Sunteți pe pagina 1din 24

The Best-Beta CAPM

(BCAPM)
(Literature:
(Literature: Paper
Paper Zou
Zou 1)
1)

Outline
The Variation-Around-Target
Preferences
Re-examination of Modern Portfolio
Theory
The Separation Theorem
Derivation of the BCAPM
Comparing the BCAPM with the CAPM
Applications
2006 by Liang Zou. All rights
reserved.

3-2

Introduction
Mean-variance (MV) CAPM has been with us for 40
years. Despite the lengthy debate over its
theoretical and empirical validity it remains a major
model for estimating cost of capital, discussing
market efficiency, and so on.
Empirical evidence suggests, however, that the
CAPM may entail pricing errors and that the market
index may not be mean-variance efficient.
Although many multifactor models are proposed,
these are empirically motivated models and are
likely to be subject to data mining. More seriously,
the factors that are reportedly useful for improving
the explanatory and predictive power of the CAPM
often
lack theoretical insights.
2006 by Liang Zou. All rights
reserved.

3-3

Motivating the BCAPM


Why improving the accuracy of
expected returns so important?
Consider the dividend-growth model
for evaluating a companys equity
value:
D

VE

kg
D First period dividend

k Required rate of return


g Dividend growth rate
2006 by Liang Zou. All rights
reserved.

3-4

What a small difference can


do
A critical problem is how to estimate the required rate of
return k. One solution is to use the CAPM and estimate
the beta of the risky cash flows.
Accuracy of k is highly important, however.
For instance, based on a current dividend yield of 2%, a
1% difference in k could lead to 50% difference in the
estimated firm equity value!
The reason is that the market capitalizes the firms future
earnings. Since there are infinitely many periods ahead,
any small improvement in the precision of expected
return (or required rate of return) per period could add up
to substantial improvement in the estimated present
value of the firm.
2006 by Liang Zou. All rights
reserved.

3-5

What is Best-Beta?
The best-beta is defined as

E
(
xx
)
m
B
E ( x m2 )
x r r0 (asset' s excess return)
x m rm r0 (market' s excess return)
2006 by Liang Zou. All rights
reserved.

3-6

What is the Best-Beta


CAPM?

The best-beta CAPM predicts the same


expected return beta relation as the
CAPM, except that the beta is different.

CAPM : E

MV

( x)

MV

E ( xm )

BCAPM : E ( x ) E ( xm )
B

MV

Cov ( x, xm ) B E ( xxm )

,
2
Var ( xm )
E ( xm )

2006 by Liang Zou. All rights


reserved.

3-7

Compare BCAPM with CAPM


Let the true expected return and beta
be given (which we do not know):

E ( x ) E ( xm )
The pricing error of the CAPM and the
BCAPM are the difference between the
true and predicted expected return,
denoted by alpha:

MV E ( x) MV E ( x m )
B E ( x) B E ( x m )

2006 by Liang Zou. All rights


reserved.

3-8

Theorem
The relation between the MV beta and
the best beta is given by

(1 )
B

2
m

MV


2
m

2
[
E
(
x
)]
2
m
where m
(0,1)
2
E ( xm )

We call m2 the "eta ratio", which is


a variation of the Sharpe ratio.

2006 by Liang Zou. All rights


reserved.

3-9

Implication of the theorem


The preceding theorem implies that
the best beta is always closer to the
true beta than the MV-beta, that is,
one of the following relations must
hold for all situations and all assets:

MV

MV

MV

2006 by Liang Zou. All rights


reserved.

B
B

3-10

Proof of the theorem


By definition,

E ( xxm ) Cov( xxm ) E ( xm ) E ( x )


2
E ( xm )
E ( xm2 )
B

Var ( xm ) MV [ E ( xm )]

2
2
E ( xm )
E ( xm )
(1 )
2
m

MV

2006 by Liang Zou. All rights


reserved.

2
m

3-11

Theorem
The preceding theorem implies

E ( x) (1 ) E
B

2
m

(1 )
B

2
m

MV

MV

thus either
B

or

MV

( x) E ( x )
2
m

MV

0
B

or MV B 0
2006 by Liang Zou. All rights
reserved.

3-12

Empirical Example
The U.S. equity annual returns on the
NYSE, AMEX, and NASDAQ have an
average risk premium of 7.67% (with
standard deviation of 16.03%) over the
period of 1952 to 1999 and 10.61%
(with standard deviation of 15.15%)
over 1983 to1999.
We then derive the eta ratio to be
0.18629 (over 1952-1999)
0.32907 (over 1983-1999)
2006 by Liang Zou. All rights
reserved.

3-13

The eta ratio as a function of


market risk premium and
volatility
1
0.8

0.6
0.4
0.2
0
0
0.1

0.2
0.3 0.5

2006 by Liang Zou. All rights


reserved.

0.4

0.3

E(xm)

0.2

0.1

3-14

Derivation of the BCAPM


Assume each investor has the
objective

max E (rp )
p

E ( rp )

With two individual parameters:


degree of risk aversion and a target
return .
This objective function can be seen as
an approximation of expected utility.
2006 by Liang Zou. All rights
reserved.

3-15

Separation Theorem
Then we can show that every one wants to
maximize the eta ratio:


2
P

E ( xP )

E ( xP2 )

In equilibrium, the common optimal risky


portfolio P must be the market portfolio and
we can then derive the BCAPM (see paper
Zou 1).
2006 by Liang Zou. All rights
reserved.

3-16

From SDF to BCAPM


Alternatively, the BCAPM can also be derived
by specifying the stochastic discount factor
(SDF) as follows:

a bxm , E ( x) 0
aE ( x) bE ( xxm )
aE ( xm ) bE ( xm2 )

E ( xxm )
E ( x)
E ( xm ) (BCAPM)
2
E ( xm )

Question: What is the difference between this


derivation and the previous derivation?
2006 by Liang Zou. All rights
reserved.

3-17

Insights
Consider a zero-price investment
strategy with a long position in any
asset i and some short position in
market and cash:

ri brm (1 b)r0

What b gives the best hedge of market


risk?
Answer: the best beta if we use the
least-squares criterion.
2006 by Liang Zou. All rights
reserved.

3-18

The least squares solution


min b E (ri brm (1 b)r0 )

The necessary and sufficient condition yields

E ( xxm )
B
b

2
E ( xm )

2006 by Liang Zou. All rights


reserved.

3-19

Further Insights
Without loss of generality, we can
always write out the equation:

xi i bi xm with E ( ) 0
The question is how we should
interpret alpha: Do we assume that it
belongs to a correctly specified linear
model, or do we assume that it
represents a source of pricing error
due to model uncertainty?
The BCAPM treats alpha as source of
error.
2006 by Liang Zou. All rights
reserved.

3-20

How many masters do you


serve?
CAPM attempts to serve two masters:
Alpha wants to be zero, and Beta wants
2
to
min E ( x bx )
b

These two masters ask you to do the


same thing if and only if happens to
be zero when you choose the optimal .
Instead, BCAPM serves only one master,
2
Beta, who wants you to

min b E ( xi bxm )

2006 by Liang Zou. All rights


reserved.

3-21

The Econometrics View


Equivalently, the MV alpha and beta are
the solution to minimizing expected
residual squares:

min ,b E ( ) b
2
i

MV

MV

The best beta and the corresponding


alpha are derived by minimizing the
combined errors:

min ,b E ( i i ) b ,
2

2006 by Liang Zou. All rights


reserved.

3-22

The Econometrics of BCAPM


When we use the historical data to estimate the
best beta and alpha, it can be shown that

Var ( ) (1 )Var (
B

2
m

MV

Var ( ) (1 ) Var (
B

2 2
m

MV

So that the BCAPM also gives more accurate


estimates of the alpha and beta.
Question: How would you compare the tstatistics of the estimated alphas of the two
models?
2006 by Liang Zou. All rights
reserved.

3-23

Concluding Remarks
MPT can be entirely re-written without much
ado. Given that investors may indeed have
personal target returns, however, the
assumption of the BCAPM could be
considered to be more realistic.
The best-beta approach could be useful for
designing more powerful (or more sensible)
tests of the CAPM-like models.
The BCAPM can be readily extended to
multifactor models, and its improvement
over the MV models become increasingly
more significant as the factor number
increases.
2006 by Liang Zou. All rights
reserved.

3-24

S-ar putea să vă placă și