Exercise Session 2, 2 October 2018

Mathematics for Economics and Finance

Prof: Norman Schürhoff
TAs: Jakub Hajda, Jimmy Lerch

Exercise 1
(a) Compute
3 1 −1 2
−5 1 3 −4
D= .
2 0 1 −1
1 −5 3 −3

(b) Calculate rank(A), |A|, tr(A), A−1 for  

1 4 7
A= 3 2 5 .
5 2 8

Solution 1
(a) We first simplify the matrix in order to make the computation of the determinant faster.
Reminder: If we add or subtract, from one row, a multiple of another row, then the determinant does not
change.
We first subtract 3 times the third row to the fourth row (row4 − 3 × row3 ).

3 1 −1 2 3 1 −1 2
−5 1 3 −4 −5 1 3 −4
D= −→ D1 =
2 0 1 −1 2 0 1 −1
1 −5 3 −3 −5 −5 0 0

Since that det(D) = det(D1 ) we compute the det(D1 ) applying the Laplace formula to the last line.

1 −1 2 3 −1 2
det(D1 ) = (−1)4+1 (−5) −1 3 −4 + (−1)4+2 (−5) −5 3 −4
0 1 −1 2 1 −1
3 1 2 3 1 −1
+(−1)4+3 (0) −5 1 −4 + (−1)4+4 (0) −5 1 3 =
2 0 −1 2 0 1
= 5 [−3 + 0 + 2 + 0 + 4 − 1] − 5 [−9 + 8 − 10 − 12 + 12 + 5] + 0 + 0
= 5 × 2 + (−5) × (−6) = 10 + 30 = 40 = det(D)
 
6 −18 6
(b) |A| = −18, rank(A) = 3, tr(A) = 11 and A−1 = − 18
1 
1 −27 16  .
−4 18 −10

1
Exercise 2
A and B are squared matrixes of order n > 2, det A and det B are their determinants. Which of the following
statements are true:

(a) if det A = det B, then det(A − B) = 0.

(b) if A and B differ only by interchange of 2 columns, then det A = det B.
(c) if det A 6= 0, then either det B = 0 or det(AB) = 0.

Solution 2
(a) Consider n =2
   
a11 a12 b11 b12
A= B= .
a21 a22 b21 b22
det A = det B V a11 a22 − a21 a12 = b11 b22 − b21 b12 .
a11 − b11 a12 − b12
det(A − B) =
a21 − b21 a22 − b22
= a11 a22 − b11 a22 − a11 b22 + b11 b22 − a12 a21 + b21 a12 + a21 b12 − b12 b21 .

as you see only under special conditions on coefficients of matrices A and B det(A − B) = 0, otherwise
det(A − B) 6= 0.
Hence if det A = det B, it does not mean det(A − B) = 0.
(b) It is not true. In fact
det A = − det B.

(c)

det(AB) = det A det B.

1) if det(AB) = 0, since det A 6= 0 =⇒ det B = 0.
2) if det(AB) 6= 0, since det A 6= 0 =⇒ det B 6= 0.
If det A 6= 0 =⇒ either det B = 0 (hence det AB = 0) or det AB 6= 0.

Exercise 3
Consider a matrix Am×n . Which of the following statements are true and which are false? Explain why.

(a) If the rows of matrix A are linearly independent, then m > n.

(b) If the rows of matrix A are linearly dependent, then m > n.
(c) If m > n, then the rows of matrix A are linearly independent.
(d) If m > n, then the rows of matrix A are linearly dependent.

Solution 3
(a) False. If the rows of matrix A are linearly independent, then m can not exceed n.

2
 (b) False. If the rows of matrix A are linearly dependent, then m and n can be in whatever relation.
If the rows of matrix A are linearly dependent, then rank(A) = row rank(A) = k, k < m. At the same
time, we do not know anything about the linear independence of the columns, k ≤ n. Thus, k < m, k ≤ n
which does not allow us to conclude anything about the relation between m and n.
(c) False. We can have at most n linearly independent rows, so if m > n, then the rows of matrix A are linearly
dependent.
(d) True. We can have at most n linearly independent rows, so if m > n, then the rows of matrix A are linearly
dependent.

Exercise 4
For what values of the constant p and q does the following system have a unique solution, several solutions, or
no solution?
For what values of the constant p and q does the following system have a unique solution, several solutions,
or no solution? 
 x1 + x2 + x3 = 2q
2x1 − 3x2 + 2x3 = 4q
3x1 − 2x2 + px3 = q.


Solution 4
Let    
1 1 1 2q
A =  2 −3 2  and b =  4q , then we rewrite the problem as Ax = b.
3 −2 p q

• Case 1: Homogeneous Case. q = 0 ⇒ b = 0.

If |A| =
6 0, which means p 6= 3, then ∃! a trivial solution 0.
If p = 3, then ∃ infinite number of solution.
• Case 2: Inhomogeneous Case. b 6= 0 (q 6= 0).
If |A| =
6 0, which means p 6= 3, then ∃ unique solution.
   
1 1 1 2q 1 1 1 2
If |A| = 0, p = 3 ⇒ rank(A) = 2, while à =  2 −3 2 4q  ∼  2 −3 2 4  and rank(Ã) = 3,
3 −2 3 q 3 −2 3 1
so no solution.

Exercise 5
Consider the following system of equations:


 w − x + y − z = b1
w − x + y + z = b2


 4w − 4x + 4y = b3
−2w + 2x − 2y + z = b4 .


(a) Give the solution set of the homogenous system.

3
 (b) Give the solution set of the inhomogenous system with b = [2, 0, 4, −3].

Solution 5
Let    
1 −1 1 −1 2
 1 −1 1 1   0 
A=  4 −4
 and b =  , then we rewrite the problem as Ax = b.
4 0   4 
−2 2 −2 1 −3
(a) To solve the homogenous system, let us first perform several operations on the rows of the following
augmented matrix as if solving the inhomogenous case:
 
1 −1 1 −1 2
 1 −1 1 1 0 
(A|b) = 
 4 −4 4
,
0 4 
−2 2 −2 1 −3
which results in the following modified system:
 
1 −1 1 0 1
] =
 0 0 0 1 −1 
(A|b)  0 0
.
0 0 0 
0 0 0 0 0

] which in our case is equivalent

Clearly, to solve for the homogenous system we have to solve the system (A|0)
to: 
w−x+y =0
z =0
Letting x and y being the so-called free variables of the problem (see that we have two equations in four
unknowns i.e. there are infinitely many solutions), we have to find the basic solutions, which constitute all
the other solutions to the problem. Let us assume that x = s and y = t. Therefore, the set of all solutions
to the homogenous system take the form of:


 w =s−t
z =0

,

 x =s
y =t


for all s, t ∈ R, which we rewrite as

   

 1 −1 

  1   0 
U = s, t ∈ R s 
 0  + t 1
  


 
0 0
 

(b) To solve the inhomogenous system, let us first remember that the solution to the homogenous system is the
] and the solution of the homogenous system U. It is easy to
sum of a particular solution of the system (A|b)
notice that the particular solution to the system (pick again x and y as free variables, set s = 0 and t = 0)
has the form of λ = (1, 0, 0, −1)0 . Thus we get:
      

 1 1 −1  
  0   1   0 
λ + U = s, t ∈ R    + s
  + t
  .

 0  0  1  
−1 0 0
 

4
Exercise 6
Let A be a matrix  
1 2 3
2 4 6
A=
−1

4 1
2 −8 −2
.
(a) Find a basis for the null space of A, N ull(A).
(b) Find a basis for the column space of A, Col(A).
(c) Verify the Rank-Nullity Theorem (Theorem A.7, p.248 of Lecture Notes).

Solution 6
(a) The null space of A is the set of all vectors x such that Ax = 0.
 
1 2 3   
2 x 0
 4 6· y = 0 
−1 4 1
z 0
2 −8 −2
     
1 2 3 1 2 3 3 0 5
 2 4 6   0 0 0   0 0 0 
 ⇔ ⇔ 
 −1 4 1   0 6 4   0 3 2 
2 −8 −2 0 −12 −8 0 0 0
The solution to the system is (x, y, z) = (5a, 2a, −3a)0 , a ∈ R and N ull(A) = (5, 2, −3)0 .
(b) The column space is the subspace that is spanned by the columns of a matrix. In other words it is the
subspace built taking all the possible (linear) combinations of the columns of a matrix.
Using the simplified matrix from the previous point
 
3 0 5
 0 0 0 
 ,
 0 3 2 
0 0 0

one can say that the first and the second columns of a matrix are linearly independent, while the third
column is a linear combination of them. The corresponding columns in the initial matrix A will present the
basis of its column space    

 1 2 

2   4 

Col(A) =    ,   .

 −1   4  
2 −8
 

(c) One needs to show that dim(Col(A)) + dim(N ull(A)) = dim(A).

dim(Col(A)) = 2, as Col(A) consists of two linearly independent vectors.
dim(N ull(A)) = 1, as N ull(A) consists of one vector.
dim(Col(A)) + dim(N ull(A)) = 3, i.e. the Theorem holds for the 3 × 4-dimensional matrix A.
(d) ALTERNATIVE SOLUTION TO (a) and (b)
The column space is the subspace that is spanned by the columns of a matrix. In other words it is the
subspace built taking all the possible (linear) combinations of the columns of a matrix:

5
 Let us find x1 , x2 , x,3 ∈ Rs.t. xi 6= 0 for at least one i and
       
1 2 3 0
2 4  6  0
x1 −1 + x2  4  + x3  1  = 0
      

2 −8 −2 0

This system of three equations leads to the following results:

5 5 2
x1 = x2 , x1 = − x3 , x2 = − x3 .
2 3 3
Therefore, the columns of A are not lineary independent. We have dim Col(A) = 2.
One can say that the first and the second columns of a matrix are linearly independent, while the third
column is a linear combination of them. The corresponding columns in the initial matrix A will present the
basis of its column space    

 1 2 

  2   4 
Col(A) =  ,  .
 −1
    4 

2 −8
 

Let us now find a basis for the null space of A, which is the set of all vectors x such that Ax = 0.
From the first part, we have
5 5 2
x1 = x2 , x1 = − x3 , x2 = − x3 .
2 3 3
Therefore, N ull(A) = (5x, 2x, −3x), x ∈ R. A basis for this space is
 
 5 
N ull(A) =  2  .
−3
 

Note that in practise, it is common to normalize the vectors of a basis, which can be done by dividing each
element by the norm of the corresponding vector.

Exercise 7
Discuss the solutions of following system for all values of a and b.

 x + 2y + 3z = 1
−x + ay − 21z = 2 .
3x + 7y + az = b


Solution 7
 
    
 x + 2y + 3z = 1  1 2 3  x 1
−x + ay − 21z = 2 ⇔ .
 −1 a −21 
 y  =  2 .
3x + 7y + az = b 3 7 a z b

| {z }
A

if det A 6= 0 =⇒ system has unique solution.

if det A = 0 =⇒ if rank(A) = rank(Â) =⇒ infinite number of solutions.
if rank(A) 6= rank(Â) =⇒ system is inconsistent.

6
 (a) Check under which conditions det A = 0.

1 2 3 1 2 3
a+2 −18
−1 a −21 1 ∗ row 1 = 0 a+2 −18 =1∗
1 a−9
3 7 a −3 ∗ row 1 0 1 a−9
= (a + 2)(a − 9) + 18 = 2a + a2 − 18 − 9a + 18 = a2 − 7a = a(a − 7)
=⇒ det A = 0 ⇐⇒ a = 0, a = 7.

(b) Check under what conditions rank(A) = rank(Â) if det A = 0.

   
1 2 3 1 1 2 3 1
 −1 a −21 2  1 ∗ row 1 −→  0 a + 2 −18 −2 ∗ row 3
2  −→
−(a + 2) ∗ row 3
3 7 a b −3 ∗ row 1 0 1 a−9 b
   
1 0 −2a + 21 −2b + 7 1 0 −2a + 21 −2b + 7
 0 0 −(a + 2)(a − 9) − 18 3 − (b − 3)(a + 2)  −→  0 0 −a2 + 7a −ab + 3a − 2b + 9 
0 1 a−9 b−3 0 1 a−9 b−3

a = 0 =⇒ det A = 0.
 
1 0 21  
1 0 21
rank(A)a=0 : 0 0 0  −→ =⇒ rank(A) = 2.
0 1 −9
0 1 −9

if rank(A) = rank(Â) =⇒ infinite number of solutions.

 
1 0 −2b + 7
21
rank(Â)a=0 :  0 0 −2b + 9 
0
0 1 b−3 −9
9
rank(Â) = 2 ⇐⇒ −2b + 9 = 0 =⇒ b = .
  2
1 0 21 −2  
1 0 21 −2
=⇒ Â a=0 = 0 0 0 0  −→ 3 =⇒ rank(Â) = 2.
b= 29 3 0 1 −9
0 1 −9 2
2

a = 7 =⇒ det A = 0.
 
1 0 7  
1 0 7
rank(A)a=7 : 0 0 0  −→ =⇒ rank(A) = 2.
0 1 −2
0 1 −2
 
1 0 7 −2b + 7
rank(Â)a=0 : 0 0 0 −7b + 21 − 2b + 9 
0 1 − 23 b−3

if rank(A) = rank(Â) =⇒ infinite number of solutions.

10
rank(Â) = 2 ⇐⇒ −7b + 21 − 2b + 9 = 0 =⇒ b = .
3
1
 
1 0 7 3
 1

1 0 7
=⇒ Â a=7 =  0 0 0 0  −→ 3
1 =⇒ rank(Â) = 2.
b= 10 1 0 1 −2
3 0 1 −2 3
3

Answer:
If a 6= 0 or a 6= 7 =⇒ det A 6= 0 =⇒ system has a unique solution.

7
 If a = 0 and b = 29 =⇒ det A = 0 =⇒ rank(A) = rank(Â) = 2 =⇒ system has infinite number of
solutions.
If a = 7 and b = 10 3 =⇒ det A = 0 =⇒ rank(A) = rank(Â) = 2 =⇒ system has infinite number of
solutions.
If a = 0 and b 6= 92 =⇒ det A = 0 =⇒ rank(A) 6= rank(Â) =⇒ system has no solution.
If a = 0 and b 6= 10
3 =⇒ det A = 0 =⇒ rank(A) 6= rank(Â) =⇒ system has no solution.

Exercise 8
(a) Let us consider the quadratic form Q given by
Q(x, y) = ax2 + bxy + cy 2 .
Show that Q is positive definite if and only if a > 0 and b2 < 4ac.
(b) Determine the definiteness of
Q = 3x21 + x22 + 6x23 − 2x1 x2 + 2x1 x3 + 4x2 x3

(c) Determine the definiteness of

Q = −x21 + 6x1 x2 − 9x22 − 2x23 .

Solution 8
(a) Recall that a quadratic form Q is PD ⇔ Dk > 0 for all leading principal minors of order k = 1, 2, . . . , n
(proposition 2.9).

b
 
2 2 T T a 2
Here we have Q(x, y) = ax + bxy + cy = x Ax, where x = (x, y) and A = b .
2 c
[1.] D1 > 0 ⇔ a > 0.
b2
[2.] D2 > 0 ⇔ det(A) > 0 ⇔ ac − > 0 ⇔ b2 − 4ac < 0.
4
Therefore, Q is PD if and only if a > 0 and b2 − 4ac < 0.

Additionally, observe that b2 − 4ac < 0 ⇔ 4ac > b2 ≥ 0 ⇒ 4ac > 0. Therefore, the two necessary and
sufficient conditions for Q to be PD also imply c > 0.
(b) In matrix notation  
3 −1 1
Q = x0 Ax, where A =  −1 1 2 .
1 2 6
Leading principle minors are:
D1 = 3 > 0.
3 −1
D2 = = 3 ∗ 1 − (−1) ∗ (−1) = 2 > 0.
−1 1
3 −1 1
1 2 −1 1 −1 1
D3 = −1 1 2 = (−1)1+1 ∗ 3 ∗ + (−1)2+1 (−1) + (−1)3+1 ∗ 1
2 6 2 6 1 2
1 2 6
= 3 ∗ (6 − 4) + 1 ∗ (−6 − 2) + 1 ∗ (−2 − 1) = 6 − 8 − 3 = −5 < 0.
D1 > 0, D2 > 0, D3 < 0 ⇒ not positive definite, not negative definite.

8
 The arbitrary minors are:
order 3 : ∆k=3 = D3 = −5 < 0 ⇒ not positive semidefinite.
order 1 : ∆11 = 3, ∆12 = 1, ∆13 = 6 ⇒ (−1)1 ∆1 < 0 ⇒ not negative semidefinite.
So, quadratic form is indefinite.
(c)   
−1 3 0 x1
Q = x0 Ax =


x1 x2 x3  3 −9 0   x2  .
0 0 −2 x3
Principal minors:
D1 = −1 < 0.
−1 3
D2 = = 9 − 9 = 0.
3 −9
−1 3 0
−1 3
D3 = 3 −9 0 = (−1)3+3 · (−2) · = 0.
3 −9
0 0 −2

Hence, neither positive definite nor negative definite.

Arbitrary minors :
m=1: fix a11 =⇒ ∆1 = −1 <0
fix a22 =⇒ ∆1 = −9 <0
fix a33 =⇒ ∆1 = −2 <0

−1 3
m=2: fix a11 , a11 =⇒ ∆2 = =0 <0
3 −9
−1 0
fix a22 , a22 =⇒ ∆2 = =2 ≥0
0 −2
−9 0
fix a33 , a33 =⇒ ∆2 = = 18 ≥ 0
0 −2
−1 3 0
m=3: fix a11 , a22 , a33 =⇒ ∆3 = 3 −9 0 =0
0 0 −2

Arbitrary minors of order 1 are negative =⇒ not positive semidefinite.

Check negative semidefiniteness:
m=1: (−1)m ∆1 ≥ 0 =⇒ ∆1 ≤ 0.
=⇒ True for matrix A
m=2: (−1)2 ∆2 ≥ 0 =⇒ ∆2 ≥ 0.
=⇒ True for matrix A
m=3: (−1)3 ∆3 ≥ 0 =⇒ ∆3 ≤ 0.
=⇒ True for matrix A =⇒ Quadratic form is negative semidefinite.

Comment: Asymmetric matrices correspond to quadratic forms in very rare and specific cases which we do
not consider in this course. Moreover, the Rules for Checking Definiteness (Lecture Notes, p.49) are valid
only for symmetric matrices. So the general rule is to reconstruct a symmetric matrix from a quadratic
form.