Lecture 7: The Arbitrage Pricing Theory

Readings:
Ingersoll Chapter 7
Cochrane Chapter 9
Huberman, A Simple Approach to Arbitrage Pricing Theory, Journal of Economic
Theory, 1982
Ross Lecture Notes
Dybvig An Explicit Bound on Individual Assets Deviations from APT Pricing in
a Finite Economy, Journal of Financial Economics, 1983
Other APT papers from the syllabus
1
APT Rosss Color Commentary (or as close as I can come to reproducing it)
The model is:
(1)

~
~
~
+ + f B a z
where,
(2)
0 )
~
( E
(3) 0 )
~
( f E
(4)
I f f E ) (
(5)
0 ) ( f E
(6)
D E ) (
a diagonal matrix
Note that (1), (2), and (3) tell us that
z a
The model says we can describe all systematic risk with this linear factor model
Only (6) has any bite to it it requires uncorrelated residuals given this structure
The theory seeks to explain relative pricing in the asset market as is done in Modigliani-
Miller theory and Black-Scholes using the absence of arbitrage
The pricing equation: ) (
1
R z R z
k
ik
K
k i
+

The pricing is in terms of exogenous
factors and the intuition is that the s carry no premium since they can be diversified away
in large portfolios.
Compare & Contrast:
The CAPM is a static model considering the general portfolio choice problem. It is an
equilibrium model that explains pricing in terms of an endogenous market aggregate (the market
portfolio). In other words, you consider risk premia as being determined by the projection of z
i

on z
m
.
A brief survey of history tells us that, prior to the development of the CAPM, the standard
intuition was:
i f i
R z E + )
~
(
People invest in risky assets if you give them a positive expected return in addition to
compensation for the time value of money, R
f
.
This implies that if you want a preference based theory, you should use concave preferences
since risk averse investors are the ones that will demand a positive premium for holding risk.
i

was thought to be a function of an assets own variance of return

2
i
.
2
Portfolio theory and the equilibrium arguments in the CAPM gave an identification of the
risk premium
i

i that said only market risk, the projection of an assets returns on the
market portfolios returns (the part of the assets return that is correlated with the market
portfolios return), is priced you ignore all the rest.
The supposed intuition of the CAPM is that idiosyncratic risk can be diversified away
leaving only systematic (market) risk to be priced.
But,
Idiosyncratic risk in the CAPM framework is defined with reference to the market portfolio,
its the residual of the regression or projection of an assets return on the markets return.
No further assumptions about the s ' are made i.e. they could be highly correlated. In fact,
they are linearly dependent since when weighted by the market value weights they must sum
to zero. So, in any large portfolio, we cannot use the law of large numbers to say this
portfolio has negligible idiosyncratic risk, contrary to our intuition.
The exception, of course, is the market (tangency) portfolio. But, then the intuition that
diversification leads to pricing based on the market portfolio is circular at best.
The APT is also a static model but it can be seen as a static version of most inter-temporal
models in finance in which the factors represent innovations in the underlying state variables.
The APT directly assumes a return structure in which the systematic and idiosyncratic
components of returns are defined a priori. Thus, the standard notion of diversification is
directly used.
The APT is based upon the absence of arbitrage and pricing is done in terms of the
exogenous factors.
Pricing Models In General:
One starts with some specification of a technology a payoff (or returns) matrix:
[ ]
si
Y Y [ ]
N
v v v
1

and the assumption of the absence of arbitrage (or something stronger like equilibrium).
One view of pricing is that from the absence of arbitrage (and so it will be true in any
equilibrium model with increasing preferences) we know that :

[ ] 0
1
>
s
p p p
s.t.
v p Y
In a complete market we can formally identify the ps as A-D state prices. In an incomplete
market, the ps still exist but technically we cant identify them as A-D prices. This approach
is currently being pursued in interesting ways.
3
Traditionally, however, people have been looking for a pricing relation like: i f
R z +
Think like an economist for a moment.
Y
is your technology it determines how money is
moved from state to state and across time in this economy.
How do we explain prices in this world? Well, whats missing so far?
Answer: Preferences of the agents in this economy.
For illustration, how does the CAPM follow in this framework?
We restrict attention to (or specify):
(1) Multivariate normality (this is sufficient a necessary condition is 2fs). Note this is
really a joint restriction on Y and v since we specify normality for Z=Y/v.
or,
(2) We can restrict preferences to be quadratic.
Either restriction we know translates to more mean and less variance being goods (i.e. you
consider u(mean, variance) as your preference restriction in the economy). This provides the
2fs and as we have seen the simplification of the general pricing equation.
The APT is a look at how we can restrict Y the technology rather than preferences.
We can interpret the returns generating model as a rank restriction on the Y matrix this will
be illustrated more concretely in what follows. In other words the factor model restricts the
nature of what can constitute the total risk of the capital market.
By considering only restrictions on the technology (restricting ourselves only to increasing
preferences) we may only consider relative prices in the asset market and we give up (by not
further restricting preferences) the ability to talk about some bigger issues i.e. what are the
risk premia or what is their relation to other aggregate or macro variables.
As an illustration lets develop an arbitrage based approach to the CAPM pricing equation or the
SML. Suppose:
i i i i
f z z
~
~
~
+ +
a 1-factor model
where, written this way,
f
~
is necessarily the unexpected innovation of some exogenous factor f f f
~ ~
+
Assume there are N risky assets and one riskless asset.
Also assume:
(1)
i f E
i
0 )
~
~
(
Consider a positive investment portfolio w ( 1 1 w ):
4
w f w z w f z w z w Z
w
+ + + + ) ( ) (
~
Set
0 w
- i.e. choose a portfolio with no systematic risk. This assumes it is not the case that
all assets have the same level of systematic risk. That is, there exist assets i, j s.t. j i

(or
that the vectors
1
and

are not collinear). Note, since relative pricing is the goal this is a
minimal assumption, on the order of i, j with j i
z z
, made in the CAPM derivation.
(2) Suppose the portfolio w is also well diversified. That is, in some metric you specify the
w
i
s are close to 1/N for all assets i.
(3) Assume either an upper bound on the variances of the

s or more simply assume that

2 2


i
i . i.e. the

s all have the same (finite) variance. Also assume that

0 )
~ ~
(
j i
E
if
j i
(uncorrelated residuals).
The variance of w is then close to
2
2
N

if w
i
is close to
N
1
i .
For example, if

,
_

2
2
1
0 ) ( ) ( 0
N
w Var w
N i

. Or,
0 w as N
0 ) ( w E 0 ) ( w Var
as N thus we can ignore the s.
(This is where we see the bite of equation (6) and where all the difficulty comes in the APT.)
Now, since we set
0 w
we have:
z w w f w z w Zw + + ) (
for large N
Economics tells us that z w must equal f
r R + 1
w s.t. 1 1 w
0 w
0 w
a well diversified, positive investment portfolio with no systematic risk must
have expected return R z w
Similarly,

s.t.
0 1
,
0
and

well diversified, we must have

. 0 z
If no money is invested, and no risk exposure assumed, we should expect no return.
Since both conditions hold, it must be that z is a linear combination of
1
and

Or,
+ 1 z
Since this must also work for R (our portfolio w):
+
R
R
And, we can write:
+ 1 R z
Further, picking an asset with = 1 we see that R z
1
where
1
z
is the expected return on
an asset with a factor loading equal to 1.
The development in the text is also intuitive:
Pick two assets with
2 1

.
5
Consider a residual risk free, positive investment portfolio ( 1 1 w ) of the two assets with
no systematic risk 0
2 2 1 1
+ w w :
w f w z w z
w
+ +
R z w z
w

or, 0 ) 1 ( R z w
In matrix form,
1
]
1

1
]
1

1
]
1

0
0
2
1
2 1
2 1
w
w R z R z

Since this matrix is singular, it must be that the first row is collinear with the second:
i i
R z ) (
Or, i i
R z +
showing our intuition was correct. (Draw a picture!)
Again considering an asset or portfolio with
1
:
+ R z
1
or, R z
1

Thus, ) (
1
R z R z
i
+
And, we get the standard intuition that for a risky asset the expected return is made up of the time
value of money and a risk premium.
In the returns equation
i i i i
f z z
~
~
~
+ +
, economics tells us that i
z
and i

should be related.
We found:
) (
1
R z R z
i i
+
What restrictions drive these results?
(1) the absence of arbitrage
(2) a one factor model of the returns generating process
Forget

for the moment.

State-by-state: s i i si
f z z +
If this holds, then each column of the Z matrix is a linear combination of 2 column vectors:
for, column j:
f z z
j j j
+ 1
(i.e. with no residual risk, the Rank(Z) = 2)
This illustrates the interpretation of the APT factor-model as a rank restriction on
Z
Here, it is a linear combination of a vector of ones and something else (f the factor).
Is it always a vector of ones?
6
Now, go back and see whats wrong.
j j
R z R z ) ( +
is only an approximate pricing result the rank restriction is only
approximate when there is residual risk. There is a whole literature on what is close Absence
of Asymptotic Arbitrage and how big can any pricing errors can be?
The APT brings in exogenously specified factors.
What are they?
What else is missing?
What are the risk premia?
i.e. How big is
R z
1
- to what does it relate?
Since there is no restriction placed on preferences other than monotonicity, we dont or cant say
anything about these questions.
Two Factor Model (with no idiosyncratic risk):
Let returns be described by:
2 1
~ ~
~
f c f b a z
i i i i
+ +
where
b
,
c
, and
1
are not collinear (so that our factor loadings are sufficiently different a
term we will use now and make more precise later).
Form a portfolio of 3 assets its return is:
c w f b w f a w Z
w
+ +
2 1
choose w so that 0 c w b w
R a w Z
w

or, 0 ) 1 ( R a w
We can write this as:
1
1
1
]
1

1
1
1
]
1

1
1
1
]
1

0
0
0
3
2
1
3 2 1
3 2 1
3 2 1
w
w
w
c c c
b b b
R a R a R a
Again, this is a singular matrix where the last 2 rows are not collinear.
So, it must be that:
i i i i
c b R a R z
2 1
+
i
Again, let
i
z
be the return on a portfolio with one unit of factor i risk.
We identify
1
1
R z and
2
2
R z and the familiar pricing formula is complete.
Unavoidable Risk:
7
Consider the two factor model:
2 1
~
f f b a z
i i i
+ +
0 i
Any positive investment portfolio has returns:
2 1
f bf w a w Zw + +
Clearly, we do not have the factor loadings on factor 2 sufficiently different across assets this is
an extreme version.
However, any portfolios with the same b are perfectly correlated and must have the same
expected return from the absence of arbitrage (here simple dominance).
Define a
o
as the expected return on any portfolio with b=0.
This is well defined by the argument just given.
Form 2 portfolios, w
+
and w
*
, with 3 assets (i, j, o).
Let:
1
*
+
+
i i
w w
2 1
*
2
*

+
+
+
o o
j j
w w
w w
By construction, the two portfolios have the properties that:
j i
o
j
o
i
b b b b
a a a a a a
2 1
2 1
*
) ( ) ( *


+ +
+ +
+
+
Now, choose
0
2 1
+
j i
b b
. Then, b
+
= b*, and it must be that a
+
= a*.
Thus,
1
1
]
1

+ ) ( ) ( ) ( 0
1 2 1
o
j
j
i o
i
o
j
o
i
a a
b
b
a a a a a a
(by solving
0
2 1
+
j i
b b
for
2
)
This must, then, hold for any choice of
1
(and its accompanying
2
). So,

j
o
j
i
o
i
b
a a
b
a a
The linear pricing relation is then (by simply rearranging this expression):

i
o
i
b a a + or,
) (
1 o
i o i i
z z b z a z +
where
o o
a z
1
z
= expected return on a portfolio with b = 1
Note: The riskless rate does not appear here, even if a riskless asset exists. This occurs because
we cannot create a risk free asset with these risky assets the absence of arbitrage therefore does
not provide a relation between R and the expected return on risky assets. In particular,
R z
o

.
8
In fact, we expect
R z
o
>
is the usual relation since you must compensate investors for their
exposure to the unavoidable factor 2 risk. Similarly,
1
z
can be greater than or less than
o
z
.
The Pricing Equation:
An interesting question to ask at this point is what can we learn from the pricing relation? Can
we put it in the context of other results we have seen?
Recall: that

) (
k
e
Z u
(i.e. the state price density is equal to the marginal utility of future
return on the optimal portfolio scaled by the lagrange multiplier or the marginal utility of current
consumption). Thus the general pricing model tells us that pricing is determined by the
covariance of an assets with the marginal utility of the return of an optimal portfolio or the its
covariance with the state price density. Recall also that when there was a set of circumstances in
which there was a simple sufficient statistic for this marginal utility we were able to derive
standard pricing equations like the CAPM (quadratic utility or normally distributed returns) or
the CCAPM (when consumption or consumption growth is sufficient for marginal utility) from
the general formula.
To make this point in a slightly different way consider the CAPM. Given the model:
(1)
] [ 1
i m
z E and bz a +
We can find constants and such that
(2)
im i
z E + ) (
.
Conversely, given (2) we can find constants a and b such that (1) holds.
This simple result follows from the standard decomposition of
] [ 1
i
z E
.
) (
) , (
) (
1
) (

E
z Cov
E
z E
i
i

Write this as

im im
m
i m i
i
bE
Var
E
E
Var
z bVar
z z Cov
E E
Var
Var
z Cov
E
z E

,
_

,
_

,
_

+
) (
) (
) (
1
) (
) (
) (
) , (
) (
1
) (
) (
) (
) , (
) (
1
) (
Where the second equality follows from
m
bz a +
. Further, if a riskless asset exists, note that
R
E

) (
1

.
From the last equation for expected return we can obviously go backward and show that there
exists an a and a b such that (1) holds. Thus the standard CAPM is equivalent to there being a
state price density that prices all assets that is linear in the return on the market portfolio. Note
this may not be a strictly positive state price density if the return on the market can be extreme.
Not too surprisingly, the same kind of result holds for the APT model as well. The pricing
model there is a multi-factor beta pricing model. From Cochrane (pg 107):
Given the model:
(1)
) ( 1
i
z E and f b a +
9
where b and f are vectors, we can find a and a vector such that:
(2)
i i
z E + ) (
where the
i
are the multiple regression coefficients of z
i
on f and a constant. Conversely, given
(2) we can find an a and a vector b such that (1) holds.
Thus these are equivalent pricing relations. This also gives us an idea of how we might look for
or select factors. The factors should be things that help model marginal utility. Thus select
variables that will proxy for how happy people are: z
m
, the business cycle, production,
consumption, input and energy prices, etc.
Asymptotic Arbitrage Opportunities (AAO):
An AAO exists if there exists a sequence of arbitrage portfolios w
n
n=2, (here we are thinking
of a sequence of economies with the number of assets, n, increasing) such that:

n
i
n
i
w
1
0 n (all are arbitrage portfolios)

>
n
i
i
n
i
z w
1
0 n (expected return is bounded away from zero)

n
i
n
j
ij
n
j
n
i
n n
w w w w
1 1
0 '
(risk disappears)
For the variance all that is technically required is that some infinite sub-sequence has variance
with a limit of zero.
We focus on variance because as
n
0 risk . So, as
n
you get a riskless payoff and
mean-variance analysis is a good approximation.
This is an extension of the notion of a riskless arbitrage with two cautions:
(1) In the simple case, a riskless arbitrage opportunity has arbitrary scale, so profit
comes with any unbounded position. With AAO, you must be careful how you
increase scale as risk must still vanish in the limit. That is, if
n
w is a sequence
defining an AAO, then
n
w is as well for any
0 >
. However,

cannot become
unbounded in an arbitrary fashion it must depend on n.
One way this can be done is to set
4
1
) ' (


n n
n
w w
Then,
n
n
n
w w
is an AAO with infinite profit in the limit




n
n
i
i
n
i n i
n
i
n
i
z w z w
1 1

(as
n
)
( ) 0
2
1
1 1 1 1 1 1
2



n
i
n
j
n
i
n
j
n
i
n
j
ij
n
j
n
i ij
n
j
n
i n ij
n
j
n
i
w w w w w w
(as
n
)
10
(2) AAO is not a preference free idea. A riskless arbitrage opportunity guarantees
infinite wealth and an infinite certainty equivalent utility of return. i.e
) ( ) ( u Zw u
n
, so an unbounded position is taken
Chebyshevs inequality
) ]
~
[ (Pr
1
t t x ob
tells us that an AAO does guarantee infinite
wealth with probability one, however an infinite certainty equivalent wealth is not guaranteed.
That is, not all investors take a position in an AAO.
A counter-example in the text demonstrates this and we can see that an opportunity to increase
wealth with no investment and vanishingly small risk does not guarantee an increase in expected
utility in the example, no investor invests in the AAO.
If a utility function is bounded below by a quadratic function, then an AAO is a good deal. But,
few nice utility functions are bounded in this way.
The Counter Example:
Let
z
z u

exp ) (
Realized returns are
} 1 , 1 , 1 { n n +
Probabilities: } 2 1 , , {
3 3 3
n n n
0 1 ] 2 1 1 [ 1
3
> + + +

n n n z
n
n
0
2
] ) [(
2 2 3 2
+

n
n n n
n
as n
1 3 1 1 1 1 3
) 2 ( ) (

+
n n n
n
e n e e e e n z Eu
As
n
, the limit of this is , so investing in this AAO reduces utility to .
Pricing With Idiosyncratic Risk:
Theorem 1: If the returns on the risky assets are given by a k-factor model with bounded
residual risk
2 2
(


i

) i
and there are no AAOs, then: There exists a linear pricing model
which gives expected returns with a mean squared error of zero that is k o
,..., ,
1
(dependent upon n) such that:

K
k
k ik o i i
b a v
1

0
1 1 2
2


n
n
i
n
v
n
Lim v
n
Lim
where n
v
is the Euclidean norm of the length of the vector v:
2
1 i
n
i n
v v


Proof: Select n assets and number them 1,n. Regress their expected returns on the b
ik
and
call the regression coefficients k

. (This is really a population exercise, formally it is a

projection of the a
i
on the space spanned by the matrix
B
and the vector
1
, the constant in the
11
regression. If there is a multi-collinearity problem, prespecify as many of the k

as is necessary
to remove the problem.)
The residuals v
i
of this regression are given by:
i k ik K o i
v b a + +
.
From the orthogonality property of the projection, we know:
0
i i
v
and
0
ik i i
b v
k
Now, consider the arbitrage portfolio:
n v
v
w
n
i n
i

The payoff on this portfolio is:
( )
i i i n
z v v n
1
) ( (
1
i k ik k i i i n
f b a v v n + +

) ( (
1
i i i i n
a v v n +

Expected profit is:
( )
i i i n
a v v n
1
] [ ) (
2 1
i ik i i i i o n
v b v v v n + +

2 1
) (
i n
v v n

n
v
n

Variance of Profit is:

n
v v n
i i n
2
2 2 1
2
) (

Now suppose that the theorem is false so that: 0

2

n
v
i
in the limit.
Then,
n
v
n
cannot go to zero and an AAO exists.
So, if no AAOs exist, then
0
n
v
n
and
0
2
2

n
v
n
v
i n
as well. QED.
The derived no arbitrage condition is 0 ) (
2
1

k ik k o i i N
b a . Each term in this sum is
non-negative. The average term is zero, thus all but a finite number must be negligible.
More precisely, if we order the assets by the size of their absolute pricing error:

n
v v
1 , then for any , no matter how small, there exists a finite N such that fewer
than N assets are mispriced by more than .
>
N N
v v v
1 1
12
With an infinite number of assets, the probability (picking one at random) of getting one which
has an error of more than is zero as the assets with this size error or worse are a finite set
which has a measure zero in an infinite set of assets.
Thus, the linear model prices most assets correctly and all with a negligible mean squared error.
It can, however, be arbitrarily bad at pricing a finite number of assets.
The Magnitude of the Residual Variance and the Pricing Bound:
Why doesnt the residual variance enter the pricing bound equation? We might expect that
assets with small residual variance should be priced very closely, after all we know that among
assets with no residual risk the pricing is exact. This is confirmed in the following theorem
(Dybvig 83).
Theorem 2: Under the conditions of Theorem 1, the pricing error must satisfy the following:
0
1
2
2

,
_

i
i
i
n
v
n
Lim

Proof: Look at a weighted regression.

Regress
i
i
a

on
i
ik
b

The residuals are given as:

i
i
i
ik
k k
i
o
i
i
v b a

+ +
Where,
0
i
ik
i
i
i
b v


k and
0
1

i i
i
i
v


13
Consider the portfolio:
2
1
2
2
2

,
_

i
i
i
i
i
n
i
v
n
v
w

Profit on the portfolio is given by:

) (
2
1
2
1
2
2
i k ik k i
i
i
i
i
f b a
v v
n


+ +

,
_

,
_

= ) (
2
1
2
1
2
2
i i
i
i
i
i
a
v v
n


+

,
_

,
_

Expected Profit is:

) (
2
1
2
1
2
2
i k ik k o
i
i
i
i
v b
v v
n + +

,
_

,
_

=
2
1
2
1
2
2
i
i
i
i
v v
n

,
_

,
_

=
n
v
i
i

,
_

,
_

2
1
2
2

Variance of Profit is:

2
4
2
1
2
2
i
i
i
i
i
v v
n

,
_

,
_

=
2
2
1
2
2
i
i
i
i
v v
n

,
_

,
_

=
N
1
Now, suppose that the theorem is false. That is,
0
1
2
2

i
i
v
n

in the limit, then

2
1
2
2
1

,
_

i
i
v
n

cannot go to zero and an AAO exists.

So, if no AAOs exist,
14
2
1
2
2
1

,
_

i
i
v
n

and

,
_

2
2
1
i
i
v
n

both 0 as n .
Priced & Unpriced Risk:
We must specify the factors more fully before we can identify what risk is priced because: we
can from any k-factors create a different set of uncorrelated factors for which only one receives a
positive factor risk premium.
The model is:
+ + f a z
with
+ 1
o
a
Now, look at an orthogonal transformation of f:
f T f

where I T T so
f T f

Then, we can write:

+ + + + f a f T a z

where, T

T a
o
+

1
We must show T is a vector with one non-zero entry and that the transformed model is valid.
The transformed model is proper since:
0 ) ( ) ( )

( T f E T f E f E and,
I IT T T f f E T T f f T E f f E ) ( ) ( )

( for any orthogonal T.
Now, choose a matrix T

,
_

1
2
1
, ) (
kxk
x T
where x is any matrix (k x k-1) whose columns are mutually orthogonal and all are orthogonal
to .
By construction, T is orthogonal I T T
So,
) , ) ( (
2
1
x T T T T I


And,
) 0 ,..., 0 , 0 , ) ((
2
1
T
The lesson being that in any equilibrium, only a portion of the uncertainty (risk) brings
compensation. This can be true even when unpriced risk is common to many or all assets. We
cant make any meaningful economic statements until priced and unpriced sources of risk are
identified. The CAPM by its preference restriction does this. In particular, we must say
something further about the nature of the factors in the returns generating model or we cannot
attach any significance to the size or the sign of the k

.
15
Fully Diversified Portfolios:
We would like to say that ) ( R z
k
k
. The problem is that all such
k
z
~
may not be priced the
same since pricing is only approximate. So,
k

would not be well defined.

A fully diversified portfolio is the limit of a sequence of positive net investment portfolios whose
weights satisfy:

<
n
i
i
n
c n w n Lim
1
2
)) ( (
i.e.. the w
i
vanish for most assets.
) ( 0
1
n i
w
for all but a finite number of assets in a fully diversified portfolio.
The important feature is of course that they have no residual risk in the limit:

,
_

n
i
i
n
n
i
i i
n
n
w n Lim w Lim
1
2
2
1
2 2
0 ) (

There may also exist less than fully diversified portfolios with no residual risk.
Theorem 3: The expected returns on all fully diversified portfolios are given correctly with
zero error (in the sense that 0 ) (
2

i i
v w so the error on any fully diversified portfolio is
negligible) by any linear pricing model satisfying Theorem 1.
Proof:
The expected return on the n
th
portfolio in a sequence is:
a
F i i k ik i K o
v w b w + +
F k Fk K o
v b + +
If n F i i
v v w
,

vanishes, the pricing is exact.
Consider
n
v
w n v w v w v
i
i i i i i F
2
2 2 2 2 2
) ( ) (


(From Cauchy-Schwartz inequality, )) ( ) ( ]) [ (
2 2 2
Y E X E XY E , but the 1
st
term is bounded
and the second term goes to zero from Theorem 1.)
So, for a fully diversified portfolio, we know o

is the return on a portfolio with no factor risk,

but it also has no residual risk since it is fully diversified. So,
R
o

must hold.
Also, for fully diversified portfolio with one source of factor risk.
i.e. b
ik
=1 and b
ij
=0
k j
, the expected return is exactly
k
k
R z +
So, R z
k
k
is now well defined.
Interpreting the factor premiums:
16
Define
) , 1 (

n n

the augmented factor loading matrix, an n x k+1 matrix.
Assume for now that it has rank = k+1 for large n. n

must have rank k or there is unavoidable

risk i.e. columns of
n

are collinear and none of the column of

n

should be collinear with a

constant vector. In either case, you cannot form vectors with one source of factor risk.
Define n Q
n n n
/

and assume the sequence Q
n
has a limit. Q
n
is k+1 x k+1 independent of n
and has a limit if each element does. We know these are all bounded since the factor loadings
are bounded.
Consider the sequence of portfolio formation problems using n = k+1, k+2, assets:
2
2
1
i
w Min such that
1
i
w 1
ik i
b w 0
ij i
b w k j
So, form well diversified portfolios with single factor risk (positive investment):
)

(
2
1
w c w w L +
where

is a vector of Lagrange multipliers and c is a vector with a 1 in the 1

st
and k
th
positions
and zeros elsewhere.
The FOCs are:
(1)

0 w
(2) w c

0
Pre-multiply (1) by

0 ) 1 ( c w

Or, c
1
)

(


Or, from (1) again
1
)

w
By construction, this portfolio has a single source of factor risk and is fully diversified if:
c Q c c c n w w n
1 1 1
)

(

)

(

is bounded.
Since c and Q
n
are finite in size this can be unbounded only if Q
n
becomes singular in the limit.
Q
n
is non-singular for finite n since n

is of full column rank.

If c is the vector i
1
the portfolio has no factor risk and is fully diversified.
Its expected return must be
R
o

.
The portfolio interpretation:
) ( R z b R a
k
ik i
+
where
k
z
is the expected return on a portfolio with one unit of factor k risk and no other risk
0 (
ij
b

k j
and fully diversified) is valid if Q
n
is non-singular in the limit.
This makes precise our loose sufficiently different phrase in an economy with residual risk.
Exact Pricing in the Linear Model:
17
Under what conditions does the total error converge to zero when is pricing exact for all
assets?
A sufficient condition:
At least one investor chooses to hold a fully diversified portfolio and each assets
idiosyncratic risk is a fair game with respect to the factors:
0 ) (
1

k i
f f E
i
Exact pricing follows from the separation result we saw previously.
When will some investor hold a fully diversified portfolio?
Is it correct to say that idiosyncratic risk is not priced and so all risk averse investors hold
well-diversified portfolios?
No this is circular reasoning.
The idea is that if the risk is diversified away by investors, then in equilibrium it will not be
priced the argument above says its not priced so no one holds it.
Example: One assets idiosyncratic risk is priced a zero factor world:
1 1 1
~ ~
+ a z
i i
a z
~ ~
+
2 i
Let the
i

~
2 i be i.i.d. if a=R, assets 2, 3, 4, are all priced correctly.
However, if R a
1
, asset 1 is not can this be sustained in an equilibrium? Yes.
By symmetry, all investors hold assets 2, 3, equally, in the limit this duplicates the riskless
asset, so a=R in any equilibrium.
Suppose ) N(0, ~
2
1

Z
e u


We know that the portfolio weight on asset 1 is:
2
1

R a
w

with 1-w put in the riskless
So, R a
1
is consistent with equilibrium.
Here, the
1
risk of asset 1 is a significant portion of all market risk.
We do have
Theorem 5: A sufficient condition for the arbitrage model to provide exact pricing in the limit
is if:
(i) returns are given by a factor model with
0 ) (
1

k i
f f E
.
(ii) The market proportion or supply of each asset is negligible
(iii) The loadings on each factor are spread evenly among the assets
(iv) No investor takes an unboundedly large position in any asset
w w
i
< * (

) i
(v) Marginal utility is bounded above zero. Then, the pricing relation
k ik o i
b a +
i
holds and the pricing errors converge to zero in the sense that:
18

n
i
i
n
v Lim
1
2
0
i.e. all errors must be negligible in the limit (its not

n
i
i
n
n
v
Lim
1
2
0 )
Proof:
(ii) and (iii) describe a fully diversified economy
(iii) - no asset is a big portion of the market portfolio
(iii) - assures that each factor affects many assets and that each factor must therefore make an
identifiable contribution to market risk no multi-collinearity in the columns of n

In a fully diversified economy, all investors can simultaneously hold fully diversified portfolios
with factor loadings spread among their portfolios in any form.
(iv) and (v) guarantee an interior optimum to each investors choice problem
(iv) - says risk aversion doesnt vanish so variance is always disliked
(v) - nonsatiation expected returns are always liked
(i) the residuals dont matter for pricing in the limit, so you get a bound no the
0
i
v
has
the same role as in the discussion of the separating distributions.
19