Bachelor in Economics

Bachelor in Business Administration

Mathematics II

Problem set 2: Matrix calculus

Solutions

Departament d’Economia i d’Història Econòmica

 Part I. Matrices and determinants

1. Consider the following matrices:

 
  −1 0  
1 0 1 0 1
A1 = A2 =  2 1  A3 =
0 1 0 1 −1
8 4
   
  1 0 1 1 3 1
−1 2
A4 = A 5 = 2 1 1  A6 = 0 1 0 
1 1
8 4 −2 0 0 −5

Compute A1 + A4 , A3 + A5 , A1 − A4 , A3 − A5 , A2 · A3 , A5 ·
A6 , A6 · A5 , A1 · A3 , where Ai · Aj is the matrix multiplication.
Solution
We have A1 ∈ M2x2 ; A2 ∈ M3x2 ; A3 ∈ M2x3 ; A4 ∈ M2x2 ; A5 ∈
M3x3 ;
A6 ∈ M3x3 . Then,
     
1 0 −1 2 1−1 0+2
• A1 + A4 = + = =
0 1 1 1 0+1 1+1
 
0 2
1 2
• A3 + A5 Not well defined operation since the number of
rows do not match.
     
1 0 −1 2 1 − (−1) 0 − 2
• A1 − A4 = − = =
0 1 1 1 0−1 1−1
 
2 −2
−1 0
• A3 − A5 Not well defined operation since the number of
rows do not match.
Recall that C = AB is defined if A has as many columns
as B has rows. If A is m × n and B is n × k, then C = AB
is m × k. Then,

2
 • A2 · A3 ∈ M
 3x3 
−1 0  
1 0 1
A2 · A3 =  2 1  · =
0 1 −1
8 4
 
−1 · 1 + 0 · 0 −1 · 0 + 0 · 1 −1 · 1 + 0 · −1
=  2 · 1 + 1 · 0 2 · 0 + 1 · 1 2 · 1 + 1 · −1 
8 · 1 + 4 · 0 8 · 0 + 4 · 1 8 · 1 + 4 · −1
 
−1 0 −1
= 2 1 1 
8 4 4

• A5 · A6 ∈ M
 3x3     
1 0 1 1 3 1 1 3 −4
A5 · A6 = 2 1 1  · 0 1 0  = 2 7 −3
8 4 −2 0 0 −5 8 28 18
 
1 · 1 + 0 · 0 + 1 · 0 1 · 3 + 0 · 1 + 1 · 0 1 · 1 + 0 · 0 + 1 · −5
= 2 · 1 + 1 · 0 + 1 · 0 2 · 3 + 1 · 1 + 1 · 0 2 · 1 + 1 · 0 + 1 · −5
8 · 1 + 4 · 0 − 2 · 0 8 · 3 + 4 · 1 − 2 · 0 8 · 1 + 4 · 0 − 2 · −5

• A6 · A5 ∈ M3x3

    
1 3 1 1 0 1 15 7 2
A6 ·A5 = 0 1 0 ·2 1 1  =  2 1 1
0 0 −5 8 4 −2 −40 −20 10
• A1 · A3 ∈ M2x3

     
1 0 1 0 1 1 0 1
A1 · A3 = · =
0 1 0 1 −1 0 1 −1

2. Determine x, y ∈ R such that

     
3 x 1 2 1 3 7 3 7
+2 =
1 2 0 5 x 4 11 y 8

3
 Solution
Let us first, compute the sum on the left-hand side of the equa-
tion:
         
3 x 1 2 1 3 3 x 1 4 2 6 7 2+x 7
+2 = + =
1 2 0 5 x 4 1 2 0 10 2x 8 11 2 + 2x 8
The problem is thus, to find the values of x and y solving
   
7 2+x 7 7 3 7
=
11 2 + 2x 8 11 y 8
That is, we need to solve: x + 2 = 3 and 2 + 2x = y. The
solution is x = 1 y = 4
3. Solve the equation X · A + B = C, where
 
0 2 0    
4 −2 1 1 −3 5
A = 3 0 −3 , B = , C=
5 1 −3 −2 4 −6
0 1 2

4
 Solution
Since A ∈ M3x3 and B, C ∈ M2x3 if follows that X ∈ M2x3 .
Let  
a b c
X=
d e f
The left-hand side X · A + B, is
 
  0 2 0  
a b c 4 −2 1
· 3 0 −3 + =
d e f 5 1 −3
0 1 2
   
3b 2a + c −3b + 2c 4 −2 1
+ =
3e 2d + f −3e + 2f 5 1 −3
 
3 · b + 4 2 · a + c − 2 −3 · b + 2 · c + 1
3 · e + 5 2 · d + f + 1 −3 · e + 2 · f − 3
Therefore, the equation is
   
3 · b + 4 2 · a + c − 2 −3 · b + 2 · c + 1 1 −3 5
=
3 · e + 5 2 · d + f + 1 −3 · e + 2 · f − 3 −2 4 −6
The solution of the equation is
−3 1 7
a= , b = −1, c = , d = 4, e = − , f = −5
4 2 3
and accordingly, the matrix X is
 −3 1

−1
X = 4 −7 2 .
4 3 −5

4. Find the matrix X that solves the following equation:

     
1 1 4 −2 6 4
·X · =
3 4 −3 2 22 14

Solution

5
 We need X ∈ M2x2 . Let us denote this matrix as
 
a b
X= .
c d
Then, the problem is to solve
     
1 1 a b 4 −2
· · =
3 4 c d −3 2
   
4a + 4c − 3b − 3d −2a − 2c + 2b + 2d 6 4
=
12a + 16c − 9b − 12d −6a − 8c + 6b + 8d 22 14

that gives as solution: a = 5, b = 6, c = 7, d = 8, with:

 
5 6
X=
7 8
Alternatively, we can invert the matrices. Namely, AXB = C
means XB = A−1 C and X = A−1 CB −1 . Then,
 −1    −1
1 1 6 4 4 −2
X= · ·
3 4 22 14 −3 2

5. Find the matrices B that commute to the following matrix A:

 
0 1 0
A = 0 0 1
0 0 0

Solution
To find the matrix X that commutes to matrix A means to find
X such that
X ·A=A·X
Since A ∈ M3x3 ⇒ X ∈ M3x3 . Let,
 
a b c
X = d e f
g h i

6
 Then,
     
a b c 0 1 0 0 1 0 a b c
X·A = A·X = d e f ·0 0 1 = 0 0 1·d e f  ⇔
g h i 0 0 0 0 0 0 g h i
   
0 a b d e f
0 d e = g
  h i
0 g h 0 0 0
It gives as solution: a = e = i, b = f, d = g = h = 0 and
hence:  
a b c
X = 0 a b 
0 0 a
Double-check:
     
a b c 0 1 0 0 a b
X · A = 0 a b  · 0 0 1 = 0 0 a
0 0 a 0 0 0 0 0 0
     
0 1 0 a b c 0 a b
A · X = 0 0 1 · 0 a b  = 0 0 a
0 0 0 0 0 a 0 0 0
6. Find the matrix X that solves the equation A · X = X · A,
where:  
0 0 0
A = 1 1 0
0 0 0
Solution
look for X· A= A · X.
Proceed as above to   
a b c 0 0 0 0 0 0
X · A = A · X = d e f  · 1 1 0 = 1 1 0 ·
g h i 0 0 0 0 0 0
 
a b c
d e f  ⇔
g h i

7
    
b b 0 0 0 0
⇔  e e 0  = a + d b + e c + f 
h h 0 0 0 0
Solution: a = e − d, b = h = 0, f = −c:
 
e−d 0 c
X= d e −c
g 0 i
Double-check:
     
e−d 0 c 0 0 0 0 0 0
X ·A= d e −c · 1 1 0 = e e 0 =
g 0 i 0 0 0 0 0 0
     
0 0 0 e−d 0 c 0 0 0
A·X = 1 1
 0 ·
  d e −c = e
  e 0
0 0 0 g 0 i 0 0 0
7. Calculate the 2 × 2 matrixes with A2 = A:
Solution
Let  
a b
A=
c d
Then,    
a b a b
A·A= ·
c d c d
so that we want to solve
 2   
a + bc ab + bd a b
=
ca + dc bc + d2 c d

Solution: a2 + bc = a, ab + bd = b, ca + dc = c, bc + d2 = d.
For instance:
d(1−d)
Sol 1: a = 1 − d, b = b, c = b , d = d.
Sol 2: a = 1, b = 0, c = c, d = 0.

8
Sol 3: a = 0, b = 0, c = 0, d = 0.
Sol 4: a = 0, b = 0, c = c, d = 1.

8. Compute the following matrix determinants:

1 0 −3 −1 1 1 2 0 1
a) −1 6 7 ; b) 1 −1 1 ; c) 1 1 0
0 1 1 1 1 −1 2 0 −2
−4 −6 −2 1 0 3
d) 2 4 2 ; e) −1 1 1
−2 −3 −1 2 1 −1

Solution
Recall Sarrus rule:
a11 a12 a13
a21 a22 a23 = a11 a22 a33 + a12 a23 a31 + a13 a21 a32
a31 a32 a33
− a13 a22 a31 − a12 a21 a33 − a11 a23 a32

9
 1 0 −3
[a)] −1 6 7 = (1 · 6 · 1) + (−1 · 1 · −3) + 0 − [0 + (1 · 7 · 1) + (1 · 0 · −1
0 1 1
−1 1 1
[b)] 1 −1 1 = [−1 + 1 + 1] − [−1 − 1 − 1] = 4
1 1 −1
2 0 1
[c)] 1 1 0 = −6
2 0 −2
−4 −6 −2
[d)] 2 4 2 = 0
−2 −3 −1
1 0 3
[e)] −1 1 1 = −11
2 1 −1

9. Solve the following equations

x 2x + 1 2x + 1 4 8 6
a) 2x + 1 3x − 1 4x = 0; b) 5 7 12 = 0
3x − 1 4x 6x − 1 3 −1 x

Solution

x 2x + 1 2x + 1
a) 2x + 1 3x − 1 4x = −6x2 + 3x
3x − 1 4x 6x − 1
and
1
−6x2 + 3x = 0 ⇔ x = {0, }
2
4 8 6
b) 5 7 12 = −12x + 180
3 −1 x

10
 and
−12x + 180 = 0 ⇔ x = 15

10. Verify that the following equations are true

a+b a a
a) a a+b a = b2 · (3a + b)
a a a+b
a−b−c 2a 2a
b) 2b b−a−c 2b = (a + b + c)3
2c 2c c−a−b

Solution

a+b a a
[a)] a a+b a | = 3ab2 + b3 = b2 (3a + b)
a a a+b

a−b−c 2a 2a
[b)] 2b b−a−c 2b =
2c 2c c−a−b
6abc+3ab2 +a3 +3ac2 +3b2 c+b3 +3ba2 +3bc2 +3ca2 +c3 = (a+b+c)3

11. When possible, find the inverse of the following matrices

1
   
1 −1 2 3 2
−4 83
   
3 8 3
a) ; b) ; c) −2 3 1 ; d) −2 1 0
5 − 43
  
11 2
3 −1 1 0 −1 5

11
 Solution
−1
−4 −3
    
3 8 4 −8 1 3 32
a) = =
5 − 34 169 −5 3 169 20 −12
−1
−4 83
    3 1
−3 2 −11 −
b) = 8 = 3256 14 3
11 2 112 − 3 4 112 28
3 −1 −1 −1 −1 3
 
 1 1 −
 −1 2 1 2 1 
1 −1 2  
−1  −2 3
 1 3 1 −2 
c) −2 3 1 = − − 
15  1 1 2 1 2 1 
3 −1 1 
 −2 3

1 3 1 −2 
−
3 −1 −1 −1 −1 3
 
−4 1 7
1 
−1 1 1
15
7 2 −1
−1
2 13
  
3 15 −31 −1
1 
d) −2 1 0 = 30 45 −2
107
0 −1 5 6 9 21

12. Compute the rank of the following matrices

3 2 31
   
    −2 3 1
2 3 2 3 1
a) ; b) ; c)  1 −1 0 ; d) −2 1 0
−4 −6 1 2 3
0 1 1 0 −1 5

Solution
Use Gauss, i.e. the row echelon form to find the rank (See
example d)). After that use the rule: The rank of a matrix with
more rows than columns is “# of columns” minus “# of Zero-
rows”. If more columns than rows, then rank is “# of rows”

12
 minus “# of Zero-rows”.
 
2 3
[a)] rk =1
−4 −6
 
2 3 1
[b)] rk =2
1 2 3
 
−2 3 1
[c)] rk  1 −1 0 = 2
0 1 1
3 2 13
 

[d)] rk −2 1 0 = 3
0 −1 5

13. Compute the rank of the following matrices as a function of the

parameter a
     
1 2 3 a 1 0 a 1 1
a) 1 a 1 ; b) 2 2 a  ; c) 1 a 1 ;
2 3 4 1 a −1 1 1 a

Solution

a)  
1 2 3
det 1 a 1 = −2a + 2
2 3 4
Then, −2a + 2 = 0 ⇔ a = 1. Therefore,
  (
1 2 3
2 if a = 1
rk 1 a 1 =
2 3 4 3 if a 6= 1

b)  
a 1 0
det 2 2 a  = −a3 − a + 2
1 a −1

13
 Then, a3 − a + 2 = 0 ⇔ a = 1. Therefore,
  (
a 1 0
2 if a = 1
rk 2 2 a  =
1 a −1 3 if a 6= 1

Applying Gauss we have,

     
a 1 0 1 a −1 1 a −1
rk 2 2 a  = rk 2 2 a  = rk 0 2a − 2 −2 − a =
1 a −1 a 1 0 0 a2 − 1 −a
 
1 a −1
rk 0 2a − 2 −2 − a  = 3
0 0 2 − a3 − a

c)  
a 1 1
det 1 a 1 = a3 − 3a + 2
1 1 a
Then, a3 − 3a + 2 = 0 ⇔ a = {1, −2}. Therefore,
 
1 if a = 1

a 1 1 
rk 1 a 1 = 2 if a = −2
1 1 a

3 if a 6= {1, −2}


14. [Sydsaeter-Hammond (1995, Example 12.21)] Consider three

firms competing in a market whose market shares on January
1, 2017 were m1 = 0.2, m2 = 0.6, m3 = 0.2. Along the year
2017 the following changes in market shares occurred:

• Firm 1 keeps 85% of its customers, while losing 5% to

Firm 2, and 10% to Firm 3;
• Firm 2 keeps 55% of its customers, while losing 10% to
Firm 1, and 35% to Firm 3;

14
 • Firm 3 keeps 85% of its customers, while losing 10% to
Firm 1, and 5% to Firm 2.

(i) Define the vector of market shares.

(ii) Define the transition matrix.
(iii) Compute the vector of markets shares on January 1, 2018.

Solution

(i) Define the vector of initial market shares.

 
0.2
m = 0.6
0.2

(ii) Define the transition matrix.

 
0.85 0.10 0.10
T = 0.05 0.55 0.05
0.10 0.35 0.85

(iii) Compute the vector of markets shares on January 1, 2018.

    
0.85 0.10 0.10 0.2 0.5
T s = 0.05 0.55 0.05 0.6 = 0.35
0.10 0.35 0.85 0.2 0.40

15. An employer want to estimate the performance of the employ-

ees based on the scores obtained in an ability test and in a re-
sponsibility test. The data obtained is collected in the following
table, where y stands for performance, x1 for ability, and x2 for
responsibility.
The employer thinks that a linear model of the type

y = a + b1 x1 + b2 x2

15
 y x1 x2
1 4 2.5
3 5.5 3
5 6 3.8

will be accurate enough, and plans to use ordinarily least squares.

This means that the vector of parameters b is obtained as the so-
lution of the following computation:
b = (X 0 X)−1 X 0 Y
where X is the matrix of values of the variables x1 and x2 ; X 0 is
the transpose of matrix X, and Y is the vector of performance.
Compute the parameters b1 and b2 .
Solution
First, we define matrices X and Y . These are
   
4 2.5 1
X = 5.5 3  ; Y = 3
6 3.8 5
so that,  
0 4 5.5 6
X =
2.5 3 3.8
Then,  
082.25 49.3
XX=
49.3 29.7
and  
0 −1 2.40 −3.89
(X X) =
−3.89 6.66
Also,  
50.5
X 0Y =
30.5
Finally the estimated values of the parameters b1 and b2 are
   
b1 −0.49
= (X 0 X)−1 X 0 Y =
b2 1.63

16
 Part II. Systems of linear equations

16. Classify and if possible, solve the following systems of equa-

tions:
 
x + 4y + 8z = 0
 2x − 5y + 3z = −12

(a) −2x + 5y − 2z = 3 (b) 3x + 2y − 5z = 1
 
3x + 7y + z = 2 7x − 4y + 2z = 0
 
 

 x + 2y − 3z = −7 x − 3y + z = −13

(c) 3x + 2y + 5z = −1 (d) 2x + 4y + 3z = 47
 
4x − y + z = 25 3x + 5y − 2z = 44
 

x − y + z = 6

(e) x + y − z = 2

x + y + z = 12


Solution

a) Write the system in matrix form

 
1 4 8 |0
−2 5 −2 |3
3 7 1 |2
We can obtain an equivalent system by
• redefining row 2 as: r2 = r2 + 2r1
• redefining row 3 as: r3 = 3r1 − r3
to obtain  
1 4 8 | 0
0 13 14 | 3 
0 5 23 | −2
Next, we can obtain an equivalent system by
• redefining row 3 as: r3 = 5r2 − 13r3

17
 to obtain  
1 4 8 | 0
0 13 14 | 3 
0 0 −229 | 41
Note that  
1 4 8
rk −2 5 −2 = 3
3 7 1
and  
1 4 8 | 0
rk 0 13 14 | 3  = 3
0 0 −229 | 41
Since the ranks coincide, and given that the number of
equations equals the number of unknowns, the Roche-Frobenius
theorem guarantees a unique solution. Solving the triangu-
lar system, we obtain
60 97 41
x=− , y=− , z=−
229 229 229
(b) Write the system in matrix form
 
2 −5 3 | −12
A = 3 2 −5 | 1 
7 −4 2 | 0

Then, det(A) = 95 6= 0, rk(A) = (A|b) = 3 ⇒ unique

solution:
x = 2, y = 5, z = 3
(c) Write the system in matrix form
 
1 2 −3 | −7
A = 3 2 5 | −1
4 −1 1 | 25

18
 Then, det(A) = 74 6= 0, rk(A) = (A|b) = 3 ⇒ unique
solution:
175 241 16
x= , y=− , z=−
37 37 37
(d) Write the system in matrix form
 
1 −3 1 | −13
A = 2 4 3 | 47 
3 5 −2 | 44

Then, det(A) = −64 6= 0, rk(A) = (A|b) = 3 ⇒ unique

solution:
x = 5, y = 7, z = 3
e) Write the system in matrix form
 
1 −1 1 | 6
A = 1 1 −1 | 2 
1 1 1 | 12

Then, det(A) = 4 6= 0, rk(A) = (A|b) = 3 ⇒ unique

solution. Now use Cramer:
6 −1 1 1 6 1 1 −1 6
2 1 −1 1 2 −1 1 1 2
12 1 1 16 1 12 1 12 1 1 12 20
x= = ; y= = ; z= =
det(A) 4 det(A) 4 det(A) 4
Therefore, the solution is

x = 4, y = 3, z = 5

17. Classify and if possible, solve the following systems of linear

19
equations:
 
2x + 3y − 7z = −1
 3x + 2y − 4z = 1

(a) 3x + 4y − 6z = 5 (b) 5x − y − 2z = 2
 
5x − 2y + 4z = −7 x + 3y − z = 3
 

2x − 5y + 3z = −12
 (
x − 2y + z = 3
(c) 3x + 2y − 5z = 1 (d)
 3x + y − 5z = 2
7x − 4y + 2z = 0


3x + 5y = 1

(e) 2x − y = 23

2x + 25y = −6


Solution
a) Write the system in matrix form
 
2 3 −7 | −1
A = 3 4 −6 | 5 
5 −2 4 | −7
Then, det(A) = −64 6= 0, rk(A) = (A|b) = 3 ⇒ unique
solution:
x = −1, y = 5, z = 2
b) Write the system in matrix form
 
3 2 −4 | 1
A = 5 −1 −2 | 2
1 3 −1 | 3
Then, det(A) = −37 6= 0, rk(A) = (A|b) = 3. Solving
using Cramer rule, we obtain,
1 2 −4 3 1 −4 3 2 1
2 −1 −2 5 2 −2 5 −1 2
3 3 −1 −37 1 3 −1 37 1 3 3 37
x= = ; y= = ; z= =
det(A) −37 det(A) 37 det(A) 37
20
 Therefore, the solution is

x=y=z=1

c) Write the system in matrix form

 
2 −5 3 | −12
A = 3 2 −5 | 1 
7 −4 2 | 0

Then, det(A) = 95 6= 0, rk(A) = (A|b) = 3 ⇒ unique

solution:
x = 2, y = 5, z = 3
d) Write the system in matrix form
 
1 −2 1 | 3
A=
3 1 −5 | 2
Applying Gauss we can write
   
1 −2 1 | 3 1 −2 1 | 3
∼
3 1 −5 | 2 0 −7 8 | 7

Given that, rk(A) = (A|b) = 2 while there are 3 un-

knowns. Therefore, the Roche-Frobenius theorem tells us
that the system is undetermined. The solution will depend
of (3 − 2) = 1 parameter. Call this free variable z and
solve to obtain:
9z 8z
x= + 1, y = − 1, z = z.
7 7
e) Write the system in matrix form
 
3 5 | 1
A = 2 −1 | 23 
2 25 | −6

21
 Applying Gauss we can write
     
3 5 | 1 3 5 | 1 3 5 | 1
2 −1 | 23  ∼ 0 13 | −67 ∼ 0 13 | −67 
2 25 | −6 0 26 | −29 0 0 | −105
Given that rk(A) = 2 6= (A|b) = 3, the Roche-Frobenius
theorem tells us that the system is incompatible, and thus
has no solution.
18. Classify and solve the following systems of linear equations as
a function of the parameter a:
 
2x + y + az = 4
 ax + y + z = 1

(a) x + z = 2 (b) x + ay + z = 2
 
x+y+z =2 x + y + az = 3
 

x + 2y + 3z = 0

(c) x + ay + z = 0

2x + 3y + 4z = 2


Solution
a) Write the system in matrix form:
 
2 1 a | 4
(M |b) = 1 0 1 | 2
1 1 1 | 2
Then, det(M ) = a − 2, det(M ) = 0 ⇔ a = 2. Hence,
(
a = 2 ⇒ rk(M ) = 2 = rk(M |b) (Two equal columns) Infinite solutions
a 6= 2 ⇒ rk(M ) = 3 = rk(M |b) ⇒ Unique solution
b) Write the system in matrix form:
 
a 1 1 | 1
(M |b) = 1 a 1 | 2
1 1 a | 3

22
 Then, det(M ) = a3 − 3a + 2; det(M ) = 0 ⇔ a = 1, −2.
Hence, (see problem 13(c)),



 if a = 1 then ⇒ rk(M ) = 1 6= 2 = rk(M |b)




 ⇒ Incompatible

if a = −2 then ⇒ rk(M ) = 2 6= 3 = rk(M |b)


 ⇒ Incompatible




 if a 6= {1, −2} then ⇒ rk(M ) = 3 = rk(M |b)

⇒ Compatible and determinate

c) Write the system in matrix form:

 
1 2 3 | 0
(M |b) = 1 a 1 | 0
2 3 4 | 2
Then, det(M ) = −2a + 2; det(M ) = 0 ⇔ a = 1. Hence,
(see problem 13(a)),



 if a = 1 then ⇒ rk(M ) = 2 6= 3 = rk(M |b)

⇒ Incompatible


 if a 6= 1 then ⇒ rk(M ) = 3 = rk(M |b)

⇒ Compatible and determinate

19. Consider a two good economy (j = 1, 2) where two individu-

als(i=a,b) aim at consuming the following bundles:

(qa1 , qa2 ) = (10, 8); (qb1 , qb2 ) = (6, 12)

These individuals are endowed with incomes (wa , wb ) = (50, 60).

Determine, using matrix algebra, whether there exists a price
system p1 , p2 ) such that those consumption bundles are feasi-
ble.
Solution

23
We can describe the two consumers through their budgets con-
straints as

p1 qa1 + p2 qa2 = wa
p1 qb1 + p2 qb2 = wa

that is

10p1 + 8p2 = 50
6p1 + 12p2 = 60

or in matrix form
    
10 8 p1 50
=
6 12 p2 60
The problem to solve is
   −1    1 1
 
p1 10 8 50 6 − 9 50
= = 1 5 =
p2 6 12 60 − 12 36 60
 50 60   150−120   5 
6 − 9 18
= −150+300 3
= 25
50
− 12 + 300
36 36 6

24