FIE400E - Investments

Topic 6 - Arbitrage Pricing Theory and Multifactor Models

Francisco Santos

Norwegian School of Economics

Outline

Arbitrage Pricing Theory – APT

Multifactor models of risk and return

Readings: BKM, chapters 10 and 13

CAPM only prices systematic risk --> ß --> if ß+ expected return +

RF + market portfolio --> Passive Strategy

Francisco Santos (NHH) FIE400E - Investments 1 / 27

Arbitrage Pricing Theory – APT
Given the failure of the CAPM in empirical tests, other models were
developed.

One of them is the Arbitrage Pricing Theory (APT) model which is a

single-factor model.

This model was developed by Stephen Ross.

The centerpiece of this model is the concept of arbitrage:

I a situation where the investor can construct a zero investment portfolio
with a sure profit.

I Arbitrage: sure gain with no risk.

I Since no investment is required, an investor can create large positions

to secure large levels of profit.

I In efficient markets, profitable arbitrage opportunities should quickly

disappear.
Francisco Santos (NHH) FIE400E - Investments 2 / 27
Arbitrage Pricing Theory – APT
Single-Factor Model
Recall that under a single-factor model:

ri = E [ri ] + βi F + ei
same assumptions as in SIM

E [ri ] is the expected value of the return on security i.

F and ei are random variables.

ei measures firm-specific surprises.

F is a macroeconomic factor that measures macro shocks.

βi measure the sensitivity of firm i to macro shocks.

F ∼ N (0, σF2 ) and ei ∼ N (0, σe2i ) .

Cov (ei , ej ) = 0 for i 6= j and Cov (F , ei ) = 0 for ∀i.

Francisco Santos (NHH) FIE400E - Investments 3 / 27
Arbitrage Pricing Theory – APT
Single-Factor Model

The return of portfolio p is given by:

rp = E [rp ] + βp F + ep

where ßF = Marco surprises

P
I E [rp ] = i wi E [ri ]
P Stocks:
I βp = i w i βi Risk from macrosurprises
Risk from firm specific surprises
P
I ep = i wi ei well deversified portfolio (only macro risk)

The variance of portfolio p is given by:

σp2 = βP2 σF2 + σe2p

with σe2p = Var ( i wi ei ) = i wi2 ei2 and lim σe2p = 0.

P P
n→∞

Francisco Santos (NHH) FIE400E - Investments 4 / 27

Arbitrage Pricing Theory – APT
Returns as a Function of the Systematic Factor
Consider a well diversified portfolio A with βA = 1.

The expected return is 10% (no surprise on the common-factor F=0).

The return on the portfolio is given by:

rA = E [rA ] + βA F = 10% + 1F
Francisco Santos (NHH) FIE400E - Investments 5 / 27
Arbitrage Pricing Theory – APT
Returns as a Function of the Systematic Factor
Consider a single stock S with βS = 1.

The expected return is 10% (no surprise on the common-factor F=0).

The undiversified stock is subject to idiosyncratic risk.
The return on the portfolio is given by:

rS = E [rS ] + βA F + eS = 10% + 1F + eS

Francisco Santos (NHH) FIE400E - Investments 6 / 27

Arbitrage Pricing Theory – APT
Returns as a Function of the Systematic Factor

Consider two well diversified portfolio A and B.

If two well diversified portfolios have the

same risk (ß) --> the E(r) needs to be the
same --> otherwise arbitrage possibility!!

βA = βB = 1.
The return on the portfolio A: rA = 10% + 1F .
The return on the portfolio B: rB = 8% + 1F
Can this return patterns coexist for long?
Francisco Santos (NHH) FIE400E - Investments 7 / 27
Arbitrage Pricing Theory – APT
Returns as a Function of the Systematic Factor

Can this return patterns coexist for long? NO!

No matter the realization of the systematic factor F , portfolio A

outperforms portfolio B – arbitrage opportunity.

How to exploit it?

Buy the portfolio with the highest return and sell the return with
lowest return.

Example:
I Buy 1M NOK of A and short 1M NOK of B – zero investment.
I From the long position in A : (10% + 1F )1M NOK .
I From the short position in B : −(8% + 1F )1M NOK .
I Net proceeds: 2% × 1M NOK = 20.000 NOK .

Francisco Santos (NHH) FIE400E - Investments 8 / 27

Arbitrage Pricing Theory – APT
Portfolios with Different Betas

But do we need portfolios with same betas?

Consider the previous portfolio A (βA = 1 and E [rA ] = 10%).

Add a portfolio C with βC = 0, 5 and E [rC ] = 6%.

Consider also a risk-free asset with rf = 4%.

Is there an arbitrage opportunity?

Francisco Santos (NHH) FIE400E - Investments 9 / 27

Arbitrage Pricing Theory – APT
Portfolios with Different Betas

Yes!

Note that:
I rC = 6% + 0, 5F
I Using A and the risk-free asset, we can construct a portfolio D that
has the same beta as C :
F We invest 50% on the risk-free asset plus 50% on portfolio A.

I Given these weights, the expected return on portfolio D is:

F E [rD ] = 0, 54% + 0, 510% = 7%

I Thus, rD = E [rD ] + βD F = 7% + 0, 5F .

We should buy portfolio D and sell portfolio C .

Francisco Santos (NHH) FIE400E - Investments 10 / 27

Arbitrage Pricing Theory – APT
Always buy the P with more return and sell those with less return --> profit is the delta y
between the two P --> Arbitrage between lines with the same ß!

To preclude arbitrage opportunities, the expected return on

all well-diversified portfolios must lie on this straight line.
APT only applies to well diversified portfolios!

Francisco Santos (NHH) FIE400E - Investments 11 / 27

Arbitrage Pricing Theory – APT
The One-Factor Security Market Line
Now consider the market portfolio M, a well-diversified portfolio.
Let us measure the systematic factor as the risk premium on that
portfolio.
This yields an SML relation equivalent to the CAPM, but only for
well-diversified portfolios:

E [rp ] = rf + (E [rM ] − rf )βp

Francisco Santos (NHH) FIE400E - Investments 12 / 27

APT and CAPM

The APT does not require that the benchmark portfolio in the SML is
the market portfolio – any well-diversified portfolio will do.

Even if the index portfolio (in CAPM) is not a precise proxy of the
true market portfolio, as long it is sufficiently diversified, the SML
relationship still holds according to the APT.

APT applies to well diversified portfolios and not to individual stocks.

APT cannot rule out a violation of the expected return-beta

relationship for any particular asset.

For this, we need the CAPM – should hold for every security.

Francisco Santos (NHH) FIE400E - Investments 13 / 27

Multifactor models

The return on security i is now a function of several factors:

K
X
ri = E [ri ] + βi,z Fz + ei
z=1

where
I There are K factors.
I βi,z denotes the factor loading of security i on factor z.
I Fz denotes factor z.
I ei still denotes firm-specific events.

Francisco Santos (NHH) FIE400E - Investments 14 / 27

Multifactor models
Multifactor SML

The expected return on security i is now given by the new SML:

K
X
E [ri ] = rf + βi,z RPz
z=1

where
I RPz denotes the risk premium for factor z: RPz = E [rFz − rf ].

Francisco Santos (NHH) FIE400E - Investments 15 / 27

Multifactor models
Example

Consider a two-factor model where:

I E [rF1 ] = 10%
I E [rF2 ] = 12%

Let the risk free rate be 4%.

Now consider a portfolio A with βA,1 = 0, 5 and βA,2 = 0, 75

If this model is the true one, what should be the total return on
portfolio A?

Francisco Santos (NHH) FIE400E - Investments 16 / 27

Multifactor models
Example

E [ri ] = rf + βi,1 RP1 + βi,2 RP2

The risk premium attributable to risk factor 1 should be the risk
premium on factor 1 adjusted for the exposure to factor 1:
I βA,1 E [rF1 − rf ] = 0, 56% = 3%.

The risk premium attributable to risk factor 2 should be the risk

premium on factor 2 adjusted for the exposure to factor 2:
I βA,1 E [rF1 − rf ] = 0, 758% = 6%.

The total return on the portfolio A should be: 4% + 3% + 6% = 13%.

Francisco Santos (NHH) FIE400E - Investments 17 / 27

The Fama-French Three-Factor Model for performance analysis

if alpha <> --> skill: fund creates value

--> model is wrong

The Fama-French (FF) 1993 three-factor model:

alpha = 0 if
model is correct
ri − rf = αi + bi (rM − rf ) + si SMB + hi HML + ei
Market Size Value
where
I SMB: Small Minus Big – the return of a portfolio of small stocks in
excess of the return on a portfolio of large stocks.

I HML: High Minus Low – the return of a portfolio of stocks with high
book-to-market ratio in excess of the return on a portfolio of stocks
with a low book-to-market ratio.

if + --> dann fund is betting von the first letter S/H

if - --> dann fund is betting on the last letter B/Low BooktoMarket

Francisco Santos (NHH) FIE400E - Investments 18 / 27

The Fama-French Three-Factor Model

The FF3F model adds two firm-characteristics variables to the market

index model.

Firm size and book-to-market predict deviations of the average stock

returns to the ones predicted by CAPM.

Fama and French argue that these variables may proxy for
yet-unknown more-fundamental variables.

For example, high book-to-market firms are more likely to be in

financial distress and small stocks may be more sensitive to changes
in the business conditions.

Thus, these variables may capture sensitivity to macroeconomic risk

factors that is not captured by β.

Francisco Santos (NHH) FIE400E - Investments 19 / 27

The Fama-French Three-Factor Model

How do we construct SMB and HML?

Easy way: Ken French provides the returns on SMB and HML, as well
as all the data needed to construct them (plus a tone of other things)
on his website – link to website.

How does it work:

I Sort industrial firms by market cap and by book-to-market.

I SMB is constructed as the difference in returns between the smallest

and largest third of firms.

I HML is constructed as the difference in returns between the high and

low book-to-market firms

Francisco Santos (NHH) FIE400E - Investments 20 / 27

The Fama-French Three-Factor Model
Davis, Fama, and French (2000) – Testing the Model

They construct portfolios by sorting firms into three size groups and
three book-to-market groups:
Size
Book-to-Market
Small Medium Big
High S/H M/H B/H
Medium S/M M/M B/M
Low S/L M/L B/L

For each of these nine portfolios, Davis, Fama, and French estimate:

ri − rf = αi + bi (rM − rf ) + si SMB + hi HML + ei

Francisco Santos (NHH) FIE400E - Investments 21 / 27

The Fama-French Three-Factor Model
Davis, Fama, and French (2000) – Testing the Model

--> here sign that model is not doing well if firm is bettin gon Small and L

rxyz-rf = +2% + 0,8 - 0,45 SMB + 0,33 HML

tstat 3,27 +14,28 -3,21 +0,88
>> Exposed to market / Big firms / no bet on value factor / --> alpha =2% --> here skill and alpha is good (not if bet on Small and low
& creates value

if alpha does not disapear in long term --> it is a sign for risk and represents risk compensation (example Winners/Losers)

Francisco Santos (NHH) FIE400E - Investments 22 / 27

The Fama-French Three-Factor Model
Davis, Fama, and French (2000) – Testing the Model

Intercepts are not significantly different from zero.

I Except for the small and low boot-to-market firms.

Betas to the market are very close to 1.

Big firms load negatively on the SMB.

Small firms load positively on the SMB.

High book-to-market firms load positively on the HML.

Low book-to-market firms load negatively on the HML.

R 2 s are high (above 0.91).

Francisco Santos (NHH) FIE400E - Investments 23 / 27

FF3F + Momentum

A fourth factor has emerged and added by many to the FF3F model –
the momentum factor.

Jegadeesh and Titman uncovered a tendency for good or bad

performance of stocks to persist over several months.

Carhart (1997) formally added the momentum factor to the FF3F –

why this also is called Carhart four-factor model.

Subsequent work showed this behavior for many asset classes.

Francisco Santos (NHH) FIE400E - Investments 24 / 27

FF3F + Momentum
rmnd -rf = 0,01 + 1,1 + 0,6 SML - 0,2 HML + 0,3 WML
tstat -2,51 6,51 2,21 -3,21 2,88
>> exp. to market, SMALL, Low B/M, Winners --> alpha significant --> here destorys value
Winners/
Losers

ri − rf = αi + bi (rM − rf ) + si SMB + hi HML + ui UMD

up/down
+ ei
where
I UMD: UP minus Down – the return of a portfolio of stocks that
performed well in the recent past in excess of the return on a portfolio
of stocks that performed badly in the recent past.

A lot of evidence that momentum is relevant, but still no very good

idea why it is relevant.

How is momentum reflecting a risk-return trade-off?

Francisco Santos (NHH) FIE400E - Investments 25 / 27

FF3F + Momentum + Liquidity Buy illiquid, sell liquid
>> alpha was positive --> comp. risk
if alpha is disapearing after publish --> mispricing

ri − rf = αi + bi (rM − rf ) + si SMB + hi HML + ui UMD + li Liq + ei

Liquidity as a factor – Pastor and Stambaugh (2003):

Francisco Santos (NHH) FIE400E - Investments 26 / 27

The Fama-French Five-Factor Model
alpha destroying value if negativ

>> small/ high B-M/ weak / no bet on Invet / no bet on Momentum (WML)
robust conservative
minus minus
weak aggressive
ri − rf = αi + bi (rM − rf ) + si SMB + hi HML + ri RMW + ci CMA + ei

where
I RMW: Robust minus Weak – the return of a portfolio of stocks with
robust profitability in excess of the return on a portfolio of stocks with
weak profitability.

I CMA: Conservative minus Aggressive – the return of a portfolio of

stocks of low investment firms (conservative) in excess of the return on
a portfolio of stocks with high investment (aggressive).

Francisco Santos (NHH) FIE400E - Investments 27 / 27