First Semester, 2024-2025

THE UNIVERSITY OF HONG KONG

DEPARTMENT OF STATISTICS AND ACTUARIAL SCIENCE

STAT3904 CORPORATE FINANCE FOR ACTUARIAL SCIENCE

Tutorial 5: Mean-Variance Portfolio Theory (Part II)

1 Key Learning Points

In the test and examination, candidates are expected to:

LP1 Derive the equation of the efficient frontier

(a) when there are two risky assets by mixing (Tutorial 4);
(b) when there are n ≥ 3 risky assets by mixing the minimum variance portfolio and
another efficient portfolio.

LP2 Find the tangency portfolio.

LP3 Find the equation of the efficient frontier when a risk-free asset is introduced.

2 Review of Key Concepts

Matrix notations will become handy in the multi-asset case.

• Portfolio rate of return: RP = xT R, where R = (R1 , . . . , Rn )T is the vector of returns of

the n assets;
• Expected rate of return: µP = E xT R = xT E(R) = xT µ;

T
• Variance of rate of return: σP2 = Var xT R = Cov xT R, xT R = xT Cov(R, R) xT =

xT Σx.

2.1 Mixing the MVP and Another Portfolio

Problem How to find efficiently the efficient (minimum variance) portfolio with a pre-specified
expected rate of return µ0 , with µ0 ≥ µMVP ?
Think! Armed with the global minimum variance portfolio (MVP), denoted as P , and an-
other portfolio on the minimum variance set (MVS), denoted as Q, we can construct
the required efficient portfolio by applying two-fund theorem.
S&AS: STAT3904 Corporate Finance for Actuarial Science 2

Two-fund Theorem. Every portfolio on the MVS is a mixture of two given mini-
mum variance portfolios, which can be taken, for convenience, as P and Q.

Action! Step 1. Construct the global minimum variance portfolio P by normalizing Σ−1 1.
Step 2. Construct another portfolio Q on minimum variance set by normalizing
Σ−1 µ.
Step 3. Calculate µP = xTP µ and µQ = xTQ µ.
Step 4. Let α be the weight invested in P (so (1 − α) is the weight invested in
Q). Solve the mean equation αµP + (1 − α)µQ = µ0 for α.
Step 5. The composition of the required portfolio is given by

αxP + (1 − α)xQ .

• This method is more preferable than solving the constrained minimization problem minx σP2
subject to the constraints that xT µ = µ0 and xT 1 = 1 by the method of Lagrange mul-
tipliers. 1

• This method involves the inversion of the variance-covariance matrix Σ. Therefore, it

is advised to recall some commonly used matrix inversion formulas, especially for 2 × 2,
diagonal and block diagonal matrices.
 2 
σ1 ρσ1 σ2
Case 1 There are just two risky assets, i.e. n = 2. Inverting Σ = is easy
ρσ1 σ2 σ22
by the general formula
 −1  
a b 1 d −b
= .
c d ad − bc −c a

Case 2 n ≥ 3 and Σ is simply a diagonal matrix of the form

   
σ12 0 ··· 0 0 1/σ12 0 ··· 0 0
2
0
 σ2 · · · 0 0 
 0
 1/σ22 ··· 0 0 

Σ =  ... .. . . .. .. , with inverse Σ−1 =
 ..
. .. .. .. ..
.
   
. . . . . . . .
2 2
   
0 0 · · · σn−1 0  0 0 ··· 1/σn−1 0 
0 0 ··· 0 σn2 0 0 ··· 0 1/σn2

Case 3 n ≥ 3 and Σ is a block matrix of the form

   −1 
Σ1 0 −1 Σ1 0
Σ= , with inverse Σ = .
0 Σ2 0 Σ−1
2
1
(For those who are interested) The explicit solution of the minimum variance portfolio for a fixed expected
rate of return µ0 is given by
bΣ−1 1 − aΣ−1 µ + µ0 (cΣ−1 µ − aΣ−1 1)
x= ,
bc − a2
µT Σ−1 µ
where a = µT Σ−1 1, b = µT Σ−1 µ and c = 1T Σ−1 1. If µ0 = µT Σ−1 1
, then

µT Σ−1 µ T −1
 
x∝ · 1 Σ 1 − µ Σ 1 Σ−1 µ ∝ Σ−1 µ
T −1
µT Σ−1 1
whose expected rate of return equals µ0 .
S&AS: STAT3904 Corporate Finance for Actuarial Science 3

 
5 0 0
Example. The inverse of Σ = 0 18 −18 is
0 −18 144

5−1
   
02×1 1/5 0 0
−1
Σ−1 = 

18 −18 = 0 4/63 1/126 .
01×2
−18 144 0 1/126 1/126

Exercise 1. (Proof of the Two-Fund Theorem) Let x = (x1 , . . . , xn )T and y = (y1 , . . . , yn )T

be two minimum variance portfolios. Show that βx + (1 − β)y, β ∈ R is also a minimum
variance portfolio.

Proof. Let the expected rates of return of x and y be µx and µy respectively. Then x (resp.
y) solves the constrained minimization problem
1 T
minw w Σw
2
s.t. wT 1 = 1,
w T µ = µx (resp. µy ).

Using the method of Lagrangian multiplier, x and y satisfy the first-order conditions
y y
 
 Σx − λx1 1 − λx2 µ = 0,  Σy − λ1 1 − λ2 µ = 0,
T
x 1 = 1, and yT 1 = 1,
xT µ = µx , yT µ = µy .
 

Multiplying the first set of equations by β and the second set by 1 − β and summing them
up, we have
 x y x y
 Σ(βx + (1 − β)y) − (βλ1 + (1 − β)λ1 ) 1 − (βλ2 + (1 − β)λ2 ) µ = 0,
(βx + (1 − β)y)T 1 = 1,
(βx + (1 − β)y)T µ = βµx + (1 − β)µy .


This shows that the portfolio βx + (1 − β)y solves the following optimization problem:
1 T
minw w Σw
2
s.t. wT 1 = 1,
wT µ = βµx + (1 − β)µy ,

which implies that βx + (1 − β)y is also a minimum variance portfolio.

S&AS: STAT3904 Corporate Finance for Actuarial Science 4

2.2 Tangency Portfolio

• How to Find the Tangency Portfolio?
Similar to the MVP, the tangency portfolio (a.k.a. the efficient portfolio of risky assets)
can be obtained by normalizing
 
µ1 − rf
 µ2 − rf 
Σ−1 (µ − rf 1) = Σ−1  ..  ,
 
 . 
µn − rf

where µ is the vector of the expected rates of return and rf is the risk-free rate.
µ∗ − rf
• The tangency portfolio M is the feasible portfolio which maximizes the slope
σ∗
among all pairs (σ ∗ , µ∗ ) in the feasible region constructed by the risky assets. After
finding the tangency portfolio xM , hence its expected rate of return µM and standard
deviation σM , we can plug these two values in the following equation to get the equation
of the efficient frontier.
µM − rf
Equation of the Efficient Frontier: µ = rf + σ, σ ≥ 0. (1)
σM

3 Problems

Attempt ALL FOUR questions. Marks for past paper questions are shown in square brackets.

1. Qualitative Inferences: Changes in Expected Return STAT2807

In a market with 3904 risky assets, every entry of the inverse of the variance-covariance 10-11
matrix is positive. A student has calculated the mean and variance of the global minimum Test
variance portfolio to be µ∗ and (σ ∗ )2 respectively, and the corresponding point has been (Slightly
marked on the σ-µ plane. Adapted)

Later on, the student found that µ2 , the expected return of asset 2, was wrongly copied
as 18% from the question paper, where the true value should be 10%. How should the
position (on the σ-µ plane) of the point representing the global minimum variance portfolio
be adjusted?
[Total: 6 marks]

Solution. Since every entry of Σ−1 is positive and x∗ is the normalization of the solution to
Σx = (1, . . . , 1)T , we have x∗i > 0 for all i = 1, 2, . . . , 3904. Moreover, x∗ depends only on
Σ, but not on µ = (µ1 , . . . , µ3904 )T . Our conclusions are:

• There is no change in x∗ , and hence no change in σP .

• However, the decrease in µ2 results in a decrease in µP (Note: the fact that x∗2 > 0
guarantees this).
S&AS: STAT3904 Corporate Finance for Actuarial Science 5

2. Mean Variance Portfolio, multiple-asset case STAT2807

Assume that there are only three risky stocks with rates of return R1 , R2 and R3 respectively. 09-10
The variance-covariance matrix and the expected rates of return are Exam

     
0.2 0 0 µ1 0.3
Σ=  0 0.5 0  and  µ2 = 0.4 .
 
0 0 0.4 µ3 0.4

(a) Find the global minimum-variance portfolio. What are the variance and expected rate
of return of this portfolio? [4 marks]
(b) Find another efficient portfolio. What is the covariance between the return of the
portfolio in (a) and the return of this portfolio? [9 marks]
(c) For any portfolio on the efficient frontier with standard deviation and expected rate of
return σ and µ respectively, find the relationship between σ and µ. [6 marks]

[Total: 19 marks]

Solutions. (a) By normalizing

  1    
1 0.2
0 0 1 5
−1   1
Σ 1 =0 0.5
0   1 =
  2 ,
1
1 0 0 0.4
1 2.5

we can get the composition of the global minimum-variance portfolio:

 ∗   10 
x1 19
x∗ = x∗2  =  19
4
.
∗
x3 5
19

As a result, the mean and variance of the rate of return of the minimum-variance
portfolio are
33
µP = (x∗ )T µ = = 0.347368,
95
 2  2  2
2 T ∗ 10 4 5 2
σP = x Σx = × 0.2 + × 0.5 + × 0.4 = = 0.105263.
19 19 19 19

(b) Normalizing   1    
µ1 0.2
0 0 0.3 1.5
−1   1
Σ µ2 =  0 0.5
0   0.4 = 0.8 ,
 
1
µ3 0 0 0.4
0.4 1
we obtain a portfolio lying on the minimum variance set:
 ∗   15 
y1 33
y ∗ = y2∗  =  33
8
.
∗
y3 10
33
S&AS: STAT3904 Corporate Finance for Actuarial Science 6

The mean and variance of the rate of return of such a portfolio are
39 13
µ = (y ∗ )T µ = and σ 2 = (y ∗ )T Σy ∗ = .
110 121
As µ > µP , this portfolio is efficient. By the result in Question 3 of Tutorial 4, the
covariance between the return of the global minimum-variance portfolio in (a) and the
portfolio thus constructed is
10 15 4 8 5 10 2
(x∗ )T Σy ∗ = × 0.2 × + × 0.5 × + × 0.4 × = = 0.105263.
19 33 19 33 19 33 19

(c) By the two-fund theorem, any efficient portfolio can be constructed as a linear combi-
nation of two given efficient portfolios, which we take as those in (a) and (b). In this
two-asset market, the mean vector and variance-covariance matrix are
 33  2 2 
95 19 19
39 and 2 13 .
110 19 121

Mixing the two portfolios, we have

33 39
µ = α+ (1 − α)
95 110
2 2 13 2
σ2 = α + (1 − α)2 + 2α(1 − α) .
19 121 19
We can eliminate α to obtain
 2  2
2 2 247 418 13 247 418
σ = − µ + 1− − µ
19 5 3 121 5 3
  
4 247 418 247 418
+ − µ 1− − µ
19 5 3 5 3
26 88 380 2
= − µ+ µ
5 3 9
 2
2 380 33
= + µ− ,
19 9 95
33
with µ ⩾ 95
.

3. Mean Variance Portfolio, multiple-asset case STAT2807

There are 4 assets with rates of return r1 , r2 , r3 and r4 , respectively. The variance-covariance 06-07
matrix and the expected rates of return are Exam

     
1 0 0 0 µ1 0.5
0 2 0 0 µ2  0.7
Σ= 0 0 3 0 , µ = µ3  = 0.9 .
    

0 0 0 4 µ4 1.0

(a) Find the minimum-variance portfolio. [10 marks]

S&AS: STAT3904 Corporate Finance for Actuarial Science 7

(b) Find another efficient portfolio by setting µ = 1, where µ is the portfolio return rate.
[2 marks]
(c) If the risk-free rate is rf = 0.3, find the efficient portfolio of risky assets. [5 marks]

[Total: 17 marks]

Solution. (a) Normalizing

 −1       
1 0 0 0 1 1 0 0 0 1 1
0 2 0 0 1 0 1/2 0 0  1 1/2
   =   =  ,
0 0 3 0 1 0 0 1/3 0  1 1/3
0 0 0 4 1 0 0 0 1/4 1 1/4
 T
∗ 12 6 4 3
we find that the minimum-variance portfolio is x = , , , .
25 25 25 25
(b) The normalization of
 −1     
1 0 0 0 1 0 0 0 0.5 0.5
0 2 0 0
 µ = 0 1/2 0 0  0.7 = 0.35 ,
    

0 0 3 0 0 0 1/3 0  0.9  0.3 
0 0 0 4 0 0 0 1/4 1.0 0.25
 T
5 1 3 5
which is x = , , , , is a portfolio on the minimum variance set. By the
14 4 14 28
two-fund theorem, the efficient portfolio with expected rate of return µ = 1 can be
expressed as a mixture of this portfolio and the minimum-variance portfolio in (a).
The expected rates of return of this portfolio and the minimum-variance portfolio are
respectively
12 6 4 3
(0.5) + (0.7) + (0.9) + (1) = 0.672
25 25 25 25
and
5 1 3 5
(0.5) + (0.7) + (0.9) + (1) = 0.725.
14 4 14 28
Solving 0.672α + 0.725(1 − α) = 1 yields α = −275/53, which means the composition
 T
∗ 104 16 184 179
of the required portfolio is αx + (1 − α)x = − , , , .
371 53 371 371
(c) Normalizing
 −1     
1 0 0 0 1 0 0 0 0.2 0.2
0 2 0 0
 (µ − rf 1) = 0 1/2 0 0 
 0.4 =  0.2  ,
    

0 0 3 0 0 0 1/3 0  0.6  0.2 
0 0 0 4 0 0 0 1/4 0.7 0.175
we find the composition of the tangency portfolio as
 T
8 8 8 7
x= , , , .
31 31 31 31
S&AS: STAT3904 Corporate Finance for Actuarial Science 8

4. A Question from SOA: Given the Full Distributions of Returns SOA

An investment house has provided an investor with the following: Course 6
Spring
2005 Q10
Scenario Probability Fund A Return Fund B Return
(Adapted)
1 0.5 25% 20%
2 0.3 10% −20%
3 0.2 −30% 25%

The annual T-bill return is 3% per annum.

(a) Calculate the correlation coefficient between Fund A and Fund B using the given sce-
narios.
(b) Determine the tangency portfolio, Portfolio M . Calculate the expected return and
standard deviation of Portfolio M .
(c) Calculate the slope of the efficient frontier supported by T-bills and Portfolio M .
(d) Another investment house has developed a portfolio, Portfolio Q, using Fund A and
Fund B. The expected return of Portfolio Q is 10% and the standard deviation is 12%.
Should the investor invest in Portfolio Q rather than Portfolio M ? Why or why not?

Solution. Instead of giving you µ and Σ, this question provides even the full distributions of
the returns of the two assets, modelled as discrete random variables.

(a) For A:
• E[RA ] = 0.5(25%) + 0.3(10%) + 0.2(−30%) = 9.5%
• E[RA
2
] = 0.5(25%)2 + 0.3(10%)2 + 0.2(−30%)2 = 522.5%2
• Var(RA ) = 522.5 − 9.52 = 432.25%2
Similarly, for B:
• E[RB ] = 0.5(20%) + 0.3(−20%) + 0.2(25%) = 9%
• E[RB
2
] = 0.5(20%)2 + 0.3(−20%)2 + 0.2(25%)2 = 445%2
• Var(RB ) = 445 − 92 = 364%2
Finally we need the covariance:

Cov(RA , RB ) = E[RA RB ] − E[RA ]E[RB ]

= 0.5(25%)(20%) + 0.3(10%)(−20%) + 0.2(−30%)(25%) − (9.5%)(9%)
= −45.5%2 .

The correlation coefficient between Fund A and Fund B is

Cov(RA , RB ) −45.5
p =p = −0.1147 .
Var(RA )Var(RB ) (432.25)(364)
 2 2

432.25% −45.5%
(b) We are given µ = (9.5%, 9%)T and Σ = . Normalizing
−45.5%2 364%2
    
−1 364%2 45.5%2 6.5% 2639
Σ (µ − rf 1) = = ,
45.5%2 432.25%2 6% 2889.25
S&AS: STAT3904 Corporate Finance for Actuarial Science 9

 T
116 127
we obtain the composition of the tangency portfolio M : , . Its expected
243 243
return and standard deviation are
116 127
µM = (9.5%) + (9%) = 9.2387%
243 243
and
s 2  2   
116 127 116 127
σM = (432.25%2 ) + (364%2 ) +2 (−45.5%2 ) = 13.2372% .
243 243 243 243

µM − rf 9.2387% − 3%
(c) The required slope is = = 0.4713 .
σM 13.2372%
(d) Since Portfolio Q has a higher rate of return but a lower standard deviation, it is more
efficient than Portfolio M and should be invested instead.

********** END OF TUTORIAL 5 **********