Financial Markets and Investments

Exercises Suggested Solutions

Raquel M. Gaspar

December 2020

These are suggested solutions to the exercises in the Booklet of Exercises. They are not
typo and/or error free.
I would very much appreciate if you would write down a list of typos and/or errors identified
during study and hand it in at the exam, for future correction. The same applies to typos in
the slides or any other material I have distributed during the course.

Contents
1 Mean–Variance Theory 2
1.1 Return and Diversification of Risk . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Investment Opportunity Sets and Efficient Frontiers . . . . . . . . . . . . . . . . 5
1.3 Portfolio Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4 International Diversification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2 Portfolio Selection Models 23

2.1 Constant Correlation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Single-Index Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Multi-Index Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3 Selecting the Optimal Portfolio 42

3.1 Expected Utility Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2 Alternatives Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4 Equilibrium in Financial Markets 66

4.1 CAPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2 APT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5 Portfolio Management 80

6 Miscellaneous 82

1
1 Mean–Variance Theory

1.1 Return and Diversification of Risk

Exercise 1.1.

(a) Expected return is the sum of each outcome times its associated probability. Expected
returns:
R̄1 = 16% × 0.25 + 12% × 0.5 + 8% × 0.25 = 12%
R̄2 = 6%
R̄3 = 14%
R̄4 = 12%
Standard deviation of return is the square root of the sum of the squares of each outcome
minus the mean times the associated probability. Standard deviations:
h i 12
2 2 2
σ1 = (16% − 12%) × 0.25 + (12% − 12%) × 0.5 + (8% − 12%) × 0.25 = 2.83%
σ2 = 1.41%
σ3 = 4.24%
σ4 = 3.27%
(b) Covariance of return between Assets 1 and 2
σ12 = (16 − 12) × (4 − 6) × 0.25 + (12 − 12) × (6 − 6) × 0.5 + (8 − 12) × (8 − 6) × 0.25 = −4
The variance/covariance matrix for all pairs of assets is:
 
0.0008 -0.0004 0.0012 0
 -0.0004 0.0002 -0.0006 0 
V =
 0.0012

-0.0006 0.0018 0 
0 0 0 0.00107

−4
Correlation of return between Assets 1 and 2: ρ12 = 2.83×1.41 = −1.
The correlation matrix for all pairs of assets is:
 
1 -1 1 0
 -1 1 -1 0 
ρ=
 1

-1 1 0 
0 0 0 1

(c)

Portfolio Expected Return

A 1/2 × 12% + 1/2 × 6% = 9%
B 13%
C 12%
D 10%
E 13%
F 1/3 × 12% + 1/3 × 6% + 1/3 × 14% = 10.67%
G 10.67%
H 12.67%
I 1/4 × 12% + 1/4 × 6% + 1/4 × 14% + 1/4 × 12% = 11%

2
 Portfolio Variance
A (1/2)2 × 0.0008 + (1/2)2 × 0.0002 + 2 × 1/2 × 1/2 × (−0.0004) = 0.00005
B 0.00125
C 0.00046
D 0.0002
E 0.0007
F (1/3)2 × 0.0008 + (1/3)2 × 0.0002 + (1/3)2 × 0.0018 + 2 × 1/3 × 1/3 × (−0.0004)+
+2 × 1/3 × 1/3 × 0.0012 + 2 × 1/3 × 1/3 × (−0.0006) = 0.00036
G 0.0002
H 0.00067
I (1/4)2 × 0.0008 + (1/4)2 × 0.0002 + (1/4)2 × 0.0018 + (1/4)2 × 0.00107 × +
+2 × 1/4 × 1/4 × (−0.0004) + 2 × 1/4 × 1/4 × 12 + 2 × 1/4 × 1/4 × 0+
+2 × 1/4 × 1/4 × (−0.0006) + 2 × 1/4 × 1/4 × 0 + 2 × 1/4 × 1/4 × 0 = 0.00027

(d)-(e)

We can conclude that as-

sets A, B and D are nor
efficient, as well as portfo-
lios b and f . In all these
cases we can find portfolios
with lower or equal risk and
higher or equal expected re-
turn.

Figure 1: Exercise 1.1 – Representation of the assets and several portfolios in the space (σp , R̄p ).

Exercise 1.2.

(a) The formula for the variance of an equally weighted portfolio (where Xi = 1/N ∀i =
1, . . . , N securities) is
" N  #  
X σ2 N N
2 1 i N − 1 X X σ ij  = 1 σi2 − σij + σij (1)
 
σH = + 
N i=1 N N i=1 j=1,j6=i
N (N − 1) N
| {z } | {z }
σi2 σij

where σi2 is the average variance across all securities, σij is the average covariance across
all pairs of securities, and N is the number of securities. Using the above formula with
σi2 = 0.005 and σij = 0.001 we have:

N 5 10 20 50 100
2
σH 0.0018 0.0014 0.0012 0.00108 0.00104

(b) As the number of securities (N) approaches infinity, an equally weighted portfolio’s vari-
ance (total risk) approaches a minimum equal to the average covariance of the pairs of

3
 √
securities in the portfolio, which is 10. Therefore the risk is σM V = 0.001 = 3.16%. Hav-
ing a risk only 10% higher than the minimum variance portfolio means σH ≤ 3.16 × 1.1 =
2
3.48% ⇐⇒ σH = 0.00121. To know how many securities a portfolio must have to respect
this condition we need to solve the inequality:

2 1  2 
σH = σi − σij + σij ≤ 0.001211
N
1
(0.005 − 0.001) ≤ 0.001211 ⇔ N ≥ 19.05
N
Thus, the portfolio must have, at least, 20 securities.
(c) No, the average covariance works as an asymptote to the variance of any portfolio. As N
increases the variance of a portfolio converges to that limit, but would only reach it at
infinity.

Exercise 1.3.

(a) If the portfolio contains only one security, then the portfolio’s average variance is equal
to the average variance across all securities, σ̄j2 . If instead an equally weighted portfolio
contains a very large number of securities, then its variance will be approximately equal
to the average covariance of all pairs of securities in the portfolio σ̄kj . Therefore, the
fraction of risk that of an individual security that can be eliminated by holding a large
portfolio is expressed by the following ratio:

σi2 − σij
D=
σi2

The above ratio is equal to 0.6(60%) for Italian securities and 0.8(80%) for Belgian secu-
rities.
(b) Setting the above ratio equal to those values and solving for σij gives σij = 0.4σi2 for
Italian securities and σij = 0.2σi2 for Belgian securities.
If the average variance of a single security, σi2 , in each country equals 0.005, then σij =
0.4σi2 = 0.4 × 0.0050 = 0.002 for Italian securities and σij = 0.2σi2 = 0.2 × 0.005 = 0.001
for Belgian securities. Using Equation (??) with σi2 = 0.005 and either σij = 0.002 for
Italy or σij = 0.001 for Belgium we have:

2 2
Portfolio Size (N securities) Italian σH Belgian σH
5 0.0026 0.0018
20 0.00215 0.0012
100 0.00203 0.00104

Exercise 1.4.

(a) The diversification ratio measures, in percentage, how much of the average asset variance
can be diversified away by building portfolios.
In this case we have

σi2 − σij 0.0046619 − 0.0007058

D= = = 84.86%.
σi2 0.0046619

4
 (b) The formula for an equally weighted portfolio’s variance is

2 1  2 
σH = σi − σij + σij
N

where σi2 is the average variance across all securities, σij is the average covariance across
all securities, and N is the number of securities. The average variance for the securities in
the table is 0.0046619 and the average covariance is 0.0007058. We want the volatility to
2
be lower than 2.83%, i.e. the variance σH ≤ 0.02832 = 0.0008. Using the above equation
and solving for N gives:
1
0.0008 ≥ (0.0046619 − 0.0007058) + 0.0007058
N
0.0000942N ≥ 0.0039561
N ≥ 41.997

Since the portfolio’s variance decreases as N increases, holding 42 securities will provide
a variance less than 0.0008, so 42 is the minimum number of securities required.

1.2 Investment Opportunity Sets and Efficient Frontiers

Exercise 1.5.

(a) We know that σA = 9% and σB = 15%. We also know that securities A and B are
combined in order to override the portfolio risk, which is only possible when ρ = −1.
Therefore, the weight of each asset in portfolio of zero risk is given the equation system
( (
xA + xB = 1 xB = 1 − xA
2 2 2 2
⇔ 2
2 2 2
σA xA + σB xB + 2xA xB σAB = 0 σA xA + σB (1 − xA ) + 2xA (1 − xA ) σAB = 0

xB = 1 − xA
(

2 ⇔
0.0081x2A + 0.225 1 − 2xA + x2A + 2xA (1 − xA ) (−0.0135) = 0

xA = 0.625 xB = 1 − 0.625 = 0.375

Therefore, xA = 62.5% and xB = 37.5%.

(b) If the null risk portfolio has a return of 7.5%, we know its composition is

7.5% = 5%xA + R̄B (1 − xA )

and from (a) we also know that xA = 62.5%. Thus,

0.075 − 0.05 × 0.625
R̄B = = 11.67%
0.375

(c) The statement is TRUE. Asset B is the one with the highest expected return and risk.
From above we see the zero-risk portfolio requires a positive investment in asset B (of
37.5%). Any portfolio with lower weight in B has a negative Sharpe ratio (slope in mean-
variance space). Thus, short selling of asset B to invest more than 100% in asset A is also
necessarily inefficient.

5
 Given the perfect negative cor-
relation, we know that geometri-
cally the investment opportunity
set (IOS) is defined by two seg-
ments of lines each passing by
each of the two risky securities
and with a common y-cross at the
zero risk portfolio. When short-
selling is forbidden the drashed
portions are not feasible.

Figure 2: Exercise 1.6 – two risky assets ρ = −1. IOS with (full + dashed lines) and without
(full) shortselling.

Exercise 1.6.

(a)-(b) From Exercise 1.1 we know R̄1 = 12%, R̄2 = 6%, σ1 = 2.83%, σ2 = 1.41% and ρ12 = −1.
We can get the IOS analytical expression to the equations by: (i) first finding the expected
return of the combination with zero risk, and then (ii) using the basic assets 1 and 2 to
find the slopes of the two lines.
(i) The minimum variance portfolio is the one without risk, σp = 0. Analytically,
2
0 = σ12 x21 + σ22 (1 − x1 ) + 2x1 (1 − x1 ) σ12
√
σp + σ2 σ2 2 1 2
x1 = = =√ √ = ⇒ x2 =
σ1 + σ2 σ1 + σ2 8+ 2 3 3

Thus, the portfolio has 33.33% of security 1 and 66.67% of security 2. The expected
return is X 1 2
R̄M V = xi R̄i = × 12% + × 6% = 8%
3 3
R̄1 − 8% R̄2 − 8%
(ii) The slopes of the two lines are given by = 1.41 and = −1.41,
σ1 σ2
respectively.
(
8% + 1.41σp σp ≤ 2, 83%
So, the IOS is given by R̄p = .
8% − 1.41σp σp ≤ 1.41%

(c) All portfolios in the segment line with positive slope dominate those in the negative slope
segment line, since risk averse investors will prefer from a set of two portfolios with the
same risk, the one with highest return. Therefore, the efficient frontier in the positive
slope segment line, i.e. R̄p = 8% + 1.41σp for σp ≤ 2.83%.
(d) If shortselling is allowed the derivations in (a)-(b) still stand, the only different is that
in the representation of the IOS the entire lines should be considered. I.e. in the above
figure the dashed segments would also be feasible.
The efficient set would, thus, be represented by the entire upper line.

IOS: R̄p = 8% ± 1.41σp and EF: R̄p = 8% + 1.41σp .

6
 In particular, all combinations of 1 and 2 that require shortselling of asset 2 to invest
more than 100% in 1 are efficient.

Exercise 1.7.

(a) We start by determining expected returns, variances and covariances of the two assets.
1
R̄1 = E (R1 ) = × (0.2 + 0.14 + 0.08) = 14%
3
1
R̄2 = E (R2 ) = × (0.16 + 0.12 + 0.08) = 12%
3
h
2 i 1 h 2 2 2
i
σ12 = E R1t − R̄1 = × (0.2 − 0.14) + (0.14 − 0.14) + (0.08 − 0.14) = 0.0024
√ 3
σ1 = 0.0024 = 4.90%
h
2 i 1 h 2 2 2
i
σ22 = E R2t − R̄2 = × (0.16 − 0.12) + (0.12 − 0.12) + (0.08 − 0.12) = 0.001067
√ 3
σ2 = 0.001067 = 3.236%


 1
σ12 = E R1t − R̄1 R2t − R̄2 = × [(0.2 − 0.14) (0.16 − 0.12) + (0.08 − 0.14) (0.08 − 0.12)]
3
= 0.0016
σ12 0.0016
ρ12 = =√ √ = +1
σ1 σ2 0.0024 0.001067
Thus, the returns of the two securities are perfectly positively correlated, thus, the in-
vestment opportunity set (IOS), when shortselling is allowed, is given by two lines: one
connecting the two risky securities, and the line with symmetric slope.

R̄1 − R̄2 R̄1 − R̄2

IOS (i) : R̄p = R̄2 − σ2 + σp
σ1 − σ2 σ −σ
| {z } | 1 {z 2}
y-cross slope
0.14 − 0.12 0.14 − 0.12
= 0.12 − 0.03236 + σp
0.049 − 0.03236 0.049 − 0.03236
= 0.08 + 1.2247σp
and
R̄p = 0.08 − 1.2247σp .

Although the negative slope line is not efficient it still belongs to the IOS.
When shortselling is not allowed, the IOS is only the segment of the line that passes by
the two risky assets

IOS (ii) : R̄p = 0.08 + 1.2247σp for 1.41% ≤ σp ≤ 3.236%

(b) The minimum variance portfolio, when shortselling is forbidden – scenario (ii) – involves
placing all funds in the lower risk security (asset 2). Consequently, the expected return
is R̄M V = R̄2 = 12% and risk is σM V = σ2 = 3.236%.
If short sales were allowed – scenario (i) – than σp = 0 and R̄p = 8%. Moreover, the
weights of the MV portfolio is,
σp − σ2
x1 = = −200% ⇒ x2 = 1 − x1 = 1 − (−2) = 300%
σ1 − σ2

7
 (c) As known, the efficient frontier is the investment opportunity set and investor are consid-
ered to be risk averse. For the two scenarios we have:
EF (i): R̄p = 8% + 1.2247σp
EF (ii) = IOS (ii) : R̄p = 0.08 + 1.2247σp for 1.41% ≤ σp ≤ 3.236%
(d) If we have a riskless asset with Rf = 10% then The investment opportunity set becomes:
– IOS (i): When shortselling is allowed
without any bound, the theoretical answer
would be the entire space σp , R̄p .
In a real life situation, there will be an extreme combination, E, where one takes the
highest possible shortselling position in asset 2. In that case the IOS would be the
R̄E − 0.1
entire area bellow the straight line R̄p = 0.1 + σp .
σE
– IOS (ii): When shortselling is not allowed the IOS is the cone limited by the lines
0.14 − 0.1
R̄p = 0.1 ± σp .
0.049
The efficient frontier (EF) becomes:
– When shortselling is allowed – scenario (i) – the efficient frontier would be the straight
line that has y-cross at 10% and has the highest possible slope.
In a real life situation ,where eventually there would be a limit to how much one can
shortsell of asset 2, it would be combinations of the riskless asset with the portfolio
with that extreme, E portfolio,
R̄E − 0.1
EF (i): R̄p = 0.1 + σp .
σE
– When shortseliing is not allowed – scenario (ii) the efficient frontier would be given
by combinations of the riskless asset with asset 1,
0.14 − 0.1
EF (ii): R̄p = 0.1 + σp .
0.049

Figure 3: Exercise 1.7 – Two perfectly correlated assets. Efficient frontier with (full + drashed)
and without (full) shortselling.

Exercise 1.8.

(a) Similar to (b) but with ρ = −1 (see slides).

(b) As discussed in Exercise ??, the investments opportunity set generated by two assets with
perfect negative correlation is given by two line segments. An alternative to the solution
presented there is to deduce directly the equation(s) R̄p = f (σp ).

8
 Starting with some transformation in σp equation:
q
σp = x21 σ12 + x22 σ22 − 2x1 x2 σ1 σ2
q
2
= (x1 σ1 − x2 σ2 )
= ± |x1 σ1 − x2 σ2 |
= ± |x1 σ1 − (1 − x1 ) σ2 |

With additional transformations, we get an equation to x1

σp + σ2
σp ± σ2 = ±x1 (σ1 − σ2 ) ⇔ x1 = ±
σ1 + σ2
Replacing x1 in the expected return equation for a two assets portfolio, we get

R̄p = x1 R̄1 + (1 − x1 ) R̄2

 
σp + σ2 σp + σ2
=± R̄1 + 1 ± R̄2
σ1 + σ2 σ1 + σ2
±σp R̄1 + σ2 R̄1 ± σp R̄2 − σ2 R̄2
= R̄2 +
σ1 + σ2
   
R̄1 − R̄2 R̄1 − R̄2
= R̄2 + σ2 ± σp
σ1 − σ2 σ1 − σ2
The first term in the right side in the intersection in the y’s axis and the second term
is the slope, which can be positive or negative, giving origin to the two expected line
segments.
(c) Equal to (d), replacing the generic ρ by 0 (see slides).
(d) Solved in class.
(e) Real life correlations are not perfect, so real life correlations are ρ 6= −1 or ρ 6= +1.
Returns laso tend to be correlate with one another, so ρ 6= 0. All other values may occur,
but for financial assets positive correlations are more common than negative.

Exercise 1.9.

(a) The investment opportunity sets are represented in the Figure ?? below.
(b) – When ρ = +1 , the least risky “combination” of securities 1 and 2 is security 2 held
alone (assuming no short sales). This requires xM
1
V
= 0 and xM
2
V
= 1, where the x0 s
are the investment weights. The standard deviation of this “combination” is equal
to the standard deviation of security 2: σM V = σ2 = 2%.
– When ρ = −1 , we can always find a combination of the two securities that will
completely eliminate risk, and we this combination can be found by solving xM 1
V
=
σ2 MV 2% 2
σ1 +σ2 . So, x1 = 5%+2% = 7 , and since the investment weights must sum to 1,
xM2
V
= 1 − x1 = 1 − 27 = 75 . So a combination of 27 invested in security 1 and 57
invested in security 2 will completely eliminate risk when ρ equals -1, and σM V will
equal 0.
– When ρ = 0 e, the minimum-risk combination of two assets can be found by solving
σ22 4% 4 4
xM
1
V
= σ2 +σ MV
2 . So, x1 = 25%+4% = 29 , and xM
2
V
= 1 − x1 = 1 − 29 = 25
29 . When
1 2
ρ equals 0, the expression for the standard deviation of a two-asset portfolio is
q
2
σp = x21 σ12 + (1 − x1 ) σ22

9
 4
Substituting 29for x1 in the above equation, we have
s 
2  2
4 25
σM V = × 0.0025 + × 0.0004 = 1.86% .
29 29

(c) (i) Both for ρ = −1 nd ρ = 0 the minimum variance portfolios remain the same.
However, for ρ = +1 if shortselling is allowed we can fully eliminate risk. Since, in
the case, we have σp = |x1 σ1 + (1 − x1 )σ2 |, setting σM V = 0, we obtain
0.02
0 = xM
1
V
×0.05+(1−xM
1
V
)0.02 ⇔ xM
1
V
=− = −66.67%, xM
2
V
= 166.67%.
0.03
(ii) When we have ρ = ±1 there is a combination of 1 and 2 that fully eliminates risk,
thus there is a risk-free investment or a “ficticious” riskless asset. The return of the
zero risk combinations give us the appropriate risk-free return Rf .

ρ = −1 : Rf = xM
1
V
R̄1 + xM
2
V
R̄2
2 5
= × 10% + × 4% = 5.71%
7 7
ρ = +1 : Rf = xM
1
V
R̄1 + xM
2
V
R̄2
= −0.6667 × 10% + 1.6667 × 4% = 0% .

In the case of ρ = 0 the minimum risk combination portfolio has positive volatility
sigmaM V = 1.86%, thus there is no risk free investment.

Figure 4: Exercise 1.9 – blue line ρ = −1, red line ρ = 0 and grey line ρ + 1, full lines (no
shortselling), dashed lines (shortselling required).

Exercise 1.10. If the risk-less rate is 10%, then the risk-free asset dominates both risky assets
both in terms of risk and return. It offers as much or higher return than each of the risky assets,
for zero risk. Assuming the investor prefers more to less and is risk averse, the only efficient
investment is 100% investment in the risk-free asset.

10
Exercise 1.11.

(a) When there is a risk-free asset that can be used for both lending and borrowing we know
the efficient frontier is a straight line tangent to the investment opportunity set of risky
assets. Thus, there is only one efficient portfolio made only of risky assets - the so called
tangent portfolio. See Figure ??.
To find this unique efficient portfolio we need to maximize Sharpe’s Ratio of all portfolios
formed with assets A, B and C. From the first order conditions of this maximisation
problem, result the following equation system:

 0.11 − Rf = 0.0004zA + 0.001zB + 0.0004zC


0.14 − Rf = 0.0010zA + 0.0036zB + 0.003zC


0.17 − Rf = 0.0004zA + 0.003zB + 0.0081zC


The Z-vector, for each given value for RF and the unrestricted tangent portfolios are:

Rf = 6% Rf = 8% Rf = 10%
zA 351.0067 185.2348 19.4631
zB -104.3624 -52.6845 -1.0070
zC 34.8993 21.4765 8.0537
xA 124.67% 120.26% 73.42%
xB -37.07% -34.20% -3.80%
xC 12.40% 13.94% 30.38%
Tangent Portfolio
Expected Return 10.63% 10.81% 12.71%
Standard Deviation 1.28% 1.35% 3.20%
Sharpe ratio 3.611 2.081 0.8474
Efficient Frontier R̄p = 0.06 + 3.611σp R̄p = 0.08 + 2.081σp R̄p = 0.1 + 0.8474σp

(b) If there is no credit to invest in risky assets nothing changes in the efficient frontier for
risk levels lower or equal to σT , however for σp > σT the efficient thing to do are the
combinations on the envelop hyperbola.
The hyperbola delimiting the IOS of the risky assets is given by

AR̄p2 − 2B R̄p + C A = 10 V −1 1
σp2 = where B = R̄0 V −1 1
AC − B 2
C = R̄0 V −1 R̄
For our concrete example we get
σp2 = 1.6450R̄p2 − 0.3426R̄p + 0.018 .

The efficient frontier is, thus, given by

(
R̄p = Rf + SRT × σp σp ≤ σT
2 2
σp = 1.6450 R̄p − 0.3426R̄p + 0.018 σp > σT
where in the expression above we should replace for the appropriate values of Rf , SRT
and σT , according to each scenario.
(c) If shortselling is forbidden we know we are not going to invest in asset B, since the optimal
would be to short sell it.
We can solve the problem numerically, imposing xi ≥ 0 to all i = A, B, C, or, in this case
the problem reduces to a two-asset case and find the two-asset tangent portfolios.
Either way, we get

11
 Rf = 6% Rf = 8% Rf = 10%
xA 93.77% 89.61% 68.83%
xC 6.23% 10.39% 31.17%

Tangent Portfolio
Expected Return 11.37% 11.62% 12.87%
Standard Deviation 2.07% 2.2% 3.39%
Sharpe ratio 2.592 1.648 0.8471
Efficient Frontier R̄p = 0.06 + 2.592σp R̄p = 0.08 + 1.648σp R̄p = 0.1 + 0.8471σp

(d)
(i) We use the same Z-vectors as in the unrestricted case, to get the Lintner portfolios.

Rf = 6% Rf = 8% Rf = 10%
xA 71.60% 71.41% 68.24%
xB -21.29% -20.31% -3.53%
xC 7.11% 8.28% 28.24%

LintnerP
Portfolios
xi 57.42% 59.38% 92.95%
xf 42.58% 40.62% 7.05%
Expected Return 8.66% 9.67% 12.52%
Standard Deviation 0.74% 0.80% 2.97%
Sharpe ratio 3.611 2.081 0.8474

Note that Lintner portfolios have the same Sharpe ratios as unrestricted tangent
portfolios. They can always be interpreted as a combination of deposit with the
(unrestricted) tangent portfolio.
(ii) Since, none the original tangent portfolios requires more than 50% shortselling, they
all satisy this restriction.
(ii) For the case of Rf = 10% this limit is satisfied and nothing changes.
For Rf = 6% and Rf = 8% the limit is not satisfied by the original tangent portfolios,
thus, we know that we will now get xB = −25%. The remaining weight we can get
numerically (for instance using excel solver).
The table below show the results.
Rf = 6% Rf = 8% Rf = 10%
xA 115.27% 112.82 % 73.42%
xB -25.00% -25.00% -3.80%
xC 9.73% 12.18% 30.38%
Tangent Portfolios (limited 25% shortselling)
Expected Return 10.83% 10.98% 12.71%
Standard Deviation 1.42% 1.47% 3.20%
Sharpe ratio 3.413 2.021 0.8474

12
Figure 5: Exercise 1.11 – Efficient Frontiers (green) when shortselling is allowed (with and with-
out borrowing). Outer hyperbola (blue) is the envelop hyperbola when we consider investment
without constraints in the three assets A, B, C. Inner hyperbola is the two-assets hyperbola
for assets A and C where the full line represents the no shortselling segment and the dashed
line the portfolios that require shortselling. Top left image: Rf = 6%. Top right: Rf = 8%.
Bottom image: Rf = 10%

13
Exercise 1.12.

(a) Since the given portfolios, A and B, are on the efficient frontier, and we know ρAB =
+5/6 = 0.8333. We can determine their covariance σAB = ρAB σA σB = 0.002 and the MV
portfolio can be obtained by finding the minimum-risk combination of the two portfolios:
2
σB − σAB 0.0016 − 0.002 1
xM
A
V
= 2 2 = =−
σA + σB − 2σAB 0.0036 + 0.0016 − 2 × 0.002 3
 
1 4
xM
B
V
= 1 − xM
A
V
=1− − =
3 3
This gives R̄M V = 7.33% and σM V = 3.83%.
Also, since the two portfolios are on the efficient frontier, the entire efficient frontier can
then be traced by using various combinations of the two portfolios, starting with the MV
portfolio and moving up along the efficient frontier (increasing the weight in portfolio A
and decreasing the weight in portfolio B).
Since xB = 1 − xA the efficient frontier equations are:
(
R̄p = xA R̄A + (1 − xA ) R̄B
2 2
2
σp2 = x2A σA + (1 − xA ) σB + 2xA (1 − xA ) σAB
⇔
(
R̄p = 0.10xA + 0.08 × (1 − xA )
2
σp2 = 0.0036x2A + 0.0016 (1 − xA ) + 2xA (1 − xA ) 0.002
⇔
σp2 = 3R̄p2 − 0.44R̄p + 0.0176 (hyperbola equation)
Since short sales are allowed, the efficient frontier will extend beyond portfolio A and out
toward infinity. The efficient frontier appears as shown in Figure ?? (full blue line).
(b) If there is a risk-free asset that can be used for both deposit and borrowing, then we know
the efficient frontier is a straight line passing by the risk-free asset and tangent to the
hyperbola given by combinations of any two efficient portfolios. So, Its equations is given
by R̄p = Rf + SRT σp where SRT is the Sharpe ratio of the tangent portfolio.
The tangent portfolio is the combination of the two efficient portfolios that has the highest
Sharpe ratio. From the FOC we find
   
−1 4.5455 12.5%
Z = V (R̄ − Rf 1) = ⇒ XT =
31.8182 87.5%
and we get
 

10%
R̄T = XT0 R̄ = 12.5% 87.5% = 8.25%
8%
  

0.0036 0.002 12.5%
σT2 = XT0 V XT = 12.5% 87.5% = 0.00171875
0.002 0.0016 87.5%
⇒ σT = 4.15%
R̄T − Rf 0.0825 − 0.02
SRT = = = 1.5076 .
σT 0.0415
So the efficient frontier in this case is given by
R̄p = 0.02 + 1.5076σp ,
the straight green line in Figure ??.

14
Figure 6: Exercise 1.12 – Efficient Frontier with (green line) and without the risk-free asset
(full blue line on the upper part of the hyperbola.

(c) Since the tangent portfolio does not require shortselling the Lintner portfolio is the tangent
portfolio itself and noting changes. When shortselling is limited a la Lintner nothing
changes in term
(i) If A and B are still feasible this means they are portfolios without any shortselling
position. In addition, since the tangent portfolio does not require shortselling of A
nor B, we have the guarantee it remains feasible. The minimum variance portfolio,
however, requires short-selling of asset A and it would be no longer feasible. In this
case the MV portfolio would be portfolio B itself. The envelop hyperbola would
be the same but limited below by B above by A. Beyond these points the efficient
frontier would be composed by parts of sets of hyperbolas in the interior of the
general envelop hyperbola.
(ii) If one of the original efficient portfolios is not longer feasible, that means that port-
folio would require shortselling of some risky asset. Without two efficient portfolios
we would not be able to derive the envelop hyperbola equation. Since we have no
information about the basic risky assets in this market, we cannot derive the new
efficient frontier, but it would be contained in the interior of the previously derived
hyperbola.

15
1.3 Portfolio Protection

Exercise 1.13.

(a) (i) Since Rf = 8% we know that for RL = 6% to minimize the probability of returns
lower than 6% is equivalent to set that probability to zero, i.e. to deposit 100% of
our wealth.
(ii) If we have Rf = RL = 8% and Gaussian returns, all efficient portfolio have the same
probability of returns lower than 8% .
(iii) If we have RL = 10% > Rf = 8% and Gaussian returns the optimal turns out to be
to leverage up as much as possible (borrowing as much as possible) to invest more
than our wealth in the tangent portfolio.
(b) Portfolios with the highest return-at-risk (RaR) are also the safest portfolio according
to Kataoka. We, thus, are interested in portfolio returns with the probabilities lower or
equal to 10%, i.e. in the worst 10% scenarios,

Pr (Rp ≤ RL ) ≤ 10%
 
Rp − R̄p RL − R̄p
Pr ≤ ≤ 10%
σp σp
 
RL − R̄p
Φ ≤ 10%
σp
RL − R̄p
≤ Φ−1 (10%)
σp
R̄p ≥ RL − Φ−1 (10%)σp
R̄p ≥ RL + 1.2816σp

So the Kataoka lines are given by R̄p = RL + 1.2816σp and the goal is to maximize RL
along the EF.
From Exercise ?? recall that for Rf = 8%, the EF is given by R̄p = 0.08 + 2.081σp . Since
the slope of the EF is higher than the slope of the Kataoka lines, the maximum RL will
be the highest expected return portfolio. That is, the lowest RaR portfolio turns out to
require extreme leverage (borrowing as much as possible) to invest more than our wealth
in the tangent portfolio.
(c) Following the exact same steps as in (b) we get

Pr (Rp ≤ 10%) ≤ 10% ⇔ R̄p ≥ 0.1 + 1.2816σp .

Since the EF has a lower y-cross and a higher slope R̄p = 0.08 + 2.081σp , we need to find
the crossing point.
0.1 − 0.08
0.1+1.2816σp = 0.08+2.081σp ⇔ σp = = 2.5% ⇒ R̄p = 13.21%.
2.081 − 1.2816

(d) In (a) we deal with the Roy criterion, in (b) with the Kataoka criterion and in (c) with
the Telser criterion. See Figure ?? for a graphical representation of the previous answers.

16
Figure 7: Exercise ?? – Efficient Frontier (red) and safety first portfolios determined in (a)–(c).
Dashed grey lines are Kataoka lines for different RL . All portfolios in the efficient frontier has
the have probability of returns lower than RL = 8%. Dark grey dashed segment of line on the
EF identify all portfolios that satisfy the Telser restriction.

Exercise 1.14.

(a) All combinations of A and B satisfy

   
Rp − R̄p 5% − R̄p 5% − R̄p
Pr (Rp ≤ 5%) = Pr ≤ =Φ
σp σp σp

since, it must be less or equal than 15% we have

 
5% − R̄p
Φ ≤ 15% ⇔ R̄p ≥ 5% − Φ−1 (15%) σp ⇔ Rp ≥ 5% + 1.0364 σp .
σp
From Exercise ?? we also know all combinations of A and B are given by the hyperbola

σp2 = 3R̄p2 − 0.44R̄p + 0.0176.

From Figure ?? we clearly see that there is no combination of A and B that satisfies the
safety condition.
(b) The combination that maximizes the likelihood of getting returns above 5% is the one
that minimizes the probability of returns lower or equal to 5%, i.e. it is the Roy portfolio
with RL = 5%. The Roy portfolio can be determine as a tangent portfolio, where RL acts
as a ficticious risk-free rate. In this case we get
    
−1
  909.091 −1136.364 5% 11.3636
Z=V R̄ − RL 1 = =
−1136.364 2045, 454545 3% 4.5455
⇓
 
Roy 71.43%
X =
28.57%

17
 The Roy portfolio is a concrete combination of A adnd B, so it belongs to the hyperbola.
It has R̄Roy = 9.43% and σ Roy = 5.28%. Thus its probability of returns lower than 5% is
 
5% − 9.43%
Φ = Φ (−0.894) = 20.1% .
5.28%

See the representation of the Roy portfolio in Figure ??.

(c) The combination with the highest 15% quantile is the Kataoka portfolio for an α = 15%.
For a fixed α we have lines with a fixed slope equal to −Φ−1 (α). In this case we have
−Φ−1 (α) = 1.0364. For find the Kataoka portfolio we need to find the hyperbola point
with the exact same slope.
From the Kataoka lines we get

∂ R̄p
R̄p = RL + 1.0364 σp ⇒ = 1.0364
∂σp Kataoka

From the hyperbola equation, σp2 = 3R̄p2 − 0.44R̄p + 0.0176 , and considering only its
upper part (the efficient part, we have
q
+0.44 + 0.442 − 4 × 3 × (0.0176 − σp2 )
R̄p =
6
and differentiating w.r.t. σp

∂ R̄p 1 1
− 1
= × 0.442 − 4 × 3 × (0.0176 − σp2 ) 2 × (−2σp )
∂σp hyperbola 6 2

Matching the slopes of the Kataoka lines with the hyperbola slope

∂ R̄p ∂ R̄p
=
∂σp
Kataoka ∂σp hyperbola
1
− 1
1.0364 = − 0.442 − 4 × 3 × (0.0176 − σp2 ) 2 σp
6
2 1
−1 2
1.0364 = 0.442 − 4 × 3 × (0.0176 − σp2 ) σp
36
1 2
1.0741 0.442 − 12(0.0176 − σp2 ) =


σ
36 p
σp2 = 0.002203 ⇒ σ Kataoka = 4.05%

Which implies and expecte return of

p
+0.44 + 0.442 − 4 × 3 × (0.0176 − 0.002203)
R̄p = = 8.9%.
6
 
45%
Finally, in terms of composition we have X Kataoka = .
55%
(d) In (a) we deal with the Telser criterion, but in this case there was no feasible portfo-
lio satisfiying the safety condition. In (b) we address Roy’s safety criterion and in (c)
Kataoka’s.
(e) If the returns were not Gaussian we could do the same type of computations but using
the correct distribution function.

18
Figure 8: Exercise ?? – No Telser portfolio feasible. Representation of the two efficient portfolios
A and B and the Roy and Kataoka portfolios.

1.4 International Diversification

Exercise 1.16. Diversification means combine different assets with different risk profiles such
that we can manage to decrease our risk exposure while maintaining our return. Of course,
diversification is only possible if the assets in the portfolio are not perfectly positively correlated
(ρ = 1). Actually, the most idyllic scenario would perfectly negatively correlation (ρ = −1)
among assets since it would allow us to cancel an important portion of portfolio’s risk: the
specific or idiosyncratic risk. Let,
N
X N X
X N
σP2 = x2i σi2 + xi xj σij
i=1 i=1 j=1
i6=j

1
If xi = N then
N  2 N X N   
2
X 1 X 1 1
σH = σi2 + σij
i=1
N i=1 j=1
N N
j6=i

Factoring out 1/N from the first summation and (N − 1)/N from the second and simplifying
yields
N  N N 
1 X σi2
 
2 (N − 1) X X σij
σH = +
N i=1 N N i=1 j=1
N (N − 1)
j6=i
1 N −1 1
= σ̄i2 + σ̄ 2 − σ̄ij + σ̄ij


σ¯ij =
N N N i
This is a quite realistic representation of what occur when we invest in a portfolio of assets.
The contribution to the portfolio variance of the variance of the individual securities goes to

19
zero as N gets very large. However, the contribution of the covariance terms approaches the
average covariance as N gets large. Actually, if we let N → ∞, it cames

2 1
σ̄i2 − σ̄ij + σ̄ij = σ̄ij


lim σH = lim
N →∞ N →∞ N

Thus, as said before, the individual risk of securities can be diversified ways. Of course the
higher the number os securities in the portfolio, the better the diversification. If we only
consider a domestic market, the available number of tradable securities is lower than when we
also consider external markets. Therefore, the major effect of diversification is to allow for a
better diversification. However, this is at a price, which is exchange rate risk.

Exercise 1.17.

(a) The return due to exchange-rate changes (RX ) is equal to f xt /f xt−1 − 1, where f xt is
the foreign exchange rate at time t expressed in terms of the investor’s home currency
per unit of foreign currency. Let f xt be the exchange rate expressed in terms of dollars
and f x∗t be the exchange rate expressed in terms of pounds. These two rates are simply
reciprocals, i.e., f x∗t = 1/f xt . So from the table in the problem we have:

∗
(1 + RX ) (1 + RX )
Period (for US investor) (for UK investor)
1 2.5/3 = 0.833 3/2.5 = 1.200
2 2.5/2.5 = 1.000 2.5/2.5 = 1.000
3 2/2.5 = 0.800 2.5/2 = 1.250
4 1.5/2 = 0.750 2/1.5 = 1.333
5 2.5/1.5 = 1.667 1.5/2.5 = 0.600

The total return to a U.S. investor from a U.K. investment is

(1 + RU S ) = (1 + RX ) (1 + RU K )

And the total return to a U.K. investor from a U.S. investment is

(1 + RU K ) = (1 + RX ) (1 + RU S )

So,
– Return to US investor
Period From US investment From UK investment
1 10% (0.833)(1.05) − 1 = −12.5%
2 15% (1)(0.95) − 1 = −5.0%
3 −5% (0.8)(1.15) − 1 = −8.0%
4 12% (0.75)(1.08) − 1 = −19.0%
5 6% (1.667)(1.1) − 1 = 83.3%
Average 7.6% 7.76%
– Return to UK investor
Period From UK investment From US investment
1 5% (1.2)(1.1) − 1 = 32.0%
2 −5% (1)(1.15) − 1 = 15.0%
3 15% (1.25)(0.95) − 1 = 18.75%
4 8% (1.333)(1.12) − 1 = 49.3%
5 10% (0.6)(1.06) − 1 = −36.4%
Average 6.6% 15.73%

20
 (b) The standard deviation of return is given by
v
uX Ri − R̄i 2
uN

σ= t
i=1
N

Thus,
– For US investor
s
2 2 2 2 2
(10 − 7.6) + (15 − 7.6) + (−5 − 7.6) + (12 − 7.6) + (6 − 7.6)
σU S =
5
= 6.95%
s
2 2 2 2 2
(−12.5 − 7.76) + (−5 − 7.76) + (−8 − 7.76) + (−19 − 7.76) + (83.3 − 7.76)
σU K =
5
= 38.06%

– For UK investor
s
2 2 2 2 2
(5 − 6.6) + (−5 − 6.6) + (15 − 6.6) + (8 − 6.6) + (10 − 6.6)
σU K =
5
= 6.65%
s
2 2 2 2 2
(32 − 15.73) + (15 − 15.73) + (18.75 − 15.73) + (49.3 − 15.73) + (−36.4 − 15.73)
σU S =
5
= 38.06%

Exercise 1.18. In general, we should hold non-domestic (N ) securities instead of domestic

securities(D) when foreign investment is more attractive than domestic investment. What
happens when the following inequality holds
R̄N − RF R̄D − RF
> ρN,D
σN σD
Specifically, for an US investor
R̄N − RF R̄U S − RF
> ρN,U S
σN σU S
R̄U S and R̄N , σN and σN,U S for the foreign countries are given in the problem and summarized
below:

R̄N (%) σN σN,U S

Austria 14 24.50 0.281
France 16 17.76 0.534
Japan 14 25.70 0.348
UK 15 15.59 0.646

We also know that R̄U S = 20%, σU S = 13.59 and RF = 6%. Thus, we have

R̄N −RF R̄U S −RF

σN σU S ρN,U S
Austria 0.327 0.289
France 0.563 0.550
Japan 0.311 0.358
UK 0.577 0.665

21
For Austria and France, the above inequality holds, so a US investor should consider those
foreign markets as attractive investments; for Japan and the UK, the above inequality does not
hold, so a US investor should not consider those foreign markets as attractive investments.

Exercise 1.19. The formula to find the minimum-risk portfolio of two assets is get by taking
the first derivative of the portfolio variance w.r.t. x1 and equal 0, which gives

σ22 − σ1 σ2 ρ1,2
xM
1
V
=
σ12 + σ22 − 2σ1 σ2 ρ1,2

where x1 is the investment weight for asset 1 and x2 = 1 − x1 .

(a) For equities, σU S = 13.59%, σN = 16.70% and ρN,U S = 0.423. So the minimum-variance
portfolio is:

(19.0%)2 − 15.39% × 19.0% × 0.423

xM V
US = = 0.6771
15.39%)2 + (19.0%)2 − 2 × 15.39% × 19.0% × 0.423
xM
N
V
= 1 − xM V
U S = 0.3266

(b) For bonds, σU S = 6.92%, σN = 12.875% and ρN,U S = 0.527. So the minimum-variance
portfolio is:

(12.875%)2 − 6.92% × 12.875% × 0.527

xM V
US = = 0.9924
(6.92%)2 + (12.875%)2 − 2 × 6.92% × 12.875% × 0.527
xM
N
V
= 1 − xM V
U S = 0.0076

(c) For T-bills, σU S = 1.068%, σN = 10.057% and ρN,U S = −0.220. So the minimum-variance
portfolio is:

(10.057%)2 + 1.068% × 10.057% × 0.220

xM V
US = = 0.9673
(1.068%)2 + (10.057%)2 + 2 × 1.068% × 10.057% × 0.220
xM
N
V
= 1 − xM V
U S = 0.0327

22
 2 Portfolio Selection Models

2.1 Constant Correlation Model

Exercise 2.1.

(a)-(b) The only assumption of the Constant Correlation Model is that the correlation between
any pair of securities is constant, such that ρij = ρ∗ ∀i, j. This is an unrealistic assumption
that may lead to introduction of model risk. On the other hand, it allows us to decrease
the number of parameters one needs to estimate to use MVT. So, the use of CCM may
lead to a considerable reduction in estimation risk. It also allows us to use cut-off methods
to find tangent portfolios.

Exercise 2.2.

(a) Yes, since all pairwise correlations are the same, this is the ideal scenario to use CCMs.
In this case we have zero model risk.

(b) If short sales are allowed, all securities will be included in the optimal portfolio. Assuming
constant correlation we can apply the cut-off method that consists in
1. Rank all securities accordingly to Sharpe’s Ratio
2. Calculate the Cut-Off point
3. Compute Z and the weights X.
In Table ?? below, given that the riskless rate equals 4%, the securities are ranked in
descending order by their excess return over standard deviation. To calculate the cut-off
point C ∗ we need a general expression that give us Ci . This expression is
N
X R̄i − RF
ρ
Ci =
1 − ρ + iρ i=1 σi

where ρ is the correlation coefficient - assumed constant for all securities. The subscript
i indicates that Ci is calculated, using data on the first i securities. Each Ci is calculated
as follows
1
X R̄10 − RF
ρ 0.5 12 − 4
C1 = = × =2
1 − ρ + 1ρ i=1 σ1 0 1 − 0.5 + 1 × 0.5 2
2  
ρ X R̄3 − RF 0.5 12 − 4 12 − 4
C2 = = + =2
1 − ρ + 2ρ i=1 σ3 1 − 0.5 + 2 × 0.5 2 4
..
.

Since short-sales are allowed, we include all securities, which implies that the cut-off rate
th
is given by the C rate of the last security. In this exercise, C ∗ = C 10 = 1.41.
The last step to find the optimal portfolio is to calculate Zs, which is given by
 
1 R̄i − RF ∗
zi = −C
(1 − ρ) σi σi

23
 N
R̄i −RF P R̄i −RF ρ
Security Rank i R̄i − RF σi σi 1−ρ+iρ C zi xi
i=1
10 1 8 4.00 4.00 0.50 2.00 2.59 189.22%
3 2 8 2.00 6.00 0.33 2.00 0.30 21.68%
6 3 5 1.67 7.67 0.25 1.92 0.17 12.69%
9 4 6 1.50 9.17 0.20 1.83 0.05 3.44%
4 5 10 1.43 10.6 0.17 1.77 0.01 0.48%
1 6 6 1.20 11.8 0.14 1.69 -0.08 −6.00%
5 7 2 1.00 12.8 0.13 1.60 -0.41 −29.59%
7 8 1 1.00 13.8 0.11 1.53 -0.81 −59.17%
8 9 4 1.00 14.8 0.10 1.48 -0.20 −14.79%
2 10 4 0.67 15.47 0.09 1.41 -0.25 −17.97%

Table 1: Exercise ??- Efficient Portfolio

Then,
   
1 R̄10 − RF 1 12 − 4
z1 = − C∗ = − 1.41 = 2.59
(1 − ρ) σ10 σ10 (1 − 0.5) 2 2
   
1 R̄3 − RF ∗ 1 12 − 4
z2 = −C = − 1.41 = 0.30
(1 − ρ) σ3 σ3 (1 − 0.5) 4 4
..
.
Zi
Finally, to find the weights Xs and since xi = PN
zi
, we have
i=1

z1 2.59
x1 = 10
= = 1.8922
P 1.37
zi
i=1
z2 0.3
x2 = 10
= = 0.2168
P 1.37
zi
i=1
..
.

Table ?? presents all previous calculations and the efficient portfolio, T A .

(e) The efficient portfolio, T A found in part (b) is the unique efficient portfolio we have with
a risk-free rate of 4%, being the tangent portfolio between the capital market line and
the efficient frontier of risky assets. Applying the formulas for portfolio’s expected return
and risk, we have R̄T = 18.907% and σT = 3.297%. Now, we can draw the capital market
line, which is the efficient frontier in this case (see Figure ??)

Exercise 2.3.

(a) The efficient frontier is the line from RF and is tangent to the efficient frontier of risky
assets. It is similar to Figure ??.
(b)

(i) In Table ??, given that the riskless rate equals 5%, the securities are ranked in
descending order by their excess return over standard deviation. To calculate the

24
 Figure 9: Exercise ?? - Efficient Frontier

cut-off point C ∗ we need a general expression that give us Ci . This expression is

N
X R̄i − RF
ρ
Ci =
1 − ρ + iρ i=1 σi

where ρ is the correlation coefficient - assumed constant for all securities. The
subscript i indicates that Ci is calculated, using data on the first i securities. Each
Ci is calculated as follows
1
X R̄1 − RF
ρ 0.5 15 − 5
C1 = = × = 0.5
1 − ρ + 1ρ i=1 σ1 1 − 0.5 + 1 × 0.5 10
2  
ρ X R̄2 − RF 0.5 15 − 5 20 − 5
C2 = = + = 0.6667
1 − ρ + 2ρ i=1 σ2 1 − 0.5 + 2 × 0.5 10 15
..
.

With no short sales, we only include those securities for which R̄i −R σi
F
> Ci . Thus,
only securities 1, 2, 5 and 6 (the four highest ranked securities in the above table) are
in the optimal (tangent) portfolio. We could have stopped our calculations after the
first time we found a ranked security for which R̄i −R σi
F
< Ci , (in this case the fifth
highest ranked security, security 4), but we did not so that we could demonstrate
that R̄i −R
σi
F
< Ci for all of the remaining lower ranked securities as well.
Since security 6 (the fourth highest ranked security, where i = 4) is the last ranked
security in descending order for which R̄i −Rσi
F
> Ci , we set C ∗ = C4 = 0.78
The last step to find the optimal portfolio is to calculate Zs, which is given by
 
1 R̄i − RF
zi = − C∗
(1 − ρ) σi σi

25
 N
R̄i −RF P R̄i −RF ρ
Security Rank i R̄i − RF σi σi 1−ρ+iρ C zi xi
i=1
1 1 10 1.00 1.00 0.5000 0.5000 0.0440 0.2375
2 2 15 1.00 2.00 0.3333 0.6667 0.0293 0.1581
5 3 5 1.00 3.00 0.2500 0.7500 0.0880 0.4749
6 4 9 0.90 3.90 0.2000 0.7800 0.0240 0.1295
4 5 7 0.70 4.60 0.1667 0.7668 - -
3 6 13 0.65 5.25 0.1429 0.7502 - -
7 7 11 0.55 5.80 0.1250 0.7250 - -

Table 2: Exercise ??(b)(i) – Efficient Portfolio (short-selling not allowed)

Then,
   
1 R̄1 − RF 1 15 − 5
z1 = − C∗ = − 0.78 = 0.0440
(1 − ρ) σ1 σ1 (1 − 0.5) 10 10
   
1 R̄2 − RF 1 20 − 5
z2 = − C∗ = − 0.78 = 0.0293
(1 − ρ) σ2 σ2 (1 − 0.5) 15 15
...

Finally, to find the weights Xs and since xi = PNzi , we have

i=1 zi

z1 0.0440
x1 = 4
= = 0.2375
P 0.1853
zi
i=1
z2 0.0293
x2 = 4
= = 0.1581
P 0.1853
zi
i=1
..
.

Table ?? presents all previous calculations and the efficient portfolio without short-
selling. Since i = 1 for security 1, i = 2 for security 2, i = 3 for security 5 and
i = 4 for security 6, the tangent portfolio when short sales are not allowed consists
of 23.75% invested in security 1, 15.81% invested in security 2, 47.49% invested in
security 5 and 12.95% invested in security 6.
(ii) When short-selling is allowed, we set the cut-off rate to C ∗ = 0.725 ir order to include
all securities in our efficient portfolio (see Table ??). The Zs and the weights Xs are
calculated as before. Concretely we have
   
1 R̄1 − RF ∗ 1 15 − 5
z1 = −C = − 0.725 = 0.0550
(1 − ρ) σ1 σ1 (1 − 0.5) 10 10
..
.
   
1 R̄2 − RF ∗ 1 12 − 5
z5 = −C = − 0.725 = −0.0050
(1 − ρ) σ2 σ2 (1 − 0.5) 10 10
..
.

And we can determine all weights xi = PNzi z . For this concrete case we have
i=1 i
PN
z
i=1 i = 0.2061. See Table ?? for concrete weight values.

26
 N
R̄i −RF P R̄i −RF ρ
Security Rank i R̄i − RF σi σi 1−ρ+iρ C zi xi
i=1
1 1 10 1.00 1.00 0.5000 0.5000 0.0550 0.2661
2 2 15 1.00 2.00 0.3333 0.6667 0.0367 0.1776
5 3 5 1.00 3.00 0.2500 0.7500 0.1100 0.5322
6 4 9 0.90 3.90 0.2000 0.7800 0.0350 0.1703
4 5 7 0.70 4.60 0.1667 0.7668 -0.0050 -0.0242
3 6 13 0.65 5.25 0.1429 0.7502 -0.0075 -0.0363
7 7 11 0.55 5.80 0.1250 0.7250 -0.0175 -0.0847

Table 3: Exercise ??(b)(ii) - Efficient Portfolio (short-selling allowed - Standard Definition)

(iii) When shotselling is allowed, but limited a la Linter, the Zs are the same as above
(compare zi in Tables ?? and ??, however the weights are determined as xi =
PNzi ,
|z |
i=1 i

z1 0.05500
x1 = 7
= = 0.2062
P 0.2667
|zi |
i=1
..
.
z2 −0.0050
x5 = 7
= = −0.0187
P 0.2667
|zi |
i=1
..
.

Table ?? presents all previous calculations and the efficient portfolio with short-
selling (Lintner definition).

N
R̄i −RF P R̄i −RF ρ
Security Rank i R̄i − RF σi σi 1−ρ+iρ C zi xi
i=1
1 1 10 1.00 1.00 0.5000 0.5000 0.0550 0.2062
2 2 15 1.00 2.00 0.3333 0.6667 0.0367 0.1376
5 3 5 1.00 3.00 0.2500 0.7500 0.1100 0.4124
6 4 9 0.90 3.90 0.2000 0.7800 0.0350 0.1312
4 5 7 0.70 4.60 0.1667 0.7668 -0.0050 -0.0187
3 6 13 0.65 5.25 0.1429 0.7502 -0.0075 -0.0281
7 7 11 0.55 5.80 0.1250 0.7250 -0.0175 -0.0656

Table 4: Exercise ??(b)(iii) - Efficient Portfolio (short-selling allowed - Lintner Definition)

(c) If the risk-free asset does not exist, their are an infinite number of efficient portfolios of
risky assets. Determine all these portfolios imply the calculation of the efficient frontier,
which can be done using pretty sophisticated matricial equations, which are outside the
scope of this course. Nevertheless, we have a different and easier way to do this calculation.
We just need to assume the existence of a fictitious risk-free rate of return to find an
efficient portfolio. Then we assume a second fictitious frontier to have a second efficient
portfolio. Now, with these two portfolios we can find any other portfolio applying the
Efficient Portfolios Theorem and we can, also, derive the representative equation of the
hyperbole that corresponds to the efficient frontier.

27
2.2 Single-Index Model

Exercise 2.4.

(a) The β of Security A is lower than 1 therefore it is considered a defensive stock. On the
other side, security B has a β higher than 1, so that it is an aggressive stock.
(b) (i) To compute the portfolio’s β we proceed as follows
X
βp = xi βi ⇔ βp = xA βA + xB βB = 0.25 × 0.75 + 0.75 × 2 = 1.6875

(ii) Using the single-index model (SIM), the portfolio’s risk is

X h i
2 2
σp2 = βp2 σm
2
+ x2i σei
2
= 1.68752 × 0.252 + (0.25) × 0.02 + (0.75) × 0.03
√
σp2 = 0.177978 + 0.018125 = 0.1961 σp = 0.1961 = 44.28%.

(c) A portfolio with A and B, which risk equals the market risk is a portfolio, which risk
equals the market risk, thus σp2 = βp2 σm
2
= 0.252 = 0.0625. To calculate the weight of
stock A (XA )we need to solve the portfolio variance equation in order to XA . To do so
we us first need to compute the return’s variance for stock A and B and the covariance
between this returns using the single-index model:
2 2 2 2
σA = βA σm + σeA = 0.752 × 0.252 + 0.02 = 0.0552
2 2 2 2
σB = βB σm + σeB = 22 × 0.252 + 0.03 = 0.28
2
σAB = βA × βB × σM = 0.75 × 2 × 0.252 = 0.09375

Then,

σp2 = x2A σA
2
+ (1 − xA )2 σB
2 2
+ 2xA (1 − xA ) × σAB
0.252 = 0.0552x2A + 0.28(1 − xA )2 + 2 × 0.75 × 2 × 0.252 xA (1 − xA )
0.0625 − 0.028 = (0.0552 + 0.28 − 2 × 0.09375)x2A + 2 × (0.09375 − 0.28)xA
0 = 0.1477x2A − 0.3725xA + 0.2175
√
0.3725 ± 0.37242 − 4 × 0.1477 × 0.2175
xA = ⇔ xA = 160.39% ∨ xA = 91.85%
2 × 0.1477
There are two possible solutions to xA , nevertheless just one makes sense, since just one
is efficient. Such solution is xA = 91.85%. The β of this portfolio is βP = xA βA + (1 −
xA )βB = 0.9185 × 0.75 + 0.0815 × 2 ≈ 0.85.

(d) In part (c) we calculated the stocks variance using SIM. When we compare these results
2SIM 2SIM
with the new data we realize that σA = 0.0552 6= 0.1 and σB = 0.28 ≈ 0.3. Thus,
the SIM does not seems to hold when we use it with stock A, despite it seems to be a
good approximation when applied to stock B.

Exercise 2.5.

2
(a) The covariance between stock B and the market portfolio is σBM = βB βM σM = 1.125 ×
2
1 × 0.4 = 0.18

28
 (b) If the Single Index Model (SIM) holds, the portfolio variance is as follows
n
X
σp2 = βp2 σm
2
+ x2i σεi
2

i=1
| {z }
systematic variance | {z }
residual variance

Thus, the residual variance in this homogenous portfolio (in a homogenous portfolio each
security weight is given by 1/N , where N is the number of securities, in this case xi =
1/2 = 0.5) is
Xn
2
σeH = x2i σεi
2
= 0.52 × 0.1 + 0.52 × 0.15 = 0.0625
i=1

(c) Since the covariance between the residual variances of security A and B are not zero, the
single-index model does not apply. Therefore, the residual variance calculated in part (b)
is not the effective residual variance of a homogeneous portfolio, which is given by the
modern portfolio’s theory. Thus, for two securities, the variance is

σe2H = x2A σe2B + x2B σe2B + 2xA xB σeA eB

= 0.52 × 0.1 + 0.52 × 0.15 + 2 × 0.5 × 0.5 × 0.1
= 0.1125

2
2
(d) As seen in part (b), the systematic risk, under SIM, is σe2Syst = βp2 σm
P
, so βH = x i βi =
i=1
0.5 × 0.875 + 0.5 × 1.125 = 1. Thus,
2 2 2
σsyst H
= βH σm = 12 × 0.42 = 0.16

(e) (i) Total risk for each individual security calculated with SIM or with Portfolio Theory
is the same as long as SIM’s assumptions hold, namely that σei M = 0. In this case
nothing is said about this, therefore anything definitive can be said.
(ii) In the general case, total risk for a portfolio computed under SIM or Markowitz
assumptions is the same, as long as SIM’s assumptions hold, namely that σei ej = 0.
However, this is not the case when we use securities A and B to construct a portfolio,
since σeA eB = 0.1. Actually, under Markowitz total variance is σp2 = βp2 σm
2
+x2A σe2B +
2 2
xB σeB + 2xA xB σeA eB = 0.16 + 0.1125 = 0.2725 and under SIM total variance is
n
σp2 = βp2 σm
2
x2i σεi
2
P
+ = 0.16 + 0.0625 = 0.2225. Thus, their total risk is also
i=1
different.

Exercise 2.6.

(a) This exercise is based on the single-index model, more precisely in the market model,
which a positive correlation between any given stock returns and the market returns,
such that the return on a stock can be written as

Ri = ai + βi Rm

The term ai represents that component of return insensitive to the return on the market,
i.e. it represents specific risk. The term ai can be broken into two components: alphai
that denotes the expected value for ai ; and εi representing the random element of ai ,
which expected value is zero (Eεi = 0). Then ai = αi + εi and

Ri = αi + βi Rm + εi

29
 Note that both εi and Rm are random variables with standard deviations denoted by σεi
and σm , respectively. The term βi Rm represent the systematic risk and measure how
sensitivity the stock’s return is to the market’s return.
The model’s main assumptions are:
– εi is uncorrelated with Rm , such that the model ability to explain stock returns
is independent of what the return

on the market happens to be. More formally
cov (εi Rm ) = E (εi − 0) Rm − R̄m = 0
– vei is independent of ej for all values of i and j. which implies that the only reason
stocks vary together, systematically, is because of a common co-movement with the
market. More formally E (εi εj ) = 0

(b) (i) The expected return is given by R̄i = ai + βi R̄m . Thus,

  

 R̄A = aA + βA R̄m 
 R̄A = 2% + 1.5 × 8% 
 R̄A = 14%

 
 

 R̄B = aB + βB R̄m
  R̄B = 3% + 1.3 × 8%
  R̄B = 13.4%

⇔ ⇔


 R̄C = aC + βC R̄m 

 R̄C = 1% + 0.8 × 8% 

 R̄C = 7.4%

 
 

 R̄ = a + β R̄  R̄ = 4% + 0.9 × 8%  R̄ = 11.2%
D D D m D D

(ii) The security variance is given by σi2 = βi2 σm

2
+ σε2i . Therefore,
 2 2 2
 2 

 σ A = βA σm + σε2A 
 σA = 1.52 × 0.0025 + 0.0009 

2
σA = 0.006525
  
 σ2 = β 2 σ2 + σ2  σ 2 = 1.32 × 0.0025 + 0.0001 2
  
σB = 0.004325
  

B B m εB B
2 2 2
⇔ ⇔


 σC = βC σm + σε2C 


2
σC = 0.82 × 0.0025 + 0.0004 


2
σC = 0.0020
  
 2
 2 2 2
 2 2
 2
σD = βD σm + σεD σD = 0.9 × 0.0025 + 0.0016 σD = 0.003625
 

2
(iii) The covariance is given by σij = βi βj σm . Therefore,
 2
σAB = βA βB σm σAB = 1.5 × 1.3 × 0.0025 σAB = 0.004875
 

  

 
 

2
σ = βA βC σm σAC = 1.5 × 0.8 × 0.0025 σAC = 0.0030
  
 AC

 
 

 
 

2
  
 σAD = βA βD σ
  σAD = 1.5 × 0.9 × 0.0025
  σAD = 0.003375

m
2
⇔ ⇔


 σBC = βB βC σm 

 σBC = 1.3 × 0.8 × 0.0025 

 σBC = 0.0026
  
2
  



 σBD = βB βD σm



 σBD = 1.3 × 0.9 × 0.0025 


 σBD = 0.002925
  
2
  

σCD = βC βD σm σCD = 0.8 × 0.9 × 0.0025 σCD = 0.0018

The covariance matrix Σ is

 
0.006525 0.004875 0.0030 0.003375
 0.004875 0.004325 0.0026 0.002925 
 
 0.0030 0.0026 0.0020 0.0018 
0.003375 0.002925 0.003625 0.0016

(c) A homogenous portfolio is a portfolio where each security weight is given by 1/n, where
n denotes the number of security. Now, n = 4, thus each security weight is 1/4 = 0.25.
N
P
(i) The portfolio’s β is the weighted average β of all , i.e βp = x i βi ,
i=1

βH = 1.5 × 0.25 + 1.3 × 0.25 + 0.8 × 0.25 + 0.9 × 0.25 = 1.125

30
 N
P
(ii) Like βP , αp is given by the weighted average α of all securities, αp = xi αi
i=1

αH = 2% × 0.25 + 3% × 0.25 + 1% × 0.25 + 4% × 0.25 = 2.5%

PN
(iii) The portfolio’s variance is σp2 = βp2 σm
2
+ i=1 x2i σε2i . Thus,
 2 X4
2 2 2 1
σH = βH σm + x2i σε2i
4 i=1
= 1.1252 × 25 + 0.252 (0.0009 + 0.0001 + 0.0004 + 0.0016)
= 0.003352

(iv) To find the portfolio’s expected return we apply the market model using the portfo-
lio’s α and β. Therefore,

R̄P = αP + βP R̄m = 2.5% + 1.125 × 8% = 11.5%

(d) Using the suggested adjustment to find the β of the following period, we have

β2A = 0.343 + 0.677β1A = 0.343 + 0.677 × 1.5 = 1.3585

β2B = 0.343 + 0.677β1B = 0.343 + 0.677 × 1.3 = 1.2231

β2C = 0.343 + 0.677β1C = 0.343 + 0.677 × 0.8 = 0.8846

β2D = 0.343 + 0.677β1D = 0.343 + 0.677 × 0.9 = 0.9523

(e) Applying the Vasiček technique with the provided data and knowing the Vasiček β is
given by
σβ2 σβ̄2
β2i = 2 1i 2 β̄1 + 2 1 2 β1i
σβ̄ + σβ1i σβ̄ + σβ1i
1 1

we have
σβ21A σβ̄21 0.0441 0.0625
β2A = β̄ 1 + β1A = ·1+ · 1.5 = 1.2932
σβ̄2 + σβ21A σβ̄2 + σβ21A 0.0441 + 0.00625 0.0441 + 0.0625
1 1

σβ2 σβ̄21 0.1024 0.0625

β2B = 2 1B 2 β̄1 + 2 2 β1B = ·1+ · 1.3 = 1.1137
σβ̄ + σβ1B σβ̄ + σβ1B 0.1024 + 0.0625 0.1024 + 0.0625
1 1

σβ2 σβ̄21 0.0324 0.0625

β2C = 2 1C 2 β̄1 + 2 β1C = ·1+ · 0.8 = 0.8683
σβ̄ + σβ1C σβ̄ + σβ21C 0.0324 + 0.0625 0.0324 + 0.0625
1 1

σβ21D σβ̄21 0.04 0.0625

β2D = β̄1 + β1D = ·1+ · 0.9 = 0.9390
σβ̄2 + σβ21D σβ̄2 + σβ21D 0.04 + 0.0625 0.04 + 0.0625
1 1

Exercise 2.7.

(a) The covariance between any two securities can be written as




σij = E Ri − R̄i Rj − R̄j

Substituting for Ri , R̄i , Rj and R̄j yields


 

σij = E (αi + βi Rm + εi ) − αi + βi R̄m + εi (αj + βj Rm + εj ) − αj + βj R̄m + εj

31
 Simplifying by canceling the α’s and combining the terms involving β’s yields



σij = E βi Rm − R̄m + εi βj Rm − R̄m + εj

Carrying out the multiplication

2 
 

σij = βi βj E Rm − R̄m + βj E εi Rm − R̄m + βi E εj Rm − R̄m + E (εi εj )

From the single-index model assumptions we know

2 2
E Rm − R̄m = σm


E εi Rm − R̄m = 0


E εj Rm − R̄m = 0

And from the data in the problem

k = cov (εi εj ) = E [εi εj ] − E [εi ] E [εj ] = E [εi εj ]

| {z } | {z }
0 0

Thus,
2
σij = βi βj σm +k

(b) The general equation for the portfolio variance is

N
X N X
X N
σp2 = Xi2 σi2 + Xi Xj σij (2)
i=1 i=1 j=1
j6=i

From the Single-Index Model we know that

σi2 = βi2 σm
2
+ σε2i (3)

and
2
σij = βi βj σm
However, in this case, the covariance among the returns residuals is K and, therefore,
2
σij = βi βj σm +k (4)

as calculates in part b. Applying (??) and (??) in (??) we get

N
X N X
N

X
σp2 = Xi2 βi2 σm
2
+ σε2i + 2


Xi Xj βi βj σm +k
i=1 i=1 j=1
j6=i

32
 Doing some transformations we finally have
N
X N X
N

X
σp2 = Xi2 βi2 σm
2
+ σε2i + 2


Xi Xj βi βj σm +k
i=1 i=1 j=1
j6=i

N
X N X
X N N
X N X
X N
= Xi2 βi2 σm
2
+ 2
Xi Xj βi βj σm + Xi2 σε2i + + Xi Xj k
i=1 i=1 j=1 i=1 i=1 j=1
j6=i j6=i

N X
X N N
X N X
X N
2
= Xi Xj βi βj σm + Xi2 σε2i + Xi Xj k
i=1 j=1 i=1 i=1 j=1
j6=i
 
N
! N
! N N X
N
X X X X
2
= Xi βi Xi βi σm + Xi2 σε2i + k  Xi Xj 
 
i=1 i=1 i=1 i=1 j=1
| {z }| {z } j6=i

βP βP
 
N
X N X
X N
= βP2 + Xi2 σε2i + k  Xi Xj 
 
i=1 i=1 j=1
j6=i

Exercise 2.8.

(a) This is a standard portfolio selection exercise, in which we have to choose the tangent
portfolio between the capital market line and the efficient frontier of risky assets. The
solution for this problem involves solving the following system of simultaneous equations
in order to Zi , ∀i > 0

R̄1 − RF = Z1 σ12 + Z2 σ12 + Z3 σ13 + · · · + ZN σ1N





R̄2 − RF = Z1 σ21 + Z2 σ22 + Z3 σ23 + · · · + ZN σ2N






R̄3 − RF = Z1 σ31 + Z2 σ32 + Z3 σ32 + · · · + ZN σ3N



 ..
.





 2 2
R̄N − RF = Z1 σN N + Z 2 σN 2 + Z 3 σN 3 + · · · + Z N σN


which can be written using matricial notation

Z = V −1 (R − RF 1)

where Σ−1 is the inverse covariance matrix, R is a column vector with the securities
returns, RF is a scalar and 1 is a column vector of 1s. The Zs are proportional to the
optimum amount to invest in each security. Then the optimum proportions to invest in
stock k is Xk , where
Zk
Xk = N
P
Zi
i=1

Thus, we need to calculate the covariance matrix and then invert it. To find each pair of
covariances we can use the variance and covariance

definitions used in the Single-Index
Model σi2 = β 2 σm
2
+ σε2i and σij = βi βj σm2
. Thus, for security 1 and for the pair 1, 2 it
comes
σi2 = β 2 σm
2
+ σε2i = 12 × 10 + 30 = 40
2
σij = βi βj σm = 1 × 1.5 × 10 = 15

33
 Proceeding similarly for the other securities we arrive to the covariance matrix
 
0.004 0.0015 0.002 0.0008 0.001 0.0015
 0.0015 0.00425 0.003 0.0012 0.0015 0.00225 
 
 0.002 0.003 0.008 0.0016 0.002 0.003 
V =  0.0008

 0.0012 0.0016 0.00164 0.0008 0.0012  
 0.001 0.0015 0.002 0.0008 0.003 0.0015 
0.0015 0.00225 0.003 0.0012 0.0015 0.00325

Then the inverse matrix is

 
317.109 −36.505 −24.337 −38.939 −24.337 −73.010
 −36.505 417.863 −54.758 −87.613 −54.758 −164.274 
 
−1
 −24.337 −54.758 213.495 −58.408 −36.505 −109.516 
V =  −38.939

 −87.613 −58.408 906.547 −58.408 −175.225 

 −24.337 −54.758 −36.505 −58.408 463.495 −109.516 
−73.010 −164.274 −109.516 −175.225 −109.516 671.453

And R̄ − RF 1 is
     
15% 1 0.1

 12% 


 1  
  0.07 

 11%   1   0.06 
R̄ − RF 1 =   − 5%  = 

 8% 


 1  
  0.03 

 9%   1   0.04 
14% 1 0.09

Finally,
 
18.983

 2.711 

−1
 −6.526 
Z = V (R − RF 1) =  

 −4.410 

 −1.526 
25.422

Since short sales are not allowed, we need to use the cut-off method to know how many
securities will show up in the tangent portfolio.
We known we will not invest in any security with a negative ”Z” and we may even not
invest in some of the securities with positive Z.
In this case it turns out the optimal portfolio will have 3 securities, i.e., securities 1, 2 and
6 (which are the first three in the ranking and for which we have zi > 0). Thus, summing
P3
over the Zs from these three securities we get i=1 zi = 47.116 and the weights to invest
in each security are
18.983 2.711 25.422
z1 = = 0.4029 , z2 = = 0.0575 , z6 = = 0.5396
47.116 47.116 47.116
P6
(b) If short sales are allowed, using the standard definition, i=1 Zi = 34.623 and the weights
to invest in each security are
18.983 2.711
x1 = = 0.5483 x2 = = 0.0783
34.623 34.623
6.526 4.41
x3 = − = −0.1885 x4 = − = −0.1283
34.623 34.623
1.526 25.422
x5 = − = −0.0441 x6 = = 0.7343
34.623 34.623

34
 P6
Using Lintner definition, i=1 |zi | = 59.609 and the weights to invest in each security are

18.983 2.711
x1 = = 0.3185 x2 = = 0.0485
59.609 59.609
6.526 4.41
x3 = − = −0.1095 x4 = − = −0.0745
59.609 59.609
1.526 25.422
x5 = − = −0.0256 x6 = = 0.4265 .
59.609 59.609
Summing up all weights we can conclude only 58.08% is imvested in risky assets, which
means the remaining 41.92% is invested in the riskless asset.

(c) If the risk-free asset does not exist, their are an infinite number of efficient portfolios of
risky assets. Determine all these portfolios imply the calculation of the efficient frontier,
which can be done using pretty sophisticated matricial equations, which are outside the
scope of this course. Nevertheless, we have a different and easier way to do this calculation.
We just need to assume the existence of a fictitious risk-free rate of return to find an
efficient portfolio. Then we assume a second fictitious frontier to have a second efficient
portfolio. Now, with these two portfolios we can find any other portfolio applying the
Efficient Portfolios Theorem and we can, also, derive the representative equation of the
hyperbole that corresponds to the efficient frontier.

2
Exercise 2.9. We know βi can be written as σim /σm . We also know that σim = ρim σi σm .
Then,
ρim σi σm ρim σi
βi = 2
= (5)
σm σm
Since we have constant correlation ρ∗ between each pair of securities we should be to express
ρim as a function of ρ∗ . If the Single-Index Model holds, then σij = βi βj σm2
that can be
rewritten as follows
ρim σi σm ρjm σj σm
σij = βi βj σj2 = 2
× 2
2
× σm = ρim ρjm σi σj
σm σm

From statistics we have σij = ρij σi σj . If we let correlations to be constant, then σij = ρ∗ σi σj .
If correlations are constant and the Single-Index Model holds, we have

ρ∗ σi σj = ρim ρjm σi σj

ρ∗ = ρim ρjm

As the correlation is constant between each pair of securities we must have ρim = ρjm . Then,

ρ∗ = ρim ρim = ρ2im

and, p
ρim = |ρ∗ | (6)
Finally, using (??) into (??), we have
p
|ρ∗ |
βi = σi
σm
So, if correlations are constant and equal to ρ∗ , then, under the Single-Index Model assumptions,
each security β is a proportion
√ of its volatility. This proportion is constant and equal to all
|ρ∗ |
securities and defined as σm .

35
Exercise 2.10. Accordingly to the Single-Index Model, the expected return and risk are given
by
R̄i = αi + βi R̄m

σi2 = 2 2
βA σ + σε2A
| {zM} |{z}
Systematic Variance Specific Variance
q
σi = 2 σ2 + σ2
βA M εA

Therefore, the table can be filled using the equations. Notice that to calculate systematic risk
we assume specific risk to be zero. On the other hand, when we calculate the specific risk we
assume that systematic risk is zero.

Invest Expected Return Systematic Risk Specific Risk q Total Risk

p p
A αA + βA R̄m 2 σ2
βA σε2A 2 σ2 + σ2
βA
M m εA
p p q
B αB + βB R̄m 2 σ2
βB σε2B 2 σ2 + σ2
βB
M m εB
p p q
C αC + βC R̄m 2 σ2
βC σε2C 2 σ2 + σ2
βC
M m εC
p q
Port K αK + βK R̄m 2 σ2
βK σ εK 2 σ2 + σ2
βK
M m εK

which give us

Invest Expected Return p Systematic Risk Specific Risk √ Total Risk

A 2% + 1.5 × 20% = 32% 1.52 × (10%) = 15% 4% 15%2 + 4%2 = 15.52%
p √
B 4% + 0.8 × 20%20% 0.82 × (10%)2 = 8% 3% 8%2 + 3%2 = 8.54%
p √
C 6% + 0.4 × 20% = 14% 0.42 × (10%)2 = 4% 2% 4%2 + 2%2 = 4.47%
p √
Port K 4.6 + 0.74 × 20 = 19.4 0.742 × (10%)2 = 7.4% 1.56% 7.4%2 + 1.56%2 = 7.56%

For the portfolio K, αK , βK and σεK are as follows

3
X
αK = xi αi = 2% × 0.2 + 4% times0.3 + 6% × 0.5 = 4.6%
i=1
3
X
βK = xi βi = 1.5 × 0.2 + 0.8 × 0.3 + 0 − 4 × 0.5 = 0.74
i=1
v
u 3
uX p
σ εK =t x2i σε2i = 4%2 × 0.22 + 3%2 × 0.32 + 2%2 × 0.52 = 1.56%
i=1

2.3 Multi-Index Model

Exercise 2.11. Let us start with multi-index model with 3 correlated indexes I1∗ , I2∗ and I3∗ :

Ri = a∗i + b∗i1 × I1∗ + b∗i2 × I2∗ + b∗i3 × I3∗ + ci (7)

36
To reduce a general three-index model to a three-index model with orthogonal indexes we need
first to set I1∗ = I1 . Then, since I1∗ and I2∗ are correlated, we can express I2∗ in terms of I1 ,
defining an index I2 which is orthogonal to I1 as follows

I2∗ = γ0 + γ1 × I1 + dt

The part from I2∗ that is independent of I1∗ and adds new information to it is given by the
residuals in the linear regression, such that I2 = dt . Thus

I2 = dt = I2∗ − (γ0 + γ1 × I1 )

I2∗ = γ0 + γ1 × I1 + I2

Substituting the above expression into equation ?? and rearranging we get:

Ri = (a∗i + b∗i2 × γ0 ) + (b∗i1 + b∗i2 × γ1 ) × I1 + b∗i2 × I2 + b∗i3 × I3∗ + ci

The first term in the above equation is a constant, which we can define as a0i . The coefficient
in the second term of the above equation is also a constant, which we can define as b0i1 . We can
then rewrite the above equation as:

Ri = a0i + b0i1 × I1 + b∗i2 × I2 + b∗i3 × I3∗ + ci (8)

This model is equivalent to equation ??, but with two orthogonal indexes, I1 and I2 , and a
third index I3∗ that can be explained by I1 and I2 , through a linear regression

I3∗ = θ0 + θ1 × I1 + θ2 × I2 + et

As before, all new information due to I3∗ is captured by the residuals et . Therefore,

I3 = et = I3∗ − (θ0 + θ1 × I1 + θ2 × I2 )

I3∗ = θ0 + θ1 × I1 + θ2 × I2 + I3

Substituting the above expression into equation ?? and rearranging we get:

Ri = (a0i + bi3 × θ0 ) + (b0i1 + bi3 × θ2 ) + (b∗i2 + bi3 ∗ θ2 ) × I2 + b∗i3 × I3 + ci

In the above equation, the first term and all the coefficients of the new orthogonal indices
are constants, so we can rewrite the equation as follows, getting a three-index model with
orthogonal indexes:
Ri = ai + bi1 × I1 + bi2 × I2 + bi3 × I3 + ci
Where ai = a∗i + b∗i2 × γ0 + bi3 × θ0 , bi1 = b∗i1 + b∗i2 × γ1 + b∗i3 × θ1 , bi2 = b∗i2 + b∗i3 × θ2 and
bi3 = b∗i3 .

Exercise 2.12.

(a) In a three-index model we have:

Ri = ai + bi1 × I1 + bi2 × I2 + bi3 × I3 + ci

Since E [Ci ] = 0, we

E [Ri ] = ai + bi1 × E [I1 ] + bi2 × E [I2 ] + bi3 × E [I3 ]

(b) To derive the variance we need to recall three assumptions of a multi-index model
1. the indexes are uncorrelated: E [Ii Ij ] = E [Ii ] E [Ij ]

37
 2. the specific factors of each security are independent: E [ci cj ] = 0
3. For any security, each index factors are independent of the specific factors of that
same security: E [Ii ci ] = 0
2 2
4. E [ci ] = σci
h
2 i
Now we can apply the variance formula σi2 = E Ri − R̄i , such that
h

2 i
σi2 = E ai + bi1 × I1 + bi2 × I2 + bi3 × I3 + ci − ai + bi1 × I¯1 + bi2 × I¯2 + bi3 × I¯3
h

2 i
= E b1i I1 − I¯1 + b2i I2 − I¯2 + b3i I3 − I¯3

Carrying out the squaring, noting that the indices are all orthogonal with each other and
using the stated assumptions gives us
σi2 = b2i1 σI1
2
+ b2i2 σI2
2
+ b2i3 σI3
2
+ σc2i

(c) Here weapply exactly

the same

 reasoning that we used in part b. Covariance is given by
σij = E Ri − R̄i Rj − R̄j .Thus,
ai + bi1 × I1 + bi2 × I2 + bi3 × I3 + ci − ai + bi1 × I¯1 + bi2 × I¯2 + bi3 × I¯3 ×





σij = E
× ai + bi1 × I1 + bi2 × I2 + bi3 × I3 + cj − aj + bj1 × I¯1 + bj2 × I¯2 + bj3 × I¯3
b1i I1 − I¯1 + b2i I2 − I¯2 + b3i I3 − I¯3 ×
"



#
=E
× b1j I1 − I¯1 + b2j I2 − I¯2 + b3j I3 − I¯3

Carrying out the squaring, noting that the indices are all orthogonal with each other and
using the stated assumptions gives us
2 2 2
σij = bi1 bj1 σI1 + bi2 bj2 σI2 + bi3 bj3 σI3

Exercise 2.14. To build such model, we can use all kind of economic explanatory factors,
such as, GDP growth rate, inflation rate, interest rate, or firms characteristics that proxies risk
factors as size, book to market ratio, sales/equity ratio, price/earnings or a market factor. For
example, Fama and French (1992 and 2003) developed in the early 90s a three factor model,
whose factors were variables built to capture size, the relation between book-value and market-
value and the market return. Earlier, late 80s, Burmeister, McElroy (1987 and 1988) and other
found that five variables are sufficient to describe security returns: two variables were related
to the discount rate used to find the present value of cash flows; one related to both size of
the cash flows and discount rates; one related only to cash flows; and a remaining variable that
captures the impact of the market not incorporated in the first four variables.

Exercise 2.15.

(a) By definition the risk-free asset does not have any risk, so that the sensitivity to risk
factors must be zero. Thus, bF 1 = 0 ∧ bF 2 = 0
(b) From the presented two-index model we know the expected return of any security is
R̄i = ai + bi1 R̄I1 + bi2 R̄I2
The above model is valid for any security including security B that is explained by factor
2, since bi1 = 0. Thus, we have
R̄B = aB + bB2 R̄I2

9.5 = −0.1 + 1.2R̄I2

9.6
R̄I2 = =8
1.2

38
(c) The expected return of security A is

R̄A = aA + bA1 R̄I1 + bA2 R̄I2

= 0.2 + 1.2 × 15 − 0.15 × 8
= 17

(d) Total risk as measured by standard deviation is

q
2 + b2 σ 2 + σ 2
σi = b2i1 σI1 i2 I2 ci

And the systematic risk is measure by

q
σi = 2 + b2 σ 2
b2i1 σI1 i2 I2

Thus, the risk of A, B and C is

q p
σA = b2A1 σI1
2 + b2 σ 2 =
A2 I2 1.22 × 252 − 0.152 × 52 = 30
q p
σB = b2B1 σI1
2 + b2 σ 2 =
B2 I2 0.82 × 252 + 02 × 52 = 20
q p
σC = b2C1 σI1
2 + b2 σ 2 =
C2 I2 02 × 252 + 1.22 × 52 = 6

(e) Variance and covariance are measured, respectively, by

σi2 = b2i1 σI1

2
+ b2i2 σI2
2
+ σc2i

σi2 = bi1 bj1 σI1

2 2
+ bi2 bj2 σI2

Applying the data in the exercise,

2
σA = b2A1 σI1
2
+ b2A2 σI2
2
+ σc2A = 1.22 × 252 − 0.152 × 52 + 52 = 925.56
2
σB = b2B1 σI1
2
+ b2B2 σI2
2
+ σc2B = 0.82 × 252 + 02 × 52 + 22 = 404
2
σB = b2C1 σI1
2
+ b2C2 σI2
2
+ σc2C = 02 × 252 + 1.22 × 52 + 12 = 37
2 2
σAB = σBA = bA1 bB1 σI1 + bA2 bB2 σI2 = 1.2 × 0.8 × 252 − 0.15 × 0 × 52 = 600
2 2
σAC = σCA = bA1 bC1 σI1 + bA2 bC2 σI2 = 1.2 × 0 × 252 − 0.15 × 1.2 × 52 = −4.5
2 2
σBC = σBC = bB1 bC1 σI1 + bB2 bC2 σI2 = 0.8 × 0 × 252 + 0 × 1.2 × 52 = 0

So that, the covariance matrix is

 
925.56 600 −4.5
 600 404 0 
−4.5 0 37

(f) (i) To find the minimum variance portfolio (mvp) we need to take the derivative and
equal to 0 of the portfolio variance in order to XB , which is the weight of security
B in the mvp. Since securities B and C are not correlated and, therefore, ρBC = 0,
we have
2 2
σV2 = XB2 2
σB + (1 − XB ) σC

39
 Taking the derivative, equaling 0 and solving for XB

∂σV2 2 2 2
= 2XB σB + 2 (1 − XB ) (−1) σC =0
∂XB
2
σC
XB = 2 + σ2
σB C

Consequently,
2
σC 37
XB = 2 + σ 2 = 404 + 37 = 0.084
σB C
XC = 1 − XB = 1 − 0.084 = 0.916

Finally the portfolio’s risk is

q p
σV = XB 2 σ2 + X 2 σ2 = 0.0842 × 404 + 0.9162 × 37 = 0.0582
B C C

(ii) If we could invest in a risk-free security, the mvp would be 100% composed with the
risk-free security, since, of course, it is impossible to build a portfolio with less risk
then the risk-free security.
(g) (i) This is a standard portfolio selection exercise, in which we have to choose the tangent
portfolio between the capital market line and the efficient frontier of risky assets.
The solution for this problem involves solving the following system of simultaneous
equations in order to Zi , ∀i = A, B, C
 2
 R̄A − RF = ZA σA + ZB σAB + ZC σAC


 2
R̄B − RF = ZA σBA + ZB σB + ZC σBC
 2
 R̄C − RF = ZA σCA + ZB σCB + ZC σC



Applying the data in the problem,

 

 17 − 5 = 925.56ZA + 600ZB − 4.5ZC 
 ZA = 0.041525

 

12.5 − 5 = 600ZA + 404ZB ⇔ ZB = −0.04311
 
 9.5 − 5 = −4.5ZA + 37ZC  ZC = 0.12667

 


PC
Then, i=A Zi = 0.12509. Therefore, the weights of the tangent portfolio are

ZA 0.041525
XA = PC = = 0.332
i=A Zi 0.12509
ZB −0.04311
XB = PC = = −0.3446
i=A Zi 0.12509
ZC 0.12793
XC = PC = = 1.0126
i=A Zi 0.12509

Finally, the portfolio’s expected return is

C
X
R̄T = Xi R̄i = 0.332 × 17 − 0.3446 × 12.5 + 1, 0126 × 9.5 = 10.96
i=A

40
 The portfolio’s variance is

σT2 = X 0 ΣX
  

925.56 600 −4.5 0.332
= 0.332 −0.3446 1.0126  600 404 0   −0.3446 
−4.5 0 37 1.0126
= 47.61

And portfolio’s risk is

σT = 6.9

(ii) The capital market line is

RT − RF
R̄i = RF + σi
σT
10.96 − 5
=5+ σi
6.9
= 5 + 0.86σi

41
3 Selecting the Optimal Portfolio

3.1 Expected Utility Theory

Exercise 3.1. A fair game is a game where the initial investment equals the expected value
of the payoff, i.e., where we have E(W ) = W0 .
We also know the utility functions of risk neutral investors are linear, while utility functions
of risk averse are concave and of risk lovers are convex functions. See general shapes of utility
function in Figure ??

Figure 10: Exercise ?? – Shape of utility functions for risk (1) lovers, (2) neutral and (3) averse.

By definition of linear, concave and convex functions we have.

(a) For any a and b and p ∈ [0, 1] if the utility function U is linear we have
U (pa + (1 − p)b) = pU (a) + (1 − p)U (b) ⇔ U (E (W )) = E (U (W )) ,
| {z }
W0

thus, we conclude that any risk neutral investor would be indifferent between entering or
not a fair game.
(b) For any a and b and p ∈ [0, 1] if the utility function U is concave we have
U (pa + (1 − p)b)p ≥ U (a) + (1 − p)U (b) ⇔ U (E (W )) ≥ E (U (W )) .
| {z }
W0

So, investors with concave utilities do not enter fair games.

(c) For any a and b and p ∈ [0, 1] if the utility function U is convex we have
U (pa + (1 − p)b)p ≤ U (a) + (1 − p)U (b) ⇔ U (E (W )) ≤ E (U (W )) .
| {z }
W0

So, investors with convex utilities enter fair games.

42
Exercise 3.2.

(a) For the investor with utility U (W ) = −W −1/3 we compute the expected utility of both
investments,

E [U (WA )] = 0.25U (4) + 0.5U (6) + 0.25U (8) = −0.5576

1 1 1
E [U (WB )] = U (4) + U (6.2) + U (8) = −0.5581
3 3 3
and conclude that investor 1, A  B.
(b) For U (W ) = −W −0.1 we get,

E [U (WA )] = 0.25U (4) + 0.5U (6) + 0.25U (8) = −0.8386

1 1 1
E [U (WB )] = U (4) + U (6.2) + U (8) = −0.8387
3 3 3
and conclude that also investor 2, A  B.
(c) Both investors have power utility, thus

U (W ) = −W −α forα > 0
0 −α−1
U (W ) = αW >0
U (W ) = −α(α + 1)W −α−2 < 0
00

U 00 (W ) α(α + 1)W −α−2 1+α 1+α

ARA(W ) = − 0
= −α−1
= ⇒ ARA0 (W ) = − <0
U (W ) αW W W2
1+α
RRA(W ) = W =1+α ⇒ RRA0 (W ) = 0
W
they prefer more to less and they are risk averse with decreasing absolute risk aversion
and constant relative risk aversion.
So, they always keep the same proportion of wealth invested in risky assets. Despite
their similarities in terms of profiles, investor 1 has α = 1/3 = 0.3(3) while investor 2
has α = 0.1, so their coefficients of RRAs are of 1.3(3) and 1.1, respectively, and we can
conclude investor 1 has a higher degree of risk aversion than investir 2.

Exercise 3.3.
Since the coin is tossed twice the game can be summarised by the scheme below.
 
1
1000 × 2 × 2 = 4000
4

1000 × 2

 
1
1000 1000 × 0.05 × 2 = 100
2

1000 × 0.05

 
1
1000 × 0.05 × 0.05 = 2.5
4

43
For log utility we have U (W ) = log (W ), and we have
1 1 1
E [U (Game)] = U (4000) + U (100) + U (2.5) = 4.6051
4 | {z } 2 | {z } 4 | {z }
8.295 4.6051 0.916

Since U (100) = 4.60651, we know the certainty equivalent of the game is C = 100 and, thus,
the investor would be willing to pay up to 900 to avoid the situation.

Exercise 3.4.
 
1
W0 + X
2

W0

 
1
W0 − X
2

(a) (i) For W0 = 1000 and X = 250, the expected utility of the game, and the associated
certainty equivalent, for each of the investor are:
1 1
E [U (Game)] = U (W0 + X) + U (W0 − X)
2 2
1 1
U (W )) = ln (W ) E [U (Game)] = log (1250) + log (750)
2 2
1 1
= (7.13) + (6.62) = 6.875
2 2
ln (C) = 6.875 ⇒ C = 967.78
1
1
V (W )) = 1 − e−0.001W 1 − e−0.001×1250 + 1 − e−0.001×750


E [U (Game)] =
2 2
1 1
= (0.7135) + (0.5276) = 0.62055
2 2
1 − e−0.001C = 0.62055 ⇒ C = 969.03
Investor 1 is willing to pay 1000 − 967.78 = 32.22 and investor 2 is willing to pay
1000 − 969.03 = 30.97.
(ii) The expected utility of the game is
1 1
max E [U (Game)] =U (W0 + X) + U (W0 − X)
X 2 2
the value X that maximizes expected utility is given by the first-order-condition
(F.O.C)
1 0 1
U (W0 + X) − U 0 (W0 − X) = 0
2 2
For both investors we get
U (W ) = ln(W )
1 1 1 1 1
U 0 (W ) = : − =0 ⇔ X=0
W 2 W0 + X 2 W0 − X
V (W ) = 1 − e−0.001W
0.001 −0.001(W0 +X) 0.001 −0.001(W0 −X)
U 0 (W ) = 0.001e−0.001W : e − e =0 ⇔ X=0
2 2

44
 Which is not surprising as risk averse investors would rather not enter fair games
(no matter the W0 or X).
(b) The optimal X = 0 does not change. The amount investors are willing to pay to avoid
the game, however, does depend on the initial wealth
1 1
U (W )) = ln (W ) E [U (Game)] = log (100250) + log (99750)
2 2
1 1
= (11.5154) + (11.5104) = 11.5129
2 2
ln (C) = 11.5129 ⇒ C = 99997.45
1
1
V (W )) = 1 − e−0.001W 1 − e−0.001×100250 + 1 − e−0.001×99750


E [U (Game)] =
2 2
1 1
= (0.9999) + (0.9999) = 0.9999
2 2
1 − e−0.001C = 0.9999 ⇒ C = 99999.99

As the wealth increases the curvature of both utility functions decrease and so they are willing
to pay less to avoid the game.

Exercise 3.5.

(a) Starting from an initial wealth of W0 = 50, the final outcome may be W = 25 or W = 75,
with equal probability.
Given the utility function, we have
1 1
If he enters the game : E [U (Game)] = U (25) + U (75)
2 2
1
25 − 0.005(25)2 + 75 − 0.005(75)2

=
2
= 34.375
If he does not enter the game : U (50) = 50 − 0.005(50)2 = 37.5

So, he chooses not to play the game.

(b) To be indifferent between playing the same or not we need the expected utility of the
game to be the same as the utility of not playing the game. Let us assign a probability p
to the outcome 75 and (1 − p) to 25. We, thus have

p 25 − 0.005(25)2 + (1 − p) 75 − 0.005(75)2 = 37.5

   

46.875 p + 21.875(1 − p) = 37.5

p = 62.5%

(c) The certainty equivalent of the game is the fixed amount that would make the investor
indifferent between playing the game or nor.
In this case we have

U (C) = E [U (Game)]
C − 0.005C 2 = 34.375
p
−1 ± 1 − 4 × (−0.005) × (−34.375) 1 ± 0.5590
C= =
2 × (−0.005) 0.01
⇒ C = 44.1

45
Exercise 3.6.
From the ranking of the projects, X  Y  Z, we know E(UX ) > E(UY ) and E(UY ) > E(UZ ).
Using a second order Taylor approximation of the RTF we also have
1
E(U ) = f (σ, R̄) ≈ R̄ − RRA(W0 )(R̄2 + σ 2 ) .
2
For each project we get
1
fX (30%, 20%) ≈ 0.2 − RRA(W0 )(0.32 + 0.22 ) = 0.2 − 0.065RRA(W0 )
2
1
fY (35%, 15%) ≈ 0.15 − RRA(W0 )(0.152 + 0.352 ) = 0.15 − 0.0725RRA(W0 )
2
fZ (5%, 8%) ≈ 0.08 − half RRA(W0 )(0.082 + 0.052 ) = 0.08 − 0.00445RRA(W0 ) .

and it musty hold

( (
fX (30%, 20%) > fY (35%, 15%) 0.2 − 0.065 RRA(W0 ) > 0.15 − 0.0725 RRA(W0 )
⇔
fY (35%, 15% > fZ (5%, 8%) 0.15 − 0.0725 RRA(W0 ) > 0.08 − 0.00445 RRA(W0 )

Solving the system we get 1.06 > RRA(W0 ) > −6.67, so any investor with RRA(W0 ) within
that range would have the suggested ranking of projects. In particular for risk neutral investors,
with RRA(W0 ) = 0, we also get X  Y  Z.

Exercise 3.7.

(a) The preferred investment will be the one with the highest level of expected utility. Thus,
we have to calculate the utility in each state of economy for the three investments. Given
the utility function U (W ) = 20W − 0.5 ∗ W 2 we get,
For investment A:

U (5) = 20 ∗ 5 − 0.5 × 52 = 87.5

U (6) = 20 ∗ 6 − 0.5 × 62 = 102
U (9) = 20 ∗ 9 − 0.5 × 92 = 139.5

For investment B:

U (4) = 20 ∗ 4 − 0.5 × 42 = 72
U (7) = 20 ∗ 7 − 0.5 × 72 = 115.5
U (10) = 20 ∗ 10 − 0.5 × 102 = 150

For investment C:

U (1) = 20 ∗ 1 − 0.5 × 12 = 19.5

U (9) = 20 ∗ 9 − 0.5 × 92 = 139.5
U (18) = 20 ∗ 18 − 0.5 × 182 = 198

Therefore, the expected utility for each investment is

E [U (WA )] = 87.5 × 1/3 + 102 × 1/3 + 139.5 × 1/3 = 109.67

E [U (WB )] = 72 × 1/4 + 115.5 × 1/2 + 150 × 1/4 = 113.25
E [U (WC )] = 19.5 × 1/5 + 139.5 × 3/5 + 198 × 1/5 = 127.20

So, Investment C is preferred because it has the highest level of expected utility.

46
(b) As before, the preferred investment will be the one with the highest level of expected
utility, so that we have to calculate the utility in each state of economy for the three
investments, now considering the new utility function U (W ) = − √1W .
For investment A:
1
U (5) = − √ = −0.4472
5
1
U (6) = − √ = −0.4082
6
1
U (9) = − √ = −0.3333
9
For investment B:
1
U (4) = − √ = −0.5
4
1
U (7) = − √ = −0.3750
7
1
U (10) = − √ = −0.3162
10
For investment C:
1
U (1) = − √ = −1
1
1
U (9) = − √ = −0.3333
9
1
U (18) = − √ = −0.2351
18
Therefore, the expected utility for each investment is

E [U (WA )] = −0.4472 × 1/3 − 0.4082 × 1/3 − 0.3333 × 1/3 = −0.3963

E [U (WB )] = −0.5 × 1/4 − 0.3780 × 1/2 − 0.3162 × 1/4 = −0.3930
E [U (WC )] = −1 × 1/5 − 0.3333 × 3/5 − 0.2357 × 1/5 = −0.4471

With this new utility function, Investment B is preferred because it has the highest level
of expected utility.
(c) For investments A and B be indifferent, using the first utility function, their expected
utility must equal. Therefore, what must be the probability π associated to payoffs 4 and
10 of investment B to have such equality?

A ∼ B ⇐⇒ E [U (WA )] = E [U (WB )]

Thus
E [U (WA )] = E [U (WB )]

109.67 = 72 × π + 115.5 × (1 − 2π) + 150 × π

π = 0.648
Since we must have 0 ≤ π ≤ 0.5, otherwise the new probabilities would not be between 0
and 1, this means investor 1 will never be indifferent between investments A and B. He
always prefer B to A .

47
 (d) For investments B and C be indifferent, using the second utility function, their expected
utility must be the same. In part c we vary the probability associated to certain payoffs,
now we allow for a change in the lowest payoff of these two investments, which is 1 for
Investment C. So,
B ∼ C ⇐⇒ E [U (WB )] = E [U (WC )]
Thus
E [U (WB )] = E [U (WC )]

−0.3963 = U (x) × 1/5 − 0.3333 × 3/5 − 0.2357 × 1/5

U (x) = −0.7456
Since U (x) = − √1x we finally have

1
U (x) = − √
x
1
−0.7456 = − √
x

x = 1.7987

Exercise 3.8. (a) To analise the investor behaviour towards risk we need to study its utility
function and its economics proprieties, which is done taking the first and the second
derivative. With the utility function U (W ) = −w−1/2 and assuming W > 0, we have
1 −3/2
U 0 (W ) = W
2
Since W > 0 it comes U 0 (W ) > 0, which means the investor prefers more to less. This
attribute is known as nonsatiation. The second derivative is
3
U 00 (W ) = − W −5/2
4
Which smaller than 0, so that the investor shows risk aversion.
(b) Absolute aversion is calculated by taking the first derivative of a measure of absolute
aversion that is
U 00 (W )
ARA(W ) = − 0
U (W )
Therefore,
3 −5/2
U 00 (W ) 4W 3 −1
ARA(W ) = − = 1 = W
U 0 (W ) 2W
−3/2 2
And,
3
ARA0 (W ) = − W −2
2
0
Since ARA (W ) < 0, the investor exhibits decreasing absolute risk aversion. In practical
terms, this means the investor increases the amount of money invested in risky assets
when her wealth increases.
Relative aversion is a similar to absolute aversion, but its calculated in proportional terms.
So, we need to take the first derivative of a measure of relative risk aversion that is

W U 00 (W )
RRA(W ) = −
U 0 (W )

48
 Therefore,
3 −5/2
W U 00 (W ) 4W W 3
RRA(W ) = − = 1 =
U 0 (W ) 2 W −3/2 2
And,
RRA0 (W ) = 0
Since RRA0 (W ) = 0, the investor exhibits constant relative risk aversion. In practical
terms, this means the percentage invested in risky assets remains constant when her
wealth increases.

Exercise 3.9.

(a) Since U (W ) = ae−bW , we have U 0 (W ) = −abe−bW and U 00 (W ) = ab2 e−bW . To have

a risk averse investor we need U 00 (W ) = ab2 e−bW < 0. Since, e−bW > 0 and b2 is
positive, then a must be negative (a < 0). On the other hand, to respect the nonsatiation
assumption we need U 0 (W ) = −abe−bW > 0. Again e−bW > 0. Because a < 0 we have
−a > 0, which implies a positive b.

(b) (i) If the investor decides not to do the risky investment, he keep the 1000 and has an
utility of E [U (Invest)] = ae−b1000 .
If he decides do do the risky investment he faces
 
1
1500
2

1000
 
1
700
2

and his expected utility from the investment is

1 −b1500 1 −b700 e−b500 + e+b300

E [U (Invest)] = ae + ae = ae−b1000
2 2 2
To compare the utility of not investing with the expected utility of the investment
−b500
+e+b300
we need to compare 1 with e 2 , which does not depend on a, but only on
b. The investor chooses the risky investment when

E [U (Invest)] > U (1000)

−b500
e + e+b300
ae−b1000 > ae−b1000
2
e−b500 + e+b300 > 2 .

(ii)

U (C) = E [U (Invest)]
1 1
e−Cb = e−b1500 + e−b700
2 2 
1 −b1500 1 −b700
−Cb = ln e + e
2 2
1 −b1500
+ 12 e−b700


ln 2 e
C=−
b

49
 The certainty equivalent of a risky investment is the certain (fixed) amount that
makes the investor indifferent between keeping that fixed amount or entering the
risky investment. It can also be interpreted as the maximum amount the investor
would be willing to “pay” to enter the risky investment.
ln( 1 e−b1500 + 12 e−b700 )
(iii) For b = 0.01 we have C = − 2 0.01 = 769.28. Since it is less than 1000
we can conclude that in this case the investor will not do the risky investment.

Exercise 3.10.

(a) See Figure ??.

Figure 11: Exercise ?? – Utility function for relevant wealth levels (W < 50).

(b) To describe this investor behaviour towards risk we need to study the following properties

– Nonsaciation
– Risk attitude (risk aversion)
– Absolute risk aversion
– Relative risk aversion

The investor respects the nonsaciation assumption if U 0 (W ) > 0. Since

U 0 (W ) = 50 − W

This propriety is respected if and only if W < 50.

To study the second property we take te second derivative

U 00 (W ) = −1 < 0

50
 Consequently, the investor shows risk aversion for the feasible values for wealth (W ∈ ]0, 50[).
Geometrically, in the allowed domain, the function is increasing and concave, being a
parable turned down (see Figure ??).
About absolute risk aversion we known
U 00 (W ) −1 −2
ARA(W ) = − = (50 − W ) ARA0 (W ) = (50 − W ) >0
U 0 (W )
Thus, this investor exhibits an increasing absolute risk aversion, i.e. when her wealth
increases she will invest a small amount of money in risky assets.
About relative risk aversion we have
W U 00 (W ) −1 50
RRA(W ) = − = W (50 − W ) RRA0 (W ) = 2 >0
U 0 (W ) (50 − W )
Thus, this investor exhibits an increasing relative risk aversion, i.e. when her wealth
increases she will invest a small percentage of her wealth in risky assets.
(c) This investor will chose the project with higher expected utility. Thus for investment X,
we have for each state of economy
1 1
U (10) = 50W − W 2 = 50 × 10 − × 102 = 450
2 2
1 1
U (40) = 50W − W 2 = 50 × 40 − × 402 = 1, 200
2 2
1 1
U (25) = 50W − W 2 = 50 × 25 − × 252 = 937.5
2 2
For investment Y,
1 1
U (20) = 50W − W 2 = 50 × 20 − × 202 = 800
2 2
1 1
U (40) = 50W − W 2 = 50 × 40 − × 402 = 1, 200
2 2
1 1
U (45) = 50W − W 2 = 50 × 45 − × 452 = 1, 237.5
2 2
Thus, expected utilities are
3
X
E [U (WX )] = Pi U (WXi ) = 0.1 × 450 + 0.2 × 1, 200 + 0.7 × 937, 5 = 941.25
i=1

3
X
E [U (WY )] = Pi U (Wyi ) = 0.05 × 800 + 0.9 × 1, 237.5 + 0.05 × 1200 = 1, 181.88
i=1

As E [U (WY )] > E [U (WX )], we have Y  X, i.e. investor’s choice should be project Y .
(d) The risk premium π is the amount the investor is willing to pay to insure against risk,
such that this is a measure of absolute risk aversion. The risk premium is calculated as
π = E [W ] − c where c is the certain equivalent. The certain equivalent is the amount
received with certainty that has the same utility than a lottery
U (c) = E [U (W )] (9)
Thus, for Investment X, we have πX = E [WX ] − cX , where
3
X
E [WX ] = Pi WXi = 0.1 × 10 + 0.2 × 40 + 0.7 × 25 = 26.5
i=1

51
 To find cX we need to use (??)

U (cX ) = E [U (WX )]
1
50cX − c2X = 941.25
2
cX = 74.85 ∨ cX = 25.15

Since cX must be in the range of possible values for WX we have cX = 25.15. Finally, the
risk premium is
πX = E [WX ] − cX = 26.5 − 25.15 = 1.35
Similarly for Investment Y , we have πY = E [WY ] − cY , where
3
X
E [WY ] = Pi WYi = 0.05 × 20 + 0.9 × 40 + 0.05 × 45 = 39.25
i=1

To find cY we use again (??)

U (cY ) = E [U (WY )]
1
50cY − c2Y = 1181.88
2
cY = 38.33 ∨ cY = 61.67

Since cY must be in the range of possible values for WY we have cY = 38.33. Finally, the
risk premium is
πY = E [WY ] − cY = 39.25 − 38.33 = 0.92
As expected the risk premium for investment X is higher due its higher risk level.

Exercise 3.11.

(a) To discover the investor’s attitudes towards risk we can draw her utility function. To do
so we need as many points as we can. From the data in the problem we already have two
points {(R, U ) : (0%, 0) (10%, 10)}.
We also have data on three investment projects and their certain equivalents, CX = 10%,
CY = 20% and CZ = 30%, that can give us another three points.
Thus, for project X
U (CX ) = E [U (RX )]

U (10%) = 0.5U (0%) + 0.5U (30%)

5 = 0.5U (30%)

U (30%) = 10
For project Y we have

U (CY ) = E [U (RY )]

U (20%) = 0.4U (10%) + 0.6U (30%)

U (20%) = 0.4 × 5 + 0.6 × 10

U (20%) = 8

52
 Figure 12: Exercise ?? - Utility Function

Finally, for project Z we have

U (CZ ) = E [U (RZ )]

U (10%) = 0.2U (−10%) + 0.8U (20%)

5 = 0.2U (−10%) + 0.8 × 8

U (−10%) = −7

With five points we can draw the utility function (see Figure ??) and observe the function
is increasing and concave, therefore for equal increases in return the marginal utility is
decreasing. Thus, the investor is risk averse.
(b) The risk premium associated with each of the projects is given by π = E(R) − C, where
C is the certainty equivalent. We thus have

E(RX ) = 15% =⇒ πX = 15% − 10% = 5%

E(RY ) = 22% =⇒ πX = 22% − 20% = 2%
E(RZ ) = 14% =⇒ πX = 14% − 10% = 4%

(c) The previous answer is based on the expected utility theorem and the utility function
proprieties. The expected utility theorem states the rational rules to order different
investment projects and basically it claims that the decision criterion is the maximization
of the expected utility.

(d) To rank the three projects we need to compute their expected utilities. Using the results
from (a) we get

E(U (RX )) = 5 E(U (RY )) = 8 E(U (RZ )) = 5 ,

so the investor prefers project Y to the other two projects and is indifferent between X
and Z, i.e. Y  X ∼ Z
(e) We know consider a game that pays 30% with probability h and 0% with probability
(1 − h). We need to find the probability level h that makes the investor indifferent

53
 between each project and this game.
hX U (30%) + (1 − hX )U (0%) = E(U (RX ))
10hX = 5
hX = 0.5

hY U (30%) + (1 − hY )U (0%) = E(U (RY ))

10hY = 8
hY = 0.8

hZ U (30%) + (1 − hZ )U (0%) = E(U (RZ ))

10hZ = 5
hZ = 0.5
From the above we get the exact same ranking as before: Y  X ∼ Z.

Exercise 3.12.

(a) To find the absolute and relative risk aversion coefficients we first need to take the first
and second derivative of the utility function
4 4
U 0 (W ) = > 0 ∧ U 00 (W ) = − 2 < 0
W W
Thus, she respects the nonsatiation assumption and is risk averse. About absolute and
relative risk aversion we know
U 00 (W ) − 42 1 1
ARA(W ) = − 0
=− W
4 = ⇒ ARA0 (W ) = − 2 < 0, ∀ W > 0
U (W ) W
W W

W U 00 (W ) − 42
RRA(W ) = − 0
= −W W
4 = 1 ⇒ RRA0 (W ) = 0
U (W ) W

Therefore, the investor exhibits decreasing absolute risk aversion and constant relative
risk aversion, i.e. as her wealth increases she always keeps the same proportion invested
in risky assets.
(b) We consider three projects X, Y, Z with only two possible outcomes, 201 and 1, and for
each of them we know E(WX ) = 101, E(WY ) = 61 and E(WZ ) = 71.
(i) Let us consider pX to be the real probability of the outcome 201 in project X and
(1 − pX ) to be the real probability of the outcome 1. Likewise use pY and pZ when
dealing with the other two projects. Then we have
E(WX ) = 101 ⇔ 201pX + (1 − pX ) = 101 ⇔ pX = 0.5
E(WY ) = 61 ⇔ 201pY + (1 − pY ) = 61 ⇔ pY = 0.3
E(WZ ) = 71 ⇔ 201pZ + (1 − pZ ) = 71 ⇔ pZ = 0.35

(ii) Using the probabilities from (i) we can determine the expected utility associated
with each project. We have,
E [U (WX )] = (1 − pX )U (1) + pX U (201) = 0.5 × 2 + 0.5 × 23.2132 = 12.6066
E [U (WY )] = (1 − pY )U (1) + pY U (201) = 0.7 × 2 + 0.3 × 23.2132 = 8.3640
E [U (WZ )] = (1 − pZ )U (1) + pZ U (201) = 0.65 × 2 + 0.35 × 23.2132 = 9.4246
and the ranking is X  Z  Y .

54
 Figure 13: Exercise ?? - Indifference Curves

(iii) The certainty equivalent of project X, CX is the certain amount that gives the
investor the same utility as the expected utility of project X. Likewise for CY and CZ
for projects Y and Z, respectively. The risk premia is defined as πX = E(WX ) − CX
and likelwise for πY , πZ .
12.6066−2
U (CX ) = E [U (WX )] ⇔ 2 + 4 ln(CX ) = 12.6066 ⇔ CX = e 4 = 14.1774
8.3640−2
U (CY ) = E [U (WY )] ⇔ 2 + 4 ln(CY ) = 8.3640 ⇔ CY = e 4 = 4.9086
9.4246−2
U (CZ ) = E [U (WZ )] ⇔ 2 + 4 ln(CZ ) = 9.4246 ⇔ CZ = e 4 = 6.3991
therefore, πX = 101 − 14.1774 = 86.8225, πY = 61 − 4.9086 = 56.0914 and πZ =
71 − 6.3991 = 64.6009 .
(c) Since the new utility function is a linear transformation of the original function
V (W ) = 2U (W ) − 4 = 2 (2 + 4 ln W ) − 4 = 4 + 8 ln −4 = 8 ln W
and taking into account that the new information on expected payoffs is irrelevant because
what matters are expected utilities, the three projects are now ordered exactly in the same
way: X  Y  Z.

Exercise 3.13.

(a) To study her risk profile we need to take the first and the second derivative of the utility
function W − 6W 2 with W < 1/12. So,
U 0 (W ) = 1 − 12W > 0 for W < 1/12, and U 00 (W ) = −12 < 0 .
Thus, the investor prefers more to less, as long as W < 1/12, and his risk averse. The
indifference curves are plotted in Figure ??.

(b) Absolute and relative risk aversion are as follows

U 00 (W ) 12 144
ARA(W ) = − 0
= ARA0 (W ) = 2 >0
U (W ) 1 − 12W (1 − 12W )
W U 00 (W ) 12W 12
RRA(W ) = − = RRA0 (W ) = 2 >0
U 0 (W ) 1 − 12W (1 − 12W )

55
 Therefore, the investor exhibits increasing absolute and relative risk aversion, i.e. as her
wealth increases she reduces the amount and the proportion invested in risky assets.
(c) While absolute risk aversion measures the variation in the amount invested in risky assets
as a function of wealth, the relative risk aversion measures the change in the proportion
invested in risky assets provoked by a variation in wealth.

Exercise 3.14.

(a) The risk tolerance function (RTF) f (σ, R̄) is nothing but the mean-variance representation
of the expected value of the utility function U (W ).
Utility functions are defined in terms of final wealth, while RTF are defined in terms of
returns, but we can always write W = W0 (1 + R). For some utility functions we may not
get a closed-form expression for f (σ, R̄), that only happens in special cases or whenever
returns follow a distribution for which R̄ and σ are sufficient statistics.
Indifference curves are level curves of the RTF, i.e., curves along which the expected
utility is constant f (σ, R̄) = K.
(b) For R̄ = exp(0.7σ) + K we have
 
∂ R̄
= 0.7 exp(0.7σ) > 0
∂σ IC
It is only possible to keep the same K level of expected utility is higher risk levels are
associated with higher expected returns, so we can conclude the investor is risk-averse.
(c) If the efficient frontier is given by R̄ = 0.05 + 0.8σ, then to find the investor optimal we
must find the point where the slopes of the indifference curves and the efficient frontier
are the same.
   
∂ R̄ ∂ R̄
=
∂σ IC ∂σ EF
0.7 exp(0.7σ ∗ ) = 0.8
0.8


∗ log 0.7
σ = = 0.1907
0.7

Exercise 3.15. Solved during lectures.

Exercise 3.16.

(a) For a two assets portfolio the risk is

2
σP2 = XA
2 2 2
σA + (1 − XA ) σB + 2XA (1 − XA ) σAB
In this case we know σAB = 0 and we pretend σP2 = (9.22%)2 . Thus,
2
σP2 = XA
2 2 2
σA + (1 − XA ) σB
2
0.0085 = (10%)2 XA
2
+ (20%)2 (1 − XA )

P1 : XA = 0.9 ∧ XB = 0.1 ∨ P2 XA = 0.7 ∧ XB = 0.3

Since,
R̄P1 = 0.9 × 8% + 0.1 × 12% = 8.4%

R̄P2 = 0.7 × 8% + 0.3 × 12% = 9.2%

Only P2 is efficient. Therefore, {(XA , XB ) ; (0.7, 0.3)} and RP = 9.2%.

56
(b) For a two assets portfolio the return is
R̄P = XA R̄A + (1 − XA ) R̄B
In this case we want to find a portfolio with a return of 11%, so
R̄P = XA R̄A + (1 − XA ) R̄B

11% = 8%XA + 12% (1 − XA )

XA = 0.25 ∧ XB = 0.75
Consequently, the portfolio’s variance is
2
σP2 = XA
2 2 2
σA + (1 − XA ) σB
= 0.252 × (10%)2 + 0.752 × (20%)2
= 0.023125
and its risk is σP = 15.21%.
(c) To find the tangent portfolio between the capital market line and the efficient frontier of
risky assets we have to solve the following system of simultaneous equations in order to
Zi , ∀i > 0
R̄1 − RF = Z1 σ12 + Z2 σ12 + Z3 σ13 + · · · + ZN σ1N




R̄2 − RF = Z1 σ21 + Z2 σ22 + Z3 σ23 + · · · + ZN σ2N






R̄3 − RF = Z1 σ31 + Z2 σ32 + Z3 σ32 + · · · + ZN σ3N



 ..
.





 2 2
R̄N − RF = Z1 σN N + Z 2 σN 2 + Z 3 σN 3 + · · · + Z N σN


which can be written using matricial notation

   

8% − 4% 0, 01 0 0
R̄ − RF 1 = ΣZ ⇔ 1̄2% − 4% =  0 0, 04 −0, 03  Z
1̄5% − 4% 0 −0, 03 0, 0625

where we have used σBC = ρBC σB σC = −0.6 × 20% × 25% = −0.03.

Solving the above equation we get
Z = Σ−1 R̄ − RF 1

where Σ−1 is the inverse covariance matrix, R̄ is a column vector with the securities
returns, RF is a scalar and 1 is a column vector of 1s. Applying this last equation
    
100.0000 0.0000 0.0000 8% − 4% 4
Z = Σ−1 (R − RF 1) =  0.0000 39.0625 18.7500   12% − 4%  =  5.1875 
0.0000 18.7500 25.0000 15% − 4% 4.25
The Zs are proportional to the optimum amount to invest in each security. Then the
optimum proportions to invest in stock k is Xk , where
Zk
Xk = N
P
Zi
i=1

Thus,      
XA 4/13.4375 29.77%
 XB  =  5.1875/13.4375  =  38.60% 
XC 4.25/13.4375 31.63%

57
(d) The tangency portfolio’s return is
3
X
R̄T = Xi R̄i = 0.2977 × 8% + 0.386 × 12% + 0.3163 × 15% = 11.76%
i=1

Since securities A and B and A and C are not correlated, the risk calculation is simplified

σT2 = σA
2 2
XA 2
+ σB 2
XB 2
+ σC XC2 + 2XB XC σBC
= 0.01 × 0.29772 + 0.04 × 0.38602 + 0.0625 × 0.31632 + 2 × 0.3860 × 0.3163 × (−0.03)
= 0.005773

Thus, the portfolio’s risk is σT = 7.60%.

The efficient frontier is given by the line:

R̄T − RF 11.76% − 4%
R̄P = RF + σP = 4% + σP = 4% + 1.022σP
σT 7.59%

(e) (i) The indifference curves are give by R̄ = 0.5σ 2 + 0.965σ + 0.01K, and we have,

∂ R̄IC
= σ + 0.965 > 0, for all σ > 0.
∂σ


Since the indifference curves are upward slopping in the space σ, R̄ , we can conclude
the investor is risk averse.
(ii) The investment decision criterion is to maximize the investor’s expected utility sub-
ject to the efficient frontier. In this case we are given indifference curves, of each
K level of expected utility. So we just need do equal the slopes of the indifference
curves to the slope of the efficient frontier to find the optimal portfolio’s risk. Let us
denote the optimal portfolio with the letter P . Therefore,

∂ R̄EF ∂ R̄IC
=
∂σ ∂σ
1.022 = σP + 0.965

σP = 5.7%

Remember that this optimal portfolio is compose by risk free and portfolio T, so that
its risk is σP = XT σT . Therefore, the weight of portfolio T in the optimal portfolio
is
σP 5.7%
XT = = = 0.75
σT 7.60%
And, of course, XF = 1 − XT = 1 − 0.75 = 0.25. Thus, she must invest 75% in
portfolio T, which corresponds to

XA = 0.2233
 
0.2977 
0.75XT = 0.75  0.3860  =⇒ XB = 0.2895
0.3163

XC = 0.2372


and 25% in the risk free asset. Therefore, she will invest

XA = 89, 302
 
0.2233 

 =⇒ XB = 115, 814
 0.2895  
Investment = 400, 000  0.2372  
 XC = 94, 884
0.25

XF = 100, 000


58
 (iii) From the indifference curves R̄ = 0.5σ 2 + 0.965σ + 0.01K we know K is the fixed
expected utility level , for the three portfolios under analysis we have

R̄T = 0.5σT2 + 0.965σT + 0.01KT

11.76% = 0.5(7.60%)2 + 0.965(7.60%) + 0.01KT =⇒ KT = 4, 137

2
R̄O = 0.5σO + 0.965σO + 0.01KO
9.82% = 0.5(5.70%)2 + 0.965(5.70%) + 0.01KO =⇒ KO = 4, 156

R̄F = 0.5σF2 + 0.965σF + 0.01KF

4% = 0.5(0%)2 + 0.965(0%) + 0.01KF =⇒ KF = 4

from what we can conclude the investor preferences are O  T  F .

(f) (i) The RTF is nothing

but the expected value of the utility function, with domain in
the space σ, R̄ . For the log utility we have

E (U (W )) = E (ln(W ))
= E (ln(W0 (1 + R)))
= ln(W0 ) + E (ln(1 + R))

and, for a general distribution of R, the last expectation cannot be written in terms
of σ = V ar(R) and R̄ = E(R).
(ii) Using a second-order Taylor approximation around W0 we get
1
U (W ) ≈ U (W0 ) + (W − W0 )U 0 (W0 ) + (W − W0 )2 U 00 (W0 )
2
W − W0 1 (W − W0 )2
ln(W ) ≈ ln(W0 ) + −
W0 2 W02
1
ln(W ) ≈ ln(W0 ) + R − (R2 )
2
where we used U 0 (W ) = 1/W and U 00 (W ) = −1/W 2 and W = W0 (1 + R).
The approximation to the RTF is thus
 
1 2
f (σ, R̄) ≈ E ln(W0 ) + R − (R )
2
1
≈ ln(W0 ) + R̄ − E(R2 )
2
1
≈ ln(W0 ) + R̄ − (σ 2 + R̄2 )
2

(iii) Recall the efficient frontier is

R̄P = 4% + 1.022σP

The optimum to the log investor is to maximize the approximation to his RTF which
is equivalent to
1
max R̄P − (σP2 + R̄p2 )
P 2
s.t. R̄P = 4% + 1.022σP

59
 Using the restriction in the objective function we get
1 2
max (4% + 1.022σP ) − (σP2 + (4% + 1.022σP ) )
σP 2
From the FCO we get
1
1.022 − (2σP∗ + 2(4% + 1.022σP∗ )1.022) = 0
2
1.022(1 − 0.04) − (1 + (1.022)2 )σP∗ = 0
σP∗ = 0.4799
So, the log-investor has an optimal risk level of 47.99% and thus he should invest
47.99%
x= = 631.57% =⇒ xF = −531.57% ,
7.60%
assuming he faces no limits on borrowing, the optimal is to borrow 531.57% to invest
631.57% in the tangent portfolio.
(iv) Indifference curves are curves of fixed expected utility, i.e. fixed levels of the RTF,
f (σ, R̄) = K. Using the Taylor approximation in (ii) we have
1
ln(W0 ) + R̄ − (σ 2 + R̄2 ) = K
2
Solving w.r.t. R̄ would give us a quadratic form, so in this case it is easier to solve
w.r.t. σ 2 . We get
IC : σ 2 = 2 (ln(W0 ) − K) + 2R̄ − R̄2

(v) Now we need to re-write the efficient frontier also w.r.t. σ 2 , so we can compare its
slope with the slope of the IC above.
 2
2 R̄ − 0.04
EF : R̄ = 0.04 + 1.022σ =⇒ σ =
1.022
The two curves will have the same slope at
 2  2
∂σ ∂σ
=
∂ R̄ IC ∂ R̄ EF
R̄∗ − 0.04 1
2 − 2R̄∗ = 2
1.022 1.022
2
(1.022) + 0.04
R̄∗ = = 53%
1 + (1.022)2
An expected return of 53% is only possible if we leverage a lot to invest in T ,
concretely
53% = (1 − x)4% + x ∗ 11.76% =⇒ x = 631, 57%.
As expected we get exactly the same optimum as in (iii).
(g) Any investor who is risk neutral, cares only about maximising the expected return of
investments. In the market situation of the exercise, when we can both lend and borrow
at the same rate RF without limits, it is always possible to borrow a bit more to increase
the expected return. Without loss of generality – as the investor is indifferent between all
investments with the same R̄, we can focus on the efficient frontier to show the optimal
∗
risk level is σneutral = +∞.
To see this note that
∗
max R̄P ⇔ max 4% + 1.022σp =⇒ σneutral = +∞
P σp

s.t. EF

60
 (h) In the case of the risk lover we can focus on efficient portfolios, because for any fixed risk
level, those are the ones that maximize expected return and a risk lover likes both risk
and expected return. His optimum can be understood as, first maximize risk and then
for the maximal risk maximize expected return. Or, maximize risk along the efficient
frontier.
Recall the efficient frontier can be written both in terms of R̄P = 0.04 + 1.022σP or
R̄P − 0.04
σP = .
1.022
Formally we can write

R̄P − 4% ∗
max σP ⇔ max =⇒ R̄lover = +∞
P R̄P 1.022
s.t. EF

3.2 Alternatives Techniques

Exercise 3.17.

(a) The geometric mean is given by

N
Y
Pij
R̄jG = 1 + R̄ij −1
i=1

Therefore, the geometric mean returns of the outcomes shown in Exercise ?? (assuming
an initial investment of 100) are:
3
G
Y
PiA
R̄A = 1 + R̄iA − 1 = 1.051/3 × 1.061/3 × 1.091/3 − 1 = 0.0665
i=1

3
G
Y
PiB
R̄B = 1 + R̄iB − 1 = 1.041/4 × 1.071/2 × 1.101/4 − 1 = 0.0698
i=1

3
G
Y
PiC
R̄C = 1 + R̄iC − 1 = 1.011/5 × 1.093/5 × 1.181/5 − 1 = 0.0907
i=1

Thus C  B  A.

(b) The idea of maximizing the geometric mean return to chose the optimal portfolio is
supported by two main arguments:
1. has the highest return probability of reaching, or exceeding, any given wealth level
in the shortest possible time; and
2. has the highest probability of exceeding any given wealth level over any given period
of time.

61
Exercise 3.18.

(a) To use the stochastic dominance criterion we need to calculate the accumulated probability
(first order stochastic dominance - FOSD) and the sum of accumulated probabilities
(second order stochastic dominance - SOSD). Table ?? exhibits the accumulated and sum
of accumulates probability.
Thus, using the accumulated probability we cannot find any FOSD. However, when we
consider the sum of accumulated probability, the SOSD allows us to rank the projects,
such that C  B  A.

Accumulated Probability Sum of Accumulated Probability

Return A B C A B C
4% 0.2 0.0 0.0 0.2 0.0 0.0
5% 0.2 0.1 0.0 0.4 0.1 0.0
6% 0.5 0.4 0.4 0.9 0.5 0.4
7% 0.5 0.6 0.7 1.4 1.1 1.1
8% 0.9 0.9 0.9 2.3 2.0 2.0
9% 0.9 1.0 0.9 3.2 3.0 2.9
10% 1.0 1.0 1.0 4.2 4.0 3.9

Table 5: Exercise ?? - FOSD and SOSD

(b) Any risk averse investor would choose the same ranking as above. So any utility function
with U 0 (.) > 0 and U 00 (.) < 0 would do. Log, negative exponencial, etc.
(c) Roy’s safety first criterion is to minimize P rob (RP < RL ). Then,

P rob (RA < 5%) = 0.2; P rob (RB < 5%) = 0.0; P rob (RC < 5%) = 0.0

Therefore, under this decision criterion, investments B and C are preferable than invest-
ment A, but to the investor investments B and C are indifferent, B v C  A.
(c) Kataoka’s safety first criterion is to maximize RL subject to P rob (RP < RL ) 6 α. For
α = 10%, maximum RL for each of the three possible investments is

IA : RL = 4%; IB : RL = 6%; IC : RL = 6%

As before B and C are preferable to A, but B and C are indifferent, B v C  A.

(d) Telser’s safety first criterion is maximize R̄P subject to P rob (RP 6 RL ) 6 α. In this
problem, the restriction is P rob (RP 6 0.5) 6 0.1, which excludes investment A, because
P rob (RA 6 0.5) = 0.2 what does not respect the restriction. Investments B and C
respect the restriction (P rob (RB 6 0.5) = 0.1 ∧ P rob (RC 6 0.5) = 0.0). However, these
two investments are not indifferent as before. Actually, Telser’s objective is to maximize
R̄P , so that we must chose the investment with higher expected return. Thus,
5
X
R̄B = PBi RBi = 0.1 × 5 + 0.3 × 6 + 0.2 × 7 + 0.3 × 8 + 0.1 × 9 = 7
i=1

4
X
R̄C = PCi RCi = 0.4 × 6 + 0.3 × 7 + 0.2 × 8 + 0.1 × 10 = 7.1
i=1

Then R̄C > R̄B ⇒ C  B.

62
 (e) The geometric mean is given by
N
Y
Pij
R̄jG = 1 + R̄ij −1
i=1

Therefore, the geometric mean returns of the outcomes are:

4
G
Y
PAi
R̄A = 1 + R̄Ai − 1 = 1.040.2 × 1.060.3 × 1.080.4 × 1.10.1 − 1 = 0.0678
i=1

5
G
Y
PBi
R̄B = 1 + R̄Bi − 1 = 1.050.1 × 1.060.3 × 1.070.2 × 1.080.3 × 1.090.1 − 1 = 0.0699
i=1

4
G
Y
PCi
R̄C = 1 + R̄Ci − 1 = 1.060.4 × 1.070.3 × 1.080.2 × 1.10.1 − 1 = 0.0709
i=1

Thus C  B  A.

Exercise 3.19.

(a) The solution to this exercise is similar to that one of Exercise ??. However, we now have
a continuous distribution what makes the calculations considerably more nasty if done
with bare hands and qualifies the exercise to be solved using Excel or a similar software.
So you may want to ask your instructor the excel file with the solution. Nevertheless we
present the charts with the FOSD and SOSD (see Figure ??), from which we can conclude
that none of these investments show FOSD or SOSD over the remaining ones.
(b) Recall that Roy’s safety first criterion is to minimize P rob (RP < RL ). Therefore we want
to calculate the following probabilities and rank them accordingly

Pr (RA < 5%) ; Pr (RB < 5%) ; Pr (RC < 5%)

Since, the returns follow normal distributions that are not standardised, we need to stan-
dardise them. Recall that,
RA − R̄A
= Z v N (0, 1)
σA
Then,
     
RA − R̄A 0.05 − 0.1 1 1
Pr (RA < 5%) = Pr < = Pr ZA < − =N − = 0.3694
σA 0.15 3 3
 
RB − R̄B 0.05 − 0.12
Pr (RB < 5%) = Pr < = Pr (ZB < −0.41176) = N (−0.41176) = 0.3400
σB 0.17
     
RC − R̄C 0.05 − 0.15 1 1
Pr (RC < 5%) = Pr < = Pr ZC < − =N − = 0.3694
σC 0.30 3 3

Therefore, under this decision criterion, investments B is preferable than investment A

and C, which are indifferent, B  A v C.

63
 (a)

(b)

Figure 14: Exercise ?? – first (a) and second-order (b) stochastic dominance graphs.

(c) Kataoka’s safety first criterion is to maximize RL subject to P rob (RP < RL ) 6 α. For
α = 10%, maximum RL for each of the three possible investments is:
– Investment A
P rob (RA 6 RLA ) 6 α
 
RLA − R̄A
P rob ZA 6 6α
σA
 
RLA − 0.1
P rob ZA 6 6 0.1
0.15
RLA − 0.1
1 −1.282
0.15
RLA 1 −0.0923

RLA = −0.0922

64
 – Investment B
P rob (RB 6 RLB ) 6 α
 
RLB − R̄B
P rob ZB 6 6α
σB
 
RLB − 0.12
P rob ZB 6 6 0.1
0.17
RLB − 0.12
1 −1.282
0.17
RLB 1 −0.0979

RLB = −0.0978
– Investment C
P rob (RC 6 RLC ) 6 α
 
RLC − R̄C
P rob ZC 6 6α
σC
 
RLC − 0.15
P rob ZC 6 6 0.1
0.30
RLC − 0.15
1 −1.282
0.30
RLC 1 −0.2346

RLC = −0.2345
Thus, A is preferable to B that is preferable to C, A  B  C.

(d) Telser’s safety first criterion is maximize R̄P subject to P rob (RP 6 RL ) 6 α. In this
problem, the restriction is P rob (RP 6 0.5) 6 0.1, which excludes the three investments,
since
P rob (RA 6 0.5) = 0.3694 0.1

P rob (RB 6 0.5) = 0.3400 0.1

P rob (RC 6 0.5) = 0.3694 0.1

(e) The Value at Risk is given by R̄i − Zα σi . Since we set α = 0.025 we have Z0.025 = 1.96.
Therefore,
V aRA = R̄A − 1.96σA = 0.1 − 1.96 × 0.15 = −0.196

V aRB = R̄B − 1.96σB = 0.12 − 1.96 × 0.17 = −0.2139

V aRC = R̄C − 1.96σC = 0.15 − 1.96 × 0.30 = −0.4392

Thus, A is preferable to B that is preferable to C, A  B  C.

65
4 Equilibrium in Financial Markets

4.1 CAPM

Exercise 4.1.

(a) Using the single-index model, the risk of a security i is given by σi2 = βi2 σm
2
+ σe2i , where
the first term is the systematic risk and the second term is the specific risk. Using in the
expression the values given in the problem
2 2 2
σA = βA σm + σe2A = 1.52 + 0.52 + 0.05 = 0.6125

Therefore the risk is σA = 0.783

2
(b) If the specific risk is null, then σe2C = 0. Security’s C variance is σC = 0.75. Thus, using
the single-index model the β of C is
2 2 2
σC = βC σm + σe2C
2
0.75 = βC × 0.25 + 0
βC = 1.73205

(c) From CAPM we know the return of a security is RA = Rf + β (Rm − Rf ). From the data
we know RA = 20% and security B is risk-free (β = 0), so that the risk-free interest rate
is 10%. Thus,


R̄A = Rf + β R̄m − Rf
0.2 = 0.1 + 1.5(R̄m − 0.1)
0.25
R̄m =
1.5
= 0.1667

(d) These assumptions are those of CAPM. See your notes or the textbook.

Exercise 4.2.

(a) From CAPM we know the return of a security is RA = Rf + β (Rm − Rf ) and its β
σ 2
is β = σi,m
2 . Since the market risk is 0.1, its variance is σm = 0.01. The covariance
m
between asset’s i return and the market return is given by σi,m = σi σm ρi,m . Finally,
ρi,m = 1, since security i is perfectly correlated with the market. So, using the given data,
σi,m = 0.2 × 0.1 × 1 = 0.02. Thus,
σi,m 0.02
β= 2
= =2
σm 0.01
and

Ri = Rf + β (Rm − Rf )
= 0.05 + 2 (0.1 − 0.05)
= 0.15

(b) The request line is given by the single-index model Ri = αi + βi R̄m . We know βi and R̄m .
To draw the line we need to find αi , which is given by the expression αi = Ri − βi R̄m . In
this case, αi = 0.15 − 2 × 0.1 = −0.05. The line is represented in Figure ??.

66
 Figure 15: Exercise ?? - Characteristic line

Exercise 4.3.

(a) Using CAPM to calculate the expected return

R̄X = Rf + βX R̄m − Rf = 0.07 + β (0.09 − 0.07)
σXm
βX can be found using βX = 2 .
σm Thus

σXm 0.02
βX = 2
= = 0.8
σm 0.025

Finally,
R̄X = 0.07 + 0.8 (0.09 − 0.07) = 0.086

(b) If R̄m = 0.12 then the expected return is

R̄X = Rf + βX R̄m − Rf = 0.07 + 0.08 (0.12 − 0.07) = 0.11

Since the CAPM’s expected return is lower than the market expected return, the price is
underpriced.

Exercise 4.4. To know the return of each portfolio to look for an arbitrage opportunity we
need to find each portfolio β, which is the weighted average of each security’s β, and each
portfolio’s expected return. Thus

β1 = x1A βA + x1B βB + x1C βC

= −0.5 × 1.5 + 0 × 1 + 1.5 × 0.5
=0

β2 = x2A βA + x2B βB + x2C βC

= 0 × 1.5 − 1 × 1 + 2 × 0.5
=0

67
and

R̄1 = x1A R̄A + x1B R̄B + x1C R̄C

= −0.5 × 0.12 + 0 × 0.1 + 1.5 × 0.05
= 0.015

R̄2 = x2A R̄A + x2B R̄B + x2C R̄C

= 0 × 0.12 − 1 × 0.1 + 2 × 0.05
=0

Therefore, we have two risk-free portfolios with different expected returns, implying an arbitrage
opportunity. So, without investing a single penny we can short-sale portfolio 2 and buy portfolio
1, earning an arbitrage profit of 1.5%.

Exercise 4.5.

(a) To fill the table given in the exercise we need to find βm , βc , R¯A and R¯B . By definition,
βm = 1. Since security C is risk-free, its β is null and Rf = 0.02. Thus the expected
return of securities A and B is


R̄A = Rf + βA R̄m − Rf
= 0.02 + 0.08 × 0.5
= 0.06

R̄B = Rf + βB R̄m − Rf
= 0.02 + 0.08 × (−0.1)
= 0.012

(b) Accordingly to the single-index model total risk is

σi2 = βi2 σm
2
+ σe2i
| {z } |{z}
Systematic Variance Specific Variance

2 2
Thus, for security A the systematic variance is βA σm = 0.52 × 0.042 = 0.0004 and the
2 2 2 2 2
specific variance is σeA = σA − βA σm = 0.12 − 0.0004 = 0.014. Thus, systematic risk
√ √
é 0.0004 = 0.02 and specific risk is 0.014 = 0.1183. For security B the systematic
2 2
risk is βB σm = (−0.1)2 × 0.042 = 0.000016 and the specific risk is σe2B = σB
2 2 2
− βB σm =
2
√
√ − 0.000016 = 0.014384. Thus, systematic risk é 0.000016 = 0.004 and specific risk
0.12
is 0.014384 = 0.1199.
(c) If CAPM holds any investor has always incentives to compose a portfolio with a risk-
free asset and the market portfolio. By holding the market portfolio, well diversified
by definition, the investor will eliminate the portfolio’s specific risk. If CAPM holds,
expectations are homogeneous meaning that all investors share the same expectations,
which should imply a very low level of trading. If, for some reason the expected return in
the market for a given security is the predict by CAPM, it should means the security is not
rewarding properly its systematic risk, therefore, it is not an equilibrium return and we
have an arbitrage opportunity. In this case, expectations are temporarily heterogenous,
until the market adjust to its equilibrium on the security market line.

68
Exercise 4.6.


(a) The equation for the security market line is R̄i = Rf + βi R̄m − Rf . Thus, from the
data in the problem we have:
(
(

R̄1 = Rf + β1 R̄m − Rf 0.06 = Rf + 0.5 R̄m − Rf

⇔

R̄2 = Rf + β2 R̄m − Rf 0.12 = Rf + 1.5 R̄m − Rf

Solving in order to R̄m and Rf , (

R̄m = 0.09
Rf = 0.03
Finally, the the security market line is

R̄i = 0.03 + 0.06βi

(b) Using the above security market line, an asset with a beta of 2 would have an expected
return of:

R̄i = 0.03 + 0.06βi = 0.03 + 0.06 × 2 = 0.15

(c) To exploit an arbitrage strategy we need to find a portfolio with asset 1 and asset 2 that
replicates the risk (βp = 1.2) of the given asset, but with a different return. since the β
of a portfolio is the weighted average of each security β and the weights of asset 1 and
asset 2 must sum 1, it comes
( ( (
x1 + x2 = 1 x2 = 1 − x1 x1 = 0.3
⇔ ⇔
βp = x1 β1 + x2 β2 1.2 = 0.5x1 + 1.5(1 − x2 ) x2 = 0.7

The return of this replication portfolio is Rp = 0.3 × 0.06 + 0.7 × 0.12 = 0.102. Therefore,
we have an arbitrage opportunity that can be exploited by short-selling the replication
portfolio and buying asset 3, making an arbitrage profit of 0.15 − 0.102 = 0.048.

Exercise 4.7.
Given the security market line in this problem, for the two stocks to be fairly priced their
expected returns must be:

R̄X = 0.04 + 0.08 × 0.5 = 0.08

R̄X = 0.04 + 0.08 × 2 = 0.2

If the expected return on either stock is higher than its return given above, the stock is a good
buy.

Exercise 4.8.
Given the security market line in this problem, the two funds’ expected returns would be:

R̄A = 0.04 + 0.19 × 0.8 = 0.192 > 0.1 → bad performance

R̄B = 0.04 + 0.19 × 1.2 = 0.268 > 0.15 → bad performance

Comparing the above returns to the funds’ actual returns, we see that both funds performed
poorly, since their actual returns were below those expected given their beta risk.

69
Exercise 4.9. Part (a) and Part (b) can be answered simultaneously.
The security market line is:

R̄i = Rf + β R̄m − Rf
Substituting the given values for assets 1 and 2 gives two equations with two unknowns and
solving simultaneously gives:
(
(
0.094 = Rf + 0.8 R̄m − Rf R̄f = 0.03

⇔
0.134 = Rf + 1.3 R̄m − Rf R̄m = 0.11

Exercise 4.10. [OBS: this exercise is out of place, it should be in the APT subsection]
A general equilibrium relationship for security returns must imply absence of arbitrage. In this
case we consider systematic risk to be concerned with market risk and interest rate risk. So it
would be interesting to find an expression that explain returns with two risk factors: market
risk; and interest rate risk. To do so, we need to create an arbitrage portfolio as follows:
X
XiARB × 1 = 0 (10)
i
X
aARB = XiARB ai = 0 (11)
i
X
bARB = XiARB bi = 0 (12)
i
Since the above portfolio has zero net investment and zero risk with respect to the given two-
factor model, by the force of arbitrage its expected return must also be zero:
X
R̄ARB = XiARB R̄i = 0 (13)
i

From a theorem of linear algebra, since the above orthogonality conditions (??), (??) and (??)
with respect to the XiARB result in orthogonality condition (??) with respect to the XiARB , R̄i
can be expressed as a linear combination of 1, ai and bi :
R̄i = λ0 × 1 + λ1 ai + λ2 bi (14)
We can create a zero-risk investment portfolio (without systematic risk) to find λ0 as follows:
X
XiZ = 1
i
X
aZ = XiZ ai = 0
i
X
bZ = XiZ bi = 0
i

Substituting the above equations into equation (??) gives:

X X X X
R̄Z = XiZ R̄i = λ0 XiZ + λ1 XiZ ai + λ2 XiZ bi = λ0
i i i i

Then, we can create a strictly market-risk investment portfolio to find λ1 as follows:

X
XiM = 1
i
X
aM = XiM ai = 1
i
X
bM = XiM bi = 0
i

70
Substituting the above equations into equation (??) gives:
X X X X
R̄M = XiM R̄i = λ0 XiM + λ1 XiM ai + λ2 XiM bi = λ0 + λ1
i i i i

or
λ1 = R̄M − λ0 = R̄M − R̄Z
Finally, we can create a strictly interest rate-risk investment portfolio to find λ2 as follows:
X
XiC = 1
i
X
aC = XiC ai = 0
i
X
bC = XiC bi = 1
i

Substituting the above equations into equation (??) gives:

X X X X
R̄C = XiC R̄i = λ0 XiC + λ1 XiC ai + λ2 XiC bi = λ0 + λ2
i i i i

or
λ2 = R̄C − λ0 = R̄C − R̄Z
Substituting the derived values for λ0 , λ1 and λ2 into equation (??), we have:



R̄i = R̄Z + R̄M − R̄Z × ai + R̄C − R̄Z × bi

Exercise 4.11.

(a) In the graph (see Figure ??) , the efficient frontier with riskless lending but no riskless
borrowing is the ray extending from RF to the tangent portfolio L and then along the
minimum-variance curve through the market portfolio M and out toward infinity (assum-
ing unlimited short sales). All investors who wish to lend will hold tangent portfolio L
in some combination with the riskless asset, since no other portfolio offers a higher slope.
Furthermore, unless all investors lend or invest solely in portfolio L, the market portfolio
M will be along the minimum-variance curve to the right of portfolio L, since the market
portfolio is a wealth-weighted average of all the efficient risky-asset portfolios held by
investors, and no rational investor would hold a risky-asset portfolio along the curve to
the left of L.
The expected return on a zero-beta asset is the intercept of a line tangent to the market
portfolio, and the zero-beta portfolio on the minimum-variance frontier must be below
the global minimum variance portfolio of risky assets by the geometry of the graph. Fur-
thermore, by the geometry of the graph, since the risk-free lending rate is the intercept of
the line tangent to portfolio L, and since L is to the left of M on the minimum-variance
curve, the risk-free lending rate must be below the expected return on a zero-beta asset.

(b) The zero-beta security market line is the line in the graph (see Figure ??) extend from the
expected return on a zero-beta asset through the market portfolio and out toward infin-
ity (assuming unlimited short sales). The expected return-beta relationships of all risky
securities risky-asset portfolios (including the market portfolio M and portfolio L) are
described by that line. The other line from the risk-free lending rate to portfolio L only

71
 Figure 16: Exercise ?? - Efficient Frontier

Figure 17: Exercise ?? - Zero-Beta Security Market Line

72
 describes the expected return-beta relationships of combination portfolios of the risk-free
asset and portfolio L; those combination portfolios are not described by the zero-beta
security market line.

Exercise 4.12. If the post-tax form of the equilibrium pricing model holds, then:


R̄i = RF + R̄m − RF − (δm − RF ) τ βi + (δi − RF ) τ

If the standard CAPM model holds, then:

R̄i = RF + R̄m − RF βi

Let us assume that the post-tax model holds instead of the standard model, and δm = RF .
Then, for a stock with (δi − RF ) τ > 0, if you are right and use the post-tax model, you would
correctly believe that the stock has a higher expected return than the stock’s return expected
by the other investor using the standard model.
Similarly, for a stock with (δi − RF ) τ < 0, you would correctly believe the stock has a lower
expected return than the stock’s return expected by the other investor using again the standard
model.
Therefore, if you manage two securities, one with (δi − RF ) τ > 0 and the other with (δi − RF ) τ <
0, you can swap them with the other investor. Since you both have heterogenous expectations,
each one of you will believe that are making an excess return.
Now consider a specific example using the following data for stocks A and B, the market portfolio
and the riskless asset:

βA = 1.0; δA = 8%; βB = 1.0; δB = 0%; R̄M = 14%; δm = 4%; RF = 4%; τ = 0.25

If the post-tax model holds, then you would correctly believe that the equilibrium expected
returns for the two stocks are:
( (
R̄A = 4 + ((14 − 4) − (4 − 4) × 0.25) × 1.0 + (8 − 4) × 0.25 R̄A = 15%
⇔
R̄B = 4 + ((14 − 4) − (4 − 4) × 0.25) × 1.0 + (0 − 4) × 0.25 R̄B = 13%

While the other investor using the standard model would incorrectly believe that the stocks’
equilibrium expected returns are:
( (
R̄A = 4 + (14 − 4) × 1.0 R̄A = 14%
⇔
R̄B = 4 + (14 − 4) × 1.0 R̄B = 14%

You would tend to buy stock A and sell stock B short. Of course, residual risk puts a limit to
the amount of unbalancing you would do. But by some unbalancing, you earn an excess return.
At the same time the other investor using the standard model would be indifferent between the
two stocks. If your tax factor was below the aggregate tax factor (τ lower than 0.25) then you
should buy stock B from the other investor and sell that investor stock A. The fact that this
will lead to higher after-tax cash flows for you is straightforward.

73
4.2 APT

Exercise 4.13.

(a) If APT’s model holds, returns are generated by a multi-index model such that
R̄i = λ0 + λ1 bi1 + λ2 bi2
Where,
λj is the risk premium associated to the risk factor Ij , j = 1, 2
bik is the sensitivity of security i to the risk factor Ij , j = 1, 2
To derive the equilibrium model we need to calculate λj . Since we know the expected
returns for three portfolios X, Y and Z and the sensitivity of each to the risk factors, we
can build a equation system with three equations and three variables:
  
 R̄X = λ0 + λ1 bX1 + λ2 bX2





0.16 = λ0 + λ1 1 + λ2 0.7
 0

 λ = 0.095929
R̄Y = λ0 + λ1 bY 1 + λ2 bY 2 ⇔ 0.14 = λ0 + λ1 0.6 + λ2 1 ⇔ λ1 = 0.0572816

 
 

R̄Z = λ0 + λ1 bZ1 + λ2 bZ2 0.11 = λ0 + λ1 0.5 + λ2 1.5 λ2 = 0.009709
  

Finally,
R̄i = 0.0959 + 0.0573bi1 + 0.0097bi2
(b) If this portfolio does not respect the equilibrium conditions defined in part a, we will find
an arbitrage opportunity. Thus, first we need to check the non arbitrage expected return
for portfolio W:
e
R̄W = 0.0959 + 0.0573bi1 + 0.0097bi2
= 0.0959 + 0.0573 × 1 + 0.0097 × 0
= 0.1532
e
Since, R̄W = 0.1489 > R̄W = 0.13, this portfolio W is not at equilibrium, allowing the
existence of arbitrage opportunities. The low level of the market expected return implies
that the current market price is too high, meaning portfolio W is overpriced. Thus, we
would like to short sell it and buy a fairly priced portfolio that replicates W’s cash flows
and risk. The subsequent increase in W’s supply will force its price to fall until reach a
e
non arbitrage price, such that R̄W = R̄W .
(c) Recall that APT equilibrium model with a risk-free asset is
R̄i = RF + bi1 λ1 + bi2 λ2 (15)
and that if the CAPM is the equilibrium model, it holds for all securities, as well as all
portfolios of securities. Assume the indexes can be represented by portfolios of securities.
Then, if the CAPM holds, the equilibrium return on each λj is given by the CAPM or


λ1 = βλ1 R̄m − RF


λ2 = βλ2 R̄m − RF
Substituting into Equation (??) yields



R̄i = RF + bi1 βλ1 R̄m − RF + bi2 βλ2 R̄m − RF


= RF + (bi1 βλ1 + bi2 βλ2 ) R̄m − RF
Defining βi as (bi1 βλ1 + bi2 βλ2 ) results in the expected return of R̄i being priced by the
CAPM:

R̄i = RF + βi R̄m − RF
Which is a solution with multiple factors fully consistent with CAPM.

74
Exercise 4.14.

(a) (i) As in the previous exercise, if APT’s model holds, returns are generated by a multi-
index model such that
R̄i = λ0 + λ1 bi1 + λ2 bi2
Thus, to find the equation that holds with these three securities we should proceed
as before
  
R̄X = λ0 + λ1 bX1 + λ2 bX2 0.10 = λ0 + λ1 0.5 + λ2 1 λ = 0.0675
 0

 
 

 
R̄Y = λ0 + λ1 bY 1 + λ2 bY 2 ⇔ 0.12 = λ0 + λ1 1 + λ2 1.5 ⇔ λ1 = 0.0015

 
 

R̄Z = λ0 + λ1 bZ1 + λ2 bZ2 0.11 = λ0 + λ1 0.5 + λ2 2 λ2 = 0.025
  

Finally,
R̄i = 0.0675 + 0.015bi1 + 0.025bi2
(ii) The risk-free rate is given by λ0 , thus RF = 0, 0675.
(b) Security D will be at equilibrium if its equilibrium expected return rate equals its market
expected return rate. Thus, we first need to compute the equilibrium expected return
using our APT model,
e
R̄D = 0.0675 + 0.015bi1 + 0.025bi2
= 0.0675 + 0.015 × 2 + 0.025 × 0.5
= 0.1075
e
Since, R̄D = 0.1075 < R̄D = 0.12, this portfolio D is not at equilibrium, allowing the
existence of arbitrage opportunities. The high level of market expected return implies
that the current market price is too low, meaning portfolio D is underpriced. Thus, we
would like to buy it and short sell a fairly priced portfolio that replicates D’s cash flows
and risk. The subsequent increase in D’s demand will force its price to increase until
e
reach a non arbitrage price, such that R̄D = R̄D .
(c) As long as we can manage to find the right proportions to invest in each security, it should
be possible to build the replication portfolio with securities A, B and C. This new portfolio
sensitivity to factor 1 and 2 must equal the sensitivity of security D to these same risk
factors. Since, the portfolio sensitivity is given by the weighted average of each security
sensitivity and the proportions invested in the three securities must sum 1, it comes
  
bD1 = xA bA1 + xB bB1 + xC bC1 2 = xA 0.5 + xB 1 − xC 0.5 x =1
 A

 
 

 
bD2 = xA bA2 + xB bB2 + xC bC2 ⇔ 0.5 = xA 1 + xB 1.5 + xC 2 ⇔ xB = 1

 
 

xA + xB + xC = 1 xA + xB + xC = 1 xC = −1
  

Exercise 4.15.

(a) To create an arbitrage opportunity, it must be possible to make a profit without investment
and risk, which means  3
 X
xi = 0





1=1
 3
X

xi bi,1 = 0




1=1

75
 A possible portfolio that respects these conditions is

x =1
 1


x2 = −2


x3 = 1

P3
Its expected return is R̄p = i=1 xi R̄i = 1 × 12 − 2 × 15 + 1 × 40 = 22.
(b) The equilibrium relationship associated to the arbitrage pricing model is
( (
0.10 = λ0 + λ1 × 1 λ0 = 0
⇔
0.20 = λ0 + λ1 × 3 λ1 = 0.1
Therefore, the APT line is
R̄i = 0 + 0.1bi1 = 0.1bi1
Thus, the missing value is R̄3 = 0.1bi1 = 0.1 × 3 = 0.3
If we compare the expected returns with the equilibrium returns we can conclude
– Since R̄1 = 12% > R̄1e = 10%, if you buy it you will get a return higher than what
you would receive in equilibrium because Security 1 is underpriced. Therefore you
should buy it
– Since R̄2 = 15% < R̄2e = 20%, if you buy it you will get a return lower than what you
would receive in equilibrium because Security 1 is overpriced. Therefore you should
(short) sell it
– Since R̄3 = 40% > R̄3e = 30%, if you buy it you will get a return higher than what
you would receive in equilibrium because Security 1 is underpriced. Therefore you
should buy it
(c) Without transaction costs, a linear relationship between βs and returns implies that
any point outside this line represents an arbitrage opportunity and a abnormal return.
However, if we consider transaction costs, the expected return in equilibrium must be
corrected, falling by the amount they assume. If transaction costs were not constant,
the relationship between βs and returns will not be linear at all. But, if the abnormal
return and the transaction costs occur at the same time, they may cancel or at least
be lower than transactions costs, reaching a new equilibrium outside the original line,
since one cannot earn abnormal returns. Thus, transaction costs may imply a non linear
relationship, which still respects the law of one price and the non arbitragem assumption.

Exercise 4.16.


(a) From the relationship between CAPM and APT we know that λj = R̄m − RF βλj .
Thus, to have consistency between CAPM and the data we need to observe
λ1

 βλ1 = R̄m − RF
(



λ1 = R̄m − RF βλ1


⇔
λ2 = R̄m − RF βλ2  λ2
 βλ2 =



R̄m − RF
From the data in the problem we know R̄m − RF = 0, 04, so we have to calculate λ1 , λ2
and λ3 , using the previously used equilibrium condition R̄i = λ0 + λ1 bi1 + λ2 bi2 . Thus,
  
R̄A = λ0 + λ1 bA1 + λ2 bA2 0.12 = λ0 + λ1 1 + λ2 0.5 λ = 0.1
 0

 
 

 
R̄B = λ0 + λ1 bB1 + λ2 bB2 ⇔ 0.134 = λ0 + λ1 3 + λ2 0.2 ⇔ λ1 = 0.01

 
 

R̄C = λ0 + λ1 bC1 + λ2 bC2 0.12 = λ0 + λ1 3 − λ2 0.5 λ2 = 0.02
  

76
 Finally,
λ1

 0.01
 βλ1 = R̄m − RF

 βλ1 = = 0.25



0.04
⇔
 λ2  β = 0.02 = 0.5

 βλ2 =


λ2
R̄m − RF 0.04

(b) Again, from the relationship between CAPM and APT, the β of each portfolio is given
by βi = (bi1 βλ1 + bi2 βλ2 ). Thus
  
βA = (bA1 βλ1 + bA2 βλ2 ) βA = 1 × 0.25 + 0.5 × 0.5 β = 0.5
 A

 
 

 
βB = (bB1 βλ1 + bB2 βλ2 ) ⇔ βB = 3 × 0.25 + 0.2 × 0.5 ⇔ βB = 0.85

 
 

βC = (bC1 βλ1 + bC2 βλ2 ) βC = 3 × 0.25 − 0.5 × 0.5 βC = 0.5
  

(c) Since λ0 = RF and λ0 = 0.1, then RF = 0.1

Exercise 4.17.

(a) If the APT assumptions hold then, in equilibrium, all securities are in the same plane
b1 /b2 /R̄. Thus, we can use deduce the equilibrium condition R̄i = λ0 + λ1 bi1 + λ2 bi2
solving the equation system, as before
  
R̄X = λ0 + λ1 bX1 + λ2 bX2 0.19 = λ0 + λ1 1 + λ2 0.5 λ = 0.07
 0

 
 

 
R̄Y = λ0 + λ1 bY 1 + λ2 bY 2 ⇔ 0.14 = λ0 + λ1 1.4 + λ2 0 ⇔ λ1 = 0.05

 
 

R̄Z = λ0 + λ1 bZ1 + λ2 bZ2 0.08 = λ0 + λ1 3 − λ2 1 λ2 = 0.14
  

Thus,

RF = λ0 = 0.07
R̄I1 = λ1 + RF = 0.05 + 0.07 = 0.12
R̄I2 = λ2 + RF = 0.14 + 0.07 = 0.21

e
(b) The expected return for portfolio W at equilibrium is given by R̄W = 0.07 + 0.05bi1 +
e
0.14bi2 = 0.07 + 0.05 × 1 + 0.14 × 0 = 0.12. Since E [RW ] = 0.13 > R̄w = 0.12 we know
the security is underpriced being an interesting investment to make (we should buy).
Portfolio’s W risk is similar to the risk of factor I1 (b1 = 1 ∧ b2 = 0), so that a possible
arbitrage strategy is to short sell the index factor (assuming you could do so) and buy
portfolio W .
An alternative is to form a new portfolio P using portfolios A, B and C, such that
b1 = 1 ∧ b2 = 0:
  

 bP 1 = b1 x + b1 y + b1 z 
 1 = 1x + 1.4y + 3z 
 x = −1
  
bP 2 = b2 x + b2 y + b2 z ⇔ 0 = 0.5x − z ⇔ y = 2.5

 
 

x+y+z =1 z = −0.5
  

To compose Portfolio P you would short sell X and Z to buy Y , in the proportions just
computed.
(c) To evaluate the fund’s performance, we need to compare the equilibrium expected return
e
with the actual return. The equilibrium expected return is calculated as R̄W = 0.07 +

77
 0.05bi1 + 0.14bi2 = 0.07 + 0.05 × 1.2 + 0.14 × 0.2 = 0.158. Now, to find the actual return
we can use the Sharpe’s Ratio (SR), defined as

R̄F und − RF
SR =
σF und

R̄F und = SR × σF und + RF (16)

σF und is not given, but if this portfolio is fully diversified it only faces systematic risk,
such that the correlation between the two factors is null and, therefore,

σF2 und = b21F und σI21 + b22F und σI22

= (1.2)2 (0.1)2 + (0.2)2 (0.25)2
= 0.0169

And σF und = 0.13. Applying it in (??), it comes

R̄F und = 0.75 × 0.13 + 0.07 = 0.1675

Thus, the fund has achieved a performance higher than what was expected under equi-
librium.
(d) It is possible since the indexes’ returns I1 and I2 can be explain by CAPM. In that
case, APT and CAPM are
equivalents, as shown in a previous exercise. In this case
R̄I1 = RF + βI1 R̄m − RF and λ1 = R̄I1 − RF such that

λ1 0.05
βI1 = = = 0.625
R̄I1 − RF 0.15 − 0.07

and
λ2 0.14
βI2 = = = 1.75
R̄I2 − RF 0.15 − 0.07
To calculate the portfolios’ βs we know that, in general, βi = bi1 βλ1 + bi2 βλ2 , then

βX = 1 × 0.625 + 0.5 × 1.75 = 1, 5

βY = 1.4 × 0.625 + 0 × 1.75 = 0.875

βZ = 3 × 0.625 − 1 × 1.75 = 0.125

Exercise 4.18.

(a) CAPM and APT pretend to explain expected returns, although through with quite dif-
ferent assumptions. CAPM is a general equilibrium model with very strong assumptions
like homogeneous expectations, while APT only assumes the absence of arbitrage. APT
is also a must more general model than CAPM in the sense it allows returns to be ex-
plained by a set of variables that can help to better explain returns. Nevertheless, under
certain circunstancies (APT’s risk factors being explained by CAPM’s market portfolio)
the two models are equivalent. From an empirical point of view, both models face a major
drawback. CAPM’s market portfolio is impossible to capture since it englobes all possible
and imaginable assets, including non tradable assets like our home. APT can be used
with all kind of variables, however what are the relevant variables no one really knows
and eventually we may not have databases for them.

78


(b) If APT holds, then the two indexes returns are also explained by APT R̄i = 0.07 + 0.03b1i + 0.05b2i ,
but with one singularity: each index only shows sensitivity to one risk factor b. Thus

{I1 ∴ b1 = 1 ∧ b2 = 0}

{I2 ∴ b1 = 0 ∧ b2 = 1}

and
R̄I1 = 0.07 + 0.03 × 1 + 0.05 × 0 = 0.1

R̄I2 = 0.07 + 0.03 × 0 + 0.05 × 1 = 0.12

Finally, it should be straight forward to you that RF = 0.07.


(c) If CAPM holds, then
the equilibrium condition is given by R̄ i = R F + β i R̄ m − RF ,
where βi R̄m − RF captures the systematic risk premium appropriate to security i. The
model does not reward specific risk because we assume fully diversified portfolios. Thus,
it must apply to the securities described in the problem. Using the data given it comes
(
(
0.304 = RF + 1.8 R̄m − RF RF = 0.07

⇔
0.135 = RF + 0.5 R̄m − RF Rm = 0.2

(d) Yes. CAPM and APT are equivalent if the indexes’ returns were explained by CAPM. In
this case,

R̄I1 0.10 − 0.07

 
 βI1 =  βI1 = = 0.23
(
 
R̄I1 = RF + βI1 R̄m − RF
 
R̄m − RF 0.20 − 0.07

⇔ ⇔
R̄I2 = RF + βI2 R̄m − RF  R̄I2  0.12 − 0.07
 βI =
  βI2 =
 = 0.385
2
R̄m − RF 0.20 − 0.07

and the indexes I1 and I2 βs are given by the expression βi = bi1 βI1 + bi2 βI2 . Thus,

βi = bi1 βI1 + bi2 βI2 = 0.23bi1 + 0.385bi2

79
5 Portfolio Management
Exercise 5.1.

(a) Volatility is not always judged as a good risk measure since it considers both systematic
and unsystematic risk. Actually, unsystematic or specific risk can be fully diversified,
therefore the systematic risk should be the only one rewarded, what explains why measures
of systematic risk are more often judged as better risk measures.
(b) Using standard deviation as the measure for variability, the reward-to-variability ratio for
a fund is the fund’s excess return (average return over the riskless rate) divided by the
standard deviation of return, i.e., the fund’s Sharpe ratio. E.g., for fund A we have:

R̄A − RF 14 − 3
= = 1.833
σA 6
See Table ?? for the remaining funds’ Sharpe ratios.
(c) A fund’s differential return, using standard deviation as the measure of risk, is the fund’s
average return minus the return on a naı̈ve portfolio, consisting of the market portfolio
and the riskless asset, with the same standard deviation of return as the fund’s. E.g., for
fund A we have:
   
R̄m − RF 13 − 3
R̄A − RF + × σA = 14 − 3 + × 6 = −1%
σm 5

See Table ?? for the remaining funds’ differential returns based on standard deviation.
(d) A fund’s differential return, using beta as the measure of risk, is the fund’s average return
minus the return on a naı̈ve portfolio, consisting of the market portfolio and the riskless
asset, with the same beta as the fund’s. This measure is often called “Jensen’s alpha”.
E.g., for fund A we have:



R̄A − RF − R̄m − RF × βA = 14 − (3 + (13 − 3) × 1.5) = −4%

See Table ?? for the remaining funds’ Jensen alphas.

(e) Treynor’s ratio is quite similar to Sharpe’s Ratio, but considering β as the appropriate
risk measure. E.g., for fund A we have:

R̄A − RF 14 − 3
= = 7.833
βA 1.5

(f) This differential return measure is similar to Jensen’s Alpha, except that the riskless rate
is replaced with the average return on a zero-beta asset. E.g., for fund A we have:



R̄A − R̄Z − R̄m − R̄Z × βA = 14 − (4 + (13 − 4) × 1.5) = −3.5%

(g) Fund B shows a better performance than Fund A when considering Sharpe’s Ratio. To
invert this the following relationship should hold

R̄A − RF R̄B − RF R̄B − RF R̄B − 3

> ⇔ 1.833 > = ⇔ R̄ > 10.33
σA σB σB 4
So, for the ranking to be reversed, Fund B’s average return would have to be lower than
10.33%.

80
 Sharpe Treynor Differential Jensen’s Differential Return
Fund Ratio Ratio Return (sigma) Alpha (Beta and R̄Z )
A 1.833 7.333 -1% -4% -3.5%
B 2.250 18.000 2% 4% 3.5%
C 1.625 13.000 -3% 3% 3.0%
D 1.063 14.000 -5% 2% 1.5%
E 1.700 8.500 -3% -3% -2.0%

Table 6: Exercise ?? - Answers (b to f)

Exercise 5.2. To compute Sharpe’s ratio (SR), defined as the fund’s excess return (average
return over the riskless rate) divided by the standard deviation of return, we need to know the
funds’ volatility, which we can calculate using the single index model. Thus
q
2 σ2 + σ 2 2
σ A = βA m eA = 1, 3 × 0.3 + 0.003 = 0.3938

q
2 σ2 + σ 2 2
σB = βB m eB = 0.9 × 0.3 + 0.04 = 0.336

Therefore,
R̄A − RF 0.15 − 0.05
SRA = = = 0.2539
σA 0.3938
R̄B − RF 0.09 − 0.05
SRB = = = 0.119
σB 0.336

81
6 Miscellaneous
Exercise 6.1.

(a) (i) Since in this country it is possible to both deposit and lend at the same interest
rate RF = 4%, we know the efficient frontier in the plan risk/expected return is
a straight line passing trough the risk free asset and the so-called tangent portfolio
(that is the only portfolio composed only of risky investments that is efficient). Thus,
the efficient frontier in this country is given by
R̄T − RF 4
R̄p = RF + σp ⇔ R̄p = 4% + σp
σT 3
where p is an efficient portfolio.
To check whether A is efficient or not we must see if it is on the straight line above
4 4
R̄A = 4% + σA ⇔ 8% = 4% + 3% ⇔ 8% = 8%
3 3
and we can conclude portfolio A belong to the efficient frontier and, thus, is an
efficient portfolio. The optimal portfolio for a super averse investor is the portfolio
that maximizes the risk tolerance function f (σ, R̄) = 2.272R̄ − R̄2 − σ 2 subject to the
restriction it must be an efficient portfolio so, R̄p = 4% + 43 σp . To get the optimal
portfolio we need to
4
max f (σ, R̄) = 2.272R̄p − R̄2 − σp2 s.t R̄p = 4% + σp
3
which is equivalent to the following unrestricted problem
   2
4 4
max f˜(σ) = 2.272 4% + σp − 4% + σp − σp2
3 3
The FOC of the problem is
∂ f˜
 
4 4 4
=0 ⇔ 2.272 × − 2 4% + σp − 2σp = 0 ⇔ σO = 0.96% .
∂σ 3 3 3
The expected return of the optimal portfolio O is then given by
4
R̄O = 4% + 0.96% = 5.297%.
3
To obtain the optimal portfolio’s composition we must rely on the fact the optimal
portfolio is efficient and any efficient portfolio is a combination of the risk free asset
with the tangent portfolio. Thus
R̄O = xF RF +(1 − xf ) R̄T ⇔ 5.297% = 4%xF +12% (1 − xF ) ⇔ xF = 84% .
So, the optimal portfolio for a super averse investor requires depositing 84% of the
initial amount and investing the remaining 16% in the tangent portfolio T .
(iii) If the simply averse invest 120% in the tangent portfolio that means they are lever-
aging themselves and taking a loan equivalent to 20% of their initial amount. Thus,
they are shortselling the risk free asset, i.e. xF = −20%. Their expected return is
R̄simply = 4% × (−20%) + 12% × 120% = 13.6% .
Since this point must also be on the efficient frontier we also have optimal risk level
must satisfy
4 4
R̄simply = 4% + σsimply ⇔ 13.6% = 4% + σsimply ⇔ σsimply = 7.2% .
3 3

82
 (iv) Total amount deposited = 1 million super averse × 1000 euros × 84% = 840 000
euros
Total amount of loans = 4 million simply averse × 2000 euros × 20% = 1 600 000
euros Since 1600000 6= 840000 we conclude the market is not in equilibrium.
(b) (i) We now know two efficient portfolios: T and B both belonging to the hyperbola
that by the envelop theorem is the frontier of the investment opportunity set of
combinations of risky assets. By the Merton theorem we also know two portfolios
are enough to derive the entire frontier, so the minimum variance portfolio M V is
also a combination of T and B.
The variance of any combination of T and B is given by
2
σ 2 = xT σT2 + (1 − xT ) σB
2
+ 2xT (1 − xT ) σT σB ρT B .
The minimum variance portfolio minimizes is the only with the lowest possible risk,
so it is it is the one that
2
min xT σT2 + (1 − xT ) σB
2
+ 2xT (1 − xT ) σT σB ρT B
⇔
2
min xT (6%)2 + (1 − xT ) (12%)2 + 2xT (1 − xT ) (6%)(12%)0.6
From the FOC we get
∂σ 2
=0 ⇔ (6%)2 − 2(12%)2 (1 − xT ) + 2(6%)(12%)0.6 − 4xT (6%)(12%)0.6 = 0
∂xT
⇔ xT = 107.69% ,
and the minimum variance portfolio involves shortselling portfolio B (xB = −7.69%)
to invest more than 100% in portfolio T (xT = 107.69%).
(ii) See slides from classes for the sketch.
In this case the efficient frontier has three branches: (i) a segment of a straight line
from the deposit rate to the first tangent portfolios; (ii) a portion of the envelope
hyperbola (between the two tangent portfolios) and (iii) another segment of a line for
risk levels higher than the risk of the second tangent portfolio (the tangent obtained
using the active interest rate).
∗
(iii) If the optimal risk levels do not change, then we know σsuper = 0.96% (from the
∗
exercise) and σsuper = 7.2% (from point a(iii)).
For the super averse investor nothing changes since their optimal risk level is below
the risk of portfolio T and the deposit rate did not change. So they still invest 84%
in the risk free asset and 16% in portfolio T .
For the simply averse investors we only know their optimal risk level is higher than
σT , but we do not know whether it is bellow risk level of the tangent portfolio when
we take the intersection with the yy-axis to be 7%, the portfolio usually denoted by
T 0.
We thus need first to determine portfolio T 0 . This portfolio must also be a combi-
nation of T and B and is the portfolio that
x̄T R̄T + xB R̄B − 7%
max p 2 2 s.t. xT + xB = 1.
xT ,xB xT σT + x2B σB2 + 2x x σ σ ρ
T B T B TB

The first order conditions are equivalent to the following system of linear equations
in zT , ZB and we know the z’s are proportional to the x’s,
( (
R̄T − 7% = σT2 zT + σT B zB 1̄2% − 7% = (6%)2 zT + (6%)(12%)0.6zB
2
⇔ ⇔
R̄B − 7% = σT B zT + σB zB 1̄5% − 7% = (6%)(12%)0.6zT + (12%)2 zB
( (
z̄T = 11.28472 x̄T = 83.87%
⇔
z̄B = 2.170139 x̄B = 16.13%

83
 Portfolio T 0 requires investing 83.87% in portfolio T and 16.13% in portfolio B. The
expected return and risk of T 0 are given by

R̄T 0 =83.37% × 12% + 16.13% × 15% = 12.48%

p
σT 0 = (83.37%)2 × (6%)2 + (16.13%)2 (12%)2 + 2(83.37%)(16.13%)(6%)(12%)0.6 = 6.38%

Since the optimal risk level of the simply averse is higher than σT 0 , we know simply
averse investors will take a loan to leverage themselves, even with the higher rate of
7% and invest more than 100% in T 0 .
The expected return is R̄simply = 7% + 12.48%−7%
6.38% 7.2% = 13.18% and therefore we
can see how much is the leverage:

13.18% = 7%xF + 12.48%(1 − xF ) ⇔ xF = −12.77% ⇒ xT 0 = 112.77% .

Simply averse investor take a loan to increase their capital by 12.77% and invest all
their money in portfolio T 0 .

Exercise 6.2.

(a) Since the expression for the efficient frontier is a straight line we know

R̄T − RF
R̄p = RF + σp ,
σT
which tells us that: (i) in this market there is a risk-free asset and that borrowing and
lending is possible at the exact same rate RF = 3.5%, also (ii) since the slope of the efficient
R̄T − RF
frontier equals the Sharpe ratio of the tangent portfolio we have SRT = =
σT
0.3436
(b) (i) Mr. Silva has a quadratic utility function. For his particular function we have:
50
• U 0 (W ) = 50 − 2(0.01)W > 0 for wealth levels that satisfy W < = 2500.
0.02
So, for a interval big enough around his initial wealth he prefers more to less.
• U 00 (W ) = −0.02 < 0. From this we conclude Mr. Silva is risk averse.
U 00 (W ) 0.02
• ARA(W ) = − 0 = . Evaluating this function at the initial
U (W ) 50 − 0.02W
wealth W0 = 1000 we get his absolute risk aversion coefficient before investment
0.02
ARA(1000) = 50−0.02×1000 = 0.02
30 . Taking the first derivative of the absolute
risk aversion function we get ARA0 (W ) = (50−0.02W
0.0004
)2 > 0 and we can conclude
Mr. Silva has increasing absolute risk aversion, i.e. when his wealth increases
he will decrease the amount of euros invested in risky assets.
0.02W
• RRA(W ) = ARA(W )W = . Evaluating this function at the initial
50 − 0.02W
wealth W0 = 1000 we get his relative risk aversion coefficient before investment
0.02×1000
RRA(1000) = 50−0.02×1000 = 20
30 . Taking the first derivative of the relative risk
aversion function we get RRA0 (W ) = (50−0.02W 50
)2 > 0. Not surprisingly (given
his increasing absolute risk aversion) Mr.Silva also has increasing relative risk
aversion, i.e. when his wealth increases he keeps a smaller portion of his wealth
in risky assets.
(ii) The risk tolerance function gives us for each pair of volatility and expected return,
(σ, R̄), the expected utility of an investor.

84
 To derive Mr. Silva’s risk tolerance function we need to compute the expected value
of his utility function rewriting it in terms of returns, instead of wealth. Note that
by definition of what wealth W and return R are, we get W = W0 (1 + R).

f (σ, R̄) = E [U (W )] = E [U (W0 (1 + R))]

E 50W0 (1 + R) − 0.01W02 (1 + R)2
 
=
50W0 (1 + E(R)) − 0.01W02 E (1 + R)2
 
=
 

= 50W0 (1 + E(R)) − 0.01W02 1 + 2R̄ + E(R2 )

 
| {z }
σ 2 +R̄2

Using W0 = 1000 and simplifying we have Mr.Silva risk tolerance function

f (σ, R̄) = 40000 + 30000R̄ − 10000σ 2 − 10000R̄2

(iii) To find Mr.Silva’s optimal risk level we have to maximize his risk tolerance function,
subject to the efficient frontier.

max f (σ, R̄) s.t. R̄ = 3.5% + 0.3436σ

σ,R̄

Including the restriction in the objective function we get

f (σ, R̄)|R=3.5%+0.3436σ = 40000+30000(3.5%+0.3436σ)−10000σ 2 −10000(3.5%+0.3436σ)2

This new restricted f function, depends only on σ. So to get its maximum we need
to take its first derivative w.r.t. σ and set it to zero
∂f
= 0
∂σ ∗
30000 × 0.03436 − 20000σ ∗ − 20000(0.035 + 0.03436σ ∗ )0.3436 = 0
∗ ∗
3 × 0.3436 − 2σ − 2 × 0.3436 [0.035 + 0.3436σ ] = 0
∗
σ = 23.13%

(c) (i) We start by computing the inputs to mean-variance theory

R̄1 = 0.25(−5%) + 0.5(0%) + 0.25(50%) = 11.25%

R̄2 = 0.25(10%) + 0.5(−5%) + 0.25(35%) = 8.75%
σ12 = 0.25(−5% − 11.25%)2 + 0.5(0% − 11.25%)2 + 0.25(50% − 11.25%)2 = 0.05047
⇒ σ1 = 22.46%
σ22 = 0.25(10% − 8.75%)2 + 0.5(−5% − 8.75%)2 + 0.25(35% − 8.75%)2 = 0.02672
⇒ σ2 = 16.35%
σ12 = 0.25(−5% − 11.25%)(−5% − 11.25%) + 0.5(0% − 11.25%)(−5% − 8.75%) +
+0.25(50% − 11.25%)(35% − 8.75%) = 0.03265

From before we also know there is a risk-free asset with RF = 3.5%. The tangent
portfolio is the one that maximizes the Sharpe ratio which is the same as solving a
linear system of equations in z1 , z2 which are proportional to the optimal weights
( ( (
R̄1 − Rf = σ12 z1 + σ12 z2 11.25% − 3.5% = 0.05047z1 + 0.03265z2 z1 = 1.263158
⇒ ⇔
R̄2 − Rf = σ12 z1 + σ22 z2 R̄2 − Rf = 0.03265z1 + 0.02672z2 z2 = 0.421053

Since z1 , z2 are proportional to the tangent portfolio weights we can easily find them
z1 1.263158 z2 0.421053
xT1 = = = 75% xT2 = = 25%
z1 + z2 1.263158 + 0.421053 z1 + z2 1.263158 + 0.421053

85
 The expected return as risk of the tangent portfolio are as follows
R̄T = 0.75 × 11.25% + 0.25 × 8.75% = 10.625%
σT2 = 0.752 × 0.05047 + 0.252 × 0.02672 + 2 × 0.75 × 0.25 × 0.03265 = 0.0423
σT = 20.57%
An alternative to compute the tangent portfolio’s volatility would be to use its ex-
pected return R̄T and the equation for the efficient frontier
10.625% = 3.5% + 0.3436σT ⇔ σT = 20.57% .

(ii) From before we know the optimal risk level of Mr. Silva is 23.13%. This is a point
in the efficient frontier, so the optimal portfolio expected return is
R̄∗ = 3.5% + 0.3436 × 23.13% = 11.51% .
The optimal portfolio is a particular combination of the risk-free asset and the tan-
gent portfolio. We find out the exact composition by solving
11.51% = 3.5%xF + (1 − xF )10.625% ⇔ xF = −12.45% ⇒ xT = 112.45% .
The optimal for Mr.Silva is to take a loan (of about 12.45% of his initial investment)
to leverage a bit his position and invest 112.45% in the tangent portfolio.
(iii) Yes it would change since the current optimal portfolio involves taking a loan. Pos-
sibly at the new active rate he is no longer interested in taking a loan. His new
optimum is most likely a combination of the tangent portfolio with a second portfo-
lio belonging to the hyperbola that is the frontier of the investment opportunity set
of risky assets.

Exercise 6.3.

(a) (i) We are in a scenario were the correlation between the returns of any two assets is
constant. So the tangent portfolio can be computed using a cut-off method.
However, since shortselling is allowed, one can also simply use the general mean-
variance theory. The inputs are:
   
8% 0.01 0.01 0.0125
R̄ = 12% V =  0.01 0.04 0.025 
15% 0.0125 0.025 0.0625
where all covariances are obtained by multiplying each pair of individual assets
volatility by the constant correlation of +0.5.
 
To find the tangent portfolio we need to solve the system R̄ − RF = V Z
        
5% 0.01 0.01 0.0125 z1 2.85 0.6
0.04 0.025  z2  ⇔ Z = V −1 R̄ − RF = 0.95 ⇒ X = 0.2
 
 9%  =  0.01
12% 0.0125 0.025 0.0625 z3 0.95 0.2

(ii) The expected return and risk of the tangent portfolio are:
 
8%
R̄T = X 0 R̄ = 0.6 0.2 0.2 12% = 10.22%

15%
   
0.01 0.01 0.0125 0.6
σT2 = X 0 V X = 0.6 0.2 0.2  0.01


0.04 0.025  = 0.2 = 0.0151
0.0125 0.025 0.0625 0.2
⇒ σT = 12.323%

86
 (iii) Since it is possible to deposit and borrow at the same rate RF = 3%, the efficient
frontier is a straight line tangent to the investment opportunity set of risky assets.
This line passes trough the risk-free point and the tangent portfolio, thus

R̄T − RF 10.22% − 3%
R̄P = RF + σp , in our case, R̄P = 3%+ σp ⇔ R̄P = 3%+0.586σp
σT 12.323%

(b) (i) The optimal risk level is attained at the point where the some indifference curve is
tangent to the efficient frontier. I.e., they both have the same slope at that point

∂ R̄p ∂ R̄p
=
∂σp EF ∂σp IC

∂ R̄p
The efficient frontier is R̄P = 3% + 0.586σp , and we have = 0.586 Differen-
∂σp EF
∂ R̄p
tiating the indifference curves we get = 2σp + 0.415
∂σp IC
The optimal is thus 0.586 = 2σp∗ + 0.415 ⇔ σp∗ = 8.55% .
(ii) Given the optimal risk level σp∗ = 8.55% and the efficient frontier equation, we get
the optimal expected return

R̄∗ = 3% + 0.586 × 8.55% = 8%

This is attainable by depositing part of the initial wealth and investing the remaining
in the tangent portfolio

8% = xF 3% + (1 − xF )10.22% ⇔ xF = 30% ⇒ xT = 70% .

The optimal for this investor s to deposit 30% of his wealth and to invest the re-
maining 70% in the tangent portfolio.
(c) (i) Nothing changes. It is still possible to deposit and borrow at the same rate, which
means portfolio T is the only combination of risky assets that is efficient. The exact
same portfolio T is feasible because it does not involve shortselling.
(ii) The optimal portfolio remains the same, for the same reason, portfolio T is feasible
even in a world with restrictions to shortsell.
(d) The ranking of assets according to Roy ranks higher assets with lower probability of
undesirable returns. In this case those are returns lower than RL = 5%.
When returns follow normal distributions we know that
R̄ − 5%
min Pr(R̄ ≤ 5%) ⇔ max
σ
and the ranking of the three assets is

15% − 5% 12% − 5% 8% − 5%
C: = 0.4 > B: = 0.35 > A: = 0.35
25% 20% 10%
The best, according to Roy, is C, than B, than A.

87