Artem Tsvetkov & Lammertjan Dam

FTM

Week 3: Luenberger, Chapters 7 and 8

Luenberger 7.6 (Simpleland) In Simpleland there are only two risky stocks A and B,
whose details are listed in Table

Details of Stocks A and B

Number of shares Price Expected Standard deviation

outstanding per share rate of return of return
Stock A 100 $1.50 15% 15%
Stock B 150 $2.00 12% 9%
Furthermore, the correlation coefficient between the returns of stocks A and B is ρ AB = 31 .
There is also a risk-free asset, and Simpleland satisfies the CAPM exactly
(a) What is the expected rate of return of the market portfolio?
(b) What is the standard deviation of the market portfolio?
(c) What is the beta of stock A?
(d) What is the risk-free rate in Simpleland?
Solution:
(a) Let us compute the weights in the market portfolio

Number of shares Price Market Weights in

outstanding per share capitalization market portfolio
Stock A 100 $1.50 $150 1/3
Stock B 150 $2.00 $300 2/3
Total $450

The expected rate of return on the market portfolio is thus

1 2
r M = 15% + 12% = 13%.
3 3

(b) The variance is

2
σM = w2A σA2 + 2w A w B ρ AB σA σB + w2B σB2 = 81.
σM = 9%.

1
(c)
1 2 2
σAM 3 σA + 3 ρ AB σA σB 35
βA = 2
= 2
= .
σM σM 27

(d) Simpleland satisfies the CAPM exactly, thus from the security market line equation

r A − r f = β A (r M − r f )

we get
β ArM − rA 1
rf = = 6 %.
β−1 4
■

2
Exam 1.2 (Mean variance and CAPM) Suppose the investment universe consists of only
three assets, with rates of return r1 , r2 , and r3 , respectively. The covariance matrix, V, and
vector of expected rates of return E[r ] are:
   
0.2 0 0 0.13
V= 0 0.25 −0.15 , E[r ] = 0.04
0 −0.15 0.25 0.04

There is a risk-free asset, and the risk-free rate is 0.01 per year.
(a) Find the portfolio weights of the mean-variance efficient portfolio of risky assets, in
terms of portfolio weights of asset 1, 2, and 3. (8 points)
Let us consider a more general market (so this question no longer relates to the specific
numbers given in question 1). Assume that all the assumptions underlying the CAPM
hold and that a risk-free asset exist with rate r f and denote the return on the mean-
variance efficient (market) portfolio with r M . Consider a portfolio A with a rate of return
rA.
(b) Use the equation for the security market line (SML) to proof that if returns of portfolio
A are perfectly correlated with the returns on the market portfolio, i.e. corr(r A , r M ) =
1, it implies that portfolio A is on the capital market line (CML). (6 points).
(c) The CAPM says that for the cross-section of returns it should hold that E[ri − r f ] =
β i E[r M − r f ], for all assets i. Gavin Jones argues that the CAPM theory has no con-
tent, since if we know E[ri − r f ] and E[r M − r f ] – for example by estimating these
quantities based on series of returns – we can solve this equation for β i so it will al-
ways hold and thus the CAPM will always be true. Is he right? Why or why not? (4
points).
Solution:
(a) The weights of risky assets in the mean-variance efficient portfolio are

⃗v = V −1 · ⃗r − r f .

With given values for the covariance matrix V, expected returns ⃗r, and risk-free rate
r f , the weights are
 
0.06
⃗v = 0.03 ,
0.03

and after normalizing  

0.5
⃗v
⃗ =
w = 0.25 .
∑ vi 0.25

3
(b) The Capital Market Line (CML) is given in Luenberger in equation 7.1 and the Secu-
rity Market Line (SML) in equation 7.2

E rM − r f
 
E ri − r f = σi
 
CML:
σ
 M
E ri − r f = β i E r M − r f
  
SML:

Since
cov (ri , r M ) σi
βi = 2
= corr ( r i , r M ) ,
σM σM
we see that the two equations are the same when corr (ri , r M ) = 1.
(c) This is wrong. β i is not a free parameter. It is a correlation coefficient based on the
relation between ri and r M .
■

4
Luenberger 7.7 (Zero-beta assets) Let w0 be the portfolio (weights) of risky assets corre-
sponding the minimum-variance point in the feasible region. Let w1 be any other portfo-
lio on the efficient frontier. Define r0 and r1 to be the corresponding returns.
(a) There is a formula of the form σ01 = Aσ02 . Find A. [Hint: Consider the portfolios
(1 − α)w0 + αw1 and consider small variations of the variance of such portfolios near
α = 0.]
(b) Corresponding to the portfolio w1 there is a portfolio wz on the minimum-variance
set that has zero beta with respect to w1 ; that is, σ1z = 0. This portfolio can be
expressed as wz = (1 − α)w0 + αw1 . Find the proper value of α.
(c) Show the relation of the three portfolios on a diagram that includes the feasible re-
gion.
(d) If there is no risk-free asset, it can be shown that other assets can be priced according
to the formula
ri − r z = β iM (r M − r z )
where the subscript M denotes the market portfolio and r z is the expected rate of
return on the portfolio that has zero beta with the market portfolio. Suppose that
the expected returns on the market and the zero-beta portfolio are 15% and 9%, re-
spectively. Suppose that a stock i has a correlation coefficient with the market of 0.5.
Assume also that the standard deviation of the returns of the market and stock i are
15% and 5%, respectively. Find the expected return of stock i.
Solution:
(a) Construct an efficient portfolio with weights x

x = (1 − α)w0 + αw1 .

Its variance is
σx2 = (1 − α)2 σ02 + 2α(1 − α)σ01 + α2 σ12 .
As α → 0 approaches zero, we should have

σx2 → σ02
α →0

and we use the fact that w0 has the minimum variance

∂σx2
→ 0.
∂α α→0
This is only possible if
σ02 = σ01 .
It follows A = 1.

5
(b) Portfolio wz is a combination of w0 and w1
wz = (1 − α)w0 + αw1 .
Covariance of the return of this portfolio with portfolio w1 is
σ1z = (1 − α)σ01 + ασ12 = 0.
Solving this equation we get
σ02
α= .
σ02 − σ12
(c) Since α < 0 and assuming r1 > r0 , we get
r z = (1 − α)r0 + αr1 = r0 + α(r1 − r0 ) < r0 .
So the point is below r0 .

(d) Since
ρσi
β iM = ,
σM
we get
0.5 · 0.05
ri = r z + β iM (r M − r z ) = 9% + · (15% − 9%) = 10%.
0.15
Even though this calculation answers question (d) in the exercise, it may be beneficial
to understand how the formula from this question is derived. Let us consider it step
by step.
(i) Remember how the CAPM theorem is proved in section 7.3 of the book. First, a
portfolio is constructed from the asset in question and the market portfolio. The
mean return and the standard deviation of this portfolio are easily obtained as
a function of the mean returns and the covariance matrix for the asset and the
market portfolio. If a relation between the rate and the standard deviation of
this portfolio can be found for some α, it can be used to identify ri as a function
of r M . For that Luenberger uses the fact that the market portfolio is located on
the efficient frontier. When α → 0, the portfolio approaches the market port-
folio on the mean-variance diagram. The line that the portfolio draws cannot
approach the efficient frontier at an angle. If it would be the case, the line would
overshoot the efficient frontier for negative α. This contradicts the definition
of the efficient frontier as no portfolio can be located beyond it. Thus, this line
should be tangent to the frontier. The tangent of the line is
dr α (ri − r M )σM
= 2
,
dσα α =0 σiM − σM
according to the CAPM theorem proof.

6
(ii) Let us construct a portfolio consisting of the minimum-variance portfolio w0
and portfolio w1 , as wx = αw0 + (1 − α)w1 . Since both portfolios are located on
the efficient frontier, their combination also belongs to it according to the two-
fund theorem. Following the very same reasoning as in the CAPM theorem we
get the tangent to the efficient frontier at α = 0 as

dr x r1 − r0
= σ1 .
dσx α =0 σ12 − σ02

If we draw the tangent line from point (r1 , σ1 ), it crosses the y axis at

r1 − r0 2 σ12 σ02
rY = r 1 − σ = r 0 − r1 .
σ12 − σ02 1 σ12 − σ02 σ12 − σ02

So, we can write the tangent as

dr x r1 − r0 r 1 − rY
= σ1 = .
dσx α =0 σ12 − σ02 σ1

(iii) In question (b), we found that the combination of minimum-variance portfolio

and an arbitrary portfolio w1 on the effective frontier produces a portfolio wz
with zero covariance to w1 when

σ02
α= .
σ02 − σ12

Note though that α in question (b) is defined differently than in step 1 and 2.
This value of α gives us the following mean return of this portfolio

σ12 σ02
rz = 2 r0 − 2 r1 .
σ1 − σ02 σ1 − σ02

Note that r z = rY . This allows us to express the tangent at (r1 , σ1 ) as

dr x r1 − r z
= .
dσx α =0 σ1

(iv) Since w1 is any portfolio on the efficient frontier we can set it to the market port-
folio, w1 = w M . Combining the tangent from step 3 with the tangent obtained
in step 1, we get
(ri − r M )σM r − rz
2
= M ,
σiM − σM σM
from which the formula in question (d) follows.
■

7
Luenberger 7.8 (Wizards ⋄) Electron Wizards, Inc. (EWI) has a new idea for producing
TV sets, and it is planning to enter the development stage. Once the product is developed
(which will be at the end of 1 year), the company expects to sell its new process for a price
p, with expected value p = $24M. However, this sale price will depend on the market
for TV sets at the time. By examining the stock histories of various TV companies, it
is determined that the final sales price p is correlated with the market return as E[( p −
p)(r M − r M ) = $20MσM 2 .

To develop the process, EWI must invest in a research and development project. The
cost c of this project will be known shortly after the project is begun (when a technical
uncertainty will be resolved). The current estimate is that the cost will be either c = $20M
or c = $16M, and each of these is equally likely. (This uncertainty is uncorrelated with
the final price and is also uncorrelated with the market.) Assume that the risk-free rate is
= 9% and the expected return on the market is r M = 33%.
(a) What is the expected rate of return of this project?
(b) What is the beta of this project? [Hint: In this case, note that

p−p
     
1
E (r M − r M ) = E E [( p − p) (r M − r M )] .
c c

(c) Is this an acceptable project based on a CAPM criterion? In particular, what is the
excess rate of return (+ or −) above the return predicted by the CAPM?
Solution:
(a) The stochastic return is R = p/c. Using independence of p and c, the expected rate
of return
h pi    
1 0.5 0.5 9
r = R−1 = E −1 = E p−1 = + $24 − 1 = $24 − 1 = 35%.
c c $16 $20 $160

(b) The covariance between the rate of return of the project and the market is
h p h p i i
σpM = E [(r − r )(r M − r M )] = E ( R − R)(r M − r M ) = E −E
 
(r M − r M )
  c c
1 9 9
=E E [( p − p) (r M − r M )] = 2
$20σM = σM 2
c $160 8
And
σpM 9
β pM = 2
= .
σM 8
(c)
9
r p = r f + β pM (r M − r f ) = 9% + (33% − 9%) = 36%.
8
The excess return is -1%.
■

8
Exam 2.1 (Data and APT)
(a) In considering various models and data, Luenberger discusses the problem of the
“Mean Blur”. Explain what is meant with this Mean Blur. Incorporate in your answer
what role period length (or frequency of the data) plays. (5 points)
Two stocks are believed to satisfy the two-factor APT model:

r1 = a1 + f 1 + 2 f 2 ,
r2 = a2 − f 1 + 2 f 2 .

Note that there is no idiosyncratic risk, (i.e. there are no error terms). There is also a risk-
free asset with a rate of return r f = 5% . It is known that E[r1 ] = 13% and E[r2 ] = 9%.
(b) What are the values for λ0 , λ1 , and λ2 for this model? (6 points)
Suppose there is a third asset, for which the returns can be modeled as:

r3 = a3 + 4 f 2 .

This asset has an expected return of E[r3 ] = 17%.

(c) Is there an arbitrage opportunity? If so how would you exploit it? (3 points)
(d) Suppose for a general two-factor model the expected return of the first factor is
E[ f 1 ] = 0. Gavin Jones argues that since this factor does not add to the expecta-
tion of the returns, the risk of this factor is not priced, and so the factor price should
be equal to zero. Is he right? Why or why not? (4 points)
Solution:
(b) According to the APT, we need:

E[r1 ] = λ0 + λ1 + 2λ2 = 13%

E[r2 ] = λ0 − λ1 + 2λ2 = 9%
E[r f ] = λ0 = 5%

Solving yields λ0 = 5%, λ1 = 2%, and λ2 = 3%.

(c) No, the factor structure implies that the expected return should be

λ0 + 4λ2 = 5% + 4(3%) = 17%,

which it is.
Alternatively you can show that a portfolio consisting of 100% long in factor 1 and 2
and shorting the risk-free rate 100% generates an asset with the same factor structure,
and also yields the same expected return.
(d) He is wrong. The expected value of the factor does not tell us anything about whether
or not the factor is priced. We need to calculate λ0 , λ1 , and λ2 , which depend on the

9
expected returns of the assets. Suppose a factor is priced. If we would demean the
factors (i.e. subtract the mean from the factor), its expectation is by construction equal
to zero. But in this case the ai ’s will go up in estimates of our factor model for any
asset i that is sensitive to the factors.
Mathematically, suppose factor f 1 has nonzero expectation.

r1 = a1 + b11 f 1 + b12 f 2

is equivalent to:

r1 = ( a1 + b11 E[ f 1 ]) + b11 ( f 1 − E[ f 1 ]) + b12 f 2

If we define our new factor

g1 = ( f 1 − E[ f 1 ]),
and our new intercept as
a1′ = ( a1 + b11 E[ f 1 ]),
we have
r1 = a1′ + b11 g1 + b12 f 2 .
The factor g1 has zero expectation by definition, yet the model is completely equiva-
lent to the factor model with factor f 1 , which has nonzero expectation.
■

10
Luenberger 8.2 (APT factors) Two stocks are believed to satisfy the two-factor model

r1 = a1 + 2 f 1 + f 2
r2 = a2 + 3 f 1 + 4 f 2

In addition, there is a risk-free asset with a rate of return of 10%. It is known that r1 = 15%
and r2 = 20%. What are the values of λ0 , λ1 , and λ2 for this model?
Solution:
According to APT, there are three constants λ0 , λ1 , λ2 s.t.

r1 = λ0 + 2λ1 + λ2
r2 = λ0 + 3λ1 + 4λ2

Setting λ0 = r f = 10% we have

(
15% = 10% + 2λ1 + λ2
20% = 10% + 3λ1 + 4λ2

Solving it, we get λ0 = 10%, λ1 = 2%, and λ2 = 1%. ■

11
Luenberger 8.4 (Variance estimate) Let ri , for i = 1, n, be independent samples of a
return r of mean r and variance σ2 . Define the estimates
1 n
n i∑
r̂ = ri
=1
n
1
s2 = ∑
n − 1 i =1
(ri − r̂ )2 .

Show that E[s2 ] = σ2 .

Solution: Using the definition of the estimate for the variance and mean rate, we come
to the following relation:
n
1
s2 = ∑
n − 1 i =1
(ri − r̂ )2
n 
1 
n − 1 i∑
2 2
= r i − 2r̂ × r i + r̂
=1
!
n n n
1
=
n−1 ∑ ri2 − 2r̂ × ∑ ri + ∑ r̂2
i =1 i =1 i =1
!
n
1
=
n−1 ∑ ri2 − 2r̂ × nr̂ + nr̂2
i =1
!
n
1
=
n−1 ∑ ri2 − nr̂2
i =1
 !2 
n n
1  1
= ∑
n − 1 i =1
ri2 −
n ∑ ri 
j =1

Returns ri are iid random variables with mean µ = E[ri ] and variance σ2 = E [ri − E[ri ]]2 .
The following relation is used throughout.
h i
E [ x − E [ x ]] = E x2 − (E[ x ])2 .
2

12
We compute the expectation of the variance estimate and show that it is equal to σ2 .
h i
E s2 =
" #
n
1
n − 1 i∑
=E (ri − r̂ )2
=1
  !2  
n n
1  1
= E
n − 1 i =1∑ ri2 −
n j∑
ri  
=1
 " #2 
n n
1  h i 1
= ∑
n − 1 i =1
E ri2 − E ∑ ri 
n j =1
  " " ##2 " #!2 
n n n n
1  1
= ∑
n − 1 i =1
E [ri − E[ri ]]2 + (E[ri ])2 − E ∑ ri − E ∑ ri
n
+ E ∑ ri 
j =1 j =1 j =1
  " # " #!2 
1  n 2 1 n n

n − 1 i∑ ∑ i ∑ ri  
2
= ( σ + µ ) − var r + E
=1
n j =1 j =1
 
1 1  
= nσ2 + nµ2 − nσ2 + n2 µ2 = σ2 .
n−1 n

This proves that the estimation is unbiased. ■

13