Professional Documents
Culture Documents
Samizdat: In Defense of the CAPM Author(s): Jack L. Treynor Reviewed work(s): Source: Financial Analysts Journal, Vol. 49, No. 3 (May - Jun., 1993), pp. 11-13 Published by: CFA Institute Stable URL: http://www.jstor.org/stable/4479643 . Accessed: 08/02/2013 07:22
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
.
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.
CFA Institute is collaborating with JSTOR to digitize, preserve and extend access to Financial Analysts Journal.
http://www.jstor.org
This content downloaded on Fri, 8 Feb 2013 07:22:28 AM All use subject to JSTOR Terms and Conditions
Samizdat
In Defense of the CAPM
Thus he won't bear systematic risk unless it competes successfully with residual risk for inclusion in his portfolio. But if he includes enough positions (long or short) in his portfolio, he can drive the portfolio's residual risk to zero. Now, if residual risk is priced, this offers him an infinite reward-to-riskratio. In order for market risk to compete successfully, it must offer an infinite expected return. Otherwise the investor will choose not to hold any systematic risk, and the market won't clear. The market model takes the assumptions of the Diagonal Model and the result of the Vasicek and McQuown model and concludes that the actual, ex post systematic return-surprise plus risk premium-of every asset will be proportional to the asset's market sensitivity. Both Vasicek and McQuown and the market model thus assume a single systematic in its various factor;but the CAPM forms (Sharpe, Treynor, Lintner, Mossin, Black) makes no assumption about factor structure. In particular, it does not make the one-factor, Diagonal Model assumption that Vasicek and McQuown and the market model make.
which they are all averse. Shortselling is permitted. There is thus nothing about factor structure in the CAPM's assumptions. But there is also nothing about factor structure in the CAPM's conclusions. Factor structure is about surprise-in particular, the nature of correlations between surprises in different assets. In the Diagonal Model, a single factor accounts for all correlation. In richer, more complex factor structures, correlations in asset surprises can be due to a variety of pervasive, market-wide influences. The conclusion of the CAPM-that an asset's expected return is proportional to its covariance with the market portfolio-makes no assertions about the surprise in that asset's return. Because expected return and surprise are mutually exclusive elements in ex post, actual return, there can be no conflict between the CAPM and factor structure.
This point is the key to the second complaint. In order to demonstrate that APT'sprincipal conclusion-that an asset's risk premium depends only on its systematic factor loadings-also holds for the CAPM,we can assume a perfectly general factor structure without fear of tripping Actually, a CAPMthat abandons on either the assumptions of the the assumptions of homogeneous CAPM or its conclusions. expectations (as Lintnerdid) and riskless borrowing and lending Consider a market in which abso(as Black did) doesn't make many lute surprise for the ith asset, xi, other assumptions. All it does as- obeys: sume is that asset risk can be adequately expressed in terms of Eq. 1 variances and covariances (with other assets); that these measures exist; that investors agree on Xi= 2,8ijuj + ei, these risk measures; and that investors can trade costlessly to in- where u1 are the systematic faccrease expected portfolio return tors in the market and ei is a or reduce portfolio variance, to residual unique to the ith asset.
z
D
-J
z
-J
z zL
11
This content downloaded on Fri, 8 Feb 2013 07:22:28 AM All use subject to JSTOR Terms and Conditions
The absolute surprise for the market as a whole,x, is simplythe sum of absolutesurprisesfor the individualassets:
x = Exi = 28ijuj + Eei.
Figure A
The CAPM asserts that any risk premiumof the ith assetdepends only on its covariancewith the market-i.e., on the expectation of the followingproduct:
Eq. 2
( f ijuj + ei)(E,BhkUk
+ Ieh).
Bearingin mind thatthe u's and e's are surprise, and assuming that all covariancesbetween e's and between u's and e's are zero, Amongthe manypossiblechoices of orthogonal factors, we can we have for this expectation: choose one in which only one factorcorrelateswith the market Eq. 3 portfolio.Assignthe factorindex to the factorsso thatthe firstterm in the bracketed expressionis the + ei2], E[EZp ijphkUjUk termfor thatfactor. Then variance moment we can for the ignore or the other systematicterms and focus on the firstandlasttermsin Eq. 4 the covarianceexpression,Equathe respective tion (7). Factoring terms into standard deviations 2] + Si2, 2 3ij[2!3hkI'jk
gives us: Eq. 8
(8i11)(2I3h1lr1) O(jk = E[UjUk] + + (si)(s)
To keep our exposition simple, we chose a special set of systemSame Order of Magnitude atic factors. Had we chosen differr1; ent systematic factors, the market + ... + (si)(si) WHOM 1)(13jlfl) portfolio could have been correlated with most or all of them, t ~~~~~~t Orders of Magnitude Bigger rather than merely one. Then the systematic portion of an asset's covariance with the marketwould aticpartzero by choosinga set of have non-zero terms for more factorsthatare orthogonal.Then than one factor. Spreading it the covariance expression be- across several factors, however, won't diminish the magnitude of comes: the aggregate systematic portion. Whether one systematic factor or Eq. 7 several are correlated with the market portfolio, then, the systematic portion of an asset's cova+ A223n20f222 Pill&10hcr11 riance with the market will be 2 + ...+ Si. orders of magnitude bigger than the unique portion. If we drop the unique term and relax the assumption that only one systematic factor correlates with the market, then the expression within brackets in Equation (4) will in general have a different value for each of the systematic factors. But the value doesn't depend on i-the index that distinguishes between individual assets. What is outside the brackets is the factor loadings 3ij for the ith individual asset on the factors uj in the factor structure. But this is the principal conclusion of APTthat an asset's risk premium should depend only on its factor loadings. Any corollaries that flow from this APT result also flow from the CAPM. One of the differences between APTand the CAPM is the specifiof how systematic cation by CAPM factors should be priced in relation to each other. The expression in brackets in the first term of Equation (4) is, to some constant of proportionality, the price of risk. The value of the expression, hence that price, is generally going to be different for different factors.APTcan't specify the value of this expression; therefore it can't specify how factors will be priced in relation to each other. What the APTdoes do is weaken the assumptions necessary to
where
Eq. 5
and
z
Eq. 6
Si2=
E[ei2].
z
D
with the marketportfolio (absolute surprise with absolute surprise) has two terms-one for
systematic factors and one for the asset's unique surprise. A key issue is the relative size of these two terms.
z
-J
Considerthe first term. One factor reflects the absolute factor weight of the asset,the other the absolutefactorweightof the market portfolio.The formeris of the same order of magnitudeas the standarddeviationof the asset's uniquerisk;the latteris ordersof magnitudelarger (see FigureA). This is because all the systematic risk in the marketportfolio'sabsolute gain or loss is reflectedin the market's factor weight, whereas the unique risk in the second term is that for a single asset.But this meansthatthe systematicproductis ordersof magnitude larger than the unique product.
This content downloaded on Fri, 8 Feb 2013 07:22:28 AM All use subject to JSTOR Terms and Conditions
reach this conclusion. APT appeals to the older, better established tradition of arbitrage arguments exemplified by Modigliani and Miller's 1958 paper. APTjoins the idea of specifying security risk in terms of a factor structure (Farrar's 1960 doctoral thesis at Harvard; Hester and Feeney's con-
temporaneous work at Yale) with Modigliani and Miller's arbitrage argument. Some may argue that, in effect, APT substitutes systematic risk factors for Modigliani and Miller's risk classes. Others may argue that the real provenance of APT is not Modigliani and Miller,but ratherVasicek and
McQuown, with multiple risk factors substituted for their single risk factor. This writer has no desire to take a position in such controversies. His only purpose is to clear up certain misunderstandings regarding the CAPM.
Post-Modern
Portfolio
Theoy
e
Comes Of Age
Protective Put
PMPT
2 M MPT
Retum
For a demonstration program and additional information contact: SPONSOR-SOFTWARE SYSTEMS, INC.
* 212/580-9703
(Fax)
This content downloaded on Fri, 8 Feb 2013 07:22:28 AM All use subject to JSTOR Terms and Conditions