Вы находитесь на странице: 1из 4

# Capital Asset Pricing Model

Dan Saunders
Regressions
A regression is the statistical procedure used to t a linear trend to data. Regression analysis
is most useful when several variables are observed simultaneously. In much of economics and
nance, variables cannot be controlled for like in a laboratory. Therefore, regressions become
important because they allow for statistical control over multiple explanatory variables.
Typically, a regression plot of just one independent and dependent variable make look like
this
A simple regression refers to such a case where there are only two variables. Generally,
this model can be written down as follows
y =
0
+
1
x +
where N(,
2
) represents the error term. The most straightforward interpretation is
that y is deterministically related to x through
1
, except for a possible dierence in means

0
, and some randomized error term , which will capture the remaining, unexplainable
1
variation. As some may remember from a previous course in statistics, the best way to
estimate the slope is by ordinary least square (OLS), which gives the following formula

1
=
1
n

i
(x
i
x)(y
i
y)
1
n

i
(x
i
x)
2
This formula should look familiar, since its the sample mean estimator for the population
parameter for the slope

1
=
E
_
_
x E(x)
__
y E(y)
_
_
E
_
_
x E(x)
_
2
_ =
Cov(y, x)
V ar(x)
Under certain assumptions, which we will implicitly make in the CAPM, this process for
estimation has all of the desirable properties that other popular estimators share; such as
the arithmetic mean x.
Asset and Portfolio Betas
Assuming that the CAPM is true, we know the following must hold for any two assets i, j
and market portfolio m
E(r
i
) r
f

i,m
=
E(r
j
) r
f

j,m
In particular, suppose asset j is the market portfolio m. Then we can re-write this expression
as
E(r
i
) r
f

i,m
=
E(r
m
) r
f

2
m
After a few algebraic steps, we nd the equation for the security market line
E(r
i
) = r
f
+

i,m

2
m
..

i
_
E(r
m
) r
f
_
Notice that we could rewrite the beta as

i
=
Cov(r
i
, r
m
)
V ar(r
m
)
Thus, if we do not know the correct value of for a given asset, then we should collect
observations of returns and regress r
i
onto r
m
in order to generate a statistically unbiased,
ecient estimate.
2
Calculating Beta in class
While it is helpful to recall regression analysis, it would take too much time to actually have
students perform regressions. Hence, most of the time you will be asked to calculate using
population parameters. Recall that we can rewrite the covariance as Cov(r
i
, r
m
) =
i

i,m
.
Then we can rewrite the denition of the true value of beta as

i
=

i

i,m

2
m
=

i

i,m
Often, the equation on the right hand side will be useful in calculating the value of beta
on homework or on exams. Just remember that this is the theoretical value of beta. In
real-world applications, it is necessary to sample and estimate this equation.
What does Beta mean?
In an interview given by William F. Sharpe (as in the Sharpe ratio), there is a somewhat
famous quote in which he explains the signicance of and the CAPM
But the fundamental idea remains that theres no reason to expect reward just for bearing risk.
Otherwise, youd make a lot of money in Las Vegas. If theres reward for risk, its got to be special.
Theres got to be some economics behind it or else the world is a very crazy place. I dont think dierently
about those basic ideas at all.
Lets put this all together. Beta represents a special type of risk that is rewarded;
specically, the systematic risk of the nancial markets. The remaining unsystematic, or
idiosyncratic, risk is not compensated with higher returns, because this risk can be diversied
away. Therefore, large undertakings of risk, in general, do not present the possibility of higher
rewards.
Recall that there is a pure time value to money. For now, lets simplify interpretation by
thinking of the risk-free rate r
f
as representing this pure opportunity cost. Lets look at the
security market line once more
E(r
i
) = r
f
+
i
_
E(r
m
) r
f
_
Then we see that the expected return to asset i starts with the baseline time value of money
r
f
, which is independent of risk. Then, the excess return above and beyond the risk-free
rate is given by E(r
m
) r
f
. Since represents all of the market risk contained in asset
i, we multiply the excess return by beta to nd the total excess return aorded to asset i,
specically because of its systematic risk.
In order to measure beta, we regress the returns of any given asset (or portfolio) on
the returns of the market portfolio. The resulting slope coecient represents the level of
market risk present within that security. On the other hand, the standard deviation of an
asset
i
still represents total risk. However, it is systematic risk, not total or unsystematic
risk, which is compensated with higher returns. To see this, lets look at an example.
3
RWJ Ch.11 #34
Assume that the CAPM is true. We are given r
f
= .04 and E(r
m
) r
f
= .075. We are also
given the following information about two assets
state probability r
1
r
2
bad .15 .09 .30
medium .55 .42 .12
good .30 .26 .44
Lets compare the expected returns with the systematic, unsystematic, and total risk of each
of these two assets.
For asset 1 we nd
E(r
1
) = .15(.09) + .55(.42) + .30(.26) = .3225

1
=
_
.15(.09 .3225)
2
+ (.42 .3225)
2
+ (.26 .3225)
2
_
1/2
= .1205
Using the SML we nd
.3225 = .04 + .075
1
=
1
= 3.77
For asset 2 we nd
E(r
2
) = .15(.30) + .55(.12) + .30(.44) = .1530

2
=
_
.15(.30 .1530)
2
+ .55(.12 .1530)
2
+ .30(.44 .1530)
2
_
1/2
= .2368
Using the SML we nd
.1530 = .04 + .075
2
=
2
= 1.51
The idea here is quite simple. Asset 1 has a higher beta and, therefore, a higher expected
return. This is because the higher beta represents a higher level of systematic risk in asset 1.
On the other hand, asset 2 has a higher standard deviation, meaning it has greater total risk.
This implies that asset 2 must also have greater unsystematic risk than asset 1. However,
unsystematic risk is not the special kind of risk William Sharpe was referring to, and, just
as in Las Vegas, it is not rewarded at all.
4