Lecture 5: Generalized Risk and Return

Readings:
Ingersoll Chapter 5 (also see chapters 8 and 9)
LeRoy & Werner Chapter 10
Merton Chapter 2
Dybvig & Ross Portfolio Efficient Sets, Econometrica, 1982
Rothschild & Stiglitz Increasing Risk I: A Definition, Journal of Economic
Theory, 1970
Rothschild & Stiglitz Increasing Risk II: Its Economic Consequence, Journal of
Economic Theory, 1971
Dybvig Distributional Analysis of Portfolio Choice, Journal of Business, 1988
Dybvig & Ingersoll Mean-Variance Theory in Complete Markets, Journal of
Business, 1982
1
After having considered the special, well-known, and well (over?) used case of mean-variance
analysis, lets step back to the general problem again.
First, Risk: The CAPM makes the assumption that the variance of the return on a portfolio
measures its risk.
Consider the idea that risk is the combination of properties of a set of random outcomes that
change the evaluation of
)]
~
( [ x u E
away from
) (x u
for concave utility u( ). Two aspects of
this definition to highlight are: (1) that what alters this evaluation (or is, on net, disliked) is, of
course, dependent upon the utility function used in the evaluation. Thought of this way, risk is
necessarily a property defined for a class of utility functions. This definition of risk also (2)
conveys only the notion of dispersion of outcomes. This requires that we correct for the mean
(location) when we talk about risk. It also means that all other aspects (good and bad) of the
distribution (beyond location) are lumped into risk.
More formally
Definition: If uncertain outcomes x
~
and
y
~
have the same location (expectation), then x
~
is
said to be weakly less risky than
y
~
for the class of utility functions U if no individual with a
utility function in U prefers
y
~
to x
~
. That is:
)]
~
( [ )]
~
( [ y u E x u E
u U. x
~
is strictly
less risky if the inequality is strict for some u U.
For some restricted classes of utility functions (some U), this ordering is complete (i.e. for all
pairs of random variables x
~
and
y
~
, with the same mean, either x
~
is weakly less risky than
y
~

or
y
~
is weakly less risky than x
~
or both).
Example: Quadratic utility: as we have seen, variance is the measure of risk. Any quadratic
utility function can be written as:
u(z) =
2
2
z
b
z
we restrict z to be
b
z
1

so that
0 ) ( z u
Expected utility (for any distribution for which mean and variance are defined) is written:
) (
2
)] ( [
2 2
+ z
b
z z u E
.
So, for any x
~
and
y
~
with
y x
,
)]
~
( [ )]
~
( [ y u E x u E
iff
2 2
y x

.
The completeness of this ordering is unfortunately not a common property across different
(broader) classes of utility functions. Consider the example introduced by Ingersoll:
x
~
= 0 with prob. =
y
~
= 1 with prob. =
8
7

4 with prob. = 9 with prob. =
8
1
For these simple distributions we easily calculate that:
2 y x
, and
2
x
= 4 <
2
y

= 7.
2
Thus, for any investor with quadratic utility, x
~
is preferred to
y
~
(it provides higher expected
utility). However, for x x u ) ( ,
y
~
provides higher expected utility. Thus, for any class of
utility function that includes both quadratic and square root utility, x
~
and
y
~
cannot be ranked.
Further, note that it is not the case that variance is the appropriate measure of risk whenever the
ordering is complete. Consider the class of cubic utility functions defined as
3
) ( cz z z u with
c > 0, where we restrict outcomes to be bounded between zero and
2
1
) 3 (

c .
Expected utility is written (translating the non-central moment to the central moments):
) 3 ( )] ( [ )] ( [
2 3 3 3
z z m c z z E c z z u E + + ,
where
3
m is the third central moment ] ) [(
3
z z E that we examined before. (Note that for
this class of utility functions variance and skewness are both disliked.)
If
y x
and x
~
is preferred to
y
~
then it must be that
0 )] ( ) ( 3 [
3 3 2 2
> +
x y x y
m m x c
.
Thus,
3 2
3 m z + is the proper measure of risk for this class of utility functions. So, even for
classes of utility functions that imply a complete ordering of random variables, variance is not a
universal measure of risk.
For the general class of all risk averse (concave) utility functions, the ordering is obviously
incomplete. We need to find a way to reduce the scope of the problem if we are to say more, i.e.,
restrictions on distributions or restrictions on utility functions (which are rarely thought to be
particularly interesting or valuable).
Mean Preserving Spreads: For illustration, lets take a look at a particular relation between x
~

and
y
~
that allows a comparison for all u( ) in U (the class of all risk averse utility functions).
Intuitively, if we take x
~
and add mean zero noise to it we should wind up with something less
attractive to risk averse agents. It turns out its not quite that simple. Why?
A random variable x
~
is less risky than
y
~
if g(
y
~
) can be obtained from f( x
~
) by the
application of a series of mean preserving spreads (MPS) (where f() and g() are density
functions).
Definition:
3
elsewhere
t d x d
t d x d
t c x c
t c x c
for
for
for
for
x s
+

< <

+ < <
+

< <

+ < <

'

0
) (

where:
) ( ) ( d d c c t > 0

> 0 t d c t c < < +

> 0 d t d < +
From a uniform distribution we might get the following:
So, we have four (non-overlapping) intervals of non-zero value with the middle two negative and
the outside two positive.
Note that:

0 ) ( dx x s
so what is added is subtracted elsewhere
and,

0 ) ( dx x xs
mean zero (really addition of s(x) doesnt change mean)
Thus, if f(x) is a density function for x, then f(x) + s(x) is also a density function that gives the
same expected value. (As long as you dont violate non-negativity for the resulting density.)
An exercise you should work through in Ingersolls text shows that if
y
~
has density g() = f() +
s(), where s() is a mean preserving spread, then
y
~
is riskier than x
~
(whose density is f()) for
the class of all concave utility functions. The proof simply compares expected utility of x
~
and
y
~
for general concave utility functions (lots of Jensens inequality applications). Transitivity
implies this works for a series for MPSs as well.
4
Rothschild & Stiglitz Theorems on Risk:
The notion of
y
~
differing from x
~
by the addition of a series of MPSs, implying more risk is
very intuitive,
y
~
has a more disperse, noisier distribution than x
~
. This can be made more
precise, more general, and less cumbersome to deal with.
Theorem 1: For the concave class of utility functions, outcome x
~
is weakly less risky than
outcome
y
~
iff
y
~
is distributed like
~ ~
+ x where
~
is a fair game with respect to x
~
.
That is:

~ ~ ~
+ x y
d
and
0 ]
~
[ x E
x
Note that the law of iterated expectations tells us
y x
in this case.
Proof (sufficiency): If

~ ~ ~
+ x y
d

)] ( [ y u E )] ( [ + x u E
equivalence of distributions
]} ) ( [ { x x u E E +
law of iterated expectations
} )] ( [ { x x E u E +
Jensens inequality
)] ( [ x u E
fair game property
Given this result, we can easily see that variance is a valid measure of risk for the class of all
concave utility functions when we restrict ourselves to normal random variables.
For any
b) , N( ~
~
y
and
a) , N( ~
~
x
, with b a, we can always write:

~ ~ ~
+ x y
d
where
a) - b N(0, ~
~

is independent of x
~
. Since independence is
stronger than the fair game property
~
is therefore also a fair game with respect to x
~
.
We can also show the related result that if
y
~
is riskier than x
~
we know it must have a larger
variance this must be true because the quadratic utility functions are members of the concave
class and for quadratic utility functions variance measures risk.
Alternatively: If
0 ]
~
[ x E
then, Cov
0 )
~
,
~
( x
(in fact Cov
0 )
~
),
~
( ( x g
for all functions
g())
Therefore,

~ ~ ~
+ x y
d
tells us that:
)
~
( y Var
=
) ( + x Var
=
) , ( 2 ) ( ) ( x Cov Var x Var + +
=
) ( ) ( Var x Var +

) (x Var
When location differs, we can correct by simply de-meaning and comparing x x
~
vs.
y y
~
or
comparing
y
~
vs.
). (
~
x y x +
This, however, shifts around the distributions and is itself
somewhat cumbersome and even a little odd when our goal is to describe a risk/return tradeoff.
A closely related concept to riskier is second order stochastic dominance SSD. This concept
incorporates the correction for location within the comparison.
x
~
displays weak 2
nd
order stochastic dominance over
y
~
if:

~
~
~ ~
+ + x y
d
where
0
~

and,
0 ] [ + x E

, x
5
Theorem 2:
y
~
is preferred to x
~
by no strictly increasing concave utility function iff x
~
SSD
y
~
Note: SSD incorporates the correction for location, so the class of utility functions for which
this holds must be narrower.
Proof (sufficiency):
Theorem 1 provides
)] ( [ )] ( [ + + + x u E x u E
for all concave utility
functions.
Because

is non-positive, 1
st
order stochastic dominance provides that:
)] ( [ )] ( [ + x u E x u E
for strictly increasing utility functions.
Thus
)] ( [ )] ( [ y u E x u E
for all increasing concave utility functions (since
y
~
and
+ + x
have the same distributions they have the same expected utility).
The relation between riskiness and SSD is strong but they are not identical.
If
y x
and x x
~
is less risky than
y y
~
we know that x
~
2
nd
order stochastically dominates
y
~
since we can write

~
) (
~ ~
+ + x y x y
d
with
0 ) | ( x E
(from riskier) so that
0 )) ( | ( + x y x E
where
x y
is a degenerate non-positive random variable.
However, if x
~
SSD
y
~
while we know that
y x
we cannot say that
y y
~
is riskier than
x x
~
. For one thing, they may not even be ranked. Further if they are, it is possible that x x
~

is riskier than
y y
~
. Let
U(0,1) ~
~
y
and
U(2,8) ~
~
x
. x
~
dominates
y
~
, and so it FSDs and
SSDs
y
~
, but x x
~
is riskier than
y y
~
.
The difference lies in the correction for location. In SSD it is built in, we compare x
~
and
y
~
,
while in riskier we compare x x
~
to
y y
~
. An x
~
with a big mean and dispersion versus a
y
~
with a small mean and small dispersion may be evaluated differently than x x
~
vs
y y
~
.
After a deviation to review the general portfolio problem and results on portfolio choice,
where this is heading is a general study of the efficient set of portfolios. SSD will play a
large role.
Table: x and y are random variables defined on [a, b] (see Ingersoll pg 123)
Concept: Utility Condition Random Variable
Condition
Distributional
Condition
Dominance )]
~ ~
( [ )]
~ ~
( [ w y u E w x u E + +
for
any random variable w
~
for
all increasing utility

~
~ ~
+ x y
0
~

Note: = not
d

y x
~ ~

Outcome by outcome
FOSD )]
~
( [ )]
~
( [ y u E x u E
for all
increasing utility functions

~
~ ~
+ x y
d
0
~

F(x), G(y) are
distribution functions:
G(t) F(t) t
Prob(xt) Prob(yt)
t
6
SOSD
)]
~
( [ )]
~
( [ y u E x u E
for all
increasing concave utility
functions

~
~
~ ~
+ + x y
d
0
~

0 ] [ + x E

, x

t
a
ds s F s G )] ( ) ( [
0 ) ( t
t
on average G>F
Riskier
)]
~
( [ )]
~
( [ y u E x u E
for all
concave utility functions

~ ~ ~
+ x y
d
0 ] [ x E
t t 0 ) (
0 ) ( b
7
Now, Portfolio choice: To quickly review: The general portfolio problem can be cast as: an
agent chooses a portfolio to maximize his/her expected utility of returns:
)]
~
( [
w w
z u E Max

N
i
si i w
z w w z z
1
~ ~ ~
s.t. 1 1 w (i.e.
1
i
w
)
Some other restrictions we might (but wont) use include:

0 w
- no short sales
0
~
z - limited liability for the primitive assets
1 * z - non-negative wealth constraint (bankruptcy)
Form the Lagrangian:
L =
) 1 1 ( )]
~
( [ w z u E
w
+
FOCs:

]
~
*)
~
( [ z z u E
w
L
with
0
and z* = zw*
or, for each asset:

]
~
*)
~
( [
i
i
z z u E
w
L
for all assets i
1 * 1

w
L

The concavity of u() and linearity of the constraint implies that the FOCs are necessary and
sufficient for a maximum.
The FOC must hold for the riskless asset, if it exists:
] *) ( [ R z u E
So, we can subtract this from the general condition to find:
0 )]
~
*)( ( [ R z z u E
i
i
in the presence of a riskless asset. This will help us examine some results concerning portfolio
formation in this general context.
Theorem: If a solution w* exists for a strictly concave utility function and a set of assets Z
then the probability distribution of its return is unique and if there are no redundant assets then
the portfolio w* is unique as well.
Proof: For each feasible w define new control variables s

by Zw or
si i s s
Z w Zw

S s ,..., 1
and define the function v( ), a derived utility function over
state returns:

s
si i i s s S
Z w u u E v ) ( )] ( [ ) ,..., (
1

The portfolio choice problem can then be written:
8
) ,... (
1 S
v Max
s

s.t. w with
si i i s
Z w
s
This is a standard convex optimization problem with an objective function that is concave in the
choice variables (note that E(u(z
w
)) is not concave in w) and a set of linear constraints. Thus, the
optimum (*), if it exists, is unique. The vector * describes the distribution of returns across
the states, so write:
* = Zw*
If there are no redundant assets Z has a unique left inverse Z Z Z L
1
) ( and
* * L w
is
then also unique.
We can say something specific about choices made by agents in this very general context but not
much:
Theorem: The optimal portfolio for a strictly risk averse, non-satiated investor will be the
riskless security if and only if
R z
j

j = 1, 2, , N.
Proof: When a riskless asset exists, w* satisfies the FOC:
0 )] *)( ( [ R z z u E
j
If
R z
j


j
then
1 * R z
will satisfy the FOC. Since u() is strictly concave we know the
probability distribution of the return is unique. If there are no redundant assets the portfolio is
unique. (Strictly risk averse agents hold only risk free portfolios or the risk free asset. We also
see why the strictly concave is required for uniqueness of the solution in the theorem above.)
If R z * we can write the FOC as
0 ) ( ) ( R z E R u
j
0 ) ( > R u
from strictly monotone u(), so if z*=R is optimal, it must be that
R z
j


j
This generalizes our earlier two-asset (one risky, one riskless) result.
Finally, Efficient portfolios: the efficient set of portfolios those portfolios for which there are
no other portfolios with the same or greater expected return and less risk. Alternatively, those
portfolios which are not second order stochastically dominated.
More specifically
Definition: A portfolio w is efficient if E w
)])} ( [ ) ( )( ( {
1 1 ,
w z u E Max w z Eu U u R w E
w R w
N
N


i.e. if there exists a utility function U u (strictly monotone, concave class), for which w solves
the investors problem:
)] ( [ w z u E Max
w

such that 1 1 w budget constraint
N
R w w is a portfolio of the traded assets a restriction whose
strength is determined by the structure of Z
We use U, the strictly monotone, concave class, because if we the use monotone, concave class
E is trivially all
N
R w with 1 1 w since this class allows a constant utility function, making
it uninteresting and a useless definition.
The following theorem shows that: Efficient portfolios are those for which there are no other
portfolios with the same or greater expected return and less risk (those that are not SSDed). But,
9
not the set for which there is not a less risky portfolio with the same mean return this set
conceptually includes the lower limb of the CAPM hyperbola.
Theorem: For some efficient portfolio k with returns given by
k
e
z
~
if
k
e
k
e
z z
~
is riskier than
w w
z z
~
(for any portfolio w), then
w
k
e
z z > .
Note: It is not true in general that
w
k
e
z z < implies that w w
z z
~
is riskier than
k
e
k
e
z z
~
since
riskier is an incomplete ordering.
Proof: If
k
e
k
e
z z
~
is riskier than
w w
z z
~
, then
)]
~
( [ )]
~
( [
k
e w w
k
e
z z z u E z u E + U u (by the definition of riskier).
If
w
k
e
z z , then
)]
~
( [ )]
~
( [
w
k
e w w
z u E z z z u E + so, )]
~
( [ )]
~
( [
w
k
e
z u E z u E U u (monotonicity).
But, this contradicts the assumption that
k
e
z
~
is an efficient portfolios return, so it must
be that
w
k
e
z z > holds.
From this, we get two corollaries:
Corollary 1 :
R z
k
e
>
for all efficient portfolios since all efficient portfolios are weakly
riskier than the risk free asset.
Corollary 2 : The riskier is an efficient portfolio, the higher is its expected return. In
other words, there is a risk/return tradeoff.
Now lets consider an alternative way to characterize the efficient set using a simple complete
markets example with two states (so we can draw it). Since the market is complete, we can find a
portfolio that has any distribution of wealth across the two states that we would like. Therefore,
we can make W
1
and W
2
(wealth in states 1 and 2: W
o
z
1
* and W
o
z
2
*) the choice variables.
)] ( [ ) ( ) (
2 2 1 1 ,
2 1
W u E W u W u Max
W W
+

subject to: p
1
W
1
+ p
2
W
2
= W
o
(budget constraint is written using the unique state prices)
L: =
) ( ) ( ) (
2 2 1 1 2 2 1 1
W p W p W W u W u
o
+ +
The FOCs give:
0 ) (
1 1 1
p W u or,

) (
1 1
1
W u
p

0 ) (
2 2 2
p W u or,

) (
2 2
2
W u
p

Or,

) (
s
s
s
s
W u p

= state price density just as we saw before.
10
Note: Under risk neutrality, state prices are proportional to the actual probabilities (or the state
price density is constant across the states). Thus, again we see that a single risk neutral agent in
the economy (who faces no restrictions on short sales or borrowing) trades to set prices in such a
way that is constant. Any risk averse agents in the economy (in equilibrium) will then also
trade so that
) (
s
W u
is constant across all states. Thus, their optimal choice must be to hold
riskless positions. Said another way: the risk neutral agent trades so that there is no reward for
risk bearing (all assets have E(z)=R) so no risk averse agent bears any risk.
Examine the set of efficient (optimal) portfolios for this two state example:
W
2
k W W +
2 2 1 1

or, 1
2
1
2
2
W
k
W

slope=
2
1
p
p

riskless asset
o
W W p W p +
2 2 1 1
or, 1
2
1
2
2
W
p
p
p
W
W
o

r
d Simply assume:

45 line f slope=
2
1

2
1
2
1
p
p
>

W
1
Line with slope =
2
1
p
p
is the budget constraint
Line with slope =
2
1

is a line of constant expected wealth

Draw an indifference curve for an arbitrary risk averse utility function. How do we know that
the constant expected wealth line is tangent to the indifference curve at the 45 line?
It must be from r outward on the line, you stay at the same expected wealth but add risk if you
move in either direction. Thus, the constant expected wealth line through r must be tangent as it
must be below the indifference curve away from r on either side.
Which of the possible consumption bundles will be chosen by a rational agent?
Consider point d in the picture. By concavity of the utility function, d is no better than r
(same expected wealth but more risk) and by strict monotonicity f is dominated by d since d
has larger W
1
and W
2
. So, by analogy all points on the budget line below r are dominated by
r, the riskless asset (they all give less expected wealth and more risk).
Rational choices are above the point r on the budget line in this picture. Thus, only choices
with W
2
W
1
are optimal. Moving in this direction gets more risk and higher (not lower)
expected wealth. Where will the indifference curve be tangent to the budget constraint?
11
Depends on u but we know it will be above point r because we assumed the budget line had a
steeper slope than the constant expected value line (that is,
2
1
2
1
2
1
2
1

> >
p
p
p
p
or,
2
2
1
1

p p
>
or,
2 1
> ).
Thus, the state price per unity probability of wealth in state one is greater than in state 2.
Choosing W
2
W
1
is, in this sense, a cost minimizing choice (for a given distribution of
wealth, choosing W
2
W
1
has lower cost than the reverse). Given state independent vN-M
utility we will show later that cost minimization in this way is really the only restriction
imposed by maximizing behavior.
We can also find the result: if
1
>
2
then W
2
W
1
from the FOC of the investors decision
problem in this example:
s s
W u ) (
So, if
1
>
2
we require ) ( ) (
2 1
W u W u >
, or, since u is weakly decreasing,
2 1
W W .
We can generalize this two state example:
Theorem (Dybvig J of B 88): Assume a complete market and equally probable states. For the
strictly monotone, concave class of utility functions U, w is an efficient portfolio if and only if,
for all states r, s:
r s s r
Zw Zw >
Proof (sketch): Suppose w is an efficient portfolio, then it follows from the FOC (
) ) (
s s
Zw u
and the concavity of u() that r s s r
Zw Zw >
(the 2
nd
inequality is weak
since we dont require strict concavity). The contra-positive of this is: s r s r
Zw Zw >
.
Now suppose that s r s r
Zw Zw >
or equivalently r s s r
Zw Zw >
. Graphically this
says:

Zw
Choose any function g() which has g(Zw
s
)=
s
for
s
> 0. We know that this function is
positive and weakly decreasing so re-label it
. ) (
s s
Zw u
Integrate this function and you get a strictly monotone concave function u() such that it satisfies
the FOC at the candidate w. Note that we have found a u() scaled so that the Lagrange
multiplier

equals 1. But, we can always do this since u() is unique only up to a positive
affine transformation. The constant of integration from the step u u is the rest of this
transformation. This is the equivalence between maximizing behavior and cost minimization.
12
Problem: Look back to the two state complete markets example. Prove that if state prices
are proportional to the actual probabilities then any choice not on the 45 line is riskier than a
bundle on the 45 line.
Instead of the formal proof of sufficiency for the last theorem, to provide a little intuition, look at
the situation of a complete market where
S s
1

s=1,,S (an unnecessary simplification).
Now, since there are S states, there are S! ways to order or assign the lottery outcomes to states
(which state generates greater wealth than which). This means that there is some cheapest way
to order the lottery. Suppose that one of the cheapest ways does not assign outcomes in reverse
order to the state price density. Then states r, s such that s r
>
but s r
w w >
. Now, switch
the outcomes for states r and s. The change in cost is:
2
) )( ( ) )( ( ) ( ) ( S w w p p w w w p w p w p w p
s r r s s r r s s s r r r s s r
+ +
which is a negative number implying a cost decrease with no change in expected utility for
state independent utility of wealth and so a contradiction.
Example Verifying the efficiency of a given portfolio:
The last theorem can be generalized and Ingersoll uses the general result to build a numerical
example illustrating the idea that verifying the efficiency of a given portfolio requires only that
we check the ordering of returns it assigns across states. It is, therefore, also true that all
portfolios which impose the same orderings on their returns as that of an efficient portfolio are
efficient, i.e., some strictly increasing concave utility function sees that portfolio as optimal.
If the matrix Z has no redundant assets then the rank of Z is NS.
When N = S the market is complete and thus there is a unique vector that supports Z. Thus
there is only one ordering of the state contingent payoffs that can represent an efficient portfolio.
If N < S, the market is incomplete, is not unique (there are N equations in S unknowns in the
supporting equation) so not all efficient portfolios need have the same orderings of returns. An
example from the text will help illustrate this here it has been changed by putting things in
terms of the s where the book uses marginal utilities see the FOC for the translation:
Consider the market characterized by:
1
1
1
]
1

6 . 0 0 . 3
5 . 1 2 . 1
4 . 2 6 . 0
Z
where
3
1
3 2 1

S=3 3! or 6 potential orderings of returns. With 2 assets, only 4 orderings are feasible.
13
The feasible set is:

<
< <
< <
>
) 1 , 2 , 3 (
) 1 , 3 , 2 (
) 3 , 1 , 2 (
) 3 , 2 , 1 (
3
1
1
7
3
1
3
1
5
3
1
7
3
5
3
1
w
w
w
w
Market possibilities state with lowest return is 1
st

(Write returns in each state as a function of w
1
and 1 w
1
then graph Zw
s
s = 1, 2, 3 against w
1
.)
Now look at an optimizing investors FOC to consider efficiency:

s
si s s
z z u *) (
for i=1, 2 Let
s s
u z u *) (
Since

s
s
u

, this FOC can also be written

1 ) (
i
z E
or
1
si s s
z
for i=1, 2
Here, these two equations are:
1 0 . 1 4 . 2 .
3 2 1
+ +
1 2 . 5 . 8 .
3 2 1
+ +
This has solutions of:
11
5
3 11
21
1

11
30
3 11
38
2
+ These represent efficient portfolios.
3 3

Graph these three lines:
1

3 2 1
, ,
2

Thus, four orderings are efficient.

14
From high to low (i.e. state with highest is listed first):
(2, 3, 1) (2, 1, 3) (1, 2, 3) (1, 3, 2)
The reverse of the final ordering is not feasible (see above), we cant find a portfolio of these two
assets that gives returns that are lowest in state 1 and highest in state 2. The first three represent
all the portfolios with
3
1
1
> w . So, any such portfolio is optimal for some agent. Of the feasible
orderings, only (3, 2, 1) (
3
1
1
< w ) isnt efficient. This isnt a practical way to approach the issue
but it helps us understand the next set of theorems and the discussion of systematic risk to come.
Homework problem: Now change the probabilities to
2
1
1

3
1
2

6
1
3

.
Which portfolios are optimal portfolios? Does the resulting change in the answer make sense
i.e. explain the difference between the two cases using the characteristics of the two assets..
Relate this to the mean variance efficient portfolios. Is this reasonable?
Convexity of the Efficient Set:
In studying the relation of the CAPM to the more general model it is interesting to ask: is the
efficient set convex? Why? To answer this question we use the technology developed above.
Again, label the set of efficient portfolios E and call the set of marketed assets M (the feasible
set of portfolios).
We define an Arrow-Debreu world, or complete market, as one in which the positive
supporting state prices are unique.
In complete markets , and so the ordering of the
s
s across states, is unique. Thus, we have:
} { } { } {
2 2 1 1 L L
s r s r s r
z z z z z z z z z M E
where a pair of states (r, s) is considered in these restrictions, (r, s) = (r
i
, s
i
), for some i
s r
<
and L is the number of distinct pairs
) , (
s r

with
s r

.
If we label states so that

3 2 1

then E is the intersection of the sets of returns patterns
with
2 1
z z if
1 2
> and the set with 3 2
z z
if 2 3
>
etc. All these intersected with the
set of marketed assets. We then have the following:
Theorem: If M is a complete market then E is a convex set.
Proof: E is the intersection of convex sets and so is convex.
15
Theorem: In a risk neutral market E=M so E is necessarily convex.
Proof: In this case
s
is constant across states. Thus, all orderings of returns across states are
efficient. From the FOC we know s r s r
z z <
, but also s r r s
z z <
.
Thus, both orderings are efficient if
s r

.
Alternatively: All assets have the same expected return and so all possible portfolios may be held
by maximizing risk neutral agents.
In general,
). (

E M E
This second set is the union over all valid s of

E
.
That is,
} {
l l
s r l
z z z E

where (r, s) = (r
l
, s
l
) for some l
l l
s r
<
for a given .
In incomplete markets we must account for the multiple orderings possible because there are
different valid s.
Two technical results are stated without proof:
1. For U and a finite state space, E is the union of a finite number of closed convex sets and
hence is closed.
2. E is connected i.e. E cannot be represented as the union of two separate (disjoint) sets.
Definition: A market is said to exhibit k-fund separation (kfs) if M z z z
k
,..., ,
2 1

such that } 1 , {
1 1

i
k
i
i
i
k
i
w z w z z E
i.e. the efficient set is contained in a set spanned by k mutual funds
Theorem: E is convex whenever M exhibits two-fund separation (2fs).
Proof: From 2fs, E is contained in a line, and since E is connected, E is convex since
connectedness implies convexity in R
1
.
Thus, we now have three special cases where we know E is convex and so we know that the
market portfolio is an efficient portfolio. However, one is uninteresting from a risk-return
perspective and the other two are actually incompatible (see Dybvig & Ingersoll). Now, we
present a counter-example from Dybvig & Ross that shows E is not convex in general.
Theorem: The efficient set is not necessarily convex and kfs (with k 3) does not guarantee
convexity.
Proof: Shown by counter-example. Assume 3 assets and 4 states. We could have less trivial 3fs
by splitting one state into many indistinguishable states and introducing fair gambles with
respect to these new states as new primary assets.
16
Let
4
1
4 3 2 1

Let Z =
1
1
1
1
]
1

48 50 50
48 38 52
48 52 44
72 82 66
The valid set of price vectors is:
)} 39 , 40 , 28 , 21 ( ), 59 , 40 , 68 , 1 {( /
} 0 {
4
span R
+
We can divide by 8,088 or 6,648 respectively to get
1 p Z
.
Thus, we have two orderings on p and, since
s
s

4
1

, on .
Look at
1 1
Zw z
2 2
Zw z with ) 0 , 0 , 1 (
1
w and ) 0 , 1 , 0 (
2
w
Since ) 50 , 52 , 44 , 66 (
1
z has an opposite ordering to (1, 68, 40, 59) and
) 50 , 38 , 52 , 82 (
2
z has an opposite ordering to (21, 28, 40, 39) both asset 1 and asset 2 are
efficient portfolios.
However, ) 50 , 45 , 48 , 74 (
2 2
1
1 2
1
+ w w is not efficient since it isnt in the opposite order to
either valid p ( ).
Alternatively, consider
) 4 . 2 , 4 . , 1 ( * w
. Because vN-M agents view equally probable states
symmetrically, we know that for every strictly monotone vN-M (state independent) utility
function that:
*)] ( [ zw u E )] 2 . 45 , 48 , 4 . 50 , 74 ( [u E
)] 4 . 50 , 2 . 45 , 48 , 74 ( [u E
(state independent utility)
)] 50 , 45 , 48 , 74 ( [u E >
(strictly monotone utility)
)] ( [
2 2
1
1 2
1
w w u E +
Thus, a convex combination of two efficient portfolios w
1
and w
2
is not an efficient portfolio
since it is dominated by another portfolio for every monotone state independent utility agent.
Thus, E is not a convex set.
2
w
Picture: E is the shaded area

2 2
1
1 2
1
w w +
1
w

17
Systematic & Non-Systematic Risk
How do we understand these concepts using the Rothschild-Stiglitz notion of riskiness and
our definition of efficiency?
Analogous to the CAPM and Beta, we have seen that the way in which the risk or variability
of an assets returns affects the risk/expected return of an individuals optimal portfolio is
through its correlation with the state price density (rather than its correlation with Z
m
.)
o Equivalently, the correlation between an investors marginal utility of the return
on his/her optimal portfolio and any assets return.
That is, if an assets returns have an inverse ordering across the states of nature as does the
marginal utility of z* () then it has a similar correlation with the marginal utility of z* as
does z* itself and much of that assets variability contributes to the value of the portfolio
think of CAPM and correlation with the market return.
Systematic risk is the notion of an individual assets contribution to the risk of an efficient
portfolio (i.e. what portion of an assets risk is priced).
Consider marginal utility of returns as a random variable: )
~
(
~ k
e k
z u u
Now, consider a conceptual regression:
ik k ik ik i
u z
~ ~ ~
+ +
which defines the systematic risk and non-systematic risk of asset i with respect to efficient
portfolio k and utility function u( ).
Consider
) (
)
~
, (
k
i k
ik
u Var
z u Cov

Normalize ik

by
) (
)
~
, (
k
k
e k
kk
u Var
z u Cov

to remove any influence of the scale of the utility function
we are using: so define
)
~
, (
)
~
, (
k
e k
i k
kk
ik
ik
z u Cov
z u Cov
b

.
ik
b
provides us with a measure of the systematic risk of asset i with respect to efficient portfolio
k that is independent of the scale of the utility function, u().
This measure possesses a portfolio property that the b of a portfolio is the weighted average
of the bs of the assets in the portfolio (using the portfolio weights)
The ordering asset i is riskier than asset j by this measure is a complete ordering and the
ordering is independent of the efficient portfolio chosen. Therefore, if we can identify an
efficient portfolio and measure marginal utility we could correct the problem we had before
of an incomplete ordering on riskiness.
If there is a riskless asset, we can write excess expected returns as being proportional to b.
Rearrange the FOC to write:
18
] [ ] [
k i k
u RE z u E
now rewrite the left hand side and rearrange the equation
) ]( [ ] [ ] [ ] , [ R z u E z u E u RE z u Cov
i k i k k i k

This holds for all assets, including
k
e
z , so:
) (
) (
) , (
) , (
R z
R z
z u Cov
z u Cov
b
k
e
i
k
e k
i k
ik

or, ) ( R z b R z
k
e ik i
+
And, since we know R z
k
e
> ,
R z
i

is positively proportional to ik
b
.
Suppose there is no riskless asset What is the equivalent of a zero-beta asset here?
The relation can come more quickly from the standard E[ z
i
] = 1 for all assets i.
Our problem is b is in terms of
k
u
, which is not something we can easily measure. In order to
see the generality of this result is lets examine some special (familiar) cases.
If utility is quadratic, u is linear in
k
e
z
~
.
Then, ik
b
becomes
ik k
e
i
k
e
ik
z Var
z z Cov
b
)
~
(
)
~
,
~
(
. Then, all we need is the efficiency of the
market portfolio. We get that from the assumed quadratic utility since, as we have seen and
will see again, it generates 2fs which implies E is convex. So, the market portfolio is in E.
If asset returns are multivariate normal
Steins lemma says )
~
,
~
cov( ))
~
( ( )
~
),
~
( (
i
k
e
k
e i
k
e
z z z u E z z u Cov
So, ik
b
again becomes
)
~
(
)
~
,
~
(
k
e
i
k
e
ik
z Var
z z Cov
b
. See comment above. Normality also implies
2fs.
The consumption CAPM (Breeden) spills out of this, as well. What we are seeing is that it
all depends on what is a sufficient statistic for marginal utility or . With quadratic utility or
multivariate normal returns, the return on an efficient portfolio is sufficient for ) (
k
e
z u
. The
assumptions of the CCAPM imply aggregate consumption is a sufficient statistic for
marginal utility.
Nonsystematic Risk:
Consider the ik

from our conceptual regression. ik

depends upon both the benchmark

portfolio and the utility function chosen. Thus, the
ik

we identified is not an unequivocal

measure of nonsystematic risk that will be agreed upon by all investors.
The one exception is a complete market or an effectively complete market. There we know that
the marginal utilities of all investors are exactly proportional ( is unique), thus for all efficient
19
portfolios and all utility functions the
k
u
s are perfectly correlated and ij ik

for all investors.
(In a pareto efficient market all investors see the same state prices, so no valuable trades can be
created. Thus, the marginal utilities must all be proportional.)
Sufficient conditions for

-risk to be nonsystematic if it is uncorrelated with the market return

(not true generally) are a pareto efficient market and that

is a fair game with respect to m

z
.
That is, if we write i
z
~
as i i i
x z
~ ~ ~
+
where
0 ]
~
[
m i
z E
and that the market is effectively
complete (which implies that E is convex), then i

is unequivocally nonsystematic risk.

i
x
~
may be systematic, nonsystematic, or a combination of both types of risk. The market is
(effectively) complete so E is convex and the market portfolio is efficient. There exists,
therefore, a m
u
for which m
z
is the efficient portfolio and from the pareto efficiency of the
market we know
) (
m m k
z u a u
for all investors k. Since is a fair game with respect to
m
z
it is also uncorrelated with
) (
m m
z u
and so uncorrelated with any
k
u
.

is therefore
recognized as nonsystematic risk by all investors and will have no price.
In particular models the notion of nonsystematic risk being risk that is uncorrelated with the state
price density can be a handy representation.
20