Lecture 12

Factor Models for Expected Returns:

Estimation and Tests

Lars A. Lochstoer
UCLA Anderson School of Management

Winter 2025

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 1 / 54

 Overview of Lecture 12

Equilibrium factor models

Time-series regression tests of factor models

1 Alpha and mispricing

2 The 25 Fama-French portfolios as test assets

1 Testing the CAPM
2 Testing the Fama-French 3-factor model

3 Alpha and mean-variance e¢ ciency (max Sharpe ratios)

4 Appendix: The Arbitrage Pricing Theory (APT)

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 2 / 54

 Risk comes in many ‡avors

we used to think that risk comes in a single ‡avor

a stock’s βi with the market factor
βi measures the ‘quantity of risk’or amount of exposure of asset i to the
market.

early empirical evidence originally supported the CAPM...now it is resoundingly

rejected

now, we know that risk comes in many ‡avors

that complicates portfolio advice
it makes performance analysis more challenging!

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 3 / 54

 Why should we expect multiple factors?

the average investor has a job

compare two stocks with the same market beta

stock A does well in a recession
stock B does poorly in a recession

CAPM says we are indi¤erent between the two stocks

are we?

some evidence for a missing recession risk factor!

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 4 / 54

 Why should we expect multiple factors?

suppose some less sophisticated investors have portfolios that are biased
towards large, growth …rms (because these are more glamorous)

then, the sophisticated investors have to overweight the small, value …rms in
their portfolio of risky assets

these sophisticated investors do not and cannot hold the market portfolio

in fact, the market portfolio is no longer e¢ cient

this gives rise to new factors (like value and size)

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 5 / 54

 Why estimate factor models?

test your model

can it account for interesting cross-sectional variation in average returns on
assets in a particular asset class?

once you have a ‘good model’

estimate cost of capital for company
estimate risk-adjusted returns on trading strategy
do performance attribution, style analysis
look for skill in risk-adjusted returns of pension funds and hedge funds

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 6 / 54

 An expected return-beta pricing model

The CAPM is the example you already know:

E [Ri ,t Rf ,t ] = βM ,i E [RM ,t Rf ,t ]

What is the property of the market portfolio?

It is mean-variance e¢ cient – it has the maximal Sharpe ratio

In fact, mathematically it is true that:

E [Ri ,t Rf ,t ] = βMVE ,i E [RMVE ,t Rf ,t ] .

where RMVE is the return to the mean-variance e¢ cient portfolio!

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 7 / 54

 The unicorn: The Mean-Variance E¢ cient Portfolio
’All’we need is to …nd the mean-variance e¢ cient portfolio.
This is what multifactor expected return models are all about
Find, say, three excess return factors that span the excess return
mean-variance e¢ cient portfolio (the ’e’superscript is for excess returns):
e e e e
RMVE ,t = a1 RF 1,t + a2 RF 2,t + a3 RF 3,t .

Then:

E Rie,t = e
βMVE ,i E RMVE ,t
= βMVE ,i E a1 RFe 1,t + a2 RFe 2,t + a3 RFe 3,t
= βMVE ,i a1 E RFe 1,t + βMVE ,i a2 E RFe 2,t + βMVE ,i a3 E RFe 3,t
= β1 E RFe 1,t + β2 E RFe 2,t + β3 E RFe 3,t

How do you estimate the betas?

Linear regression, as usual
Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 8 / 54
 Di¤erent Estimation Methods for Linear Factor Model

1 Time series regressions (this lecture)

I OLS

2 Cross-sectional regressions (next lecture)

I OLS
I Fama and MacBeth (1973)

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 9 / 54

 Single Factor Equilibrium Model: APT version
Assume there is only one factor that explains covariation in stock returns:

Rite = αi + βi ft + εit ,

where εit are uncorrelated across stocks and time, and ft is a traded factor (e.g.,
excess market returns)

Since ε-risk can be diversi…ed away in large portfolios, it cannot will not earn a
risk premium if markets are perfectly competitive
Two sources of variation in returns: βi ft and εit
I Risk premium on βi ft is βi E [ft ]
I Risk premium on εit is zero
I Thus, risk premium on stock should be E [Rite ] = βi E [ft ] and so ... αi = 0!

Let’s test this model using the 25 Fama-French portfolios!

Let the factor be the market factor

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 10 / 54

 The 25 Fama-French portfolios

Fama and French (1993) sort the data into 25 portfolios.

they construct portfolios by sorting them along two dimensions:

1 5 bins sorted on size
2 5 bins sorted on book-to-market

25 Portfolios: large spread in average excess returns

theh goal e
i is to maximize the variation in the expected excess returns E Rit vs.
E Rjte , i.e. the left hand side variables in the regression.

Then, we want to see if the factors f t can capture this variation.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 11 / 54

 Fama-French 25 b/m and size sorted portfolios

25 Fama-French portfolios. Monthly data. 1960-2015.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 12 / 54

 Testing the CAPM

the CAPM predicts that αi = 0 for all assets and that variation in the cross-section
of expected returns can be explained by variation in the market times an assets βi

one-step procedure: run a time series regression of excess returns on the factor to
estimate the βi ’s

Rite = αi + βi Rmt
e
+ εit

let’s plot the predicted excess returns R̂ie = β̂i 1

T ∑T e
t =1 Rmt against the
realized excess returns R̄ie = T1 ∑T e
t =1 Ri ,t

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 13 / 54

 Failure of the CAPM

25 Fama-French portfolios. Monthly data. 1960-2015. Plot of the predicted excess

returns β̂i T1 ΣT 1 T
t =1 Rmt against the realized average excess returns T Σt =1 Rit
e e

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 14 / 54

 CAPM alphas

CAPM α’s for the 25 Fama-French portfolios from a time series regression of
returns on the factors. Monthly data. 1960-2015.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 15 / 54

 CAPM betas

CAPM β’s for the 25 Fama-French portfolios from a time series regression of
returns on the factors. Monthly data. 1960-2015.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 16 / 54

 CAPM betas (bar chart)

CAPM β’s for the 25 Fama-French portfolios from a time series regression of
returns on the factors. Monthly data. 1960-2015.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 17 / 54

 Multifactor model

the CAPM fails to capture the cross-section of expected returns built from the FF
25 portfolios sorted on size and book-to-market.

we need more factors f t to explain variation in returns.

Fama and French: construct factors f t from returns

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 18 / 54

 Fama and French (1993) 3-factor model
The Fama/French factors are constructed using the 6 value-weight portfolios
formed on size and book-to-market.
SMB (Small Minus Big) :

SMB = 1/3(SmallValue + SmallNeutral + SmallGrowth)

1/3(BigValue + BigNeutral + BigGrowth).

HML (High Minus Low) :

HML = 1/2(SmallValue + BigValue ) 1/2(SmallGrowth + BigGrowth).

Rtm Rmt Rft , the excess return on the market, is the value-weight return on
all NYSE, AMEX, and NASDAQ stocks (from CRSP) minus the one-month
Treasury bill rate (from Ibbotson Associates).

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 19 / 54

 Time Series Regression

one-step procedure: run a time series regression of excess returns on the factor to
estimate the βi ’s

Rite = αi + βm m smb smb

i Rt + βi Rt + βhml
i Rthml + εit

De…ne the factor "risk premium" or "risk price" as λ̂ = λm , λsmb , λhml where
j
λ̂ = 1
T ∑T
t =1 Rtj for j = m, smb, hml

0
let’s plot the predicted excess returns β̂i λ̂ against the realized excess returns
R̄ie = T1 ∑T e
t =1 Rit

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 20 / 54

 FF 3-factor model

25 Fama-French portfolios. Monthly data. 1960-2015. Plot of the predicted

0
excess returns β̂i λ̂ against the realized average excess returns T1 ΣT e
t =1 Rit . The risk
prices, λ̂, are the means of the factors.
Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 21 / 54
 FF 3-factor alphas

Fama-French 3-factors α̂’s for the 25 Fama-French portfolios from a time series
regression on the factors. Monthly data. 1960-2015.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 22 / 54

 FF 3-factor betas

Fama-French 3-factors β̂’s for the 25 Fama-French portfolios from a time series
regression on the factors. Monthly data. 1960-2015.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 23 / 54

 Testing the model: Single factor
We can test the model
E [Rite ] = βi E [ft ]
by running time series regressions:

Rite = αi + βi ft + εit , t = 1, . . . , T

With i.i.d errors, the asymptotic test statistic for the pricing errors is:

" # 1
f¯ 2 1
T 1+ 2 α̂0 Σ̂ α̂ χ2N .
σ̂f

where Σ̂ denotes the covariance matrix of εt .

The CAPM is resoundingly rejected by Fama-French (1993)

This is the ’GRS test’(from Gibbons, Ross, and Shanken (1987))

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 24 / 54

 Testing the model: Multiple factors
We can test the model
E [Rite ] = βi0 E [f t ]
by running time series regressions:

Rite = αi + βi0 f t + εit , t = 1, . . . , T

With i.i.d Normally distributed errors, the exact small-sample test statistic for the
pricing errors is:
h i 1
0 1
(T N K )/N 1 + f̄ Σ̂f f̄ α̂0 Σ̂ε 1 α̂ FN ,T N K

where Σ̂ε denotes the covariance matrix of εt , Σ̂f denotes the covariance matrix
of the factors ft , f̄ is the average factor, and α̂ are the OLS estimates of α.
The Fama-French 3-factor model is also rejected...! (though not nearly at the
same signi…cance level as the CAPM)
Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 25 / 54
 Factor Models and Mean-Variance E¢ ciency:
The Data Mining Concern

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 26 / 54

 Properties of the in-sample MVE
We can, given a set of assets, easily compute the in-sample MVE portfolio:

min w0 Ω̂w such that w0 R̄ e = m

w

where w = [w1 w2 ... wN ]0 , Ω̂ is the sample variance-covariance matrix of

e R̄ e ... R̄ e 0 is the vector of sample average
excess returns, and R̄te = R̄1t 2t Nt
excess returns for each asset.

The optimal portfolio weights are (up to a constant of proportionality):

wMVE ∝ Ω̂ 1
R̄ e

The in-sample mean variance e¢ cient portfolio therefore has squared Sharpe ratio:
2
SRMVE = R̄ e 0 Ω̂ 1
R̄ e

Next slide has the derivations

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 27 / 54
 MVE derivations

Objective function in Lagrangian form:

1
min w0 Ω̂w k w0 R̄ e m
w 2

First order condition wrt w (an N 1 vector):

Ω̂w k R̄ e = 0

Thus:
wMVE = k Ω̂ 1
R̄ e .

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 28 / 54

 MVE derivations (cont’d)
Recall, wMVE = k Ω̂ 1 R̄ e .

Average excess return on MVE portfolio

0
e
R̄MVE = wMVE R̄ e = k R̄ e 0 Ω̂ 1
R̄ e

Variance of excess return on MVE portfolio:

0
e
var (RMVE ) = wMVE Ω̂wMVE = k 2 R̄ e 0 Ω̂ 1
Ω̂Ω̂ 1
R̄ e
= k 2 R̄ e 0 Ω̂ 1
R̄ e

Squared Sharpe ratio of MVE:

k R̄ e 0 Ω̂ 1 R̄ e 2
2 e
SR (RMVE ) = = R̄ e 0 Ω̂ 1
R̄ e
k 2 R̄ e 0 Ω̂ 1 R̄ e

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 29 / 54

 The GRS statistic revisited
Consider an asymptotic version (with i.i.d. residuals) of the Gibbons-Ross-Shanken
(GRS) statistic for testing whether a factor model is rejected or not:

α̂0 Σ̂ε 1 α̂
T 0 χ2 (N )
1 + R̄Fe Σ̂F 1 R̄Fe

0
Note that R̄Fe Σ̂F 1 R̄Fe is the in-sample maximum Sharpe ratio squared obtained
using the factor portfolios only.
ΣF is the K K variance-covariance matrix of the factors

Note that α̂0 Σ̂ε 1 α̂ is the maximum Sharpe ratio squared of hedged stock returns.
Recall:
h 0 e
i
E [α̂i + ε̂it ] = E Rite ^ βi RFt = α̂i for all i and
var (α̂ + ε̂t ) = Σ̂ε (an N N matrix)

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 30 / 54

 A Mean-Variance Decomposition
Since hedged stock returns are uncorrelated with the factor returns (by
construction), we have that
0
R̄ e 0 Ω̂ 1
R̄ e = α̂0 Σ̂ε 1 α̂ + R̄Fe Σ̂F 1 R̄Fe

We show this is true in a couple of slides

Then, we have that

0
α̂0 Σ̂ε 1 α̂ 1 + R̄Fe Σ̂F 1 R̄Fe + α̂0 Σ̂ε 1 α̂
0 = 0 1
1 + R̄Fe Σ̂F 1 R̄Fe 1 + R̄Fe Σ̂F 1 R̄Fe
1 + R̄ e 0 Ω̂ 1 R̄ e
= 0 1
1 + R̄Fe Σ̂F 1 R̄Fe

Thus, all alphas are zero if the maximum in-sample Sharpe ratio obtained using
the factors equals the maximum in-sample Sharpe ratio of the test assets!
In other words, if a linear combination of the factors is the in-sample
mean-variance e¢ cient portfolio, the factor model prices all assets perfectly
in the sense that all alphas equal zero
Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 31 / 54
 The failure of the CAPM revisited
The fact that the CAPM does not work means the Sharpe ratio of the market
portfolio is far from the maximum Sharpe ratio obtainable using the test assets

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 32 / 54

 A corollary on the implication of alpha

If a strategy has ’alpha’di¤erent from zero, it means it can be combined

with the factor portfolios to obtain a higher Sharpe ratio than what one
could get using the factor portfolios alone

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 33 / 54

 Implications of alpha: Implementation of Max SR Portfolio
The investable set of assets are N stocks, labelled i = 1, ..., N and K factor
portfolios, labelled j = 1, ..., K .

De…ne the factor-neutral assets as:

0 e
Ritα Rite ^
β RFt = α̂i + ε̂it , for all i

Put all of the investable asset returns in an (N + K ) 1 vector

h i0
Rt = RFe 1 t RFe 2 t ... RFe t R1t
α R α ... R α
2t Nt . Note that we only use the
K
factor-neutral assets and the factors. Otherwise, there would be collinearity
between Rie , Riα , and RFe .

Let Ω̂ denote the sample variance covariance matrix of Rt . Note that it is block
diagonal:
Σ̂F 0
Ω̂ =
0 Σ̂ε

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 34 / 54

 Implementation of Max SR Portfolio (cont’d)
Now, let’s …nd the portfolio weights that makes the maximum Sharpe ratio:

wMVE ∝ Ω̂ 1
Rt

0
Note that Rt = R̄F0 α̂0 , where R̄F is the K 1 vector of sample factor means
and α̂ is the N 1 vector of estimated alpha for the individual assets. Thus:

Σ̂F 1 0 Σ̂F 1 R̄F

wMVE ∝ Rt =
0 Σ̂ε 1 Σ̂ε 1 α̂

From our earlier results, the max Sharpe ratio squared is:
2 0 1
SRMVE = Rt Ω̂ Rt
Σ̂F 1 0 R̄F
= R̄F0 α̂0 = R̄F0 Σ̂F 1 R̄F + α̂0 Σ̂ε 1 α̂
0 Σ̂ε 1 α̂

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 35 / 54

 We’re always only one factor away from the MVE portfolio

Consider the misspeci…ed factor model

Rite = αi + βi0 ft + εit .

where αi 6= 0 for all i.

Which factor is missing?

The factor with portfolio weights

wα ∝ Σ ε 1 α

Adding this factor mechanically ensures, in-sample, that the factors span the
in-sample mean-variance e¢ cient portfolio

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 36 / 54

 The Missing Factor and Data Mining

Let’s think more about the portfolio weights of the ’…nal’factor

wα ∝ Σ ε 1 α

Let’s assume, for discussion purposes, that Σε is diagonal.

The ’…nal’factor goes long positive alpha assets and short negative alpha
assets

Seems a lot like how we do our characteristics-based factor. E.g., value and
momentum
I Go long value stocks, short growth stocks
I Go long winners, short losers, etc.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 37 / 54

 Data Mining and Economic Rationale

The fact that we know how to make our model work in-sample, for a given set of
test assets, makes introducing this factor vacuous in itself

We need:

An economic story of what risk or behavioral phenomenon the factor

represents

Out-of-sample, cross-country and/or cross-asset corroborating evidence

Thus, given how easy it is to data-mine, we are looking for economics that explain
why we would think the phenomenon persists out-of-sample

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 38 / 54

 Additional reading

I have written a note on factor model testing— time-series and cross-sectional

regressions, as well as the mean-variance math laid out in the previous slides

It’s posted on BruinLearn under ’Week 9’and may serve as a useful background
reading in addition to the slides

I’ve also posted the Fama-French (1993) paper, as well as the Gibbons, Ross,
Shanken (1987) paper.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 39 / 54

 APPENDIX

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 40 / 54

 Arbitrage Pricing Theory (APT) of Ross (1976)

assume the data are generated by a multi-factor model

Rite = αi + βi 1 f1t + βi 2 f2t + . . . + βiK fKt + εit , i = 1, . . . , N

= αi + βi0 f t + εit

the APT does not identify what the factors are

the factors f t could be traded assets, macro variables, or latent factors

assume the errors are uncorrelated with each factor:

E εit (fjt f jt ) = 0 8 i, j

some textbooks/authors (implicitly) assume the factors are uncorrelated with one
another...an orthogonal factor model.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 41 / 54

 Arbitrage Pricing Theory of Ross (1976)

assumptions:

1 disturbances are independent of the factors:

Cov[εit , fjt ] = 0

2 εit is independent of εjt :

Cov[εit , εjt ] = 0
This implies Σε is diagonal.

this is ‘like’the multiple factor model

contribution of Ross (1976) and APT: derive equilibrium implications

see also Chamberlain and Rothschild (1983)

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 42 / 54

 Arbitrage Pricing Theory of Ross (1976): Equilibrium

Suppose that excess returns are generated by a linear factor model:

Rite = ai + bi 1 f1t + bi 2 f2t + . . . + biK fKt + εit , i = 1, . . . , N

and assume no risk-free arbitrage opportunities exist.

Then, there exist risk prices λj for each factor such that the expected return on
any security j can be stated as:

E [Rite ] = λ0 + bi 1 λ1 + bi 2 λ2 + . . . + biK λK , i = 1, . . . , N

The theory puts no restrictions on these risk prices, except when the factors are
traded assets.
(for details; see Chapter 9 of Cochrane (2005) and Chapter 6 of Campbell, Lo, and MacKinley
(1997))

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 43 / 54

 Example with two factors
assume now the data are generated by a two-factor model:

Rite = αi + βi 1 f1t + βi 2 f2t + εit , i = 1, . . . , N

suppose we build an equally weighted wi = 1 portfolio using NP assets

NP

NP NP NP NP
1 1 1 1
e
Rpt =
NP ∑ αi +
NP ∑ βi 1 f1t +
NP ∑ βi 2 f2t +
NP ∑ εit
i =1 i =1 i =1 i =1
NP
1
= αp + βp1 f1t + βp2 f2t +
NP ∑ εit
i =1

in a ‘well-diversi…ed’portfolio, the residual risk disappears because of the Law

of Large Numbers

e
only systematic risk is left: Rpt αp + βp1 f1t + βp2 f2t

well-diversi…ed ) weights wi cannot be too extreme

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 44 / 54

 Example with two factors
in a well-diversi…ed portfolio, the residual risk disappears
only systematic risk is left:

Rite αi + βi 1 f1t + βi 2 f2t , i = 1, . . . , P

1 zero aggregate risk portfolio

1 construct a portfolio with β1 = 0 and β2 = 0
2 the expected excess return on this portfolio is 0

2 factor-1-mimicking portfolio
1 construct a zero-investment portfolio with β1 = 1 and β2 = 0
2
e ]
the expected (excess) return on this portfolio is: E [R1t

3 factor-2-mimicking portfolio
1 construct a zero-investment portfolio with β1 = 0 and β2 = 1
2
e ]
the expected return on this portfolio is: E [R2t

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 45 / 54

 Recovering Risk Prices

The expected return on asset i is:

E [Rite ] λ0 + βi 1 λ1 + βi 2 λ2 ,

1 zero aggregate risk portfolio

1 construct a portfolio with β1 = 0 and β2 = 0
2 the expected excess return on this portfolio is λ0 = 0

2 factor-1-mimicking portfolio
1 construct a zero-investment portfolio with β1 = 1 and β2 = 0
2
e ]
the expected (excess) return on this portfolio is the risk price λ1 = E [R1t

3 factor-2-mimicking portfolio
1 construct a zero-investment portfolio with β1 = 0 and β2 = 1
2
e ]
the expected return on this portfolio is the risk price λ2 = E [R2t

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 46 / 54

 APT Pricing for Asset B
Consider another asset B with factor loadings given below:
e
RBt constant + .5 f1t + .5 f2t

construct a portfolio A using the portfolios that ‘mimick’the factors:

1 invest .5 in factor-1-mimicking portfolio
2 invest .5 in factor-2-mimicking portfolio
3 invest 1 .5 .5 in risk-free asset

then the expected return on this portfolio A is:

E [RAt ] Rft + .5λ1 + .5λ2

hence the expected return on asset B should be equal to:

E [RBt ] Rft + .5λ1 + .5λ2

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 47 / 54

 APT Pricing for Asset D
Consider another asset D. How do we price this asset?
e
RDt constant + .75 f1t + .75 f2t

construct a portfolio C :
1 invest .75 in factor-1-mimicking portfolio
2 invest .75 in factor-2-mimicking portfolio
3 invest 1 .75 .75 in risk-free

then the expected return on this portfolio C is:

E [RCt ] Rft + .75λ1 + .75λ2

hence the expected return on asset D should be equal to:

E [RDt ] Rft + .75λ1 + .75λ2

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 48 / 54

 APT pricing with two factors
we assume the data are generated by a two-factor model:

Rite = αi + βi 1 f1t + βi 2 f2t + εit , i = 1, . . . , N

in a well-diversi…ed portfolio, the residual risk disappears

only systematic risk is left

the quantity of risk is determined by the loadings βi 1 and βi 2

all investments must be on a plane in the space of (E [Rit ] , βi 1 , βi 2 )

hence, we can …nd the price of risk for each factor (λ1 , λ2 ) such that:

E [Rite ] = βi 1 λ1 + βi 2 λ2 , i = 1, . . . , N

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 49 / 54

 APT in Equilibrium

Pricing in equilibrium: There exist risk prices λj for each factor such that the
expected return on any security i can be stated as:

E [Rite ] = βi 1 λ1 + βi 2 λ2 + . . . + βiK λK , i = 1, . . . , N

very general

relative pricing: price one asset relative to others

no need to measure the return on the total wealth portfolio (or ‘the market’)

the theory does not actually tell you which factors to use!
I for a larger set of securities, you probably need more factors

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 50 / 54

 Testing the model: Robust test (Optional Material)

White covariance matrix of residuals allows for non-normal errors terms with
time-varying volatility

Asymptotic test

A little cumbersome in terms of notation

Have to set up the system as a big panel regression

Here goes...

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 51 / 54

 Testing the model: Robust test

rt = [r1t r2t ... rNt ]0 , εt = [ε1t ε2t ... εNt ]0

N 1 N 1
2 3
1 0 0
6 Rm,t
e 0 0 7
6 7
6 0 1 0 7
6 7
6 0 R e 0 7
xt = 6 m,t 7,
(N K ) N 6 .. .. . . .
. 7
6 . . . . 7
6 7
4 0 0 1 5
0 0 e
Rm,t
β = [ α1 β1 α2 β2 αN βN ] 0 ,

where K = 2 and the β vector is obtained by running the N univariate regressions.

Then, f ( β) = xt εt is a (N K) 1 vector.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 52 / 54

 Robust test (cont’d)

Referring back to asymptotics slides in Lecture 2 and the Asymptotics Note on

CCLE, let the sample mean of the moment condition be:
T
1
gT β̂ =
T ∑ ft β̂ = 0
t =1

From the Central Limit Theorem:

p
T gT β̂ N (0, ST )

where
T T
1 0 1
ST =
T ∑ ft β̂ ft β̂ =
T ∑ xt ε̂t ε̂t0 xt0
t =1 t =1
is an (N K) (N K ) matrix.

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 53 / 54

 Robust test (cont’d)
Then, using the Law of Large numbers and Central Limit Theorem, we have
^
β β N (0, Σ ( β))
1 1
where Σ ( β) = T1 ET [xt xt0 ] ST ET [xt xt0 ] is the (N K) (N K)
covariance matrix of β.

Remember now that ^ β is is an (N K ) 1 vector. To test the α’s, we are only

interested in the odd rows and columns of the full variance-covariance matrix,
Σ ( β ).

Call the N N covariance matrix of the α’s that result from extracting the odd
rows and columns from Σ ( β) for Σα,White . Then the asymptotic test is
1
α̂0 Σα,White α̂ χ2 (N )

Lars A. Lochstoer UCLA Anderson School of Management () Winter 2025 54 / 54