Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
261 views

Assignment 3

This document contains an assignment for a portfolio optimization course. It introduces the Markowitz framework for portfolio selection and describes how to formulate and solve several related optimization problems in MATLAB. Students are asked to: 1) maximize expected return without short selling, 2) minimize variance allowing short selling, and 3) minimize variance subject to an expected return constraint with and without short selling. They are also asked to compute the efficient frontier by solving a weighted optimization problem for different weights.

Uploaded by

Dani Wedaje
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
261 views

Assignment 3

This document contains an assignment for a portfolio optimization course. It introduces the Markowitz framework for portfolio selection and describes how to formulate and solve several related optimization problems in MATLAB. Students are asked to: 1) maximize expected return without short selling, 2) minimize variance allowing short selling, and 3) minimize variance subject to an expected return constraint with and without short selling. They are also asked to compute the efficient frontier by solving a weighted optimization problem for different weights.

Uploaded by

Dani Wedaje
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

October 6, 2006

Optimization I /MN 1
Martin Berggren

Uppsala University
Information Technology
Scientific Computing

Assignment 3: Portfolio Optimization


Submit a written report containing a short presentation of the problem, results, discussion, source code, and a
print-out of the result from your Matlab sessions. Answer all question and comment your results. If the report
is incomplete, I will return it for completion before starting to grade it. You may work alone or in groups of two
(not three). Both persons in a group should contribute to the solution and the report. Discussions between
the groups are encouraged. If you receive substantial help from another group, say so in the report. You are
not allowed to copy solutions or computer codes from others. I prefer answering questions in person or by
phone (instead of email).
The absolute deadline for submitting this years assignments is 2006-11-13. You will obtain a 1/2-point bonus
point on the final exam if a correct solution for this assignment is handed in at the latest 2005-10-24.

The Markowitz framework for portfolio selection


The return of investment instruments such as publicly-traded stocks are unknown at the time when the investment is made. Typically, there is a tradeoff between risk and return: a higher return is associated with a higher
risk. A quantitative study of such tradeoffs is made particularly simple using an analysis from the 1950s due
to Markowitz assuming that the returns of the investments are random variables with a known multivariate
normal distribution.
An asset is an investment instrument that can be bought or sold. The rate of return of the asset is the number
r satisfying X = ( + r)X , where X and X are the prices of the asset at purchase and selling, respectively.
As an example, the rate of return from deposits in a bank account is the interest rate. Assume that an investor
wants to select a portfolio of n possible assets. The investor puts the fraction wi of available funds into asset i
with rate of return ri , i = , . . . , n. We assume that all available funds are invested, so Pni= wi = . Thus, rate of
return for the portfolio will be r = Pni= ri wi .
The situation when a weight wi is negative corresponds to a short selling of the asset, that is, the investor
buys the asset, sells it to someone else, and use the amount thus received to invest in the other assets. Short
selling can significantly increase the total return of the portfolio at the price of a substantially increased risk.
When short selling is not allowed, we require that wi C , whereas there are no constraints on wi when
unlimited short selling of the asset is allowed.
The rates of returns are often not known in advance. We assume in the Markowitz framework that we
possess estimates of their expected values ri = E(ri ) and their covariances i j = E((ri ri )(r j r j)). These
estimates can for instance be obtained by analysis of previous performance of the assets. The properties of
the expectation value yields that the expected rate of return and the variance of the of the portfolio will be
r = Pni= ri wi and = Pni= Pnj= wi w ji j.
We will consider portfolios of five assets with yearly expected rates of returns and covariances according
to Table 1. Note that any matrix of covariances is symmetric and positive semidefinite.

To do
1. Formulate the linear program of maximizing the expected return of the portfolio when short selling is
not allowed. Solve the linear program by inspection (that is, do not use software or hand calculation).
What is the variance of the portfolio?
2. Formulate the quadratic program of minimizing the variance of the portfolio (short selling allowed).
Set up the KKT system and solve it using Matlab. Report the computed weights and the variance of
corresponding portfolio and compare with 1.

Table 1: Covariances and rates of returns for five assets. Note that the numberis in the table should be multiplied with
. For instance, r = . and = .. These values are borrowed from Example 6.11 of Chapter 6 in Luenberger:
Investment Science, Oxford Univ. Press (1997).

Asset

Covariances ( )
. . . .
. . .
.
. .
.
. .
.

ri ()
.
.
.
.
.

Note that Matlab can handle block matrices. Assume that A is an n-by-n matrix and b a column vector
of dimension n, and that these can be blocked as
A=

A
A

A
,
A

b
b = ,
b

where
matrix A has as many rows as matrix A and matrix A has as many rows as matrix A ,
matrix A has as many columns as matrix A and matrix A has as many columns as matrix A ;
both b and b are column vectors.
Then Matlab matrices A and b can be specified as
A = [A_11, A_12; A_21; A_22];
b = [b_1; b_2];
where A_11, A_12, A_21, A_22, b_1, and b_2 are previously defined matrices with dimensions that
match as above. Note that elements on the same row are separated with , (a separating space is also
OK), whereas a new row is indicated by ; (inserting a new row by pressing the return key is also OK).
3. An extremely risk averse investor would choose the strategy in 2 whereas an extremely risk preferring
investor would choose the strategy in 1. Most investors choose a strategy in between these extremes.
For instance, if an investor requires a particular expected rate of return r, it make sense to calculate the
particular portfolio that meets that goal at the minimum variance. Formulate the quadratic program
that minimizes the portfolio variance, subject to a precise value r of the expected rate of return for the
portfolio. Formulate the problem both for the case when short selling is and is not allowed.
4. Let and r be the portfolio variance and expected rate of return calculated as in 3 above. The efficient
frontier is the set of all possible pairs (, r). A good way of computing the efficient frontier is to solve,
for a parameter B B , the quadratic program
n
n n

min Q Q wi w ji j ( ) Q ri wi

i= j=
i=
n

subject to Q wi = , and, when short selling is not allowed,

(1)

i=

wi C , for i = , . . . , n.
(a) Which cases do the extreme values = and = correspond to?
(b) Solve (1) for = ., ., ., . . . , . using quadprog, both for the case when short selling is and
is not allowed. Plot and compare the efficient frontiers.

You might also like