American Finance Association

The Capital Asset Pricing Model (CAPM), Short-Sale Restrictions and Related Issues
Author(s): Stephen A. Ross
Source: The Journal of Finance, Vol. 32, No. 1 (Mar., 1977), pp. 177-183
Published by: Blackwell Publishing for the American Finance Association
Stable URL: http://www.jstor.org/stable/2326912
Accessed: 27/04/2010 08:51

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=black.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Blackwell Publishing and American Finance Association are collaborating with JSTOR to digitize, preserve
and extend access to The Journal of Finance.

http://www.jstor.org
THE JOURNAL OF FINANCE * VOL. XXXII, NO. 1 - MARCH 1977

THE CAPITAL ASSET PRICING MODEL (CAPM), SHORT-SALE

RESTRICTIONS AND RELATED ISSUES

STEPHEN A. Ross*

I. INTRODUCTION

THE MEAN VARIANCE CAPITAL asset pricing model (CAPM) developed by Sharpe [5]
and Lintner [3] has become a focal point for finance. Under conditions of
perfection in competitive markets and assumptions that permit us to consider only
the means and variances of returns, the CAPM provides an intuitively appealing
and empirically testable hypothesis on asset returns. In deriving the CAPM Sharpe
[5] and Lintner [3] assumed that there was a riskless asset in the investment
opportunity set, and the first significant extension of their work was by Black [1]
who showed that the assumption of a riskless asset could be dispensed with. Black's
result naturally raised a number of conjectures concerning what occurs with
alternative realistic weakenings of the underlying assumptions. For example, do the
conclusions of the CAPM still hold if short sales are restricted or if borrowing is
penalized on some assets but not on others?
This paper has two objectives. First, a very simple and straightforward approach
to the CAPM will be taken. Essentially we will show that all of the familiar results
follow directly from the observation that the market portfolio, a,m is mean-
variance efficient. Second, we will apply this viewpoint to gather together and
hopefully resolve a number of desultory questions and problems that have arisen
from the above considerations of the original CAPM. This will enable us to better
understand the strength or robustness of the CAPM results with respect to changes
in the assumptions on which it is based.
(1) It should be clear that, irrespective of the form of market institutions, if the
market portfolio is mean-variance efficient then the CAPM results hold. (In what
follows, V will denote the covariance matrix of risky assets and E the vector of
expected returns.) Since all assets are held in positive amounts in the market
portfolio, efficiency implies satisfaction of the first order conditions for maximizing
return at a given variance level,
Ei= 9+Xcov{xixm}-9+a(, (1)
where 0 and X are Lagrange multipliers, xi is the random return on the ith asset and
xm is the random return on the market portfolio. The first order conditions,
however, merely restate the CAPM. Rearranging, yields the familiar form

Ei- 9=XAJ, (2)

* Professor of Economics and Finance, University of Pennsylvania. The author is grateful to the
National Science Foundation and the Rodney L. White Center for Financial Research, University of
Pennsylvania, which supported this research. Also to Richard Roll for valuable comments.

177
178 The Journal of Finance

or since for the market portfolio,

Em-0 = XGM'

we have
2

Ej-0= lim(Em- 0).

CT2
iJm

or

Ei - 0= fi(Em -0), 3

where 8i - /i2mm2, is the "beta" for security i. Thus, efficiency implies that the
security line equation of the CAPM (3) holds. Conversely, if the market portfolio is
inefficient, the CAPM will not hold. In other words, the efficiency of the market
portfolio and the CAPM are equivalent.
(2) It follows that the route to establishing the CAPM lies in simply proving the
efficiency of the market portfolio. With this guide we can considerably weaken the
usual assumptions and simplify many proofs. The following sections will illustrate
this approach.
(3) Consider the CAPM without a riskless asset. Now, each individual portfolio,
a V, is efficient and indexing agents by v = 1,.-. , n we have
Ei -0=XvVaVP. (4)

Letting c be the proportion of wealth held by the vth agent, the market portfolio
P

is simply a convex combination of the individual portfolios, i.e.,

am= E wpapV (5)

v

Hence, aggregating (4) yields

Ej-0=XValm _Aol2M, (6)

where

)- Ea' A] (7)

and
- P1 P0 (8)
= 1wp

Of course am is indeed a portfolio and (6) simply restates the CAPM. Notice, too
that if a is any portfolio uncorrelated with the market portfolio,

aiEi= 0,
E (9)
 The Capital Asset Pricing Model 179

Eo

FIGURE 1

then 0 can be interpreted as the return on a zero-beta portfolio as in Black [1] or

Ross [4]. With a riskless asset, it is easy to see that 0 is its rate of return.
(4) Now suppose that we do not permit short sales except on a single riskless
asset. A geometric approach is best here (see Figure 1). Since all feasibly risky
portfolios are convex combination of assets, the feasible mean-variance set attain-
able solely by investments in risky assets must still be convex. It follows that
separation obtains and all agents can achieve the efficient frontier by borrowing
and lending against (in general) a single risky portfolio. In the familiar fashion,
then, this single portfolio must be the market portfolio, and since it is efficient the
CAPM holds. Notice, of course, that while all assets are not necessarily represented
in efficient portfolios of risky assets, with short sale restrictions they appear in
positive amounts in the market portfolio and, therefore, the interior first-order
conditions (1) must be satisfied. This will further imply (barring singularities in the
covariance matrix) that the efficient frontier for risky assets is strictly concave in a
neighborhood of the market portfolio.
(5) In general, with short sales constraints the first order (Kuhn-Tucker) condi-
tions take the form
Via =+ Ei+ Si, (10)

where Vi is the ith row of V, t and D are Lagrange multipliers and

Si> 0 implies that ai = 0,

180 The Journal of Finance

and

ai > O implies that i =O.

If for asset i, ai = 0 and Si> 0, then the linearity relation does not hold. In fact, it
is precisely because it fails to hold that asset i is not included in the portfolio.
(Equation 10 also disproves a common fallacy in drawing the efficient set for risky
assets with no short sales. At a switching point where the portfolio composition
changes and a new asset, say, i enters while j leaves, Si= 0 and Si= 0; since D is the
slope on the frontier, there is no "kink".)
Section 4 naturally suggests that we might try verifying the CAPM if there exists
one risky asset which can be shorted. The following counterexample uses the
(Kuhn-Tucker) conditions (10) to show that this is not the case. Assume that the
first asset may be freely shorted (observe, though, that no one in this example takes
advantage of this possibility).
Let

0
1
01
11
0=1234
1
V= ? ? 22 ; E= (1,2,3,4)

2 2 2

(It is easy to verify that V is positive definite and, therefore, an acceptable

covariance matrix.) The first order conditions for optimality for an efficient
portfolio, y, with mean return 3 are given by

Yl+21Y4 =g+ g

72+ 2Y4=g+2g

Z3+ 2I4=g+3~g

12(71 + Y2 + y3) + Y4 =g +4g

which imply that

I1 - 5
3.7 5.7
'YI 14; /2= 14;Z3 14; Z/4=

and

g=2; 28 = 7D

Now let the mean return be instead 2 ? . The first order conditions are satisfied by
a portfolio, a, with

a1-24; 2=24; 3=2; 24 4=0,

and

(a ;4' Da24
 The Capital Asset Pricing Model 181

Since the minimum variance point with short sales restrictions occurs at a I= a2
= a3= 3, with a return of 2, the first order conditions guarantee that expected
return is maximized at the respective variance levels. From (10) we also require

- (aa1+ ?2+ 3) + 0 > (a+ 4;a,

or

2 >12'

to insure that a4= 0 is optimal.

To construct an example violating the CAPM we only have to verify that the
efficiency conditions, (4), are not satisfied by both portfolios. The Y portfolio
contains all four assets and the a portfolio only the first three. Recalling that Vi
denotes the ith row of V, we have for y

V17-V27 1 E1-E2
V1-V37 2 E1-E3

and

V17-V47 3 El-E4
V1-V37 2 E1-E3

as would be implied by the CAPM, i.e., the y portfolio will satisfy (4) with the
multipliers 0 and A eliminated. On the other hand, for the a portfolio, while

Vl1a-V2a 1
Vla- V3a= 2

we also have

Vla-V4a 5 3 El-E4
Vla-V3a 2 2 E1-E3

so that the a portfolio does not satisfy the efficiency conditions (4).
Furthermore, there is nothing very special about the example chosen. If we
change the covariance matrix to the identity matrix and let all assets be shorted
except the second, then at a return of 3 all assets will be held positively, but at 6 the
first will be shorted and the second will not appear. A market portfolio can now be
constructed such that the CAPM fails on the second asset. This example should
remove any lingering conjectures that independence of shortable assets alone will
restore the CAPM.
(6) What happens if there is a riskless asset, but like other assets it cannot be
shorted? Now the relevant geometry is shown in Figure 2. Obviously, if all
investors are long in the riskless asset the situation is as before and the CAPM
holds. If not, then the range where the efficient set is not necessarily linear is
relevant. We could now construct a counterexample to the CAPM just by adding
to the example given above a riskless asset with a sufficiently low rate of return. It
 E

(lending
rote)

FIGURE 2

El (borrowing rate)

(tending
rate)

1 3~~~~~~~~~~~
FIGURE 3

182
 The Capital Asset Pricing Model 183

should also be noted, though, that since Si= 0 in (10) for all investors for an asset i
that can be shorted, the CAPM will hold for that subset of assets than can be sold
short. In the example of section 5, the CAPM is valid for the first three assets (as
though they could be shorted); however, in general, the intercept need not be at the
riskless rate.
(7) Suppose now that short sales are not absolutely excluded, but rather are
penalized as in the "real world". We fortunately don't have to be too specific about
the exact nature of the penalty; once again it may induce agents to hold portfolios
that do not contain all assets. However, the analyses above remain valid and if, for
example, there is a riskless asset that can be freely shorted, then separation obtains
and as in (4) the CAPM holds. If the riskless asset can be shorted but only at a
higher "penalty" rate and if risky assets can also only be shorted with penalty, then
the efficient frontier has two linear segments and the CAPM will, in general, fail.
This case is treated in Blume and Friend [2] and illustrated in Figure 3. Of course,
if risky assets can be freely shorted while the riskless asset is penalized the CAPM
holds by the argument of section 6 for the risky assets, but the intercept need not
be at the riskless rate.
(8) One final case is interesting and worth treating. Suppose that returns are
generated by a single factor market model with many assets, as in Ross [4]. In this
case nearly all well diversified portfolios will approximately be efficient and the
CAPM may be expected to hold in an approximate sense. The area of approxima-
tion contains an important class of problems with genuine empirical significance,
but they are beyond the scope of this paper.

II. CONCLUSION

These notes have developed a simple efficiency derivation of the CAPM and used
this approach to examine the impact of some short sale restrictions on the security
line equation. The plethora of alternative institutional restrictions that can be
envisioned makes an exhaustive treatment impossible, and while many can be
treated in a fairly straightforward fashion, some are formidable. To mention one
difficult case of particular importance, if there are collateral restrictions on short
sales and, perhaps, associated bankruptcy provisions, the CAPM will have to be
analyzed anew. The meaningful presence of financial intermediaries in such a
world will also cause problems for the simple CAPM. The viewpoint taken in these
notes, however, should facilitate the further analysis of such issues.

REFERENCES
1. FischerBlack."CapitalMarketEquilibriumwith RestrictedBorrowing."
Journalof Business,March
1973.
2. MarshallBlumeand IrwinFriend."A New Look at the CapitalAsset PricingModel."Journalof
Finance, March 1973.
3. John Lintner."The Valuationof Risk Assets and the Selection of Risky Investmentsin Stock
Portfoliosand CapitalBudgets."Reviewof Economicsand Statistics,February1965,pp. 13-37.
4. Stephen A. Ross. "The ArbitrageTheory of Capital Asset Pricing,"forthcoming,Journal of
Economic Theory, 1976.
5. WilliamSharpe."CapitalAsset Prices:A Theoryof MarketEquilibriumunderConditionsof Risk."
Journal of Finance, September1964,pp. 425-442.