Lecture 6: Arbitrage Pricing Theory

Investments

FIN460-Papanikolaou

APT

1/ 48

Overview

1. Introduction 2. Multi-Factor Models 3. The Arbitrage Pricing Theory

FIN460-Papanikolaou

APT

2/ 48

Introduction

The empirical failure of the CAPM is not that surprising:

We had to make a number of pretty unrealistic assumptions to

prove the CAPM.

For example, we made the assumption that investors had identical

preferences, had the same information, and hold the same portfolio (the market).

Also, there is the problem that identifying and measuring the

market return is difcult, if not impossible (the Roll Critique)

In this lecture we will study a different approach to asset pricing called the Arbitrage Pricing Theory or APT. The APT species a pricing relationship with a number of systematic factors.

FIN460-Papanikolaou

APT

3/ 48

Returns Are Multi-Dimensional

Barr Rosenberg, Extra Market Components of Covariance in Security Markets, Journal of Financial and Quantitative Analysis, 1974: Companies possessing similar characteristics may, in a given month, show returns that are different from the other companies. The pattern of differing shows up as the factor relation.

A set of common factors - not just a monolithic market - inuence returns

FIN460-Papanikolaou

APT

4/ 48

Comovement
Stocks in the same industry tend to move together. Example: European Banks

Source: BARRA
FIN460-Papanikolaou APT 5/ 48

Comovement
However, there are other common factors that simultaneously affect returns. Within the banking industry, the size factor is at work

Source: BARRA
FIN460-Papanikolaou APT 6/ 48

Multiple factors and the CAPM

The presence of multiple factors makes the CAPM a much more restrictive theory. Consider two sources of risk, technology and monetary policy.

Do all stocks respond the same way to technological innovation? Do all stocks respond the same way to changes in interest rates?
Assume that RM = RT + RI ,

CAPM then implies that E [Ri ] = r f + (E (Rm ) r f ) i

where

cov(Ri , Rm ) cov(Ri , RT ) + cov(Ri , RI ) = var(Rm ) var(Rm )

CAPM says that both covariances are priced exactly the same.
FIN460-Papanikolaou APT 7/ 48

Arbitrage Pricing Theory

The APT is an approach to determining asset values based on law of one price and no arbitrage. It is a multi-factor model of asset pricing. The APT is derived from a statistical model whereas the CAPM is an equilibrium asset pricing model.

To get the APT, we dont have to assume that everyone is

optimizing.

Recall that we did need this assumption to get the CAPM. This makes the APT a much more reasonable theory.
Unlike the CAPM, we need very few assumptions to get the APT.

FIN460-Papanikolaou

APT

8/ 48

APT Assumptions

The assumptions necessary for the APT are:

a) All securities have nite expected values and variances. b) Some agents can form well diversied portfolios c) There are no taxes d) There are no transaction costs

Notice that we have considerably fewer than with the CAPM! The central idea behind the APT will be that we can price some assets relative to other assets.
a) We will derive restrictions on the price of assets based on no-arbitrage. b) We will be able to say something stronger if we exclude near-arbitrage or extremely good deals.

FIN460-Papanikolaou

APT

9/ 48

Specifying Uncertainty
We need a systematic way of thinking about uncertainty or risk in a dynamic setting. We will assume that investors know exactly which states of nature can occur, and exactly what will happen in each of the states. We know which states are possible, but not which will actually occur: We only know what the probabilities of each of the states are.

You can think of this as meaning that we can always draw a

multi-branched tree for the world.

One implication is that investors know what the price of each and every security will be if the economy evolves in a particular way. With standard factor models, there are an innite number of states of nature, since the factors can take on any values.

FIN460-Papanikolaou

APT

10/ 48

No Arbitrage

Pricing restrictions in the APT come from the Absence of Arbitrage. Absence of Arbitrage in nancial markets means that NO SECURITY EXISTS WHICH HAS A NEGATIVE PRICE AND A NON-NEGATIVE PAYOFF.

Also, no security can be created which has this property.

This rule also implies that:
a) Two securities that always have the same payoff must have the same price. b) No security exists which has a zero price and a strictly positive payoff.

This is the same idea that has been applied to option pricing!

FIN460-Papanikolaou

APT

11/ 48

No Arbitrage

1. In an efciently functioning nancial market arbitrage opportunities should not exist (for very long). 2. Unlike equilibrium rules such as the CAPM, arbitrage rules require only that there just one intelligent investor in the economy.

This is why derivatives security pricing models do a better job of

predicting prices than equilibrium based models.

In this lecture, we will see how to apply the same concept to

pricing portfolios of assets.

3. If arbitrage rules are violated, then unlimited risk-free prots are possible.

FIN460-Papanikolaou

APT

12/ 48

Arbitrage - Example
Suppose that there are only two possible states for Ination and Interest Rates: high or low We know exactly how four securities will perform in each of the possible states: State/ Stock Int. Rate Ination Prob. Apex (A) Bull (B) Crush (C) Dreck (D) High Real Int. Rates High In. Low In. 5% 10% 0.25 -20 0 90 15 5% 0% 0.25 20 70 -20 23 Low Real Int. Rates High In. Low In. 0% 10% 0.25 40 30 -10 15 0% 0% 0.25 60 -20 70 36

FIN460-Papanikolaou

APT

13/ 48

Arbitrage - Example

Lets assume the prices of each of the four securities are $100 and calculate expected returns, standard deviations and correlations: Everything looks normal here, but there is a simple arbitrage opportunity lurking in these numbers!

Stock A B C D

Current Price 100 100 100 100

Expect. Return(%) 25.00 20.00 32.50 22.25

Standard Dev. (%) 29.58 33.91 48.15 8.58

A 1.00 -0.15 -0.29 0.68

Correlation Matrix B C -0.15 1.00 -0.87 -0.38 -0.29 -0.87 1.00 0.22

D 0.68 -0.38 0.22 1.00

FIN460-Papanikolaou

APT

14/ 48

Arbitrage - Example

Consider the return of an Equal-Weighted portfolio of A, B and C, and compare this with the return of D:
State/ Port. EW Port. of A,B,C Dreck (D) High Real Int. Rates High In. Low In. 23.33 15 23.33 23 Low Real Int. Rates High In. Low In. 20.00 15 36.67 36

This table shows that the return of the EW portfolio is higher in all states, this means that there is an arbitrage opportunity. What would happen to the price of D in a well functioning market?

FIN460-Papanikolaou

APT

15/ 48

Arbitrage - Example

Based on the arguments presented earlier, we will take the position that arbitrage opportunities cannot exist (for very long!). That means that there is something wrong with these prices. Our goal in this lecture is to come up with a model of security prices where:
a) If prices/returns obey this model, there is no arbitrage. b) If prices/returns fail to obey this model, there is arbitrage.

FIN460-Papanikolaou

APT

16/ 48

Arbitrage Pricing Theory

The previous example here is too restrictive. In reality:

1. An innite number of states are possible 2. There are generally a large number of factors, and a continuum of possible factor realizations.

The factor model framework gives us a systematic way of describing how expected security returns must relate to their comovement with the economy. If there is to no arbitrage in the economy, the we can also price assets relative to one another based on their comovement with these factors. This is the basis of the APT.

FIN460-Papanikolaou

APT

17/ 48

Arbitrage Pricing Theory

First we need a Factor Model (or Return Generating Process (RGP)), which is a mathematical expression for how the security returns move with economic factors:

We will call the sources of movement factors We will call the stock sensitivities to the factors factor loadings
(or, equivalently, factor betas or factor sensitivities).

We have already seen an example of such a model. It is the single-index model that we used to simplify the correlation structure between securities.

FIN460-Papanikolaou

APT

18/ 48

Factor Model

The Factor Model (or RGP) is:

r i E (r i ) = bi,1 f1 + ... + bi,K fK + e i

The fi s are common factors that affect most securities. Examples are economic growth, interest rates, and ination. We require that, for each of the factors, E ( fi ) = 0. This means that, instead of dening an f directly as economic growth, we would have to dene it as the deviation of economic growth from what was expected. Often we assume cov( fi , fj ) = 0 for i = j.

FIN460-Papanikolaou

APT

19/ 48

Factor Model

The Factor Model (or RGP) can be written as:

r i = E (r i ) + bi,1 f1 + ... + bi,K fK + e i

bi, j denotes the loading of the ith asset on the jth factor. This
tells you how much the assets return goes up when the factor is one unit higher than expected. The e i in this equation is idiosyncratic risk. For example, e i will be negative when a rms president dies, or a
rm loses a big contract.

We will assume that cov(e i , e j ) = 0 for all securities i and j.

FIN460-Papanikolaou

APT

20/ 48

Factor Model - Example

An example using this return generating process (BKM, chapter 10): Suppose that two factors have been identied for the U.S. economy: the growth rate of industrial production, IP, and the ination rate, IR. IP is expected to be 4%, and IR 6%. A stock with a beta of 1.0 on IP and 0.4 on IR currently is expected to provide a rate of return of 14%. If industrial production actually grows by 5%, while the ination rate turns out to be 7%, what is your revised estimate of the realized return on the stock?

FIN460-Papanikolaou

APT

21/ 48

Factor Model - Example

We know E (IP) = 4% and bIP = 1, E (IR) = 6%, bIR = .4, and

E (ri ) = 14%
The two factors are therefore:

fIP = (0.05 0.04) = 0.01 fIR = (0.07 0.06) = 0.01

Plug these into the return generating process gives the expected return conditional on these realization of the industrial production growth rate (IP) and the ination rate (IR):

E (r i | fIP , fIR ) = 0.14 + 1 0.01 + 0.4 0.01 = 15.4%

What about the idiosyncratic risk (ei )?
FIN460-Papanikolaou APT 22/ 48

Factor Model
Using the return generating process we can calculate the variance of this portfolio as (for two factors):

var(Ri ) = var bi,1 f1 + bi,2 f2 + e i

2 2 = b2 i,1 var ( f 1 ) + bi,2 var ( f 2 ) + 2 bi,1 bi,2 cov( f 1 , f 2 ) + e,i 2 2 = b2 i,1 var ( f 1 ) + bi,2 var ( f 2 ) + e,i

(if factors are uncorrelated)

the general formula for n factors:

n n

2 i =

j=1 k=1

bi, j bi,k j,k + 2 e,i

where j,k denotes the covariance between the jth and kth factors. (What is 2,2 ?) n 1. The systematic variance is n j=1 k=1 bi, j bi,k j,k . 2 2. The idiosyncratic variance is e,i .
FIN460-Papanikolaou APT 23/ 48

Factor Model

Under the factor model, the covariance between two stocks, i and j (for two factors)

cov(Ri , R j ) = cov(bi,1 f1 + bi,2 f2 + ei , b j,1 f1 + b j,2 f2 + e j ) = bi,1 b j,1 var( f1 ) + bi,2 b j,2 var( f2 ) + +(bi,1 b j,2 + b j,1 bi,2 ) cov( f1 , f2 ) = bi,1 b j,1 var( f1 ) + bi,2 b j,2 var( f2 ) (if factors are uncorrelated)

FIN460-Papanikolaou

APT

24/ 48

Factor Model

Examples of factor models:

a) The Fama-French 3-factor model of returns. b) The 3-factor model of the term-structure (level, slope, curvature).

The APT will give a theoretical justication for the use of empirical factor models in determining the fair rate of return. Multi-Factor Models vs the CAPM Does a well-diversied portfolio have any idiosyncratic risk?

How do our APT denitions of a systematic and idiosyncratic risk

differ from the CAPM denitions?

How does our APT denition of a well-diversied portfolio differ

from the CAPM denition?

FIN460-Papanikolaou

APT

25/ 48

Denitions
In the context of the APT, a diversied portfolio is a portfolio that carries no idiosyncratic risk:

r p E (r p ) = b p,1 f1 + ... + b p,K fK

This is dened relative to a specic factor model. We will assume that investors can form such diversied portfolios.
In the context of the CAPM, you may see different denitions:
1. A diversied portfolio will have the least variance for a given level of expected returns.
ALL minimum-variance efcient portfolios are diversied according to this denition.

2. A diversied portfolio will have zero idiosyncratic risk, i.e. it will have an R2 with the market.
ONLY the market portfolio (and combinations of this with the riskless asset, i.e. portfolios on the CAL) will be diversied according to this denition.
FIN460-Papanikolaou APT 26/ 48

Arbitrage Pricing Theory

The APT Pricing Equation is:

E (r i ) = 0 + 1 bi,1 + ... + K bi,K

This equation species that the relation between the expected return of a security and its factor loadings (bs) is linear.

The s or factor risk premia tell you how much extra return you get
for each extra unit of risk your portfolio has.

Note that there is one for each factor in the economy, plus one extra: 0 . If there is a risk-free asset, then it must be the case that portfolios with no risk (with all bs equal to zero) have a return of the risk-free rate, i.e., that 0 = r f .
How does it compare to the CAPM?

FIN460-Papanikolaou

APT

27/ 48

APT - Example
Here we will use a simple example to understand how investors price risky securities. This will allow us to better understand what each of the the elements of the Arbitrage Pricing Theory. We will try to better understand what Risk means in the APT context, and to try to better dene and understand the following terms: Factor Factor Loading Factor Risk-Premia Arbitrage To start out, lets look again at how we specify uncertainty in the APT setting. Note that this is the specication used in most of nance and
economics, including in derivatives securities pricing.

FIN460-Papanikolaou

APT

28/ 48

APT - Example
Suppose that we know that IBM is going to pay a liquidating dividend in exactly one year, and this is the only payment that it will make. However, that dividend that IBM pays is uncertain its size depends on how well the economy is doing.

If the economy is in an expansion, then IBM will pay a dividend of

140.

However, if the economy is in a recession, then its dividend will be

only 100.

If the two states are equally likely, the expected cash ow from IBM ) = 120. IBM is E (CF1
Note that, consistent with our denition of uncertainty, we know exactly what will happen to IBM in each scenario, but we dont know which scenario will occur.
FIN460-Papanikolaou APT 29/ 48

APT - Example
IBM Boom Payoff (Pr=0.5) Bust Payoff (Pr=0.5) 140 100 120 100 20%

E (CF1 ) Time 0 Price

Discount Rate

Assuming that the price of IBM is $100, we see that investors in this economy are applying a discount rate of 20% to IBMs expected cash-ows. The way they come up with the price of IBM is to take the expected

IBM ) = $120, and discount this cash ow at time 1 from IBM, E (CF1 cash-ow back to the present at the appropriate discount rate, which is apparently 20%.
Equivalently, we can say that the expected return of IBM is 20%.

FIN460-Papanikolaou

APT

30/ 48

APT - Example
Now lets consider a second security, DELL which, like IBM, will pay a liquidating dividend in one year, and which has only two possible cash ows, depending on whether the economy booms or goes into a recession over the next year. IBM Boom Payoff (Pr=0.5) Bust Payoff (Pr=0.5) 140 100 120 100 20% DELL 160 80 120 ? ?

E (CF1 ) Time 0 Price

Discount Rate

Now lets consider how investors will price DELL. Note the IBM and DELL have the same expected cash-ow, so one might guess that a reasonable price for DELL might also be $100.

FIN460-Papanikolaou

APT

31/ 48

APT - Example
However, even though the expected cash-ows for DELL are the same, we see that the pattern of cash ows across the two states are probably worse for DELL: DELLs payoff is lower in the bust/recession state, which is when we are more likely to need the money. It only does better when things are good.

The recession is when the rest of our portfolio is more likely to do

poorly, and when we are more likely to lose our job, and our consulting income is likely to be lower. We need the cash more in a recession In the boom/expansion, our portfolio will probably do better; were more likely to have a good job; well probably have more consulting income. This means that if DELL were the same price as IBM, we would buy IBM.

FIN460-Papanikolaou

APT

32/ 48

APT - Example
Therefore, to induce investors to buy all of the outstanding DELL shares, it will have to be the case that DELLs price is lower than $100. Also, note that the three are statements are equivalent:

The price investors will pay for DELL will be less than $100, even
though DELLs expected cash ows are the same as IBMs.

The discount rate investors will apply to DELLs cash ows will be
higher than the 20% applied to IBMs cash ows.

The expected return investors will require from DELL will be

higher than the IBMs expected return of 20%. Lets assume that investors are only willing to buy up all of DELLs shares if the price of DELL is $90, or, equivalently, that the discount rate that they will apply to DELL is 33.33%:

E (RDELL ) =

E (CF1DELL ) PDELL 120 90 30 = = = 0.3333 PDELL 90 90

FIN460-Papanikolaou

APT

33/ 48

APT - Example
IBM Boom Payoff (Pr=0.5) Bust Payoff (Pr=0.5) 140 100 120 $100 20% DELL 160 80 120 $90 33.33%

E (CF1 ) Time 0 Price

Discount Rate

The way to think about expected returns is that this is something that investors determine. After looking at the pattern of cash-ows from any investment, and deciding whether they like or dislike this pattern, investors determine what rate they will discount these cash ows at (to determine the price.) This rate is what we then call the expected return. Since investors accurately calculate the expected cash-ows, the average return they realize on the investment will be the discount rate.

FIN460-Papanikolaou

APT

34/ 48

APT - Example
Now lets see how all of this relates to the APT equations: 1. Calculating the business cycle factor fBC in the two states:

First, assume that we use the NBER (National Bureau of

Economic Research) business cycle indicator to construct our factor.
The indicator is one (at the end of the next year) if the economy is in an expansion, and zero if the economy is in a recession.

However, remember that for the factor, we need the unexpected

component of the business cycle.

Assuming there is a 50%/50% chance that we will be in an

expansion/recession, the expected value of the indicator is 0.5. This means that the business-cycle factor has a value of 0.5 = 1 0.5 if the economy booms, and 0.5 = 0 0.5 if the economy goes bust.

FIN460-Papanikolaou

APT

35/ 48

APT - Example
2. Calculating the factor loadings for the two securities:

The way of doing this is to run a time-series regression of the

returns of IBM on the factor. Lets do this rst for IBM:

rIBM,t = E (rIBM ) + bIBM,BC fBC,t + eIBM,t

where E (rIBM ) is the intercept and bIBM ,BC is the slope coefcient.

Here, we have only two points, so we can t a line exactly: 0.40 = E (rIBM ) + bIBM,BC 0.5 0.00 = E (rIBM ) + bIBM,BC 0.5 (boom) (bust)

Solving these gives E (rIBM ) = 0.20 (which we already knew) and bIBM,BC = 0.4. Similar calculations for DELL gives E (rDELL ) = 0.3333 and bDELL,BC = 0.8889.
FIN460-Papanikolaou APT 36/ 48

APT - Example

3. Interpreting the factor loadings.

The factor loadings bIBM,BC and bDELL,BC tell us how much risk
IBM and DELL have.

Risk means that the security moves up or down when the factor
(the business cycle) moves up and down.

Based on the way that we have characterized uncertainty, the expected return and the factor loading bi,BC of each security tell us
everything that we need to know to calculate all of the payoff, per dollar, in every state of nature.
a) This will be true for every well-diversied portfolio, in every factor model. b) One way of thinking about this is that the expected return tells us what the reward is, and the bs tell us what the risk of the security is.

FIN460-Papanikolaou

APT

37/ 48

APT - Example
4. Calculating the factor risk premia (s)

The Factor Model tells us nothing about why investors are

discounting the cash ows from the different securities at different rates. To determine how investors view the risks associated with each of the risks in the economy, we have to evaluate the APT Pricing Equation. Now that we have the factor loadings, we can determine the factor risk-premia (or s), by regressing expected returns Eri on factor loadings bik . Since there is only one factor and two stocks

E (rIBM ) = 0 + BC bIBM,BC E (rDELL ) = 0 + BC bDELL,BC

to which the solutions are 0 = 0.0909 and BC = 0.2727.

FIN460-Papanikolaou

APT

38/ 48

APT - Example

5. Interpreting the risk-premia (s):

DELL is considered to be a riskier security than IBM, in the sense

that investors discount DELLs cashows at a higher rate than IBM.

This is reected in DELLs higher loading on the BC factor. The BC is a measure of how much more investors discount a
stock as a result of having one extra unit of risk relating to the BC factor.

0 tells you how high a return investors require if a security has no

risk.

FIN460-Papanikolaou

APT

39/ 48

APT - Example
Using the set of securities that we used to calculated the s for the pricing equations, we can always construct a portfolio p with any set of factor loadings (b p,k ). The pricing equation then tells us what the expected return (or discount rate) for the portfolio of securities is. Alternatively, this equation tells us, if we nd a (well-diversied) portfolio with certain factor loadings, what discount-rate that portfolio must have for there to be no arbitrage. Finally, if we add a third security to the mix and calculate its factor loadings, the APT will tell us what its expected return must be in order to avoid arbitrage opportunities. Essentially what we doing is pricing securities relative to other securities, in much the same way we are pricing derivatives relative to the underlying security.
FIN460-Papanikolaou APT 40/ 48

Finding Arbitrages

Arbitrage arises when the price of risk differs across securities.

1. If investors are pricing risk inconsistently across securities, than, assuming these securities are well diversied, arbitrage will be possible. 2. To see this, lets extend the example to include a risk-free asset which we can buy or sell at a rate of 5%.
That is the rate for borrowing or lending is 5%.

3. Since 0 = 0.0909, we know that we can combine DELL and IBM in such a way that we can create a synthetic risk-free asset with a return of 9.09%. 4. So we borrow money at 5% (by selling the risk-free asset) and lend money at 9.09%

FIN460-Papanikolaou

APT

41/ 48

Building the Arbitrage Portfolio

To determine how much of IBM and DELL we buy/sell, we solve the equation for the weights on IBM and DELL in a portfolio which is risk-free:

wIBM bIBM,BC + (1 wIBM ) bDELL,BC = b p,BC = 0

or

wIBM =

bDELL,BC = 1.8182 bELL,BC bIBM,BC

and wDELL = (1 wIBM ) = 0.8182. This means that, to create an arbitrage portfolio, we can invest $1.8182 in IBM, short $0.8182 worth of DELL, and short $1 worth of the risk-free asset. This portfolio requires zero initial investment and has a positive payoff in all states.
FIN460-Papanikolaou APT 42/ 48

Building the Arbitrage Portfolio

To verify that this works, lets look at the payoff to the IBM/DELL risk-free portfolio in the two states: Payoff(boom) = wIBM (140/100) + (1 wIBM ) (160/90) = 1.0909 Payoff(bust) = wIBM (100/100) + (1 wIBM ) (80/90) = 1.0909 This means that the payoff from the zero-investment portfolio is $0.0409 = 1.0909 1.05 in both the boom and bust states. So this portfolio (in this simple economy) is risk free. Of course, we can scale this up as much as we like. For $1 million investment in the long and short portfolios, we would get a risk-free payoff of $40,900 (assuming prices did not move with our trades)

FIN460-Papanikolaou

APT

43/ 48

Interpreting the Arbitrage

What is going on here is that investors are pricing risk inconsistently across securities. With any pair of securities, we can calculate 0 and BC , however, each pair of securities will give you a different set of s: Security 1 IBM IBM DELL Security 2 DELL RF RF

0
0.0909 0.05 0.05

BC
0.2727 0.3750 (= (0.2 0.05)/0.4) 0.3187 (= (0.3333 0.05)/0.8889)

This means that, to nd the arbitrage, we could have 1. calculated the s using any pair of securities, and then 2. calculated the expected return (or discount rate) for the third security 3. bought the high return and sold the low return.

FIN460-Papanikolaou

APT

44/ 48

Interpreting the Arbitrage

Whenever this happens, there will be an arbitrage. If there are arbitrageurs in the economy, they will move prices until the arbitrage disappears, and risk is priced consistently. This is the argument behind the APT. Note that this assumes however that the arbitrageurs have unlimited capital and patience. We will see later that in some cases that this may be an erroneous assumption and in fact, there may be limits to arbitrage.

FIN460-Papanikolaou

APT

45/ 48

Summary

What have we learned? 1. Here, we have used a simple example to understand how investors price risky securities. 2. The idea behind the APT is that investors require different rates of return from different securities, depending on the riskiness of the securities. 3. If, however, risk is priced inconsistently across securities, then there will be arbitrage opportunities. 4. Arbitrageurs will take advantage of these arbitrage opportunities until prices are pushed back into line, risk is priced consistently across securities, and arbitrages disappear.

FIN460-Papanikolaou

APT

46/ 48

Glossary
We have dened the following terms: Factors ( fs) move up and down with the economy, and affect the future cash ows of securities. Factor loadings (bs) for each security, for each factor, tell you how much the security moves (on average, in percent) when the factor moves by 1%. Factor risk-premia: (s) for each factor, tell you how much higher a discount-rate investors apply to a security if its factor loading on a particular factor is higher by one. Arbitrage arises if risk is priced inconsistently across (well-diversied) securities. One way of thinking about this is that we can nd two well
diversied securities/portfolios, with exactly the same risk/factor-loadings, which have their cash ows discounted at different discount rates.
FIN460-Papanikolaou APT 47/ 48

Summary

The APT is an alternative model to the CAPM. It makes fewer assumptions, and as a result gets weaker predictions. In particular, the APT does not say what the systematic factors are, whereas the CAPM says that the market portfolio is the only systematic source of risk. Later: How to specify the factors?
a) Factors can be specied a priori: they could be macroeconomic variables (ex ination, output) that capture the systematic risk in the economy or portfolios proxying for these risks. b) Factors can be extracted via Principal Components or Factor Analysis.

FIN460-Papanikolaou

APT

48/ 48