0% found this document useful (0 votes)
4 views3 pages

Mutual Fund Separation Theorem Explained

1. The document discusses the optimal portfolio selection problem, where an agent aims to maximize expected consumption while minimizing risk. 2. It shows that the optimal portfolio will allocate wealth across risky assets in fixed proportions, regardless of the agent's risk preferences. This is known as the mutual fund separation theorem. 3. The proportions invested in each risky asset depend on the expected returns, variances, and covariances of the assets. Assets with higher expected returns relative to their risk will receive larger allocations in the optimal portfolio.
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
4 views3 pages

Mutual Fund Separation Theorem Explained

1. The document discusses the optimal portfolio selection problem, where an agent aims to maximize expected consumption while minimizing risk. 2. It shows that the optimal portfolio will allocate wealth across risky assets in fixed proportions, regardless of the agent's risk preferences. This is known as the mutual fund separation theorem. 3. The proportions invested in each risky asset depend on the expected returns, variances, and covariances of the assets. Assets with higher expected returns relative to their risk will receive larger allocations in the optimal portfolio.
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

1

UNIVERSITY OF CALIFORNIA Economics 202A


DEPARTMENT OF ECONOMICS Fall 2012
D. Romer


OPTIMAL PORTFOLIOS AND THE MUTUAL FUND SEPARATION THEOREM


Set-up

An agent has wealth W. Theres a riskless asset, which pays a return of zero for sure, and N risky
assets. Asset N pays return r
i
. The rs have means
1
,
2
, ,
N
and variance-covariance matrix
. is N x N, and we denote its ij element by
ij
.

Thus, the agents consumption is C = B + x
1
(1 + r
1
) + x
2
(1 + r
2
) + + x
N
(1 + r
N
), where B is the
agents holdings of the riskless asset and x
i
is

holdings of risky asset i. Using the budget
constraint, B + x
1
+ x
2
+ + x
N
= W, we can rewrite this as

C = W + x
1
r
1
+ x
2
r
2
+ + x
N
r
N
.

The agents objective function is assumed to depend only on the mean and variance of
consumption (and to be increasing in the mean and decreasing in the variance). One case where
this arises is quadratic utility, as in class. Another case is constant absolute risk aversion utility
and normally distributed returns, along the lines of Problem 8.14 in the book.

Since the individual cares only about the mean and variance of consumption, then the optimal
allocation of the individuals wealth will have the lowest possible variance of consumption given
its mean. (If not, it would be possible to lower the variance without changing the mean, which
would make the agent better off.) Thus, rather the solving the full optimization problem, we will
focus on minimizing variance for a given mean.


Case 1: Two risky assets, solved with algebra

With two risky assets, C = W + x
1
r
1
+ x
2
r
2
. Thus the mean of C is C = W + x
1

1
+ x
2

2
. To find
the variance of C, note that C E[C] = (r
1

1
)x
1
+ (r
2

2
)x
2
. Thus,

E|(C -E|C])
2
] = E|((r
1
- p
1
)x
1
+ (r
2
- p
2
)x
2
)
2
]

= x
1
2
E|(r
1
- p
1
)
2
] + x
2
2
E|(r
2
- p
1
)
2
] + 2x
1
x
2
E|(r
1
- p
1
)(r
2
- p
2
)]

= x
1
2
o
11
+ x
2
2
o
22
+ 2x
1
x
2
o
12
.

So the Lagrangian for the problem of minimizing variance subject to achieving some target level
of expected consumption, Z, is

2

I = x
1
2
o
11
+ x
2
2
o
22
+ 2x
1
x
2
o
12
+ z|Z - (w+ x
1
p
1
+ x
2
p
2
)].

The first-order conditions for x
1
and x
2
are

2x
1
-
o
11
+ 2x
2
-
o
12
= zp
1
,

2x
2
-
o
22
+ 2x
1
-
o
12
= zp
2
.

Solving these two linear equations for x
1
-
and x
2
-
gives us

(1) x
1
-
=
o
22
p
1
- o
12
p
2
o
11
o
22
- o
12
2
z
2
,

(2) x
2
-
=
o
11
p
2
- o
12
p
1
o
11
o
22
- o
12
2
z
2
.


Discussion

The Mutual Fund Separation Theorem

Notice what happens as changes that is, as the agent puts more or less weight on the mean
relative to the variance. x
1
and x
2
change in the same proportion. Thus, agents who differ in their
attitudes toward risk will hold different amounts of the riskless asset, but their mix of the risky
assets (that is, their ratio of x
1
to x
2
) will be the same.

This is the Mutual Fund Separation Theorem. We can construct an optimal mix (that is, an
optimal mutual fund) of risky assets. Depending on their risk preferences, agents will choose
different combinations of the safe asset and the mutual fund; but they will not choose different
mixes of risky assets.

An Attempt at Intuition for the Mutual Fund Separation Theorem

Consider a portfolio of risky assets. If the agent holds none of the portfolio and puts all his or her
wealth into the safe asset, his or her mean consumption is W, and its standard deviation is zero.
As the agent moves out of the riskless asset into the portfolio, both the mean and standard
deviation of consumption change linearly with the amount invested in the portfolio. Thus, as the
agent shifts out of the riskless asset into the portfolio, his or her mean consumption and its
standard deviation move along a ray (in standard deviation of consumptionmean consumption
space) from the point (0,W). Thus, each portfolio gives the agent access to a ray of points out of
(0,W) in standard deviationmean space. Every agent prefers to be on a higher ray than to be on
a lower one. So every agent chooses the portfolio with the highest slope in this space that is,
the portfolio with the highest ratio of expected excess return (return minus that on the safe asset)
to standard deviation. Agents risk attitudes then determine where on the line they choose to be
that is, how much of the portfolio they hold.

3

The Determinants of the Mix of the Two Assets

Equations (1) and (2) also show what determines how much of the two assets the agent holds.

For example, suppose
12
= 0. Then x
1
=
1
/2
11
, x
2
=
2
/2
22
. Thus whether the agent holds a
positive, negative, or zero amount of the asset is determined by whether the assets expected
excess return is positive, negative, or zero. Holdings of an asset are proportional to its expected
excess return, inversely proportional to its variance, and increasing in the importance the agent
attaches to the mean of his or her consumption relative to its variance.

Consider also a small increase in
12
starting from
12
= 0. The marginal effect is to reduce the
agents holdings of asset 1 if his or her holdings of asset 2 are positive, and to increase his or her
holdings of asset 1 if his or her holdings of asset 2 are negative.


Case 2: N risky assets, solved with linear algebra

Here C = W + Xr, where X = [x
1
x
2
x
N
] and r = [r
1
r
2
r
N
]. Thus, E[C] = W + X (where
= [
1

2

N
]), and Var(C) = XX.

The Lagrangian is L = XX + [Z (W + X)]. The ueiivative of XX with iespect to X is
(X) + X, which equals X + X, oi 2X. (To see this, consiuei the ueiivative with
iespect to x
1
. x1 appeais twice in XX, anu so theie aie two teims. The fiist teim is X
times the vectoi |1 u u u], which is X times |
11

21

N1
], which is the fiist element of
X (anu hence the fiist element of X, since is symmetiic). The seconu teim is the
vectoi |1 u u u] times X, which is |
11

12

1N
] times X, which is also the fiist element
of X. Thus the ueiivative of XX with iespect to x
1
is 2 times the fiist element of X.)

Thus, the Nx1 vectoi of fiistoiuei conuitions is 2X* = , which implies

X
-
=
z
2

-1
p.

The Nutual Funu Sepaiation Theoiem holus heie: when changes, all the elements of X*
change by the same piopoition. The intuition (such as it is) is the same as foi the case of
two assets.

You might also like