Professional Documents
Culture Documents
Theory
2017
1
Introduction
• In this topic we seek to explore the risk
and return provided by portfolios.
•
• Ultimately we want to be able to establish
portfolios that different clients with different
risk profiles would be comfortable with.
2
Optimal Portfolio Choice
U4
U3
U2
E(Rp ) U1
rf
Minimum variance opportunity set
0 p
3
Efficient Portfolio
• An efficient portfolio is a combination of
assets that has the best expected level of
return for a given level of risk.
4
Portfolio return: Expected Value and Variance
Let the return on asset 1 be R1and that on asset 2 be
R2.
E ( R p ) w2 E ( R1 ) w2 E ( R2 )
n
More generally we get: E ( R p ) w E(R )
i 1
i i
2
p
w w 2w1 w2 12
2 2
1 1
2
2
2
2
n n
i 1 i 1
n n n
w w w
2 2 2
p i i i j ij
i 1 i 1 i 1, j i 6
Example
Consider a portfolio composed of assets X and Y
with the following attributes:
Return on X Return on Y Probability
11 -3 0.2
9 15 0.2
25 2 0.2
7 20 0.2
-2 6 0.2
Assume the portfolio is split equally between the two
assets (i.e., wi=0.5).
p 4.97% 8
Correlation Coefficient 12
12
12 12 12 1 2 , 1 1
1 2 12
9
The Portfolio Opportunity Set (for a
two-asset portfolio)
• A portfolio opportunity set (POS) is a set of
all feasible portfolios with different
combinations of X and Y.
2
p
w 2
1w 22
1w1 w2 12
2
2
2
2
12
12 12 12 1 2
1 2
w w 2w1 w2 12 1 2
2
P
2
1
2
1
2
2
2
2
w1 1 w2 2
2
dE ( R p )
dE ( R p ) E ( R1 ) E ( R2 )
slope dw cons tan t
d p d p 1 2
dw
12
Thus the portfolio opportunity set when ρ12 = +1 is linear.
Y (50;16)
0 p
14
The straight line of the portfolio opportunity set when ρ12 = +1
can be calculated as follows:
E ( R p ) w1E ( R1 ) (1 w1 ) E ( R2 )
To make things more tractable we shall let E ( R p ) rp , E ( R1 ) r1 ,
E ( R2 ) r2
Thus the portfolio return can now be expressed as follows:
rp w1r1 (1 w1 )r2
We know that: p w1 1 w2 2 w1 1 (1 w1 ) 2
Solving for w1 yields: w1 p 2
1 2
Substitute w1 into the portfolio return function:
p 2 p 2 r1 r2 r1 r2
rp r1 1 r2 r2 2 p
1 2 1 2 1 2 1 2 15
Which is a straight line taking the form: rp a b p
r1 r2 r1 r2
Where; a r2 2 and b
1 2
1 2
16
Example:
Expected return Standard deviation
Asset 1 14 6
Asset 2 8 3
Calculate the portfolio opportunity set. Just substitute the
respective values into equation 11.
r1 r2 14 8 r1 r2 14 8
a r2 2 8 3 2, and b 2
1 2 63 1 2 6 3
w w 2w1 w2 1 2,
2
p
2 2
1 1
2 2
2 2
so p w1 1 w2 2
rp w1r1 (1 w1 )r2
We have to establish two straight lines for the opportunity set:
p
One equation when: w w
1 1 2 2
And the other when: p w1 1 w2 2
18
When:
p 2
p w1 1 w2 2 and so w1
1 2
And rp w1r1 (1 w1 )r2
p 2 p 2 r1 r2 r1 r2
rp r1 1 r2 r2 2 p
1 2 1 2 1 2 1 2
rp a b p
19
When:
p 2 p 2 r1 r2 r1 r2
rp r1 1 r2 r2 2 p
1 2 1 2 1 2 1 2
rp a b p
20
So the general result when ρ12 = -1 is:
p 2 p 2 r1 r2 r1 r2
rp r1 1 r2 r2 2 p
1 2 1 2 1 2 1 2
OR rp a b p
21
Example
Expected return Standard deviation
Asset 1 14 6
Asset 2 8 3
r1 r2 14 8 r1 r2 14 8 6 2
a r2 2 8 3 10, and b
1 2 63 1 2 6 3 9 3
2 2
Thus rp a b p 10 p and rp a b p 10 p
3 3
2
Thus rp a b p 10 p
3
22
The case where ρ12 = 0
Going back to our two main equations: p
2
w
2 2
1 1 w2 2 2 w1 w2 12 1 2
2 2
w w
2
p
2 2
1 1
2 2
2 2
rp w1r1 (1 w1 )r2
Solving for w1 in the portfolio variance equation and then
plugging the result into the return function yields a hyperbola
with:
rp f ( p )
2
23
The Portfolio Opportunity Set for different 12
values
E(Rp )
Y
12 1 (6;14)
(0;10) Z 12 0
C
B
A
12 1
X
(3;8)
0 p
24
Portfolio Opportunity Set: The Case of One Risky
Asset and One Risk-free Asset
Let asset 1 be a risky asset and the other asset be a risk free
asset. Let the return from the risk-free asset be rf .
w w 2w1 w f 1 f 1 f
2
p
2
1
2
1
2
f
2
f
But for the risk free asset variance is zero, so: w
2
p
2
1
2
1
p w12
1
So 2 2
1 w1 1
p
w1
1 25
Substituting w1 into the portfolio return equation
.
yields:
rp rf w1 (r1 rf ) rf p
r r
1 f
1
The resultant POS is a straight line. It
has an intercept equal to the risk free rate (rf)
and a slope equal to:
r r
1 f
1
26
The POS when there is a riskless security
E(Rp )
POS Equation
Slope of POS =
r r
1 f
1
rf
0 p
27
The Minimum Variance Portfolio
E(Rp )
Y
0 p
28
Finding Minimum Variance Portfolio (two-asset
portfolio case)
If we have two assets (X and Y) then an MVP is the
combination of X and Y that provides the portfolio with
minimum variance.
Investors who are extremely risk averse will select the MVP if
no risk free investment is available.
R p w1 Rx 1 w1 R y
xy
w (1 w ) 2w (1 w ) , but
2 2 2
1 x 1
2 2
y 1 1 xy xy xy xy x y
p
x y
w (1 w ) 2w (1 w )
2
xy x y
2 2 2 2
p 1 x 1 y 1 1
29
We minimize the portfolio variance by setting the first
derivative of portfolio variance w.r.t. w1 equal to zero:
d
2
p
2w1 2 2w1 2 xy x y 4w1 xy x y 0
2
x
2
y
2
y
dw1
Solving for w1 yields:
xy x y
2
y
w1*
2 xy x y
2
x
2
y
30
Example
Use the information in the following table to calculate the
minimum variance portfolio?
Asset X (%) Asset Y (%) Probability
11 -3 0.2
9 15 0.2
25 2 0.2
7 20 0.2
-2 6 0.2
p
2
w1
2 2
1 (1 w1 ) 2
2 2 w1 (1 w1 ) 2
12
(0.487) 2 (0.0076) (0.513) 2 (0.00708) 2(0.487)(0.513)( 0.33)(0.0872)(0.0841)
0.0018025 0.0018632 0.0012092
0.0024565
p 4.956%
1 2
p N
2
1 2 1
p N N
• Thus as N increases the portfolio standard
deviation or risk goes down.
35
Relaxing some of the above
assumptions
• The above proof is based on a number of
simplifying assumptions.
36
Portfolio Variance
n n
2
p
wi w j ij
i 1 i j
37
• Note that variance is a special type of
covariance. Variance is the covariance of
2
an asset with itself. That is, . ii i
w w
2
p
w 2
i i
2
i j ij
i 1 i 1 i j , j i
38
• Note that a 2-asset portfolio has 2 variance and 2
covariance terms. A 3-asset portfolio has 3 variance and 6
covariance terms. An N-asset portfolio has N variance
terms and N2-N covariance terms. We shall assume that
the portfolio is equally weighted: wi = wj = 1/N and that
12 32 ....... jn Cov.
2
1 n
1 n n
p i 2 ij
2
• Then, 2
i 1 N N i 1 i j , j i
1 n
( N 2
N )Cov
2
2 i
2
p
N i 1 N2
1 n 2 1 1 n 2 Cov
2 i 1 Cov 2 i Cov
2
p
N i 1 N N i 1 N
39
• Thus as N increases portfolio variance goes down.
40
Total Risk Diversification Benefits
41
Mean-Variance Frontier for N-Risky Assets
Derivation
• Up to now we have learnt how to compute the mean-
variance frontier in the following cases:
– Two risky assets
– Two assets (one risky, one risk-free)
42
Case 1: MVF - Many risky assets (no
risk free asset)
• The formal optimization problem that should be solved to
get the MVF is: for a given value of expected portfolio
return, choose portfolio proportions, wi’s to minimize risk.
• That is: n
W w1 w2 ........... wN
R R1 R2 ........... RN
1 1 1 1...........1 1
44
The return for an N-assets portfolio is:
N
R p wi Ri w1 R1 w2 R2 .................... wn RN
i 1
E ( R p ) wi E ( Ri ) [ w1 w2 ....wn ]. W E ( R) W E
i 1
.
E ( RN )
The last equality just simplifies notation. The vector of mean
returns shows up so much that I’ll call it E instead of carrying
E(R) around.
The variance of the portfolio return is:
p
2
E (R p E ( R p ))2 2
E W R W E E (W ( R E ))
2
EW (R E)(R E)W
WE ( R E )( R E )W W W
where E ( R E )( R E ) and is the variance - covariance
46
matrix of returns
Now, let’s restate our problem with this notation. Minimizing
the variance is the same as minimizing 1/2 the variance, so
we want to choose weights to:
1
min W W s.t WE ; W1 1
W 2
E 1 E E 11
1 1
1 E 1 1 1
Solving the above for and yields:
1
E E E 1
1 1
1 1
1 E 1 1 1
Which implies that:
1
E 1
1 E 1 1
Now that we know what the λ and δ multipliers are, we can go
back and find the weights,
W (E 1 ) E 1
1 1
1
E 1
1 E 1 1
So we can plug into the weights equation to get:
1
E 1
W E 1 E 1
1
1 1
W W
2
P
If you plug in W and its transpose you should be able to get:
1
E 1
p 1 1 E 1 1
2
A B E 1
Let B C 1 E 1
This implies that variance of the portfolio is:
A B
1
C B 1 C B
p 1 B C 1 AC B 2 1 B A 1 AC B 2 1 B A 1
2 1
1 C 2 2 B A
C B B A
AC B 2
1
AC B 2
k k
Min w w C
n m mn s.t. wn 1 and wn rn r
n 1 n 1
k k k
Min
w
w w C
n,m 1
n m nm s.t. wn 1 and wn rn r
n 1 n 1
Differentiating in the Presence of a
Covariance Term
• We need clarity on how to differentiate the
following term of the objective function:
1 k
2 n,m1
wn wmCmn
k n n
• Note that: wn wmCmn p wi w j ij
2
n , m 1 i 1 i j
w w
2
p i j ij
i 1 i j
3
wi w1 i1 wi w2 i 2 wi w3 i 3
i 1
3 3 3
wi w1 i1 wi w2 i 2 wi w3 i 3
i 1 i 1 i 1
wi w j ij
2
p
i 1 i j
p2 n
2 w j 1 j
w1 j 1
Now, going back to our optimization problem. First, we
formulate the Lagrange function and then calculate the first
order conditions as follows:
1 k k k
L wn wmCmn wn 1 wn rn r
2 n , m 1 n 1 n 1
L k
wn C1n r1 0
w1 n 1
L k
wn C2 n r2 0
w2 n 1
.
L k
wn Ckn rk 0
wk n 1
L k
wn 1 0
n 1
L k
wn rn r 0
n 1
To help see how the above system can be transformed into
matrix notation we rewrite the above first order conditions as
follows:
w1C11 w2C12 w3C13 ....... wk C1k r1 0
w1C21 w2C22 w3C23 ....... wk C2 k r2 0
.
.
w1Ck1 w2Ck 2 w3Ck 3 ....... wk Ckk rk 0
w1 w2 w3 ...................... wk 1
w1r1 w2 r2 ......................... wk rk r
The above system can then be restated in matrix notation
as follows:
w1 0
w2 0
.
.
. .
A 1
. .
w 0
k
1
r
Case 2: MVF - Many risky assets (with a risk free)
• In the presence of a risk-free asset and any number of risky
assets the set of efficient portfolios is a straight line. The
objective function is similar to the case of risk assets only.
• The constraint that portfolio expected return must be equal
to a certain return can now be expressed as follows:
E ( R p ) w1 E ( R1 ) w2 E ( R2 ) ..... wn E ( Rn ) rf w0
n
rf w0 wi E ( Ri )
i 1
61
So the minimization problem is now given as follows:
n n
min var (Rp) s.t. rf w0 wi E ( Ri ) & w0
i 1
w 1
i 1
i
That is, for a given expected return μ you carry out a separate
minimization problem so that as μ changes the efficient set is
traced out.
62
• That is, for a given expected return μ you carry
out a separate minimization problem so that as μ
changes the efficient set is traced out.
63
Efficient Portfolios with a Risk-Free Asset
E(Rp ) E
G
Z
rZ
rf
G
0 Z p
• Points along this ray show the expected return and risk for
portfolios comprising the risk free asset and the portfolio of
risk assets given by the chosen point on the GG frontier.
w Cov(r , r )w
n ,m0
n n n m s.t w
n 0
n 1 and wn rn r
n 0
w Cov(r , r )w
n , m 1
n n n m
• So we seek to solve the following optimization problem:
k k k
w Cov(r , r )w
n , m 1
n n n m s.t w
n 0
n 1 and wn rn r
n 0
1 k k k
L wn wmCmn wn 1 wn rn r
2 n,m 1 n 0 n 0
L
r0 0
w0
L k
wnC1n r1 0
w1 n 1
L k
wn C2 n r2 0
w2 n 1
.
L k
wn Ckn rk 0
wk n 1
L k
wn 1 0
n 1
L k
wn rn r 0
n 1
r0 0
w1C11 w2C12 w3C13 ....... wk C1k r1 0
w1C21 w2C22 w3C23 ....... wk C2 k r2 0
.
.
w1Ck1 w2Ck 2 w3Ck 3 ....... wk Ckk rk 0
w1 w2 w3 ........................... wk 1
w1r1 w2 r2 w3r3 ............. wk rk r
0 0 0 0... 0 1 r0 w0 0
0 w1 0
C11 C12 C13 ... C1k 1 r1
. . .... . . . . .
. . .... . . . . .
. . .... . . . . .
0 Ck1 Ck 2 Ck 3 ... Ckk 1 rk wk 0
1 1 1 1... 1 0 0 1
r0 r1 r2 r3 ... rk 0 0 r
k
1
Min wn wmCmn s.t. wn 1
2 n 1
1 k k
L wn wmCmn wn 1
2 n ,m1 n 1
L k
wnC1n 0
w1 n 1
L k
wnC2 n 0
w2 n 1
.
.
.
L k
wnCkn 0
wk n 1
L k
wn 1 0
n 1
w1C11 w2C12 w3C13 ....... wk C1k 0
w1C21 w2C22 w3C23 ....... wk C2 k 0
.
.
w1Ck 1 w2Ck 2 w3Ck 3 ....... wk Ckk 0
w1 w2 w3 ........................... wk 0
C11 C12 ....... C1K 1 w1 0
C
C ........ C2 K 1 w2 0
21 22
. . . . . .
. . . . .
w
CK 1 CK 2 ...... CKK 1 k 0
1 1 1 0 1
Call this Matrix A
w1 0
w2 0
. .
A 1
.
wk
0
1
Example: Calculating expected return and variance for a
minimum variance portfolio
C AA C AB C AC 1 1 4 1 2
C C BA CBB C BC 1 4 2 3 4
CCA CCB CCC 1 2 1 4 3
1 3 3
L wn wmCmn wn 1
2 n ,m 1 n 1
L k
wnC1n 0...................( FOC1)
w1 n 1
L k
wnC2 n 0...................( FOC 2)
w2 n 1
L k
wnC 33n 0..................( FOC3)
w3 n 1
L k
wn 1 0..................( FOC 4)
n 1
w1C11 w2C12 w3C13 0
w1C21 w2C22 w3C23 0
w1C31 w2C32 w3C33 0
w1 w2 w3 1
C11 C12 C13 1 w1 0
C w
C C 1
2 0
21 22 23
w3 0
C31 C32 C33
1
1
1 1 1 0
C AA C AB C AC 1 1 4 1 2
C CBA CBB
CBC 1 4 2 3 4
CCA CCB CCC 1 2 1 4 3
1 1 4 1 2 1 w1 0
1 4 2 3 4 1 w2 0
w3 0
1 2 3 4 3 1
1
1 1 1 0
Call this Matrix A
w1 0
w 0
2 1
w3 A 0
1
• Thus, w1 = 0.6538, w2 = 0.2692, w3 = 0.0769, ʎ = 0.7596.
• Now we need to calculate the expected return and the
variance for the MVP: E (rA )
E (rp ) wA E (rA ) wB E (rB ) wC E (rC ) wA wB wC E (rB )
E (rC )
1 2
0.6538 0.2692 0.0769 1
3 2
0.7115
1 1 4 1 2 0.6538
p2 W T CW 0.6538 0.2692 0.0769 1 4 2 3 4 0.2692 0.7595
1 2 3 4 3 0.0769
P2 0.875
Required Reading
• Markowitz, H. (1952). Portfolio selection, Journal of
Finance 7, 77-91.
•
• Tobin, J. (1958). Liquidity preference as behaviour
towards risk. Review of Economic Studies, February
1958, 65–86.
•
• Markowitz, H. (1991). Foundations of portfolio theory,
The Journal of Finance 46 (2), 469-477.
85