CHAPTER 10

ARBITRAGE PRICING THEORY AND MULTIFACTOR

MODELS OF RISK AND RETURN
Arbitrage pricing theory (APT) is a well-known method of estimating the price of an asset.
The theory assumes an asset's return is dependent on various macroeconomic, market and
security-specific factors.

How it works/Example:
APT is an alternative to the capital asset pricing model (CAPM). Stephen Ross developed the
theory in 1976.
The APT formula is:
E(rj) = rf + bj1RP1 + bj2RP2 + bj3RP3 + bj4RP4 + ... + bjnRPn
where:
E(rj) = the asset's expected rate of return
rf = the risk-free rate
bj = the sensitivity of the asset's return to the particular factor
RP = the risk premium associated with the particular factor
The general idea behind APT is that two things can explain the expected return on a financial
asset: 1) macroeconomic/security-specific influences and 2) the asset's sensitivity to those
influences. This relationship takes the form of the linear regression formula above.
There are an infinite number of security-specific influences for any given security including
inflation, production measures, investor confidence, exchange rates, market indices or
changes in interest rates. It is up to the analyst to decide which influences are relevant to the
asset being analyzed.
Once the analyst derives the asset's expected rate of return from the APT model, he or she can
determine what the "correct" price of the asset should be by plugging the rate into a
discounted cash flow model.
Note that APT can be applied to portfolios as well as individual securities. After all, a
portfolio can have exposures and sensitivities to certain kinds of risk factors as well.

Why it Matters:
The APT was a revolutionary model because it allows the user to adapt the model to the
security being analyzed. And as with other pricing models, it helps the user decide whether a
security is undervalued or overvalued and so he or she can profit from this information. APT
is also very useful for building portfolios because it allows managers to test whether their
portfolios are exposed to certain factors.

APT may be more customizable than CAPM, but it is also more difficult to apply because
determining which factors influence a stock or portfolio takes a considerable amount of
research. It can be virtually impossible to detect every influential factor much less determine
how sensitive the security is to a particular factor. But getting "close enough" is often good
enough; in fact studies find that four or five factors will usually explain most of a security's
return: surprises in inflation, GNP, investor confidence and shifts in the yield curve.

Arbitrage Pricing Theory (APT)

As its name implies, the Arbitrage Pricing Theory, or APT, describes a mechanism used by
investors to identify an asset, such as a share of common stock, which is incorrectly priced.
Investors can subsequently bring the price of the security back into alignment with its actual
value.

The Arbitrage Pricing Theory Model

The APT model was first described by Steven Ross in an article entitled The Arbitrage Theory
of Capital Asset Pricing, which appeared in the Journal of Economic Theory in December
1976. The Arbitrage Pricing Theory assumes that each stock's (or asset's) return to the
investor is influenced by several independent factors.

The APT Formula

Furthermore, Ross stated the return on a stock must follow a very simple relationship that is
described by the following formula:
Expected Return = rf + b1 x (factor 1) + b2 x (factor 2)... + bn x (factor n)
Where:

rf = the risk free interest rate, which is the interest rate the investor would expect to
receive from a risk-free investment. Typically, U.S. Treasury Bills are used for U.S.
dollar calculations, while German Government bills are used for the Euro
b = the sensitivity of the stock or security to each factor
factor = the risk premium associated with each entity

The APT model also states the risk premium of a stock depends on two factors:

The risk premiums associated with each of the factors described above
The stock's own sensitivity to each of the factors; similar to the beta concept
Risk Premium = r - rf = b(1) x (r factor(1) - rf) + b(2) x (r factor(2) - rf)... + b(n) x (r
factor(n) - rf)

If the expected risk premium on a stock were lower than the calculated risk premium using
the formula above, then investors would sell the stock. If the risk premium were higher than
the calculated value, then investors would buy the stock until both sides of the equation were

in balance. Arbitrage is the term used to describe how investors could go about getting this
formula, or equation, back into balance.

Factors Used in the Arbitrage Pricing Theory

It's one thing to describe the APT theory in terms of simple formulas, but it's another matter
entirely to identify the factors used in this theory. That's because the theory itself does not tell
the investor what those factors are for a particular stock or asset, and for a very good reason.
In practice, and in theory, one stock might be more sensitive to one factor than another. For
example, the price of a share of ExxonMobil might be very sensitive to the price of crude oil,
while a share of Colgate Palmolive might be relatively insensitive to the price of oil.
In fact, the Arbitrage Pricing Theory leaves it up to the investor, or analyst, to identify each of
the factors for a particular stock. Therefore, the real challenge for the investor is to identify
three items:

Each of the factors affecting a particular stock

The expected returns for each of these factors
The sensitivity of the stock to each of these factors

Identifying and quantifying each of these factors is no trivial matter, and is one of the reasons
the Capital Asset Pricing Model remains the dominant theory to describe the relationship
between a stock's risk and return.
Keeping in mind the number and sensitivities of a stock to each of these factors is likely to
change over time, Ross and others identified the following macro-economic factors they felt
played a significant role in explaining the return on a stock:

Inflation
GNP or Gross National Product
Investor Confidence
Shifts in the Yield Curve

With that as guidance, the rest of the work is left to the stock analyst.

APT versus the Capital Asset Pricing Model

As mentioned, the Arbitrage Pricing Theory and the Capital Asset Pricing Model (CAPM) are
the two most influential theories on stock and asset pricing today. The APT model is different
from the CAPM in that it is far less restrictive in its assumptions. APT allows the individual
investor more freedom to develop a model that explains the expected return for a particular
asset.
Intuitively, the APT makes a lot of sense because it removes the CAPM restrictions, and
basically states "The expected return on an asset is a function of many factors as well as the
sensitivity of the stock to these factors." As these factors move, so does the expected return
on the stock, and therefore its value to the investor.
In the CAPM theory, the expected return on a stock can be described by the movement of that
stock relative to the rest of the market. The CAPM is really just a simplified version of the
3

APT, whereby the only factor considered is the risk of a particular stock relative to the rest of
the market, as described by the stock's beta.
From a practical standpoint, CAPM remains the dominant pricing model used today. When
compared to the Arbitrage Pricing Theory, the Capital Asset Pricing Model is both elegant and
relatively simple to calculate.

10.1 MULTIFACTOR MODELS: AN OVERVIEW

1. Factor Models of Security Returns
The index model introduced earlier in Chapter 8 gives us a way of decomposing
stock variability into market or systematic risk, due largely to macroeconomic
events, versus firm-specific or idiosyncratic effects that can be diversified in large
portfolios. In the index model, the return on the market portfolio summarizes the
broad impact of macro factors. Sometimes, however, rather than using a market
proxy, it is more useful to focus directly on ultimate sources of risk. This can be
useful in risk assessment when measuring ones exposure to particular sources of
uncertainty. Factors models are tools that allow us to describe and quantify the
different factors that affect the rate of return on a security during any time period.
1. Factor Model of Security Returns
To illustrate we will start by examining a single-factor model like the one
introduced in Chapter 8. As noted there, uncertainty in asset returns has two
sources: a common or macroeconomic factor, and firm-specific events. The
common factor is constructed to have zero expected value, because we use it to
measure new information concerning the macroeconomy, which by definition,
has zero expected value.
If we call F the deviation of the common factor from its expected value, i the
sensitivity of firm i to that factor, and ei the firm-specific disturbance, the factor
model states that the actual return on firm i will equal its initially expected return
plus a (zero expected value) random amount attributable to unanticipated
economy-wide events, plus another (zero expected value) amount attributable to
firm-specific events. Formally, the single-factor model is described by Equation
10.l:
ri E(ri ) i F ei ,

10.1

where, E(ri) is the expected return on stock i. Notice that if the macro factor has a
value of 0 in any particular period (i.e., no macro surprises), the return on security
will equal its previously expected value, E(ri), plus the effect of firm-specific
events only. The non-systematic components of the return, the eis, are assumed to
be uncorrelated among themselves and uncorrelated with the factor F.
The factor model decomposition of returns into systematic and firm-specific
components is compelling, but confining systematic risk into a single factor is
5

not. Indeed, when we motivated the index model in Chapter 8, we noted that the
systematic or macro risk factor summarized by the market return arises from a
number of sources, for example, uncertainty about the business cycle, interest
rates, inflation, etc. The market return reflects macro factors as well as the
average sensitivity of firms to those factors. (When we estimate a single-index
regression, therefore, we explicitly impose an (incorrect) assumption that each stock has the
same relative sensitivity to each risk factor. If stocks actually differ in their betas relative to the
macroeconomic factors, then lumping all systematic sources of risk into one variable such as
return on the market index will ignore the nuances that better explain individual stock returns.)

How can we improve on the single-index model but still maintain the useful
dichotomy between systematic and diversifiable risk? It is easy to see that models
that allow for several systematic factors multifactor models can provide
better description of security returns. (Apart from their use in building models of equilibrium
security pricing, multifactor models are useful in risk management applications. These models give us a
simple way to measure our exposure to various macroeconomic risks, and construct portfolios to hedge
those risks.)

Lets start with a two-factor model. Suppose the two most important
macroeconomic sources of risk are uncertainties surrounding the state of the
business cycle, news of which we will again measure by unanticipated growth in
GDP, and changes in interest rates, which may be captured by the return on a Tbond portfolio. The return on any stock will respond both to sources of macro risk
and to its own firm-specific influences. Therefore, we can write a two-factor
model describing the excess rate of return on a stock i in some time period as
follows:
ri E ( ri ) iGDP GDP iIR IR ei ,

10.2

where IR denotes the unexpected changes in interest rates. The two macro factors
on the right-hand side of the equation comprise the systematic factors in the
economy. As in the single-factor model, both of these macro factors have zero
expectations: they represent changes in these variables that have not already been
anticipated. (The coefficient of each factor in Equation 10.2 measures the sensitivity of stock
returns to that factor; for this reason the coefficients are sometimes called factor sensitivities,
factor loading, or equivalently, factor betas.) An increase in interest rates is bad news

for most firms, so we would expect interest rate betas generally to be negative.
As before, ei reflects firm-specific influences.
Factor betas can provide a framework for a hedging strategy. The ides foe an
investor who wishes to hedge a source of risk is to establish an opposite factor
exposure to offset that particular source of risk. Often, futures contracts can be
used to hedge particular factor exposure.

2. A Multifactor Security Market Line

The multifactor model is no more than a description of the factors that affect
security returns. There is no theory in the equation. The obvious question left
unanswered by a factor model like Equation 10.2 is where E(r) comes from, in
other words, what determines a securitys expected rate of return. This is where
we need a theoretical model of equilibrium security returns.
In previous chapters we developed one example of such a model: the security
market line (SML) of the CAPM. The CAPM asserts that securities will be
priced to give investors an expected return comprised of two components: the
risk-free rate, which is compensation for the time value of money, and a risk
premium, determined by multiplying a benchmarks risk premium (i.e. the risk
premium offered by the market portfolio) times the relative measure of risk,
beta:
E(r) r f [E(rM ) rf ]

10.3

If we denote the risk premium of the market portfolio by RPM, then, a useful way
to rewrite equation 10.3 is:
E(r) r f RPM

10.4

We pointed out in Chapter 8 that you can think of beta as measuring the
exposure of a stock or portfolio to market-wide or macroeconomic risk factors.
Therefore, one interpretation of the SML is that investors are rewarded with a
higher expected return for their exposure to macro risk, based on the sensitivity
to this risk (beta), as well as the compensation for bearing each unit of that
source of risk (i.e., the risk premium), but are not rewarded for exposure to firmspecific uncertainty (the residual term, ei in Equation 10.1).
How might this single-factor view of the world generalize once we recognize the
presence of multiple sources of systematic risk? Perhaps, not surprisingly, a
multifactor index model gives rise to a multifactor security market line (SML)
in which the risk premium is determined by the exposure to each systematic risk
factor, and by a risk premium associated with each of those factors. Such a
multifactor CAPM was first presented by Merton (1973). For example, in a twofactor economy in which the risk exposure can be measured by Equation 10.2,
we would consider that the expected rate of return on a security would be the
sum of: 1) The risk-free rate of returns, 2) The sensitivity to GDP (i.e., the GDP
beta) times the risk premium for bearing GDP risk, and 3) The sensitivity to
interest rate risk (i.e., the IR beta) times the risk premium for bearing interest
rate risk. This assertion is expressed as follows in Equation 10.5:
7

E(r) r f GDP [E(rGDP ) r f ] IR [E(rIR ) r f ]

10.5

In that equation, GDP denotes the sensitivity of the security return to unexpected
changes in GDP growth, and RPGDP = [E(rGDP) - rf] is the risk premium
associated with one unit of GDP exposure, that is, the exposure corresponding
to a GDP beta of 1.0.
It is clear that Equation 10.5 is a generalization of the simple security market
line. In the single-factor SML, the benchmark risk premium is given by the risk
premium of the market portfolio, but once we generalize to multiple risk
sources, each with its own risk premium, we see that the insights are very
similar. However, one difference between a single- and multi-factor economy is
that a factor risk premium can be negative. For example, a security with a
positive interest rate beta performs better when rates increase, and thus would
hedge the value of a portfolio against interest rate risk. Investors might well
accept a lower rate of return, that is, a negative risk premium, as the cost of this
hedging attribute. In contrast, a more typical security that does worst when rates
increase (a negative-IR beta) adds to interest rate exposure, and therefore has a
higher required rate of return. Equation 10.5 shows that the contribution of
interest rate risk to required return for such a security would than be positive, the
product of a negative-factor beta times a negative-factor risk premium.
Another reason that multifactor models can improve on the descriptive power of the index
model is that betas seem to vary over the business cycle. In fact, some of the variables that are
used to predict beta are related to the business cycle (e.g., earnings growth). Therefore, it makes
sense that we can improve the single-index model by including variables that are related to the
business cycle.

We still need to specify how to estimate the risk premium for each factor.
Analogously to the simple CAPM, the risk premium associated with each factor
can be thought of as the risk premium of a portfolio that has a beta of 1.0 on that
particular factor, and a beta of zero on all other factors. In other words, it is the
risk premium one might expect to earn by taking a pure play on that factor. For
now lets just take the factor risk premiums as given and see how a multifactor
SML might be used.
10.2 ARBITRAGE PRICING THEORY
Stephen Ross developed the Arbitrage Pricing Theory (APT) in 1976. Like
CAPM, the APT predicts a security market line linking the expected returns to
risk, but the path it takes to the SML is quite different. Rosss APT relies on
three key propositions: 1) security returns can be described by a factor model, 2)
8

there are sufficient securities to diversify away the idiosyncratic rick, and 3)
well-functioning security markets do not allow for the persistence of arbitrage
opportunities. We begin with a simple version of Rosss model, which assumes
that only one systematic factor affects security returns.
1. Arbitrage, Risk Arbitrage, and Equilibrium
To understand the APT we begin with the concept of arbitrage, which is the act
of exploiting the mispricing of two or more securities to achieve risk-free
economic profits. An arbitrage opportunity arises when an investor can earn
riskless profit without making a net investment; in other words when the
investor can construct a zero investment portfolio that will yield a sure profit. To
construct a zero investment portfolio one has to be able to sell short at least one
asset and use the proceeds to purchase (go long on) one or more assets.
Borrowing may be viewed as a short position in the risk-free asset. Clearly, any
investor would like to take as large a position as possible in an arbitrage
portfolio.
An obvious case of an arbitrage opportunity arises when the Law of One Price is
violated. For example, consider a security that is priced differently in two markets
(and the price differential exceeds transaction costs). A long position in the
cheaper market financed by a short position in the more expensive one will lead
to a sure profit. The net proceeds are positive, and there is no risk because the
long and short positions offset each other. The Law of One Price is enforced by
arbitrageurs; if they observe a violation of the law, they will engage in arbitrage
activity simultaneously buying the asset where it is cheap and selling where it
is expensive. In the process, they will bid up the price where it is low and force
it down where it is high until the arbitrage opportunity is eliminated. (The idea that
market prices will move to rule out arbitrage opportunities is perhaps the most fundamental
concept in capital market theory. Violation of this restriction would indicate the grossest form
of market irrationality.)

The critical property of a risk-free arbitrage portfolio is that any investor,

regardless of risk aversion or wealth, will want to take an infinite position in it.
Because those large positions will force prices up or down until the opportunity
vanishes, we can derive restrictions on security prices that satisfy a "noarbitrage" condition, that is, a condition that rules out the existence of arbitrage
opportunities.
The CAPM is an example of a dominance argument, implying that all investors
hold mean-variance efficient portfolio. If a security is mispriced, then investors
will tilt their portfolios toward the underpiced and away from the overpriced
securities. Pressure on equilibrium prices results from many investors shifting

their portfolios, each by a relatively small dollar amount. The assumption that a
large number of investors are mean-variance sensitive is critical. In contrast, the
implication of a no-arbitrage condition is that a few investors who identify an
arbitrage opportunity will mobilize large dollar amounts and quickly restore the
equilibrium.
Practitioners often use the terms "arbitrage" and "arbitrageurs" more loosely
than our strict definition. "Arbitrageur" often refers to a professional searching
for mispriced securities in specific areas such as merger-target stocks, rather than
to one who seeks strict (risk-free) arbitrage opportunities. Such activity is
sometimes called risk arbitrage to distinguish it from pure arbitrage.
2. Well-Diversified Portfolios
The first to apply this concept to equilibrium security returns was Ross (1976)
who developed the Arbitrage pricing theory (APT). The APT depends on the
assumption that well-functioning capital markets preclude arbitrage
opportunities. A violation of the APTs pricing relationships will cause
extremely strong pressure to restore them even if only a limited number of
investors become aware of the disequilibrium. Rosss accomplishment is to
derive the equilibrium rates of return and risk premiums that would prevail in a
market where prices are in alignment to the extent that arbitrage opportunities
have been eliminated.
Suppose now that we construct a highly diversified portfolio with a given beta.
If we use enough securities to form the portfolio, its firm-specific or non-factor
risk can be diversified away. Only the factor (or systematic) risk remains. If we
construct an n-stock portfolio with weights wi, (wi = 1), then the rate of return
on this portfolio is as follows:
rP E(RP ) P F e P ,

10.6

where p wi i is the weighted average of the i and E(rp ) wi E(ri ) is the

expected return of the n securities,. The portfolio non-systematic component
(which is uncorrelated with risk factor, F) is eP wi ei , which similarly is a
weighted average of the ei of the n securities.
We can divide the variance of the portfolio into systematic and non-systematic
sources as wee saw in Chapter 8. The portfolio variance is:
P2 P2 F2 2 (eP ) ,

10

2
where F is the variance of the factor F, and (e ) is the non-systematic risk
of the portfolio, which is given by:
p

2 (e p ) Variance ( wi ei ) wi2 2 (ei )

If the portfolio were equally weighted, that is wi = 1/n, then the non-systematic
variance would be:
2

2 (eP , wi

1
1 2 (ei ) 1 2
1
) 2 (ei )
(ei ) ,
n
n
n
n
n

where the last term is the average value across securities of non-systematic
variance. In words, the non-systematic variance of the portfolio equals the
average non-systematic variance, (e ) , divided by n. Therefore, when the
portfolio gets large in the sense that n is large, its non-systematic variance
approaches zero. This is the effect of diversification.
2

We conclude that for the equality weighted portfolio, the non-systematic variance
approaches zero as n becomes ever larger. This property is true of portfolios
other than the equally weighted ones. Any portfolio for which the weight wi
become consistently smaller as n gets larger will satisfy the condition that the
portfolio non-systematic risk will approach zero. In fact, this property motivates
us to define a well-diversified portfolio as one that is diversified over a large
enough number of securities, with each weight (wi) small enough that for
practical purposes the non-systematic variance, 2(ep), is negligible.
Because the expected (mean) value of ep for any well-diversified portfolio is zero,
and its variance is also effectively zero, we can conclude that any realized value
of ep will be virtually zero. Rewriting Equation 10.1, we conclude that for a welldiversified portfolio, for all practical purposes, the return is:
rP E(RP ) P F ,

P2 P2 F2 p P F

Large (mostly institutional) investors can hold portfolios of hundreds and even
thousands of securities; thus the concept of well-diversified portfolios clearly is
operational in contemporary financial markets. Well-diversified portfolios,
however, are not necessarily equally weighted.
3. Betas and Expected Returns

11

Because nonfactor risk can be diversified away, only factor risk should command
a risk premium in market equilibrium. Non-systematic risk across firms cancels
out in well-diversified portfolios; one would not expect investors to be rewarded
for bearing risk that can be eliminated through diversification. Instead, only
systematic risk of a portfolio of securities should be related to its expected return.
Figure 10.1 in the textbook illustrates the difference between a single security
with a beta of 1.0 and a well-diversified portfolio with the same beta. The
expected return of the portfolio is 10%; this is where the solid line crosses the
vertical axes. At this point the systematic factor is zero, implying no macro
surprises. For the portfolio (Panel A), all the returns plot exactly on SML; there is
no dispersion around the line as in Panel B (individual stock), because the effects
of the firm-specific events are eliminated by diversification. So, the well-diversify
portfolios return is determined completely by the systematic risk factor.
Now consider Figure 10.2, where the dash line plots the return of another welldiversified portfolio, portfolio B, with an expected return of 8% and beta also
equal to 1.0. Could portfolio A and B coexist with the return patterns depicted?
Clearly not: No matter what systematic factor turns out to be, the portfolio with
higher return (A) outperforms the one with lower return (B), leading to an
arbitrage opportunity.
The example in the textbook shows that if you sell short $1 million of B and buy
$1 million of A, you would have a riskless payoff of $20,000. Your profit is riskfree because the factor risk cancels out across the long and shot positions.
Moreover, the strategy requires zero net investment. You should pursue it on an
infinitely large scale until the return discrepancy between the two portfolios
disappears. Well-diversified portfolios with equal betas must have equal
expected returns in market equilibrium, or arbitrage opportunities do exist.
In fact we can go further and show that the alpha of any well-diversified portfolio (see equation
below) must be zero, even if the beta is not zero. The proof is similar to the easy zero-beta case.
If the alphas were not zero, then we could combine two of these portfolios into a zero-beta
riskless portfolio with a rate of return not equal to the risk-free rate. But this would be an
arbitrage opportunity.
R P P P RM

10.7

What about portfolios with different betas? In this case their risk premiums must
be proportional to portfolio beta. For a proof of this property see Figure 10.3 in
the textbook. We find that portfolio D has an equal beta but a greater expected
return than portfolio C. From our previous analysis we know that this constitutes
an arbitrage opportunity.

12

We conclude that, to preclude arbitrage opportunities, the expected return on all

well-diversified portfolios must lie on the straight line from the risk-free asset in
Figure 10.3. Note that risk premiums are indeed proportional to portfolio betas.
The risk premium is depicted by the vertical arrow, which measures the distance
between the risk-free rate and the expected return on the portfolio. The risk
premium is zero for = 0 and rises in direct proportion to beta.
To prove that, suppose that portfolio V has a beta of V and an alpha of V.
Similarly, suppose that portfolio U has a beta of U and an alpha of U. Taking
advantage of any arbitrage opportunity involves buying and selling assets in
proportions that create a risk-free profit on a costless position. To eliminate the
risk we buy portfolio V and sell portfolio U in propositions chosen so that the
combination (portfolio Z) will have a beta of zero. The portfolio weights that
satisfy this condition are:
wV

- U
V
wU
,
and
V U
V U

Note that the two weights add up to 1.0 and that beta of the combination
portfolio is in fact zero:
Z wV V wU U

- U
V
V
U 0
V U
V U

Therefore, portfolio Z is risk-less: it has no exposure to the systematic factor

because its beta is zero. But the excess return of the portfolio is not zero unless
V and U equal zero.
RZ

- U
V
V
U 0
V U
V U

Therefore, unless the two alphas equal zero, the zero-beta portfolio has a certain
rate of return that differs from risk-free rate (its excess return is different from
zero.) To rule out this arbitrage opportunity it must earn only the risk-free rate.
Therefore:
E(rZ ) wV E(rV ) wU E(rU ) r f

Rearranging the above equation, we can conclude that risk premiums are indeed
proportional to portfolio betas:

13

E(rV ) rf
V

E(rU ) rf
U

We conclude that the only value for alpha that rules out arbitrage opportunities is
zero. Therefore, we rewrite Equation 10.7 setting alpha equal to zero:
RP P RM , or rP rf P [rM rf ]

Here we consider the market portfolio as a well-diversified portfolio, and

measure the systematic risk factor as the unexpected return on the market
portfolio. Because the market portfolio must be on the SML and the beta of the
market portfolio is 1.0, we can rearrange the equation above as:
E(rP ) r f P [E(rM ) r f ]

10.8

Hence, this equation is identical to SML relation of the CAPM (see Figure 10.4
in the textbook, where the intercept is rf, and the slope is E(rM) rf). We have
used the no-arbitrage condition to obtain an expected return-beta relationship
identical to that of the CAPM, without the restrictive assumptions of the CAPM.
Our demonstration suggest that despite its restrictive assumptions, the main
conclusion of the CAPM, namely, the (SML) expected return-beta relationship
should be at least approximately valid.
It is worth noting that in contrast to the CAPM, the APT does not require that
the benchmark portfolio in the SML relationship be the true market portfolio.
Any well-diversified portfolio lying on the SML in Figure 10.4 may serve as the
benchmark portfolio. For example, one might define the benchmark portfolio as
the well-diversified portfolio most highly correlated with whatever systematic
factor is thought to affect stock returns. Accordingly, the APT has more
flexibility than does the CAPM because problems associated with an
unobservable market portfolio are not a concern.
In addition, the APT provides further justification for use of the index model in the practical
implementation of the SML relationship. Even if the index portfolio is not a precise proxy for
the true market portfolio, which is a cause of considerable concern in the context of the
CAPM, we now know that if the index portfolio is sufficiently well diversified, the SML
relationship should still hold true according to the APT.

10.3 INDIVIDUAL ASSETS AND THE APT

We have demonstrated that if arbitrage opportunities are to be ruled out, each
well-diversified portfolios expected (excess) return must be proportional to its
beta. The question is whether this relationship tells us anything about the
14

expected returns on the component stock. The answer is that if this relationship
is to be satisfied by all well-diversified portfolios it must be satisfied by almost
all individual assets although the proof of this proposition is somewhat difficult.
We said almost because, according to the APT, there is no guarantee that all
individual assets will lie on SML. If for example, only one security violates the
SML, then the effect of this violation on a well-diversified portfolio will be too
small to be of importance for any particular purpose, and meaningful arbitrage
opportunities will not arise. But if many securities violate the expected returnbeta relationship, the relationship will no longer hold for well-diversified
portfolios comprising these securities, and arbitrage opportunities will be
available. Consequently, we conclude that imposing the no-arbitrage condition
on a single-factor security market implies maintenance of the expected returnbeta relationship for all well-diversified portfolios and for all but possibly a
small number of individual securities.
1. The APT and the CAPM
The APT serves many of the same functions as the CAPM. It gives us a
benchmark for rates of return that can be used in capital budgeting, security
valuation, or investment performance evaluation. Moreover, the APT highlights
the critical distinction between the non-diversifiable risk (factor risk) that
requires a reward in the form of risk premium and diversifiable risk that does
not.
The APT is an extremely appealing model. It depends on the assumption that a
rational equilibrium in capital markets precludes arbitrage opportunities. A
violation of the APT pricing relationships will cause extremely strong pressure
to restore them even if only a limited number of investors become aware of
disequilibrium. Furthermore, the APT yields an expected return-beta relationship
using a well-diversified portfolio that practically can be constructed from a large
number of securities.
In contrast, the CAPM is derived assuming an inherently unobservable "market"
portfolio. The CAPM argument rests on mean-variance efficiency, that is, if any
security violates the expected return-beta relationship, then, many investors
(each relatively small) will tilt their portfolios so that their combined overall
pressure on prices will restore an equilibrium that satisfied the relationship.
In spite of these appealing advantages, the APT does not fully dominate the
CAPM. The bottom line is that neither of these theories dominates the other. The
APT is more general in that it gets us to the expected return-beta relationship
without requiring many of the unrealistic assumptions of the CAPM, particularly
15

the reliance on the market portfolio. The latter improves the prospects for testing
the APT. But the CAPM is more general in that it applies to all assets without
reservation. The good news is that both theories agree on the expected return-beta
relationship.
A more productive comparison is between the APT and the index model. The
implication of the index model is that the market index portfolio is efficient and
that the expected return-beta relationship holds for all assets. The assumption
that the probability distribution of security returns is stationary (so that sample
period returns can provide valid estimates of expected returns and variances) and
the observability of the index make it possible to test the efficiency of the index
portfolio and the expected return-beta relationship. In contrast, the APT uses a
single-factor security market assumption and arbitrage arguments to obtain the
expected return-beta relationship for well-diversified portfolios. Because it
focuses on the no-arbitrage condition, without the further assumptions of the
market or index model, the APT cannot rule out a violation of the expected returnbeta relationship for any particular asset. For this, we need the CAPM
assumptions and its dominance arguments.
10.4 A MULTIFACTOR APT
We have assumed so far that only one systematic factor affects stock returns. This
simplifying assumption is in fact too simplistic. We have noted that it is easy to
think of several factors driven by the business cycle that might affect stock
returns: interest rate fluctuations, inflation rates, oil prices, and so. Presumably,
exposure to any of those factors will affect a stocks risk and hence, its expected
rate of return. We can derive a multifactor version of the APT to accommodate
these multiple sources of risk.
Suppose that we generalize the single-factor model expressed in Equation 10.1 to
a two-factor model:
ri E(ri ) i1 F1 i2 F2 ei

10.8a

In example 10.2, factor 1 was the departure of GDP growth from expectations,
and factor 2 was the unanticipated changes in interest rates. Each factor has a zero
expected value because each measures the surprise in the systematic variable
rather than the level of the variable. Similarly, the firm-specific component of
unexpected return, ei, also has zero expected value. Expending such a two-factor
model to any number of factors is straightforward.

16

Establishing a multi-factor APT is similar to one-factor case. But first we must

introduce the concept of a factor portfolio, which is a well-diversified portfolio
constructed to have a beta of 1.0 on one of the factors and a beta of zero on any
other factor. We can think of a factor portfolio as a tracking portfolio. That is, the
returns on such a portfolio track the evolution of particular sources of
macroeconomic risk, but are uncorrelated with other sources of risk. It is possible
to form such factor portfolio because we have a large number of securities to
choose from, and a relatively small number of factors. Factor portfolios will
serve as the benchmark portfolio for a multifactor security market line.
To generalize the arguments in Example 10.5 in the textbook for a two-factor
model, note that the factor exposures of any portfolio, P, are given by its betas,
P1 and P2. A competing portfolio, Q, can be formed by investing in factor
portfolios with the following weights: P1 in the first factor portfolio; P2 in the
second factor portfolio, and (1 - P1 - P2) in T-bills. By construction, portfolio Q
will have betas equal to those of portfolio P, and expected return of:
E(rQ ) P1 E(r1 ) P2 E(r2 ) (1 P1 P2 )rf
r f P1 [E(r1 ) r f ] P2 [E(r2 ) r f ]

10.9

Using the numbers in Example 10.5 on pp.360 we receive:

E(rQ) = 4 + 0.5(10 4) + 0.75(12 4) = 13%
Because portfolio Q has precisely the same exposures as portfolio A to the two
sources of risk, their expected returns also ought to be equal. So, portfolio A also
ought to have an expected return of 13%. If it does not, then there will be an
arbitrage opportunity (see Example 10.6 in the textbook.)
We conclude that any well-diversified portfolio with betas P1 and P2 must have
the return given in Equation 10.9 if arbitrage opportunities are to be precluded. If
you compare Equations 10.3 and 10.9 you will see that the above two-factor
equation is simply a generalization of the one-factor SML. So, we conclude that
the multifactor generalization of the security market line of the APT and the
CAPM are effectively equivalent.
The generalized APT must be qualified with respect to individual assets just as in
the one-factor APT. Equation 10.9 cannot be satisfied by every well-diversified
portfolio unless it is satisfied by virtually every security taken individually.
Equation 10.9 thus represents the multifactor SML for an economy with multiple
sources of risk. A multifactor CAPM would, at the cost of additional
assumptions, apply to any and all individual securities. As we have seen, the

17

result will be a security market line equation (a multidimensional SML) that is

identical to that of the multifactor APT. (We pointed out earlier that one application of
the CAPM is to provide fair rates of return for regulated utilities. The multifactor APT can
be used to the same ends see nearby box in the textbook on p.362.)

10.5 WHERE SHOULD WE LOOK FOR FACTORS?

One shortcoming of the multifactor APT is that it gives no guidance concerning
the determination of relevant risk factors or their risk premiums. In contrast, the
CAPM implies that the risk premium on the market is determined by the market's
variance and the average degree of risk aversion across investors. Two principles
guide us when we specify a reasonable list of factors. First, we want to restrict
ourselves to a limited number of systematic factors with considerable ability to
explain security returns. Second, we wish to choose factors that seem likely to be
important risk factors, that is, factors that concern investors sufficiently so that
they will demand meaningful risk premiums to bear exposure to those sources of
risk.
As we have seen, the CAPM also has a multifactor generalization, sometimes called the
intertemporal CAPM. This model provides some guidance concerning the risk premiums on
the factor portfolios. Moreover, recent theoretical research has demonstrated that one may
estimate an expected return-beta relationship even if the true factors or factor portfolios cannot
be identified.

The multifactor CAPM and APT are elegant theories of how exposure to
systematic risk factors should influence expected returns, but they provide little
guidance concerning which factors (sources of risk) ought to result in risk
premiums. A full-blown test of the multifactor equilibrium model, with prespecified factors and hedge portfolios, is as yet unavailable. A test of this
hypothesis would require three stages:
1. Specification of risk factors
2. Identification of portfolios that hedge these fundamental risk factors
3. Test of the explanatory factors power and risk premiums of the hedge
portfolios
1. The Chen, Roll and Ross Five-Factor Model
A step in this direction was made by Chen, Roll and Ross (1986) who chose the
following set of factors based on the ability of these factors to pain a broad
picture of the macro-economy. Their set is obviously but one of many possible
sets that might be considered:
1. % change in industrial production (IP),
2. % change in expected inflation (EI),
18

3. % change in unanticipated inflation (UI),

4. Excess return of long-term corporate bonds over long-term government
bonds (CG), and
5. Excess return of long-term government bonds over T-bills (GB).
With the identification of these potential economic factors, Chen, Roll and Ross
skipped the procedure of identifying factor portfolios (the portfolios that have
the highest correlation with the factors). Instead, by using the factors
themselves, they implicitly assumed that factor portfolios exist that can proxy
for the factors. They use these factors in a test similar to that of Fama and
MacBeth.
A critical part of the methodology is the grouping of stocks into portfolios.
Recall that in the single-factor tests, portfolios were constructed to span a wide
range of betas to enhance the power of the test. In a multifactor framework the
efficient criterion for grouping is less obvious. Chen, Roll and Ross chose to
group the sample stocks into 20 portfolios by size (market value of outstanding
equity), a variable that is known to be associated with stock returns.
They first used five years of monthly data to estimate the factor betas of the 20
portfolios in a first-pass regression. This is accomplished by estimating the
following regressions for each portfolio:
rP M rM IP IP EI EI UI UI CG CG GB GB eP

10.10a

where M stands for the stock market index. Chen, Roll and Ross used as the
market index both the value-weighted NYSE index (VWNY) and the equally
weighted NYSE index (EWNY).
Using the 20 sets of first-pass estimates of factor betas as independent variables,
they now estimated the second-pass regression (with 20 observations, one for
each portfolio) as follows:
rP 0 M M IP IP EI EI UI UI CG CG GB GB eP ,

10.10b

where the s become estimates of the risk premiums on the factors.

Chen, Roll and Ross (1986) ran this second-pass regression for every month of
their sample period, re-estimating the first-pass factor betas once every 12
months. The estimated risk premiums (the values for the parameters, s) were
averaged over all the second-pass regressions. The results of the regressions
show that the two market indexes EWNY and VWNY are not significant (their tstatistics of 1.218 and -0.633 are less than 2). Note also that the VWNY factor
19

has the wrong sign in that it seems to imply a negative market-risk premium.
Industrial production (IP), the risk premium on corporate bonds (CG), and
unanticipated inflation (UI) are the factors that appear to have significant
explanatory power (see also Chapter 13, Table 13.4 on p.447).
2. The Fama-French Three-Factor Model
An alternative approach to specifying macroeconomic factors as candidates for
relevant sources of systematic risk uses firm characteristics that seem, on
empirical ground, to proxy for exposure to systematic risk. The factors chosen are
variables that on past evidence seem to predict average returns well and therefore
may be capturing risk premium. One example of this approach is the Fama and
French (1996) free-factor model, which has come to dominate empirical research
and industry applications.
Fama and French added firm size and book-to-market ratio to the market index to
explain average returns. These additional factors are motivated by observations
that average returns on stocks of small firms and on stocks of firms with a high
ratio of book value to market value of equity have historically been higher than
predicted by the security market line of the CAPM. This observation suggests that
size or book-to-market ratio may be proxies for exposures to sources of
systematic risk not captured by the CAPM beta, and thus result in risk premium.
Fama and French (1996) proposed the following multifactor model:
Rit i iM RMt iSMB SMBt iHML HMLt eit

10.11

where:
SMB = Small Minus Big, i.e., the return of a portfolio of small stocks in excess of the
return on a portfolio of large stocks
HML = High Minus Low, i.e., the return of a portfolio of stocks with a high book-tomarket ratio in excess of the return on a portfolio of stocks with a low bookto-market ratio
Taking the difference in returns between two portfolios has an economic interpretation. The SMB
return, for example, equals the return from a long position in small stocks financed with a short
position in the large stocks. Note that this is a portfolio that entails no net investment. As a result, there
is no compensation for time value of money, only for risk, and the total return therefore may be
interpreted as a risk premium. Note that in this model the market index does play a role

and is expected to capture systematic risk originating from macroeconomic factors.

These two firm-characteristic variables are chosen because of long-standing
observations that corporate capitalization (firm size) and book-to-market ratio predict
deviations of average stock returns from levels consistent with CAPM. Fama and
20

French justify this model on empirical grounds: while SMB and HML are not
themselves obvious candidates for relevant risk factors, the argument are that these
variables may proxy for yet-unknown more-fundamental variables. For example,
Fama and French point out that firms with high ratios of book-to-market value are
more likely to be in financial distress and that small stocks may be more sensitive in
changes in business conditions, Thus, these variables may capture sensitivity to risk
factors in the macroeconomics.
The problem with empirical approaches such as the Fama and French model, which
use proxies for extra-market sources of risk, is that none of the factors in the proposed
models can be clearly identified as hedging a significant source of uncertainty. Black
(1993) points out that when researchers scan and rescan the database of security
returns in search for explanatory factors (an activity often called data-snooping), they
may eventually uncover past patterns that are due purely to chance. Black observes
that return premiums to factors such as firm size have proven to be inconsistent since
first discovered. However, Fama and French have shown that size and book-to-market
ratios have predicted average returns in various time periods and in markets all over
the world, thus mitigating potential effects of data-snooping.
10.6 THE MULTIFACTOR CAPM AND THE APT (optional) p.364

21