AIN3220

Investment and Risk Analysis

Chapter 4
Chapter 4: Mean-Variance Analysis II –
Multiple-Asset Portfolio

Question: How do you invest with more than two stocks? In particular
• What does it mean by the “best” portfolios and how to find them?
• How many such portfolios do you need for your investment?

Topics in this Chapter:

4.1 Multiple-Asset Portfolios.
4.2 Markowitz Portfolio Theory.
4.3 Global Minimum Variance Portfolio, G.
4.4 Two Fund Theorem.

References: EGBG Chapters 5–6.

AIN3220 Investment and Risk Analysis 4-1

4.1 Multiple-Asset Portfolios
As in the case of two-asset portfolio in Chapter 3, the essential elements
for the N -asset portfolio, for N > 2, basically have the same structure. For a
portfolio consisting of N assets, the portfolio weights are x1, . . . , xN . We define
the following matrices:
     
x1 r1 µ1
x =  ..  , r =  ..  , µ =  ..  = E[r],
xN rN µN
   2 
1 σ1 . . . σ1,N
1 =  ..  , and Ω =  .. . . . ..  .
2
1 σN,1 . . . σN
where
• x is the vector of portfolio weights,
• r is the vector of returns,
• µ is the vector of expected returns,
AIN3220 Investment and Risk Analysis 4-2
 • 1 is the vector of ones, and
• Ω is the variance-covariance matrix.

4.1.1 Budget constraint

• As with two assets, the budget constraint is

x1 + x2 + · · · + xN = 1 ⇔ x>1 = 1.

• The budget constraint simply states that xi represents the percentage of

your wealth allocated to asset i, for i = 1, ..., N .
• Note that the budget constraint does NOT require xi to be between 0 and
1.
• When xi < 0, you are short-selling asset i.

AIN3220 Investment and Risk Analysis 4-3

 • On the other hand, when xi > 1, you place more than 100% of your
wealth into asset i. Since the budget constraint requires the sum of xi’s to be
equal to 1, xi > 1 would readily imply that xj < 0, for some i 6= j = 1, ..., N .
• In words: if you want to invest more than more than 100% on one stock,
you need to short-sell other stocks for that!

4.1.2 Return and risk of a portfolio

• As with two assets, the return and expected return of an N asset portfolio:

rP = x>r and µP = x>µ.

• As with two assets, the variance of returns for an N asset portfolio:

σP2 = x>Ωx.

AIN3220 Investment and Risk Analysis 4-4

• Here, rP is a random variable representing the return of the portfolio. µP is
a number representing the expected return of the portfolio.
• The variance of the returns for the N -asset portfolio is denoted as σP2 .

AIN3220 Investment and Risk Analysis 4-5

4.1.3 An example of three assets
Example 4.1 Consider a portfolio of three assets. The vector of expected
returns and the variance-covariance matrix are:
   
0.10 0.20 0.15 0.17
µ = 0.20 , Ω = 0.15 0.25 0.09 .
0.30 0.17 0.09 0.35

For a portfolio P with portfolio weights x = (0.01, 0.03, 0.96)>, calculate the
expected return, µP , and the standard deviation, σP , of the portfolio P .
Answer: The portfolio return is given by rP = x>r, the expected return of
P:  
0.10
µP = E(rP ) = x>µ = (0.01, 0.03, 0.96) 0.20 = 0.295,
0.30
and the variance of P :
  
0.20 0.15 0.17 0.01
σP2 = x>Ωx = (0.01, 0.03, 0.96) 0.15 0.25 0.09 0.03 = 0.3313.
0.17 0.09 0.35 0.96
AIN3220 Investment and Risk Analysis 4-6
 √
Hence σP = 0.3313 = 0.5756.
Tip: Make sure you can verify these calculations by hand!
4.1.4 The Markowitz problem of the three assets
Question: Is the portfolio P the ‘best’ one in Example 4.1 with the small-
est risk? More specifically, is portfolio P the portfolio that achieves the ex-
pected return 0.295 with the smallest variance?

Example 4.2: We continue from Example 4.1. Find the optimal portfolio
x with µP = 0.295 and the smallest variance.

Answer: The optimal portfolio solves the celebrated Markowitz portfolio-

selection problem:
1 >
min x Ωx
x1 ,x2 ,x3 2

subject to the constraints

x>1 = 1, (4.1)
x>µ = µP = 0.295. (4.2)
AIN3220 Investment and Risk Analysis 4-7
Hence, the Markowitz portfolio-selection problem is a typical example
of an Constrained Optimization Problem!

Harry Markowitz (1927–), Nobel Prize 1990

AIN3220 Investment and Risk Analysis 4-8

4.1.5 Partial Differentiation (Review)
We first review the basics of partial differentiation from AMS 1001.
Denote f (x1, ..., xN ) to be the function with N variables, x1...xN . The partial
∂f
derivative of f with respect to xi, denoted as ∂x i
is defined as

∂f (x1, ..., xN ) f (x1, .., xi + δ, ..., xN ) − f (x1, .., xi, ..., xN )

= lim . (4.3)
∂xi δ→0 δ
∂f
In practice, finding ∂x i
can be done by taking derivative with respect to
xi while keeping other xj , for i 6= j = 1, ..., N , as constants.
Example 4.3 Let f (x1, x2, x3) = x21 + x22 + x23 + 2x1x2 + 2x2x3 + 2x1x3, find
∂f ∂f ∂f
∂x1 , ∂x2 , ∂x3 .
Answer:
∂f
• When calculating ∂x1 , we treat x2 and x3 as if they were constants and
obtain
∂f
= 2x1 + 2x2 + 2x3.
∂x1
AIN3220 Investment and Risk Analysis 4-9
• Similarly, we also have
∂f ∂f
= 2x1 + 2x2 + 2x3, = 2x1 + 2x2 + 2x3.
∂x2 ∂x3

Exercise: Let g(x1, x2, x3) = x31 +x42 +x23 +2x21x2 +3x32x3 +5x1x3. Calculate
∂g ∂g ∂g
∂x1 , ∂x2 , ∂x3 .

AIN3220 Investment and Risk Analysis 4-10

4.1.6 Constrained Optimization
Unconstrained optimizations
Consider first the minimization problem
min f (x, y)
x,y

where f (x, y) = x2 + 2y 2.
The solution can be found by first taking the partial derivative of the ob-
jective function f (x, y) with respect to its arguments x and y and setting the
results equal to zero. Hence
∂f
∂x = 2x = 0,
∂f
∂y = 4y = 0.
This pair of equations is called the first order conditions with respect to
x and y, and is represented as a system of two simultaneous equations
to be solved. In general, solving the system of simultaneous equations amounts
to us the knowledge of solving system of linear equations from AMS 1001.
In this example, the solution is quite straightforward as we can see that x∗ =
y ∗ = 0. Hence f (x∗, y ∗) = 0 and the minimum of the objective function is zero.
AIN3220 Investment and Risk Analysis 4-11
Constrained optimizations
Now consider again the optimization problem
min f (x, y)
x,y

but this time subject to the constraint g(x, y) = c, for example x + y = 1.

This ‘constrained’ optimization, can be solved using the Method of La-

grange Multipliers.

The method of Lagrange multipliers involves 3 steps:

Step 1:
We first introduce a new variable λ, called the Lagrange multiplier, and con-
struct the Lagrange function (or Lagrangian) defined by
L(x, y, λ) = f (x, y) + λ(c − g(x, y))
= x2 + 2y 2 + λ(1 − x − y).
Step 2:
We now minimize L in an unconstrained way by partially differentiating L
AIN3220 Investment and Risk Analysis 4-12
with respect to each of its three arguments (x, y, λ) and setting the result
equal to zero (to obtain the first order conditions):
∂L
∂x = 0 ⇔ 2x∗ − λ∗ = 0, (4.4)
∂L
∂y = 0 ⇔ 4y ∗ − λ∗ = 0, (4.5)
∂L
∂λ = 0 ⇔ 1 − x∗ − y ∗ = 0. (4.6)
Step 3:
Solving the set of simultaneous equations for x∗, y ∗ and λ∗ gives us the solution
to the original constrained problem. Solving Eqs. (4.4)–(4.6) yields
x∗ = 32 , y ∗ = 31 , and λ∗ = 43 .
Remarks:
• We observe that x∗ + y ∗ = 32 + 13 = 1 as required.
• The important point is that these values minimize the original function
f (x, y) subject to the constraint x + y = 1.
∗ ∗ 2 2 1 2
Hence f (x , y ) = 3 + 2 3 = 32 , and this is the lowest we can get whilst

simultaneously satisfying the constraint.

AIN3220 Investment and Risk Analysis 4-13
• We are not really interested in the value of the Lagrange multiplier λ as this
was introduced simply to allow us to find x∗ and y ∗.
• The above example is presented in scalar notation but could also be
written in matrix notation.
• IMPORTANT: You need to know

- how to set up the Markowitz problem (a constrained optimization)

and construct the Lagrange function (Step 1)
- establish the first-order conditions from scratch in scalar notation (Step
2)
- solve these first-order conditions (Step 3). Later in this chapter, we shall
show that there are formulas for these solutions!

AIN3220 Investment and Risk Analysis 4-14

4.1.7 Solving the Markowitz problem in scalar notation
Example 4.2 (cont.). The Markowitz problem in scalar notation can be
re-written as
1 > 1
min x Ωx = min 0.20x21 + 0.25x22 + 0.35x23 + 2 × 0.15x1x2
x1 ,x2 ,x3 2 x1 ,x2 ,x3 2
!
+2 × 0.17x1x3 + 2 × 0.09x2x3

subject to

x>1 = x1 + x2 + x3 = 1,
x>µ = 0.10x1 + 0.20x2 + 0.30x3 = 0.295.

Note: The Markowitz problem with only risky assets involves two con-
straints

AIN3220 Investment and Risk Analysis 4-15

Step 1:
First define the Lagrangian:
1 >
L(x1, x2, x3, λ, γ) = x Ωx + λ[1 − xT 1] + γ[µP − x>µ]
2
1
= [0.20x21 + 0.25x22 + 0.35x23 + 2 × 0.15x1x2 + 2 × 0.17x1x3 + 2 × 0.09x2x3]
2
+ λ[1 − (x1 + x2 + x3)] + γ[0.295 − (0.10x1 + 0.20x2 + 0.30x3)].
Step 2:
The optimal portfolio x∗ satisfies the first-order conditions:
∗
∂L
∂x1 = 0 ⇔ 0.20x 1 + 0.15x∗2 + 0.17x∗3 = λ∗ + 0.10γ ∗, (4.7)
∗
∂L
∂x2 = 0 ⇔ 0.15x 1 + 0.25x∗2 + 0.09x∗3 = λ∗ + 0.20γ ∗, (4.8)
∗
∂L
∂x3 = 0 ⇔ 0.17x 1 + 0.09x∗2 + 0.35x∗3 = λ∗ + 0.30γ ∗, (4.9)
∂L
∂λ = 0 ⇔ x∗1 + x∗2 + x∗3 = 1, (4.10)
∗
∂L
∂γ = 0 ⇔ 0.10x 1 + 0.20x∗2 + 0.30x∗3 = 0.295. (4.11)
Step 3:
There are 5 equations and 5 unknowns (x∗1 , x∗2 , x∗3 , λ∗, γ ∗). Solving the system
AIN3220 Investment and Risk Analysis 4-16
of equations gives us the final result:

x∗1 = −0.3333, x∗2 = 0.7167, x∗3 = 0.6167, λ∗ = 0.1067, and γ ∗ = 0.39.

The corresponding expected return and variance of the optimal portfolio are:

µx∗ = 0.295 and σx2∗ = 0.2217, σx∗ = 0.4709 < σP .

Therefore, portfolio P is NOT the portfolio that achieves the expected return
µP = 0.295 with minimum variance!

AIN3220 Investment and Risk Analysis 4-17

Remark:
Solving systems of equations by hand as in Example 4.2 is tedious and
quickly becomes unmanageable as the number of assets, N , in the
Markowitz problem increases.
Solution: Solve the first-order system in matrix form!
Example 4.2 (A revisit):
• Writing Eqs (4.7)–(4.9) in matrix form yields
   ∗    
0.20 0.15 0.17 x1 1 0.10
0.15 0.25 0.09 x∗2  = λ∗ 1 + γ ∗ 0.20
0.17 0.09 0.35 x∗3 1 0.30
Ωx∗ = λ∗1 + γ ∗µ; (4.12)
• Writing Eq (4.10) in matrix form yields
 ∗

x1∗
1 1 1 x2  = 1
x∗3
1>x∗ = 1. (4.13)
AIN3220 Investment and Risk Analysis 4-18
• Writing Eq (4.11) in matrix form yields
 ∗

x1∗
0.10 0.20 0.30 x2  = 0.1
x∗3
µ>x∗ = 0.295. (4.14)

Note: It turns out that these first-order condition equations are the exactly
same for the general Markowitz problem with any number of
assets!

AIN3220 Investment and Risk Analysis 4-19

4.2 Markowitz Portfolio Theory
4.2.1 The Markowitz problem: General formulation
• Formally, for a given fixed expected rate of return µP , we wish to find the
vector of portfolio weights x∗ that minimize the variance. The minimization
problem is
1 ∗> ∗ 1
x Ωx = min x>Ωx, (4.15)
2 x 2
subject to

the budget constraint x>1 = 1, (4.16)

the target return constraint x> µ = µ P . (4.17)

• The constrained minimization problem (4.15)–(4.17) is called the Markowitz

problem.
• The corresponding solution is called Optimal Portfolio or Mean-Variance
frontier portfolio.

AIN3220 Investment and Risk Analysis 4-20

4.2.2 Solution to the Markowitz problem – N risky assets
• Step 1: To solve the Markowitz problem, Eqs. (4.15)–(4.17), we define the
Lagrangian:

1
L(x, λ, γ) = x>Ωx + λ(1 − 1>x) + γ(µP − µ>x),
2

with Lagrange multipliers λ and γ.

• Step 2: The first-order conditions for the optimal weights, x∗, and
Lagrange multipliers, λ and γ, are:

Ωx∗ = λ1 + γµ, (4.18)

1>x∗ = 1, (4.19)

AIN3220 Investment and Risk Analysis 4-21

 µ> x ∗ = µ P . (4.20)

• Step 3:

The solutions to the first-order conditions are

C − BµP AµP − B
λ= , γ= , (4.21)
∆ ∆

and

x∗ = λΩ−11 + γΩ−1µ, (4.22)

where
AIN3220 Investment and Risk Analysis 4-22
 A = 1>Ω−11, B = 1>Ω−1µ, C = µ>Ω−1µ, ∆ = AC − B 2.
(4.23)

• Derivations (OPTIONAL):
From (4.18):
x∗ = λΩ−11 + γΩ−1µ. (4.24)
Substituting (4.24) into (4.19) and (4.20), we obtain
λ1>Ω−11 + γ1>Ω−1µ = 1, (4.25)
λµ>Ω−11 + γµ>Ω−1µ = µP . (4.26)

For convenience, we define the following scalars:

A = 1>Ω−11, B = 1>Ω−1µ, C = µ>Ω−1µ, ∆ = AC − B 2.
Then (4.25) and (4.26) become
Aλ + Bγ = 1,
Bλ + Cγ = µP .
AIN3220 Investment and Risk Analysis 4-23
• Note:

- These formulas show that the solutions to the first-order conditions in

Step 2 can be reduced to doing matrix addition, subtraction, and mul-
tiplications!
- The general method requires knowledge of Ω−1.

AIN3220 Investment and Risk Analysis 4-24

Example 4.4: Let us solve Example 4.2 in matrix form.
Answer: This is Step 3, i.e. it remains to solve the first-order conditions in
Eqs (4.7)–(4.11) using Eqs (4.21)–(4.22).
Note first that
 
14.7858 −6.9274 −5.4004
Ω−1 = −6.9274 7.6536 1.3966  . (4.27)
−5.4004 1.3966 5.1210

Hence, A, B, C, and ∆ in (4.23) now become

A = 1>Ω−11 = 5.6983, B = 1>Ω−1µ = 1.0056,

C = µ>Ω−1µ = 0.4814, ∆ = AC − B 2 = 1.7318.

When µP = 0.295, Eq (4.21) now becomes

C − BµP 0.4814 − 1.0056 × 0.295
λ= = = 0.1067, (4.28)
∆ 1.7318
AµP − B 5.6983 × 0.295 − 1.0056
γ= = = 0.39. (4.29)
∆ 1.7318
AIN3220 Investment and Risk Analysis 4-25
In terms of Eq (4.22), the optimal portfolio weight, x∗,
 
−0.3333
x∗ = λΩ−11 + γΩ−1µ =  0.7167  . (4.30)
0.6167

The variance of the optimal portfolio, σP2 , is

σP2 = x∗>Ωx∗ = 0.2217, (4.31)

or

σP = 0.4709. (4.32)

These are the same answers as those in Example 4.2!

AIN3220 Investment and Risk Analysis 4-26

4.2.3 Plotting the Minimum Variance Set (MVS) graphically
• The MVS: All portfolios that have the minimum variance for a certain
target portfolio returns, µP.
• To determine the MVS (and thus the efficient frontier), we first set a target
portfolio return, µP, and determine which portfolio with this return has
the minimum variance. This gives us one point on the MVS, the
optimal portfolio for the required return of µP.
• To generate the entire MVS, we cycle µP through all possible portfolio re-
turns (see Fig. 4.1).
• The minimum variance set (MVS) is a hyperbola that encloses all possible
portfolios of the individual assets.

AIN3220 Investment and Risk Analysis 4-27

 1

0.8

0.6

0.4 Efficient Frontier

Asset 1
mu

Asset 2

0.2 Asset 3
MVP

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

-0.2

-0.4
sigma

Figure 4.1: The blue hyperbola denotes the minimum variance set (MVS)
that encloses all possible portfolios that can be constructed by combining to-
gether assets 1, 2, and 3. The babyblue dot represents the MVP that
achieves µP = 0.295 with minimum variance 0.4709 (Example 4.4).
AIN3220 Investment and Risk Analysis 4-28
Equation for MVS

• The minimum variance, σP2 , of the optimal portfolio that achieves the
target return, µP, is

σP2 = x∗>Ωx∗ = x∗>[λ1 + γµ] = λ + γµP

yielding

AµP2 − 2BµP + C
σP2 = ; (4.33)
∆

or equivalently, solving µP in terms of σP yields

B 1p
µP = ± ∆(AσP2 − 1).
A A

AIN3220 Investment and Risk Analysis 4-29

• Therefore, the MVS is a hyperbola in mean-standard deviation space and a
parabola in mean-variance space. It is called Mean Variance Frontier.

• How can we interpret the plot of M V S?

• In Chapter 3, we found that when two assets are combined, the resulting
portfolios exist along a hyperbolic curve in mean-standard deviation space.
• When combining N > 2 risky assets, the resulting portfolios exist in a
two-dimensional region bounded by a hyperbola (see Fig. 4.1).
• The hyperbolic curve that bounds the region of possible portfolios is the
minimum variance set (MVS). The portfolios along this curve have
the minimum variance for a certain target portfolio return, µP (see
Fig. 4.2).

AIN3220 Investment and Risk Analysis 4-30

• There is a unique portfolio which has the smallest variance among all MVS,
which is called the global Minimum Variance Portfolio (MVP), G.
• The Minimum Variance Portfolio (MVP), G separates MVS into
two parts (see Fig. 4.3):

- The portfolios along the upper half of the MVS starting at the MVP
are efficient, i.e., no other assets exist in their upper-left quadrant. This
section of the MVS is known as the efficient frontier.
- The lower half of the MVS is known as the inefficient frontier.

AIN3220 Investment and Risk Analysis 4-31

 1

0.8

0.6

0.4 Asset 1
Asset 2
mu

Asset 3

0.2 Efficient Frontier

Inefficient Frontier

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

-0.2

-0.4
sigma

Figure 4.2: The blue line denotes the efficient frontier of the MVS, whereas
the red line denotes the inefficient frontier of the MVS.

AIN3220 Investment and Risk Analysis 4-32

4.3 Global Minimum Variance Porfolio, G
Global MVP
• To locate the global minimum variance portfolio, G, we first compute the
derivative of σP2 with respect to µP using (4.33),
dσP2 AµP − B
=2 .
dµP ∆
Setting this equal to zero yields the expected return of the MVP,

B
µG = .
A

Substituting this expression back into (4.33) gives

2 Aµ2G − 2BµG + C
σG =
∆
B 2 B



A A − 2B A + C
= ,
∆
AIN3220 Investment and Risk Analysis 4-33
 and simplying this yields the following for the variance of returns of the
MVP:

2 1
σG = .
A

• To compute the weights of the MVP, first substitute the above expression
for µG into (4.21), giving
B


C − BµG C − B A 1
λ= = = . (4.34)
∆ ∆
A
B
AµG − B A A − B
γ= = = 0. (4.35)
∆ ∆
After substituting these expressions of λ and γ into (4.22), we obtain the
following expression for the weights of the MVP:

Ω−11 Ω−11
xG = = > −1 .
A 1 Ω 1

AIN3220 Investment and Risk Analysis 4-34

Example 4.5 Consider assets 1, 2, and 3 in Example 4.1. The vector of
expected returns and the variance-covariance matrix are:
   
0.10 0.20 0.15 0.17
µ = 0.20 , Ω = 0.15 0.25 0.09 .
0.30 0.17 0.09 0.35
The inverse variance-covariance matrix is
 
14.7858 −6.9274 −5.4004
Ω−1 = −6.9274 7.6536 1.3966  .
−5.4004 1.3966 5.1210
What are µG and σG of the GMVP, G? What are the portfolio weights for G?

AIN3220 Investment and Risk Analysis 4-35

Answer: Since

A = 1>Ω−11 = 5.6983, B = 1>Ω−1µ = 1.0056,

C = µ>Ω−1µ = 0.4814, ∆ = AC − B 2 = 1.7318,

the mean return and standard deviation of returns for the global MVP are:
B
µG = = 0.1765,
A
r
1
σG = = 0.4189.
A
Finally, the portfolio weights for this global MVP, xG is
 
0.4314
Ω−11 Ω−11
xG = = = 0.3725 . (4.36)
A 5.6983
0.1961

AIN3220 Investment and Risk Analysis 4-36

 1

0.8

0.6

G_MVP
0.4
Asset 1
Asset 2
mu

Asset 3
0.2
Efficient Frontier
Inefficient Frontier

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

-0.2

-0.4
sigma

Figure 4.3: The purple dot represents the GMVP, G that separates the
hyperbola into an efficient part (shown in blue) and an inefficient part
(shown in red) in Example 4.5.

AIN3220 Investment and Risk Analysis 4-37

4.4 Two-Fund Theorem
• The Two-Fund Theorem states that the entire MVS can be created
by combining together any two distinct portfolios that already exist on the
MVS!
• This result suggests that an investment management company only needs
to create two funds for all of their clients.
• Which two funds should we choose? It can be any two funds on the
efficient frontier!
• To illustrate, let us fix the following two funds on the efficient frontier:

- The minimum variance portfolio that achieves the expected return µP =

0.295 in Example 4.2, i.e.
 
−0.3333
x∗ =  0.7167  .
0.6167

AIN3220 Investment and Risk Analysis 4-38

 - The global minimum variance portfolio xG in Example 4.5 that achieves
µG = 0.115, i.e.
 
0.4314
xG = 0.3725 .
0.1961
• Consider now a third portfolio on the efficient frontier with the expected
return of µ̂P. How can we find the associated efficient portfolio x̂?
• The Two-Fund theorem offers a very simple formula:

x̂ = φx∗ + (1 − φ)xG,

where

µ̂P − µG
φ= .
µP − µG

AIN3220 Investment and Risk Analysis 4-39

• Derivation (OPTIONAL): This is because the Two-Fund Theorem
states that we simply need to find γ such that x̂ = φx∗ + (1 − φ)xG that
achieves the expected return µ̂P :

x̂>µ = µ̂P
(φx∗ + (1 − φ)xG)> µ = µ̂P
φx∗>µ + (1 − φ)x> Gµ = µ̂P
φµP + (1 − φ)µG = µ̂P
φ(µP − µG) + µG = µ̂P

or equivalently,
µ̂P − µG
φ = .
µP − µG

Example 4.6
What is the minimum variance portfolio weight, x̂ that achieves the expected
return µ̂P = 0.50?

AIN3220 Investment and Risk Analysis 4-40

Answer: Since φ is
0.50 − 0.1765
φ = = 2.7295,
0.295 − 0.1765
the minimum variance portfolio weight, x̂ that achieves the expected return
µ̂P = 0.50 is
   
−0.3333 0.4314
x̂ = φx∗ + (1 − φ)xG = 2.7295  0.7167  + [1 − (2.7295)] 0.3725
0.6167 0.1961
 
−1.6559
=  1.3118  .
1.3441
The corresponding minimum variance associated with portfolio x̂, denoted as
σ̂P2 , is given as
  
0.20 0.15 0.17 −1.6559
σ̂P2 = x̂>Ωx̂ = −1.6559 1.3118 1.3441 0.15 0.25 0.09  1.3118 

0.17 0.09 0.35 1.3441

= 0.5199,
AIN3220 Investment and Risk Analysis 4-41
or

σ̂P = 0.7210.

Remark: When there is a risk-free asset, then only one fund of risky
assets is needed to generate the entire MVS. We will analyze this result and
the corresponding One-Fund Theorem later.

AIN3220 Investment and Risk Analysis 4-42

 1

0.8

0.6

G_MVP

0.4 Asset 1
Asset 2
mu

Asset 3
0.2 Efficient Frontier
Inefficient Frontier
Portfolio_Hat
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

-0.2

-0.4
sigma

Figure 4.4: The brown dot on the efficient frontier (shown in the blue)
represents the MVP that achieves µ̂P = 0.50 with minimum variance 0.7210
(Example 4.6).

AIN3220 Investment and Risk Analysis 4-43