very small portfolios with predominance of one or two
stocks. The preliminary results of this work, performed by Professor Irina
4
Dubova and four Business Administration senior students show that the
optimal solutions for 42 portfolios (one for every quarter and redefined
quarterly) tend to confirm this assertion.
An example
Suppose four stocks with the following returns:
Month
1
2
3
4
5
6
7
8
9
10
11
12
Average
Variance
Standard
Deviation
Weight
Stock 1
18.25%
2.03%
-11.15%
-14.50%
21.37%
14.28%
11.50%
-6.11%
-2.81%
-14.23%
-10.71%
15.15%
1.92%
1.68%
12.95%
Stock 2
19.13%
2.14%
-5.27%
-6.27%
4.01%
4.00%
9.40%
-4.04%
-1.07%
-8.56%
-8.83%
10.22%
1.24%
0.68%
8.23%
Stock 3
18.87%
13.42%
2.32%
-10.81%
-21.58%
12.57%
15.42%
-6.32%
5.71%
4.00%
2.12%
-24.44%
0.94%
1.83%
13.54%
Stock 4
7.59%
-7.83%
-25.11%
-10.08%
11.54%
-13.72%
-18.73%
11.50%
18.72%
9.25%
20.02%
15.72%
1.57%
2.26%
15.03%
25%
25%
25%
25%
The excess return matrix is:
Month Stock 1 Stock 2 Stock 3 Stock 4
1 16.33% 17.89% 17.93% 6.02%
2 0.11% 0.90% 12.48% -9.40%
3 -13.07% -6.51% 1.38% -26.68%
4 -16.42% -7.51% -11.75% -11.65%
5 19.45% 2.77% -22.52% 9.97%
6 12.36% 2.76% 11.63% -15.29%
7 9.58% 8.16% 14.48% -20.30%
8 -8.03% -5.28% -7.26% 9.93%
9 -4.73% -2.31% 4.77% 17.15%
10 -16.15% -9.80% 3.06% 7.68%
11 -12.63% -10.07% 1.18% 18.45%
12 13.23% 8.98% -25.38% 14.15%
The transpose matrix is found by the Excel Search and Reference function,
=TRANSPOSE(Matrix).
5
Month
1
2
3
4
5
6
7
8
9
10
11
12
Stock 1
16.33%
0.11% -13.07% -16.42% 19.45% 12.36%
9.58% -8.03% -4.73% -16.15% -12.63% 13.23%
Stock 2
17.89%
0.90% -6.51% -7.51%
8.16% -5.28% -2.31% -9.80% -10.07%
Stock 3
17.93% 12.48%
Stock 4
2.77%
2.76%
1.38% -11.75% -22.52% 11.63% 14.48% -7.26%
6.02% -9.40% -26.68% -11.65%
4.77%
9.97% -15.29% -20.30% 9.93% 17.15%
3.06%
8.98%
1.18% -25.38%
7.68% 18.45% 14.15%
By matrix multiplication and dividing by the number of observations
(n=12), then the covariance matrix is:
Stock 1
Stock 2
Stock 3
Stock 4
Stock 1
Stock 2
Stock 3
Stock 4
0.01676831 0.00936991 -0.00042371 0.00097519
0.00936991 0.00677348 0.00213887 -0.00038209
-0.00042371 0.00213887 0.01833034 -0.00758114
0.00097519 -0.00038209 -0.00758114 0.02258129
1. Define a weight vector for each stock
Stock 1
25%
Stock 2
25%
Stock 3
25%
Stock 4
25%
2. Calculate the average return for the portfolio. It is the scalar product
of the weight vector times the return vector (the return vector is the
average return for the stocks in the first table). In Excel use
SUMPRODUCT. For this level of participation the average return of
the portfolio is 1.42%.
3. Multiply the weight vector by the covariance matrix (you will obtain a
vector). Use the Excel function for matrix multiplication. In the
example.
Stock 1
Weights vector x covariance
matrix
Stock 2
Stock 3
Stock 4
0.00667243 0.004475046 0.00311609 0.00389831
4. The portfolio variance is calculated as the scalar product of the
participation or weight vector times the vector obtained in 3. In this
example the portfolio variance is 0.00454047. The portfolio standard
deviation is the square root of the variance.
5. Assume that the risk free rate is 1.5%, then construct the tangent:
tn θ =
Rm − r
m
m
∑∑α α σ
k
j
kj
k=1 i=1
6. Use Solver to maximize the tangent subject to the conditions that
the sum of the weights is one and the weights are nonnegative.
6
7. As an Excel operation (the image is in Spanish, but Excel users will
recognize the correspondent words and functions) as in figures 2
and 3
Figure 2. Setup for the optimal portfolio problem.
Figure 3. Solver for optimal portfolio.
8. When you press the Resolver (Solve) button, you get the optimal
portfolio composition.
Stock 1
Weights
Stock 2
92.18%
0.00%
7
Stock 3
0.00%
Stock 4
7.82%
This produces a portfolio return of 1.90% and a portfolio variance
0.12052486.
Bibliographic References
Benninga, Simon Z. (1997),Financial Modeling, MIT Press.
Black, F. (1972). "Capital Market Equilibrium with Restricted Borrowing."
Journal of Business 45 (July), 444-455.
Levy, Haim and Marshall Sarnat (1982), Capital Investment and Financial
Decisions, 2 nd Ed. Prentice Hall.
Merton, R. C. (1973), "An Analytic Derivation of the Efficient Portfolio
Frontier." Journal of Financial and Quantitative Analysis 7 (September),
1851-1872.
Elton Edwin J. and Martin Jay Gruber (1995), Modern portfolio theory and
investment analysis, Wiley.
8
View publication stats