Professional Documents
Culture Documents
N
X
w n rn
n=1
N
X
wn rn =
n=1
2
(r ) = var (r ) = cov
N
X
wn rn ,
n=1
N
X
w n rn ,
n=1
N
X
m,n=1
N
X
n=1
wn wm cov (rn , rm ) .
| {z }
=cnm
!
w n rn
Covariance Matrix
C is the covariance matrix of (r1 , . . . , rN ),
c1N
c2N
c3N
cNN
xm xn cnm 0
for all x RN .
m,n=1
r =
var (r ) =
N
1 X
ri ,
N
1
N
n=1
N
X
N
1 X
Eri =
m=m,
N
n=1
N
X
1
N2
n=1
n=1
2 =
2
0
N
as N .
! N
N
X
X
1
var (r ) = 2 E
(ri r ) (rj r )
N
i=1
j=1
N
X
X
X
1
1
2 +
cov (ri , rj )
cov (ri , rj ) = 2
= 2
N
N
i,j=1
i=j
i6=j
1
2
2
2
N
+
N(N
1).3
=
.7
+ .3 2 .3 2 .
N2
N
Diversification in General
Simple Example
Figure: For no short selling: the lines labeled = 1 are the lower
bounds on . The upper bound is the line labeled = 1. The set of
points (, r ) for [0, 1] are the curved line.
Mean Variance Theory
Variance Bounds
((1 )1 + 2 )2 = (1 )1 + 2
Variance Bounds
Variance Bounds
In Figure 1, the point where the two lines meet on the y-axis is
r (0 ) where 0 is s.t. (0 ) = 0 when = 1, i.e.
(1 )1 0 2 = 0
1
,
1 + 2
1
1
and so r (0 ) = 1+
r
+
1
r2 .
1
1 +2
2
0 =
case = 1
() |(1 )1 + 2 |
case = 1
For 3 Assets
Add a third asset with expected return r3 and std. dev. 3 . Let
1 equal the total allocation in assets 1 and 2, then repeat
analysis from before.
Results in more options for allocation (hyper place of R3
instead of the lower dimensional hyper lane of R2
There is a region of possible (, r ) points rather than just a
curve (See Figure 3).
In general, can find feasible sets of points (, r ) for N-many
assets, and it gives us a good idea of a portfolios
mean-variance trade-off.
Feasible Region
Figure: Assets 1 and 2 are the same as from slide 4, and for the new
asset we have r3 = .11 and 3 = .1.
n=1
n,m=1
P
P
subject to the constraints nn=1 wn = 1 and N
rn = r ,
n=1 wn
where r is our desiredP
level of return.
We say portfolio r = N
n=1 wn rn is efficient if there exists no other
portfolio r such that Er Er and (r ) < (r ).