PS Answers Fall2022 Merged
PS Answers Fall2022 Merged
PS Answers Fall2022 Merged
1 Optimization
Consider the function f defined for all (x, y) by
f (x, y) = (1 + y)3 x2 + y 2 .
Prove that f has a unique stationary point at (0, 0) which is a local minimum, but f has no global minimum.
Anwer:
FOC:
∂f
= 2(1 + y)3 x = 0, (1)
∂x
∂f
= 3x2 (1 + y)2 + 2y = 0. (2)
∂y
From (1) ⇒ x = 0 or y = −1. If x = 0, then from (2) ⇒ y = 0. If y = −1, from (2) derive contradiction.
⇒ Only (0, 0) is stationary point.
00 00 00 00
Check Hessian, f11 = 2(1 + y)3 , f22 = 6x2 (1 + y) + 2, f12 = f21 = 6x(1 + y)2 . At point (0, 0),
" #
2 0
H=
0 2
where f is a real-valued C 1 function, θ ∈ Θ is a parameter vector, and X(θ) is a closed, convex subset of RN
for all θ. Denote the gradient by Df (x; θ). Determine whether the following statements are true or false:
1
Answer:
3. False. If Df (x∗ ; θ) 6= 0, then x∗ may be a corner solution yet still a local extremum.
4. True. We want to show that if f is strictly quasi-concave and x∗ is a local max then x∗ is a global
max. By definition of local max, x∗ is a local max ⇒ there exists δ > 0 such that ∀x ∈ X(θ) when
||x−x∗ || < δ it is true that f (x∗ ) ≥ f (x). To prove by contradiction, lets assume that x∗ is not a global
maximum. Then ∃x0 ∈ X(θ) and x0 6= x∗ such that f (x0 ) > f (x∗ ). Consider xλ := λx0 + (1 − λ)x∗ ,
0 ≤ λ ≤ 1. Since X(θ) is convex and x0 , x∗ ∈ X(θ), then xλ ∈ X(θ) for λ ∈ [0, 1]. On the other side,
for λ small enough, it is clear that ||xλ − x∗ || < δ. From strict quasi-concavity definition we have:
and this holds for small enough λ. This is a contradiction with x∗ being a local maximum (since we
got that f (xλ ) > f (x∗ ) even when ||xλ − x∗ || < δ). Thus, if f is strictly quasi-concave, x∗ is not a
global max implies x∗ is not a local max. So, we arrive to a contradiction.
5. True. If f is strictly concave on X × Θ, then the value function is strictly concave. The value function
inherits this properties from the objective function.
3 Cobb-Douglas production
Consider a competitive firm that produces a single output y using two inputs x1 and x2 . The firm’s
production technology is described by a Cobb-Douglas function
β
y(x) = f (x1 , x2 ) = xα
1 x2 ,
where α + β < 1, α > 0 and β > 0. Taking as given the output price p and the input prices w1 and w2 , the
firm maximizes its profits given by
Π(x) = py(x) − wT x.
1. Does the production function exhibit decreasing, constant, or increasing returns to scale? Show.
2. Write the first-order conditions and check whether sufficient conditions for a maximum are satisfied.
3. Solve for the firm’s factor demands, giving the optimal input levels x∗i as functions of input and output
prices.
Answer:
1.
α β
f (λx1 , λx2 ) = (λx1 ) (λx2 )
= λα+β f (x1 , x2 )
So Production function f (x1 , x2 ) is hd(α + β), since α + β < 1, it exhibits decreasing returns to scale.
2
2. For max Π = py(x) − w0 x, the FOC: py 0 (x) − w = 0. It is the sufficient condition for a maximum,
x1 ,x2
if Π(x) is concave. To check this, firstly, we will show y(x) is strictly concave provided α + β <
β−1 ∂ 2 y ∂2y
∂y
1 ⇔Hessian is negative definite. ∂x 1
= αx1α−1 xβ2 , ∂x
∂y
2
= βxα
1 x2 , ∂x2 = α(α − 1)xα−2
1 xβ2 , ∂x 2 =
1 2
2
β−2
β(β − 1)xα
1 x2 , ∂x∂1 ∂x
y
2
= αβxα−1
1 x2β−1 , ⇒
!
α(α − 1)x1α−2 xβ2 αβx1α−1 xβ−1
2
H= .
αβxα−1
1 xβ−1
2 β(β − 1)xα β−2
1 x2
One way to check is to show that the leading principal minors,d1 < 0, d2 > 0,
d1 = α(α − 1)xα−2
1 xβ2 < 0 since (α − 1) < 0
2(α−1) 2(β−1) 2(α−1) 2(β−1)
d2 = |H| = αβ(α − 1)(β − 1)x1 x2 − α2 β 2 x1 x2
2(α−1) 2(β−1)
= αβx1 x2 (1 − (α + β)) > 0 since α + β < 1
which implies Hessian is Negative Definite ⇒ the Hessian of Π is Negative Definite ⇒ Π is strictly
concave, so the sufficient condition is satisfied.
3. By FOC:
∂Π
= 0 ⇒ pαx1α−1 xβ2 = w1 (1)
∂x1
∂Π β−1
= 0 ⇒ pβxα
1 x2 = w2 (2)
∂x2
1
β β 1−α−β
(1) α −1 w1 βw1 pα β w1
(2) ⇒ β x1 x2 = w2 ⇒ x2 = αw2 x1
(3), substituting (3) into (1), solve for x∗1 = w1 α w2 ,
h α α i 1−α−β 1
pβ w2
x∗2 can be obtained in the same way, x∗2 = w 2
α
β w1 .
max α ln x1 + (1 − α) ln x2 ,
x1 ,x2
1
s.t p1 x1 + p2 x2 = w. Since we have x2 = p2 (w − p1 x1 ), we can rewrite
1
maxα ln x1 + (1 − α) ln (w − p1 x1 ).
x1 p2
3
Solve it, we have
αw
x1 = ,
p1
(1 − α) w
x2 = .
p2
where
D = {(x, y, z) ∈ R3 | x2 + ay 2 ≤ a, x2 + az 2 ≤ a, a > 0}.
2. For which points (x, y, z) and parameter a does constraint qualification hold?
6. Without computations, what can you say about the solution of the optimization problem if:
Answer:
1. The objective function f (x, y, z) is continuous, and the constraint set D is closed and bounded. There-
fore by Weierstrass’ Theorem we can conclude that f (x, y, z) has a maximum on D.
The rank of the Jacobian, rank([5g1 ; 5g2 ]) can be lower than 2 if:
4
i. x = 0, the rank of the Jacobian is 0, but the constraints cannot be binding with a > 0.
Therefore the rank of the Jacobian is equal to the maximum number of binding constraints
and constraint qualification is satisfied.
ii. x 6= 0, the rank of the Jacobian is 1, but both constraints become identical (x2 ≤ a). With
only one constraint the maximum number of binding constraints is 1 and equal to the rank
of the Jacobian in this case too. Therefore the constraint qualification is satisfied.
∂L
1: ∂x = 2 − 2λ1 x − 2λ2 x = 0,
∂L
2: ∂y = 1 − 2λ1 ay = 0,
∂L
3: ∂z = a − 2λ2 az = 0,
λ1 ≥ 0, a(1 − y 2 ) − x2 ≥ 0, λ1 (a(1 − y 2 ) − x2 ) = 0,
λ2 ≥ 0, a(1 − z 2 ) − x2 ≥ 0, λ2 (a(1 − z 2 ) − x2 ) = 0.
4. from 1: x∗ = 1
λ1 +λ2
from 2: y ∗ = 1 2 2
2aλ1 ⇒ λ1 6= 0 ⇒ a(1 − y ) − x = 0
from 3: z ∗ = 1 2 2
2λ2 ⇒ λ2 6= 0 ⇒ a(1 − z ) − x = 0
1 4aλ22 − a a2
x2 = a(1 − z 2 ) = a(1 − ) ⇒ = 2 ,
4λ22 2
4λ2 λ2 (a + 1)2
4a2
⇒ 4aλ22 − a = ,
(a + 1)2
1 a
⇒ λ22 = + ,
4 (a + 1)2
s
∗ 1 a
⇒ λ2 = + ,
4 (a + 1)2
5
q
1 a
4 + (a+1)2
Then we find λ1 = a and:
1 a
x∗ = q ,
1
+ a a+1
4 (a+1)2
1
y∗ = z∗ = q .
1 a
2 4 + (a+1) 2
6. (a) The domain D is no more bounded. In this case y becomes free and there is no more optimum.
(b) The domain D is bounded and by Weierstrass we can find an optimum.
6
Mathematics for Economics and Finance (Fall 2022)
Problem Set 14: Statistics and Static Optimization
Professor: Roxana Mihet
TA: Samy
Due Dec 23
⇒ f has only one stationary point, x = (0, 0, 0). The Hessian matrix is
2 −1 2
H(x) = −1 2 1 ,
2 1 6
and the leading principal minors of the Hessian matrix are D1 = 2, D2 = 3, D3 = 4. H(x) is positive definite,
and thus f (x) is strictly convex. The point x = (0, 0, 0) is a local (and a global) minimum.
x2 + y 2 = 100.
L(x, y, λ) = ax + by + λ(100 − x2 − y 2 ).
1
Substitute from the first two into the third to get,
√
a2 + b2 a2 + b2
100 = =⇒ λ = .
4λ2 20
a 10a
x= =√ .
2λ a2 + b2
b 10b
y= =√ .
2λ a2 + b2
3 Complements or substitutes?
An agent consumes two goods, x1 and x2 , with prices p1 and p2 ,respectively. Her utility function is of the
form U (x1 , x2 ) = α(xα α
1 + x2 ), with α < 1. Verify that utility function is strictly concave. Derive the demand
function of the agent. In what direction does the demand for good 1 change if there is an increase in the
price of good 2.
max α(xα α
1 + x2 ) s.t. p1 x + p2 x2 = y.
x1, x2
L = α(xα α
1 + x2 ) + λ[y − p1 x + p2 x2 ].
FOC:
∂L
= α2 xα−1
1 − λp1 = 0
∂x1
α2
=⇒ λ = (1)
p1 x1−α
1
∂L
= α2 xα−1
2 − λp2 = 0
∂x2
α2
=⇒ λ = (2)
p2 x1−α
2
To verify that the utility function is strictly concave, check Hessian matrix for U (x1, x2 ) = α(xα α
1 + x2 ),
0 2 α−1 0 2 α−1 0 2 α−2 0 2 α−2 00 00
" U1 = α x1 , U2 = α x2 , U#11 = α (α − 1)x1 , U22 = α (α − 1)x2 , U12 = U21 = 0 ⇒ H =
α2 (α − 1)xα−2
1 0
.
0 α2 (α − 1)xα−2
2
Check the leading principal minors: D1 = α2 (α−1)x1α−2 < 0, because α < 1; D2 = α4 (α−1)2 x1α−2 xα−2
2 >
0. =⇒ H is negative definite. =⇒ U (x1, x2 ) is strictly concave.
2
To find the demand function, using (1) and (2),
α2 α2
1−α =
p 1 x1 p2 x1−α
2
x2 1−α p1
=⇒ ( ) =
x1 p2
x2 p1 1−α
1
=⇒ =( ) (3).
x1 p2
p 1 x + p 2 x2 = y
=⇒ x∗1 = x1 (p1 , p2 , y)
y
= 1
p1 + p2 ( pp12 ) 1−α
y
= α .
p1 [1 + ( pp12 ) 1−α ]
The demand for the other good is almost identical, but with the roles of p1 and p2 reversed.
p1 ( 1−α
α
−1)
∂x∗1 y
α
α−1 ( p2 ) · (− pp21 )
2
= α .
∂p2 p1 [1 + ( pp12 ) 1−α ]2
∂x∗
Because 1 − α > 0, the sign of ∂p21 is same as α.If α > 0, the goods are substitutes; If α < 0, the good are
complements, means as an increase in the price of either good reduces the demand for both.
where x1 indicates her expenditure on good 1 and x2 here expenditure on good 2. The prices for the
commodities are p1 and p2 respectively, both positive. And the consumer can not spend more than her
income I, also a positive value.
1. Assuming that consumption of either commodities must be non-negative, formulate the utilization
maximization problem of this consumer.
4. Suppose p1 = 2, p2 = 1, and I = 6. Using Kuhn-Tucker’s Theorem, or otherwise, find the pair (x1 , x2 )
that maximizes the consumer’s utility.
Answer:
3
1. The optimization problem becomes
2. The objective function u(x1 , x2 ) is continuous. The constraint set D is closed , bounded because for
every (x1 , x2 ) ∈ D, we have 0 ≤ x1 ≤ PI1 , 0 ≤ x2 ≤ PI2 ⇒ By Weierstrass’ Theorem, there must be a
maximum of u on D.
4
5 Dynamic programming: Minimizing quadratic costs
The agent solves
∞
X
β t x2t + vt2
min
{vt }
t=0
Answer: We approach this problem by using the Bellman equation and the guess and verify method.
The Bellman equation is given by
V (xt ) = Ax2t .
Replacing the guess in the Bellman, and using the transition equation, we get
n o
2
Ax2t = min x2t + vt2 + βA (2xt + vt ) .
{vt }
This result is used to find the constant A. Putting it back in the Bellman, we get,
2 2
2βAxt 2βAxt
Ax2t = x2t+ − + βA 2xt − ,
1 + βA 1 + βA
2 ! 2
2βA 2βA
Ax2t = 1+ x2t + βA 2 − x2t .
1 + βA 1 + βA
Note that we can drop the term x2t , since the Bellman is valid for all values of xt and we obtain a quadratic
equation for A, which gives the value of the constant given β,
2 ! 2
2βA 2βA
A= 1+ + βA 2 − .
1 + βA 1 + βA
5
6 Dynamic programming question on Final Exam
Consider an economy where the utility of the representative agent is given by
∞
X
β t u(Ct , Lt ), β ∈ (0, 1),
t=0
where Ct is the consumption and Lt the leisure in period t. Assume that the representative agent has the
following instantaneous utility function
L1−γ
t
u(Ct , Lt ) = ln(Ct ) + θ .
1−γ
Denote output by Yt , capital by Kt , and the number of hours worked by Ht . The production function is
given by:
Yt = AKtα Ht1−α , (1)
where A is a constant productivity parameter. The total available time in each period is normalized to 1.
So, Ht + Lt = 1. The law of motion of capital is
where δ ∈ (0, 1) is the depreciation rate and It denotes investment in period t. K0 is given. The resource
constraint in period t is given by Yt = Ct + It . We want to find the optimal paths of the choice variables for
t = 0, 1, ..., ∞.
2. List all choice variables and the required state variables in this problem.
4. Derive the necessary first order conditions and the intertemporal Euler equation expressing the marginal
rate of substitution in terms of Yt+1 and Kt+1 .
Dynamic optimization:
1. Richard Bellman (1957) stated his Principle of Optimality as follows: “An optimal policy has the
property that whatever the initial state and initial decision are, the remaining decisions must constitute
an optimal policy with regard to the state resulting from the first decision.” Bellman (1957, p83).
(1 − Ht )1−γ
V (Kt ) = max ln[AKtα Ht1−α − Kt+1 + (1 − δ)Kt ] + θ + βV (Kt+1 )
(Ht ,Kt+1 ) 1−γ
6
4. FOC with respect to Ht :
(1 − α)AKtα Ht−α
= θ(1 − Ht )−γ
Ct
⇔
Yt
(1 − α) Ht
= θ(1 − Ht )−γ
Ct
⇔
Yt θCt
(1 − α) = .
Ht (1 − Ht )γ
FOC with respect to Kt+1 :
1
= βV 0 (Kt+1 ).
Ct
Rewrite the Bellman equation for the next time step t + 1 and take the derivative with respect to Kt+1 :
Y Y
α Kt+1 +(1−δ) α Kt+1 +(1−δ)
V 0 (Kt+1 ) = t+1
Ct+1 and so we obtain 1
Ct =β t+1
Ct+1 . After rewriting obtain the following
Euler equation:
Yt+1
Ct+1 = α + (1 − δ) βCt . (3)
Kt+1
7
Mathematics for Economics and Finance (Fall 2022)
Problem Set 1 (Optional): Sets, functions, logic, proofs
Professor: Roxana Mihet
TA: Eliott Gabrielle
You do not need to hand in any solutions!
Solutions: Sept 23
1 Exercise: Sets
The consumption set of a consumer is:
Answer: y 0 is a minimal consumption threshold. Imagine the goods y is food. It is reasonable to assume
you have no real benefits to consume less than the minimum you need to survive. Consuming less is not an
admissible solution of an eventual optimization problem.
Comment: Note that in the notation y 0 the prime stands to distinguish the point y 0 from y. It has no
relation to a derivative.
2 Exercise: Functions
Let L(X, Y ) be the vector space of all linear mappings from vector space X to vector space Y . Let
T ∈ L(Y, Z),S ∈ L(X, Y ). Show that the composition of the two linear functions, T (S(.)), is linear.
1
Let x0 , x00 ∈ X, and a, b ∈ R,
3 Exercise: Functions
Prove there exists a bijection between the natural numbers and the integers. (Hint: define a function f
separately on the odd and even positive integers.)
Remember that according to our definition, a function is a bijection iff it is both one-to-one and onto.
We will prove that this function f : N → Z is a bijection, by first showing that it is one-to-one and then
showing that it is onto.
One-to-one: Suppose towards a contradiction that f (x) = f (y). Then they both must have the same sign.
Therefore either f (x) = x2 and f (y) = y2 . So f (x) = f (y) =⇒ x = y =⇒ x = y. Contradiction. The
second case is very similar, f (x) = (x+1)
2 and f (y) = y+1)
2 . So f (x) = f (y) =⇒ x + 1) = (y + 1) =⇒ x = y.
Contradiction, and thus f is one-to-one.
Onto: If y is positive, then f (2y) = y. Therefore, y has a pre-image. If y is negative, then f (a(2y + 1)) = y.
Therefore, y has a pre-image. Thus, f is onto.
Since f is a bijective function, this tells us that N and Z have the same size! Another way to describe this
mapping is: positive integers ⇐⇒ even natural numbers, negative integers ⇐⇒ odd natural numbers.
4 Exercise: Logic
Negate the statements:
4. There is a country in which every citizen has a friend who knows all the constellations.
5. In every village, there is a person who knows everybody else in that village.
Answer:
2
3. ∃(y, z) ∈ R2 , y < z, ∀x ∈∈ R such that x ≤ y or x ≥ z.
4. In every country there is a citizen who has no friend who knows all the constellations.
5. There exists a village where everybody does not know everyone else in the same village. Or in other
words: In at least one village, each person does not know everyone else. Or: In at least one village,
nobody knows everyone else.
Answer: We have A = there are 400 new bachelor students, B = at least two of them have a birthday on
the same day. Suppose the opposite is true: ¬B = there are no two people with the same day of birth. Then
we show that ¬A follows, that is ¬A = there cannot be 400 first-year students enrolled. Thus ¬B =⇒ ¬A.
Since there are at most 366 days in a year, thus the largest number of students that have distinct birthdays
is 366 — one for each day of the calendar. Thus 400 students are too many for the available 366 days, and
some of the students will share their days of birth.
6 Exercise: Induction
n(n+1)(n+2)
Prove that 1×2+2×3+...+n×(n+1) = 3 for all natural n > 1. (Hint: use mathematical induction).
2(2 + 1)(2 + 2)
1×2+2×3=8=
3
That allows us to state that the equation holds for n = k
k(k + 1)(k + 2)
1 × 2 + 2 × 3 + ... + k(k + 1) =
3
The inductive step:
We verify that the equation holds for n = k + 1 (given that it holds for n = k):
Thus we conclude that 1 × 2 + 2 × 3 + ... + n(n + 1) = n(n + 1)(n + 2) holds for all natural n > 1.
3
bond? (Hint: Compute S = 1 + k + k 2 + k 3 + .. + k n first.)
1−kn+1 1
Answer: The sum S = 1 + k + k 2 + k 3 + .. + k n = 1−k . Here, k = 1+r .
The price of the bond is the present discounted value of cash flows we get from it.
4
Mathematics for Economics and Finance (Fall 2022)
Problem Set 2: Linear algebra I
Professor: Roxana Mihet
TA: Eliott Gabrielle
You do not need to hand-in your answers
Due September 30th
1 − xn+1
1 + x + x2 + ... + xn = (1)
1−x
S = 1 + x + x2 + x3 + ... + xn
xS = x + x2 + x3 + ... + xn + xn+1
S − xS = 1 − xn+1
1 − xn+1
S=
1−x
Or by mathematical Induction:
1 − x2
1.) n = 1 implies 1 + x =
1−x
1 − xn+1
2.) Assume that it is true for n = k : 1 + x + ... + xn =
1−x
1 − xn+1 1 − xn+1 + xn+1 − xn+2 1 − xn+2
3. if n = k + 1 : 1 + x + ... + xn + xn+1 = + xn+1 = =
1−x 1−x 1−x
1
3 Exercise: Rank, determinant, trace, inverse of a matrix
1. Compute
3 1 −1 2
−5 1 3 −4
D= .
2 0 1 −1
1 −5 3 −3
5 2 8
Answer:
1. 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
4+3 4+4
+(−1) (0) −5 1 −4 + (−1) (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
−4 18 −10
2
4 Gaussian Elimination Method
For the matrix below
3 2 4
A= 2 0 2 .
4 2 3
Find its inverse using Gaussian elimination and answer if the matrix is 1) symmetric, 2) idempotent, 3)
orthogonal.
Answer:
Answer:
1. False. If the rows of matrix A are linearly independent, then m can not exceed n.
2. 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.
3. False. We can have at most n linearly independent rows, so if m > n, then the rows of matrix A are
linearly dependent.
3
4. True. We can have at most n linearly independent rows, so if m > n, then the rows of matrix A are
linearly dependent.
4
Mathematics for Economics and Finance (Fall 2022)
Problem Set 3: Linear algebra: systems of equations
Professor: Roxana Mihet
TA: Eliott Gabrielle
Due Oct 7nd
2. If the market is complete what can you say about the rank of matrix X?
919
complete? Are there any redundant assets?
135
113
5. Consider an economy with 3 states and 4 assets. The payoff matrix is: X = . Is this
3 9 15
5 5 15
market complete? Are there any redundant assets?
Answer:
1. If market is complete =⇒ m ≥ n , we have to have at least as many assets as there are states in the
economy, otherwise ( m < n ). We are not able to replicate any desired combination of payoffs.
2. If market is complete =⇒ rank(X) = n. Even if m ≥ n , It could be that rank(X) < n. In this case
we are not able to replicate all combinations =⇒ rank(X) = n is a necessary and sufficient condition.
3. There are no redundant assets if m = n . If m > n & rank(X) = n =⇒ some of the assets are linear
combination of other assets =⇒ redundant.
3 0 1
4. X = 5 1 4 n = 3, m = 3
9 1 9
1
Let us find rank(X).
3 0 1 3 0 1 3 0 1
5 1 4 −→ −7 1 0 −→ −7 1 0 −→
9 1 9 −18 1 0 12 0 0
3 0 1 0 0 1
−7 1 0 −→ 0 1 0 .
1 0 0 1 0 0
5.
1 3 5
1 1 3
X= n = 3, m = 4
3 9 15
5 5 15
2 Determinant, inverse
Calculate det(A), A−1 for
1 4 7
A= 3 2 5 .
5 2 8
Answer:
Finding determinant.
1 4 7
2 5 4 7 4 7
det(A) = 3 2 5 = (−1)1+1 + 3(−1)2+1 + 5(−1)3+1 = 2 ∗ 8 − 2 ∗ 5 − 3(4 ∗
2 8 2 8 2 5
5 2 8
8 − 2 ∗ 7) + 5(4 ∗ 5 − 2 ∗ 7) = 6 − 54 + 30 = −18
2
1 4 7 1 4 7
(−3) row 1 + row 2, (−5) row 1 + row 3
det(A) = 3 2 5 = 0 −10 −16 =
5 2 8 0 −18 −27
1 4 7 1 0 1
1
= (−2) ∗ (−9) 0 5 8 = 18 ∗ 2 0 0 2
=
0 2 3 (−2) row 3 + row 1, (− 25 ) row 3 + row 1 0 1 3
2
1 0 1 1 0 0 1 0 0
= 18 0 0 1 = 18 0 0 1 = −18 0 1 0 = −18
(− 32 ) row 2 + row 3, − row 2 + row 1
0 1 23 0 1 0 0 0 1
Inverted matrix.
Gauss Elimination Method:
1 4 7 1 0 0
(−3) row 1 + row 2, (−5) row 1 + row 3
3 2 5 0 1 0
5 2 8 0 0 1
1 4 7 1 0 0
→ 0 −10 −16 −3 1 0
∗ (−0.1)
0 −18 −27 −5 0 1
1 4 7 1 0 0
→ 0 1 1.6 0.3 −0.1 0
(−4) row 2 + row 1, 18 row 2 + row 3
0 −18 −27 −5 0 1
1 0 0.6 −0.2 0.4 0
→ 0 1 1.6 0.3 −0.1 0
1
0 0 1.8 0.4 −1.8 1
1.8
1 0 0.6 −0.2 0.4 0
→ 0 1 1.6 0.3 −0.1 0
4 10
0 0 1 18 −1 18 (−0.6) row 3 + row 1, (−1.6) row 3 + row 2
6 6
1 0 0 − 18 1 − 18
1
→ 0 1 0 − 18 1.5 − 16 ⇒
18
4 10
0 0 1 18 −1 18
6 6
− 18 1 − 18 6 −18 6
A−1 = − 18 1
1.5 − 16 = − 18 1
1 −27 16
18
4
18 −1 10 18 −4 18 −10
Method of Adjoint Matrix:
For the computation of the inverse matrix A−1 we use the following algorithm:
1
A−1 = adj(A) = (dij )
det(A)
where
1
dij = (−1)i+j det(Mji )
det(A)
1 2 5 1 6
d11 = − 18 (−1)1+1 = − 18 (16 − 10) = − 18
2 8
1 3 5 1 1
d21 = − 18 (−1)2+1 = 18 (24 − 25) = − 18
5 8
3
1 3 2 1 4
d31 = − 18 (−1)3+1 = − 18 (6 − 10) = 18
5 2
1 4 7 1 18
d12 = − 18 (−1)1+2 = 18 (32 − 14) = 18
2 8
1 1 7 1 27
d22 = − 18 (−1)2+2 = − 18 (8 − 35) = 18
5 8
1 1 4 1
d32 = − 18 (−1)2+3 = 18 (2 − 20) = − 18
18
5 2
1 4 7 1 6
d13 = − 18 (−1)1+3 = − 18 (20 − 14) = − 18
2 5
1 1 7 1
d23 = − 18 (−1)2+3 = 18 (5 − 21) = − 16
18
3 5
1 1 4 1
d33 = − 18 (−1)3+3 = − 18 (2 − 12) = 10
18
3 2
6 18 6
d11 d12 d13 − 18 18 − 18 6 −18 6
−1 1 27 16 1
A = d21 d22 d23 = − 18 − 18 = − 18 1 −27 16
18
4 18 10
d31 d32 d33 18 − 18 18 −4 18 −10
1
M 0 = In − 1 × 10
n
where 1 is an n × 1 vector of ones. It can be used to transform data to deviations from their mean.
a) Show that M 0 is symmetric idempotent.
Hint: it must satisfy (M 0 )0 M 0 = M 0 .
n
b) To obtain the sum squared deviations about the mean ( (xi − x̄)2 ), compute x0 (M 0 )0 M 0 x.
P
i=1
Answer:
4
Clearly, M 0 is symmetric (M 0 )0 = M 0 (all entries off the diagonal are equal).
1
1− n − n1 ··· − n1 1− 1
n − n1 ··· − n1
−1 .. ..
0 0 0 n 1 − n1 − n1 . −1
n 1− 1
n − n1 .
(M ) M =
.. ..
.. ..
. − n1 . − n1 . − n1 . − n1
− n1 − n1 · · · 1 − n1 − n1 − n1 ··· 1 − n1
(1 − n1 )2 + (− n1 )2 (n − 1) 1 2
+ (− n1 )2 (n
−(1 − n)n − 1)
··· − n1
= 1 − n1 = − n1
..
= − n1 1− 1
− n1 .
n
.. ..
− n1 . − n1
.
− n1 − n1 ··· 1 − n1
= M 0.
Therefore, M 0 is idempotent, M 0 M 0 = M 0 .
2.
1 0
M 0 x=(In − 11 )x
n
n
1
P
n xi
x1
i=1
n
1
P
x2 n x i
= .. − i=1
..
.
.
xn n
1
P
n xi
i=1
= x − x̄1.
0
x0 (M 0 )0 M 0 x = (x − x̄1) (x − x̄1)
x1 − x̄
x2 − x̄
= x1 − x̄ x2 − x̄ · · · xn − x̄
..
.
xn − x̄
n
X
= (xi − x̄)2 .
i=1
5
3. If we interchange the last two rows of matrix A, the determinant of A is not affected.
Answer:
1. False. It might be the case that n is smaller than m but neither columns nor rows are linearly inde-
pendent and the matrix does not have full rank. Take A = 13×2 with Rk(A) = 1 as a counterexample.
2. False. If m = n, A is indeed a square matrix, but it does not mean it has full rank. Take A = 12×2 as
a counterexample.
3. False. The determinant is multiplied by −1. Notice however that the rank is not affected.
2 −2 3
where each column is a different asset, and each row is a different scenario. Your goal is to determine
the portfolio allocation in each asset class, denoted θ ∈ R3 , given the client’s target payoff b ∈ R3 in each
scenario. In other words, you want to determine θ such that Aθ = b. To make sure that this is feasible, you
determine the following:
1. Calculate |A|.
3. Calculate A−1 .
Answer:
1. |A| = −10.
3.
1 1
2 2 − 12
A−1 = 1
− 52 1
5 5
− 51 − 35 4
5
6
4. Aθ = b ⇒ θ = A−1 b ⇒
1 1 1
θ1 = b1 + b2 − b3
2 2 2
1 2 1
θ2 = b1 − b2 + b3
5 5 5
1 3 4
θ 3 = − b1 − b2 + b3
5 5 5
1. For which values of the constants p and q does the system have a unique solution, several solutions or
no solution?
For exercises (b) and (c) assume that the rank of matrix A is equal to the size of vector x. In other words,
if x is n × 1, the rank of A has to be n. (This condition implies that (A0 A)−1 exists).
1. Show that if the solution of the system exists and is unique, then B is in the sub-space generated by
the column vectors of A (Hint: B can be rewritten as a linear combination of the column vectors of
A).
Ax = B − ,
where is constructed in such a way that B is in the sub-space generated by the column vectors of A.
To be efficient, we want to minimize the norm of . What should we do and what is the final solution?
(You can provide a graphical intuition in parallel with your mathematical results).
Answer:
1 1 1
det(A) = 2 −3 2 = −5p + 15 6= 0
3 −2 p
If p 6= 3, then the matrix is invertible and the system will admit one solution, whatever the value of q.
7
If p = 3, then we can simplify the system as follows:
1 1 1 2q 1 1 1 2q
2 −3 L
2 2q 3 − L − L =⇒ 2 −3 2 2q
2 1
3 −2 3 2q 0 0 0 −5q
If q 6= 0 and p = 3, then the system admits no solutions. If q = 0 and p = 3, then the system admits
an infinity of solutions.
The 2nd way
1 1 1 2q
Let A = 2 −3 2 and b = 4q , then we rewrite as Ax = b.
3 −2 p q
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.
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
3 −2 3 q 3 −2 3 1
rank(Ã) = 3, so no solution.
2. We know that Ax = B holds. We can rewrite A = {a1 , a2 , a3 }, where ai is the column vector i of
P3
matrix A ∀i = 1, 2, 3. So B = x1 a1 + x2 a2 + x3 a3 = i=1 xi ai
3. We want to minimize the norm of , in other words, we want to minimize the distance between vector
B, and the sub-space generated by A. Thus, we will have to project orthogonally in the sub-space of
A ⇔ A⊥
A0 = A0 (B − Ax) = A0 B − A0 Ax = 0
This gives us the set of normal equations A0 B = A0 Ax. If A’A is full rank, then we can premultiply
by (A0 A)−1 and obtain the standard OLS estimator:
x = (A0 A)−1 A0 B.
But: As we have seen in (a), the system of equations A admits for no solutions if p = 3. In this case,
det(A) = 0 and the matrix A is not invertible. This implies, that also (A0 A) is not invertible.
8
7 Cramer’s Rule
Consider the following system of equations:
5P1 + P2 − P3 = 9
−2P1 + 5P2 − P3 = 3
−2P1 − 2P2 + 14P3 = 34
9
Mathematics for Economics and Finance (Fall 2022)
Problem Set 4: Linear algebra: systems of equations
Professor: Roxana Mihet
TAs: Elliot
Due Oct 14th
x1 + x2 + x3 = 2q
2x1 − 3x2 + 2x3 = 4q
3x1 − 2x2 + px3 = q.
Answer:
Let
1 1 1 2q
A = 2 −3 2 and b = 4q , then we rewrite as Ax = b.
3 −2 p q
3 −2 3 q 3 −2 3 1
rank(Ã) = 3, so no solution.
x3
1
10, 000
d = 20, 000 be “a demand vector”, and
10, 000
0 0.5 0.5
C = 0.4 0.3 0.05 be an input matrix.
0.2 0 0.35
We assume that all produced goods not used in the subsequent production process are consumed. Thus,
X, C and d are linked by the following equation X − CX = d.
X = (I − C)−1 d
1.82 1.3 1.5 10, 000
= 1.08 2.2 1 20, 000
33, 600
So the production schedule should be $59, 200 labor, $64, 800 transportation and $33, 600 food.
1 2 6
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.
2
The arbitrary minors are:
Answer:
1. 1. In matrix notation
1 −1 −1
QID = x0 Ax, where A = −1 1 0 .
−1 0 1
Leading principle minors are:
D1 = 1 > 0.
1 −1
D2 = = 1 ∗ 1 − (−1) ∗ (−1) = 0.
−1 1
1 −1 −1
1 0 −1 −1 −1 −1
D3 = −1 1 0 = (−1)1+1 ∗ 1 ∗ + (−1)2+1 (−1) + (−1)3+1 ∗ (−1)
0 1 0 1 1 0
−1 0 1
= 1 − 1 − 1 = −1 < 0.
D1 > 0, D2 = 0, D3 < 0 ⇒ not positive definite, not negative definite.
3
The arbitrary minors are:
Principal minors:
D1 = 2 > 0.
2 −1
D2 = = 2 − (−1)(−1) = 1 > 0.
−1 1
2 −1 −1
−1 1 2 −1
D3 = −1 1 0 = (−1)1+3 · (−1) · + (−1)3+3 · 1 · = −1 + 1 = 0.
−1 0 −1 1
−1 0 1
1 0
m=2: fix a11 , a11 =⇒ ∆2 = =1 >0
0 1
2 −1
fix a22 , a22 =⇒ ∆2 = =1 >0
−1 1
2 −1
fix a33 , a33 =⇒ ∆2 = =1 > 0
−1 1
2 −1 −1
m=3: fix a11 , a22 , a33 =⇒ ∆3 = −1 1 0 =0≥0
−1 0 1
4
=⇒ True for matrix A
m=2: ∆2 ≥ 0
m=3: ∆3 ≥ 0
2 −1 −1
2. The corresponding matrix is A = −1 1 0 . It is symmetric since aij = aji . Since det A =
−1 0 1
0 =⇒ A−1 doesn’t exist, so A is not invertible. It is not orthogonal because determinant of an
orthogonal matrix must be equal ±1 (see Exercise 1.4.2 from Exercise Session 2).
3.
2−λ −1 −1
−1 1 − λ 2−λ −1
|A − λI| = −1 1−λ 0 = (−1)1+3 (−1) + (−1)3+3 (1 − λ) =0
−1 0 −1 1−λ
−1 0 1−λ
=⇒ −(1 − λ) + (1 − λ)((2 − λ)(1 − λ) − 1) = (1 − λ)((2 − λ)(1 − λ) − 2) = (1 − λ)(λ2 − 3λ) =
= λ(1 − λ)(λ − 3) = 0.
λ1 = 0
eigenvalues λ2 = 1 .
λ3 = 3
4. λ1 = 0 :
2 −1 −1 1 0 −1 !
row1 + row2 −1 1 0
(A − λI) = −1 1 0 −→ −1 1 0 −→
1 0 −1
−1 0 1 1 0 −1
! x1 0 x1 = x1
−1 1 0 −x + x = 0
1 2
−→ x2 = 0 =⇒ x2 = x1 .
1 0 −1 x1 − x3 = 0
x3 0 x3 = x1
x1 c
1
=⇒ x = x2 = c ∀c ∈ R =⇒ λ1 = 0, a1 = (1, 1, 1)0 .
x3 c
5
λ2 = 1 :
1 −1 −1 !
row1 + row2 0 −1 −1
(A − λI) = −1 0 0 −→
−1 0 0
−1 0 0
! x1 0 x1 = 0
0 −1 −1 −x − x = 0
2 3
−→ x = 0 =⇒ x2 = x2 .
2
−1 0 0 −x 1 = 0
x3 0 x3 = −x2
x1 0
=⇒ x2 = x2 = c ∀c ∈ R =⇒ λ2 = 1, a2 = (0, 1, −1)0 .
x3 −c
λ3 = 3 :
row1 − row2
−1 −1 −1 0 1 −1 !
0 1 −1
(A − λI) = −1 −2 0 (−1)row2 + row3 −→ 0 2 −2 −→
1 0 2
−1 0 −2 (−1)row3 1 0 2
! x1 0 x1 = −2x3
0 1 −1 x2 − x3 = 0
−→ =
x2 0 =⇒ x2 = x3 .
1 0 2 1 + 2x3 = 0
x
x3 0 x3 = x3
x1 −2c
=⇒ x3 = x2 = c ∀c ∈ R =⇒ λ3 = 3, a3 = (−2, 1, 1)0 .
x3 c
5. From the lecture notes we know three different ways to decompose a matrix A, namely
1. Spectral decomposition (or eigendecomposition): A = CΛC 0 , where Λ is a diagonal matrix with all
eigenvalues and matrix C contains all corresponding eigenvectors.
2. LU decomposition: A = LU , where L is a lower triangular and U is an upper triangular matrix.
3. Cholesky decomposition: A = LL0 = U 0 U , with L and U as above.
1. Spectral decomposition: the corresponding eigenvector matrix C = (a1 , a2 , a3 ) is given by
1 0 −2
C = 1 1 1 .
1 −1 1
T
1 1 1
2 0 −2 3 3 3
1 1
C −1 = adj (C) = 2 3 1 = 0 1
− 21
det C 6 2
2 −3 1 − 13 1
6
1
6
6
At the end we result in the following eigendecomposition:
1 1 1
1 0 −2 0 0 0 3 3 3
1
A = 1 1 1 0 1 0 0 − 12
2
1 −1 1 0 0 3 − 13 1
6
1
6
Remark. One may also normalize the eigenvectors and obtain an orthogonal matrix C̃ and thus the
0
0
following matrix decomposition: A = C̃ΛC̃ . The normalized eigenvectors are a1 = √13 , √13 , √13 ,
0 √ 0
a2 = 0, √12 , − √12 , a3 = − √23 , √16 , √16 . And thus the resulting decomposition is
√ 1
√1 0 − √23 0 0 0 √ √1 √1
3 3 3 3
√1 − √12
√1 √1 √1
A= 0√
0 1 0 .
3 2 6 2
√1
3
− √12 √1
6
0 0 3 − √23 √1
6
√1
6
2. The LU decomposition corresponds to the Gaussian elimination algorithm. Our first step is to find
a reduced form of A such that we are left with an upper triangular matrix.
2 row1
1
2 −1 −1 1 0 0 1 − 21 − 12 1
2 0 0
2 row1 + row2
1 1
−1 1 0 0 1 0 ∼ 0 − 21 1
1 0
2 2
∼
2 row1 + row3 row2 + row3
1
−1 0 1 0 0 1 0 − 21 1
2
1
2 0 1
1 − 12 − 12 12 0 0
1
∼ 0 − 12 12 1 0 .
2
0 0 0 1 1 1
The first part is the matrix U and the second part is L−1 . Now, we go on by finding the inverse of
L−1 that is L.
1
2 0 0 1 0 0 2row1 1 0 0 2 0 0
row2 − 2row1
1
2 1 0 0 1 0 ∼ 0 1 0 −1 1 0 ,
The second matrix is the searched lower triangular matrix L. After all the computations we end
up with
2 0 0 1 − 21 − 21
1
A = LU = −1 1 0 0 − 21
2
−1 −1 1 0 0 0
7
Matrix LL0 is symmetric by definition. Now, we can just use the lower part of the matrix and find the
searched values, i.e.
2
2 l1,1
A = −1 2 2
1 = l1,1 l2,1 l2,2 + l2,1 .
2 2 2
−1 0 1 l1,1 l3,1 l2,1 l3,1 + l2,2 l3,2 l3,3 + l3,1 + l3,2
After solving the resulting system of 6 non-linear equations with 6 unknowns we can obtain 4 possible
matrices L.One of them is √
2 0 0
L = − √12 √1 0.
2
− √12 − √12 0
n
6. One may note that An = CΛC −1 = CΛC −1 ·CΛC −1 ···CΛC −1 = CΛC −1 CΛC −1 C ···C −1 CΛC −1 =
CΛn C −1 .
n
0 0 0 0 0 0
n
Λ = 0 1 0 = 0 1 0 .
0 0 3 0 0 3n
1 1 1
1 0 −2 0 0 0 3 3 3 2 · 3n−1 −3n−1 −3n−1
Thus An = 1 1 1 0 1 0 0 1
− 12 = −3n−1 1
· 3n−1 + 1 1
· 3n−1 − 21
2 2 2 2
1 −1 1 0 0 3n − 31 1
6
1
6 −3n−1 1
2 · 3n−1 − 1
2
1
2 · 3n−1 + 21
5. Basis, subspaces
Let u1 = (3, 3, 4)0 and uk = (−1, −1, 5)0 be a basis for the subspace W of Rn .
3 1
5. Compute the corresponding residual terms ev and ew .
Answer:
1.
3 −1
A = 3 −1
4 5
.
2.
1 1
2 2 0
P = A(A0 A)−1 A0 =
1 1
0
2 2
0 0 1
8
.
3. n = 3 and dim(W ) = 2.
4.
1 1 3
2 0
2 1 2
1 1 3
Pv = 2 0 2 = 2
2
0 0 1 3 3
1 1
2 2 0 1 1
1 1
P w = 2 2 0 1 = 1
0 0 1 1 1
5.
− 21
1
ev =
2
0
0
ew = 0
Answer:
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 .
9
2. 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
5 5 2
x1 = x2 , x1 = − x3 , x 2 = − 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.
10
From the first part, we have
5 5 2
x1 = x2 , x1 = − x3 , x 2 = − 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.
−1 −1 3 1 0 1
1. Verify that x1 , x2 and x3 are eigenvectors of A and find the corresponding eigenvalues.
2. Let
B = A · A.
Show that
B · x2 = x2
and
B · x3 = x3 .
Is it true that
B · x1 = x1 ?
1.
−2 − λ −1 4
|A − λI| = 2 1−λ −2
−1 −1 3−λ
1−λ −2 −1 4 −1 4
= −(2 + λ) −2 −1
−1 3−λ −1 3−λ 1−λ −2
= −(2 + λ)(3 − 4λ + λ2 − 2) − 2(−3 + λ + 4) − (2 − 4 + 4λ)
= −λ3 + 2λ2 + 7λ − 2 − 2λ − 2 − 4λ + 2 = −λ3 + 2λ2 + λ − 2
= −(λ − 1)(λ − 2)(λ + 1) = 0
=⇒ Eigenvalues λ = 1, λ = 2, λ = −1.
11
λ=1:
(A − λI)X = 0
−3 −14 0 2 −2
−3 ∗ row 3
2 −2
0 −→ 0 −2 2
2 ∗ row 3
−1 −12 −1 −1 2
! !
0 1 −1 0 1 −1
−→ −→
1 1 −2 −1 ∗ row 1 1 0 −1
! x 0 (
0 1 −1 y−z =0 y=z
y = 0 =⇒ .
1 0 −1 x−z =0 x=z
z 0
1
Take z = 1 =⇒ 1 is an eigenvector that corresponds to λ = 1.
1
λ=2:
(A − λI)X = 0
−4 −1 4 0 3 0
−4 ∗ row 3
2 −1 −2 −→ 0 −3 0
2 ∗ row 3
−1 −1 1 1 1 −1
! !
0 1 0 0 1 0
−→ −→
1 1 −1 −1 ∗ row 1 1 0 −1
! x 0 (
0 1 0 y=0 y=0
y = 0 =⇒
0 −1 x−z =0
1 x=z
z 0
1
Take z = 1 =⇒ 0 is an eigenvector that corresponds to λ = 2.
1
λ = −1 :
(A − λI)X = 0
−1 −1 4 ! ! !
1 1 −4 1 1 −4 1 1 −4
2 2 −2 −→ −→ −→ 4 ∗ row 1
1 1 −1 −1 ∗ row 1 0 0 3 0 0 1
−1 −1 4
! x 0 (
1 1 0 x+y =0 x = −y
y = 0 =⇒ .
0 0 1 z=0 z=0
z 0
1
Take y = 1 =⇒ −1 is an eigenvector that corresponds to λ = −1.
12
1
2. x2 = −1 corresponds to λ = −1 =⇒ Ax2 = λx2 =⇒ Ax2 = −x2 .
13
Mathematics for Economics and Finance (Fall 2022)
Problem Set 5: Linear algebra, Calculus
Professor: Roxana Mihet
TAs: Madhushree A
Due Oct 21st
1 Spectral theorem
Let Q be the following quadratic form:
x1 1 −1 −1
1. Q = xT Ax where x = x2 and A = −1 2 0
x3 −1 0 2
1−λ −1 −1
2. det(A−λIn ) = 0 ⇔ −1 2−λ 0 = 0 ⇔ (1−λ)(2−λ)2 −(4−2λ) = 0 ⇔ λ3 −5λ2 +6λ = 0
−1 0 2−λ
Thus, we have 3 eigenvalues: λ1 = 0, λ2 = 2, λ3 = 3. Two are positive, and one is equal to zero. This
implies that the quadratic form Q is positive semidefinite.
3. for λ1 = 0:
1 −1 −1 1 −1 −1
(A−λIn ) = −1 2 0 row 2 and row 3 + row 1 and finally row 2+row 3 −→ 0 1 −1
−1 0 2 0 −1 1
1 −1 −1 x1 0 (
x1 = 2x2
−→ 0 1 −1 2 0 ⇒
x =
x2 = x3
0 0 0 x3 0
2c
⇒ x1 = c ∀c ∈ R =⇒ λ1 = 0
c
⇒ To get the orthonormal eigenvectors, we need to rescale the eigenvector by his norm such that the
1
√ √
new norm is 1. Putting c constant and equal to 1, The actual norm is 22 + 12 + 12 = 6, so:
2
√
6
x̂1 = c √1 =⇒ λ1 = 0
6
√1
6
for λ1 = 2:
−1 −1 −1 −1 −1 −1 x1 0 (
x1 = 0
(A − λIn ) = −1 0 0 −→ −1 0 0 x2 = 0 ⇒
x2 = −x3
−1 0 0 0 0 0 x3 0
0
⇒ x2 = −c ∀c ∈ R =⇒ λ2 = 2
c
0
−1
x̂2 = c √ 2
=⇒ λ2 = 2
√1
2
for λ1 = 3:
−2 −1 −1 −1 0 −1 −1 0 −1 x1
(A−λIn ) = −1 −1 0 row 1 minus row 2 −→ −1 −1 0 −→ −1 −1 0 x2 =
−1 0 −1 −1 0 −1 0 0 0 x3
0 (
x1 = −x2
0 ⇒
x2 = x3
0
−c
⇒ x3 = c ∀c ∈ R =⇒ λ3 = 3
c
−1
√
3
3
x̂ = c √1 =⇒ λ3 = 3
3
√1
3
• Λ = diag(λ1 , λ2 , λ3 )
Because Ĉ is an orthogonal matrix (by definition), Ĉ 0 Ĉ = In . From this statement, one may note that
0 0 3 0 0 3n
2
5. Because one of the eigenvalues is equal to 0, we can’t find an inverse.
But assuming we can, from definition 2.22 of the lecture notes, we know that the inverse matrix of A is
just ĈΛ−1 Ĉ 0 . By using the spectral decomposition in this specific case (when the matrix is symmetric),
we don’t need to do any inverse computation:
• Ĉ doesn’t change.
0
∞ 0 0
• Λ−1 = 0 1 0 .
1
0 0 3
• We know everything to compute the inverse matrix. But in our case, as told before, there is no
inverse matrix!
2 Matrix Decomposition
Let A be a symmetric matrix with
2 1 −2
A= 1 2 −2
−2 −2 5
2
1. Show that the characteristic polynomial may be written in the form p (λ) = (λ − 1) (λ − 7).
2. Show that the vector equation Ax = x reduces to a single scalar equation. Find orthonormal eigenvec-
tors x, y corresponding to the eigenvalue λ = 1.
2−λ 1 −2
det (A − λI) = 1 2−λ −2
−2 −2 5−λ
2−λ −2 1 −2 1 −2
= (2 − λ) −1 −2
−2 5−λ −2 5−λ 2−λ −2
= (2 − λ) [(2 − λ) (5 − λ) − 4] − (5 − λ − 4) − 2 (−2 + 4 − 2λ)
= (2 − λ) 6 − 7λ + λ2 − 1 + λ + 4λ − 4
= (λ − 1)2 (λ − 7).
3
2. Ax = x ⇒ (A − I) x = 0, so each vector x that solves Ax = x is at the same time an eigenvector of A
that corresponds to λ = 1.
Let’s find the eigenvector corresponding to λ = 1:
x1 0 2−1 1 −2 x1 0
(A − λI) x2 = 0 ⇔ 1 2 − 1 −2 x2 = 0
x3 0 −2 −2 5 − 1 x3 0
0 0 0 x3 0
x1 + x2
x1 + x2 − 2x3 = 0 ⇒ x3 = .
2
y1 + y2
y1 + y2 − 2y3 = 0 ⇒ y3 = .
2
We are looking for orthogonal vectors:
x̂0 y = 0.
Thus
y1
1 y1 + y2 3
√1 √1 √1 y2 = √ y1 + y2 + = √ (y1 + y2 ) = 0
3 3 3 2
y1 +y2 3 3
2
y1 + y2 y1 − y1
⇒ y1 = −y2 , y3 = = = 0.
2 2
Hence, vector y of the form:
y1
y = −y1
4
is an eigenvector corresponding to the eigenvalue 1 and y is orthogonal to x. Take y1 = 1.
Normalizing vector y results in:
1
1 √
2
y 1 √1
ŷ = =√ −1 = − 2 .
kyk 1+1
0 0
These two vectors define a hyperplane that is not changed under linear transformation Ax.
z3 0 −2 −2 5 − 7 z3 0
After a few elementary transformations we obtain the following system in reduced form:
! z1 0
1 1 1 z1 + z2 + z3 = 0
z2 0 ⇒
1
0 1 2 z2 + 21 z3 = 0
z3 0
z3
0
Take, e.g. z3 = 1, then the solution vector is given by z = (−0.5, −0.5, 1) or after normalising
0
ẑ = − √16 , − √16 , √26 .
4. Using the spectral theorem for symmetric matrices (matrix A is symmetric) we can write D = S −1 AS,
where D is a diagonal matrix with principal diagonal composed of eigenvalues and matrix S is composed
of the corresponding eigenvectors. In our case:
√1 √1 √1
1
√1 √1
1 0 0 √
3 2 6 3 3 3
0 , S = S 0 = √12
√1 −1
S= − √12 √1 , D = 0 1 − √12 0 .
3 6
√1 0 − √26 0 0 7 √1 √1 − √26
3 6 6
5
Check that
D = S −1 AS.
1
√1 √1
1
√1 √1
√ 2 1 −2 √
3 3 3 3 2 6
√1
− √12
√1
= 0 1 2 −2 3 − √12 √1
2 6
√1 √1 − √26 −2 −2 5 √1 0 − √26
6 6 3
√1 √1 1 1 √1 1
√ √ √
3 3 3 3 2 6
−1
= √12 √ 0 √13 − √12 √1
2 6
√7 √7 14
−√ √1 0 − √26
6 6 6 3
1 0 0
= 0 1 0 .
0 0 7
A−1 = SD−1 S −1
1 1 √1
1
√1 √1
√ √ 1 √
3 2 6 1 0 0 3 3 3
√1
− √12 √1 0
√1
= 1
0 2 − √12 0
3 6 1
√1 1
3
0 − √26 0 0 7
√1
6
√1
6
− √26
√1 √1 1 √1 √1 √1
√
3 2 7 6 3 3 3
√1
= − √12 1 √1
√ − √12 0
3 7 6 2
√1 0 2
− 7√ √1 √1 − √26
3 6 6 6
6
7 − 17 2
7
1 6 2
= − 7 7 7
.
2 2 3
7 7 7
4 Concavity
Is a function f (x) = 1 − x2 concave? Is it strictly concave? Do the same properties still hold for a function
f (x, y) = 1 − x2 ?
6
For strict concavity the inequality should be strict.
5 Function continuity
a) Let f, g be two continuous functions on [a, b] ⊂ R and let x0 ∈ (a, b) with f (x0 ) = g(x0 ). We then
define
f (x) if x ∈ [a, x ),
0
h(x) =
g(x) if x ∈ [x0 , b].
Answer:
a) Let us first observe that h(x) is continuous on [a, xo ) since f is continuous by definition and h is
continuous on (x0 , b] since g is continuous. The only point we have to verify is x0 . Recall the definition
of a continuous function in terms of topological limits: a function h(x) is continuous at point x0 ∈ (a, b)
if
1. h(x0 )exists
2. limx→x0 h(x)exists and
3. limx→x0 h(x) = h(x0 )
7
1. h(x0 ) = f (x0 ) = g(x0 ) = L
2. The limit exists if both sides limits exist.
– limx%x0 h(x) = limx%x0 f (x) = f (x0 ) = L
– limx.x0 h(x) = limx.x0 g(x) = g(x0 ) = L
3. limx→x0 h(x) = L = h(x0 )
b) Let us observe that h(x) is continuous on (−∞, −1) ∪ (−1, 2) ∪ (2, ∞) because f1 (x) = 1, f2 (x) = x2
and f3 (x) = x + 2 are all continuous on R. We therefore only have to verify the continuity of h(x) at
x0 = −1 and x1 = 2.
We have verified that h(x) is continuous at both points x0 and x1 . It is therefore continuous on R.
4
h(x)
−5 −4 −3 −2 −1 0 1 2 3 4 5
x
8
Mathematics for Economics and Finance (Fall 2022)
Problem Set 6: Calculus
Professor: Roxana Mihet
TA: Madhushree
Due Oct 28
1 Limits
Find the following limits
4n
1. lim ,
n→∞ n!
n2 +3n
2. lim 2 ,
n→∞ 2n +1
3. limx→0 x2 sin 1
.
x
q
4. lim 9x3 +5 .
x3 +7x
x→∞
Answer:
4n
1. We want to find: lim = lim 4∗4∗···∗4
n→∞ n! n→∞ 1∗2∗···∗n
n−4
4n 4 ∗ 4 ∗ ··· ∗ 4 4∗4∗4∗4 4
We can write = ≤ × →0
n! 1 ∗ 2 ∗ · · · ∗ n |1 ∗ 2{z
∗ 3 ∗ 4} 5
| {z }
constant
→0 because 4/5 < 1
n
4
So by the Sandwich theorem 0 ≤ lim ≤0
n→∞ n!
4n
lim =0
n→∞ n!
3
n2 +3n 1+ n
2. lim 2 = lim 1 = 12 .
n→∞ 2n +1 n→∞ 2+ n2
3. limx→0 x2 sin 1
x
1
1 ≤ sin( ) ≤ 1 ⇒ bounded
x
1
x2 ≥ 0 ∀x ⇒ −x2 ≤ x2 sin( ) ≤ x2 ⇒ −x2 ≤ f (x) ≤ x2
x
by Sandwich theorem 1
lim x2 = 0 and lim −x2 = 0 =⇒ lim x2 sin( ) = 0
x→0 x→0 x→0 x
q
4. lim x3 +7x
9x3 +5
x→∞ r
x3 (1+ x72 )
q q
x3 +7x 1 1
lim 9x3 +5 = lim x3 (9+ x53 )
= 9 = 3
x→∞ x→∞
2 Homothetic preferences
1
Suppose that in a two-commodity world, the consumer’s utility function takes the form u(x) = [α1 xρ1 + α2 xρ2 ] ρ , α1 +
α2 = 1 (CES).
1
1. Show that when ρ = 1, indifference curves become linear.
2. Show that as ρ → 0, this utility function comes to represent the same preferences as the (generalized)
Cobb-Douglas utility function u(x) = xα 1 α2
1 x2 .
3. Show that as ρ → −∞, indifference curves become “right angles”; that is, this utility function has in
the limit the indifference map of the Leontief utility function u(x1 , x2 ) = M in {x1 , x2 } .
Answer:
2. Since every monotonic transformations of a utility function represents the same preference, we shall
consider
1
ũ(x) = ln u(x) = ln(α1 xρ1 + α2 xρ2 ).
ρ
By L’Hopital’s Rule,
α1 xρ1 ln x1 + α2 xρ2 ln x2 α1 ln x1 + α2 ln x2
lim ũ(x) = lim = .
ρ→0 ρ→0 α1 xρ1 + α2 xρ2 α1 + α2
Hence
1 1
(α1 xρ1 + α2 xρ2 ) ρ ≥ (α1 + α2 ) ρ x1 .
Therefore
1 1 1
(α1 ) ρ x1 ≥ (α1 xρ1 + α2 xρ2 ) ρ ≥ (α1 + α2 ) ρ x1 .
1
Letting ρ → −∞, we obtain lim [α1 xρ1 + α2 xρ2 ] ρ = x1 by Sandwich Theorem.
ρ→−∞
3 Sandwich Theorem
Use the Sandwich theorem to prove that:
1
lim xcos =0 (1)
x→0 x
2
1
x cos ≤ |x| .
x
Hence
1
−|x| ≤ x cos ≤ |x| .
x
Since
lim −|x| = lim |x| = 0 ,
x→0 x→0
4 Sandwich theorem
Consider the function : (
1 + 2x2 if x is rational
f (x) =
1 + x4 if x is irrational
Answer: Since we are considering the limit when x gets closer to 0, then we may assume that |x| ≤ 1. In
this case, we have x4 ≤ x2 ≤ 2x2 . Hence for any x, we have
1 ≤ f (x) ≤ 1 + 2x2 .
lim f (x) = 1 .
x→0
5 Homogeneous functions
Is a linear production function homogeneous? Is it homothetic?
F (K; L) = αK + βL
where K denotes capital, L denotes labor, and α and β are strictly positive constants.
6 Homogeneous function
What is the degree of homogeneity of this function?
3
• f (x, y) = x2 + y 2
• f (x, y) = x4 + y 4 + 2
Answer:
• f (x, y) is HOM(2)
1. M is open.
2. The set of points of the set M which are isolated is not empty.
3. M is unbounded.
4. There exists a limit point of the set M which does not belong to M .
Answer:
1. False. It is not known if all the boundary points do not belong to the set.
2. False. Isolated point is a point of the set, for which there exists a neighborhood with no other points
of this set. We do not have enough information to claim if there are some isolated points in the set.
3. False. We do not have enough information to claim this. M can be either bounded or unbounded.
4. True. The boundary points that do not belong to the set are just these limit points.
8 Derivatives
Compute the derivatives:
1. y = sin 1+x
2x
2
√
2. y = ln(x + 1 + x2 )
q
(x−1)(x−2)
3. y = (x−3)(x−4)
Answer:
1.
2 · 1 + x2 − 2x · 2x 2(1 − x2 )
dy 2x 2x
= 2 cos = 2 cos
dx (1 + x2 ) 1 + x2 (1 + x2 ) 1 + x2
2.
− 1
dy 1 1 1
= √ · 1 + · 1 + x2 2 · 2x = √
dx x + 1 + x2 2 1 + x2
4
q
(x−1)(x−2)
3. the domain of the function y = (x−3)(x−4) is D = {(−∞, 1] ∪ [2, 3) ∪ (4, ∞)}
for x ∈ (4, ∞) we can take ln for both sides:
1
ln y = [ln(x − 1) + ln(x − 2) − ln(x − 3) − ln(x − 4)]
2
Take derivative for both sides:
y0
1 1 1 1 1
= + − −
y 2 x−1 x−2 x−3 x−4
s
0 1 (x − 1)(x − 2) 1 1 1 1
y = + − −
2 (x − 3)(x − 4) x − 1 x − 2 x − 3 x − 4
q
for x ∈ (−∞, 1] we can rewrite y = (1−x)(2−x)
(3−x)(4−x) , then do the same as above
q
for x ∈ [2, 3) we can rewrite y = (x−1)(x−2)
(3−x)(4−x) , then do the same as above.
9 Taylor expansion
Consider the function f (x) = ln(x2 + x − 1). Compute the exact value of f at x = 1.1 and compare it to
the third-order Taylor approximation in the neighborhood of x = 1. What is the order of the approximation
error?
Answer:
5
Mathematics for Economics and Finance (Fall 2022)
Problem Set 7: Calculus
Professor: Roxana Mihet
TA: Madhushree
Due: Nov 4
1 Derivatives
Find the following derivatives:
1+x−x2
1. y(x) = 1−x+x2
Answer:
1.
1 + x − x2 0 (1 + x − x2 )0 (1 − x + x2 ) − (1 + x − x2 )(1 − x + x2 )0
=
1 − x + x2 (1 − x + x2 )2
(1 − 2x)(1 − x + x ) − (1 + x − x2 )(−1 + 2x)
2
=
(1 − x + x2 )2
(1 − 2x)(1 − x + x2 + 1 − x + x2 )
=
(1 − x + x2 )2
2(1 − 2x)
=
(1 − x + x2 )2
2 Taylor approximation
Consider the function f (x) = ln(x2 + x − 1). Compute the exact value of f at x = 1.1 and compare it to
the third-order Taylor approximation in the neighborhood of x = 1. What is the order of the approximation
error?
Answer:
1
The error equals 0.001 ∼
0.27 = 0.37% of the solution. Generally, we define the order of the error as the order
of the derivative in the Lagrange Remainder of the Taylor series. In this example we use a third order
approximation, so the order of the error is four. However, a precise estimation of the error requires an
analysis of the remainder term.
3 Partial derivatives
Assume (
1
(x2 + y 2 ) sin x2 +y 2, x2 + y 2 =
6 0,
f (x, y) =
0, x2 + y 2 = 0,
1. Show that all first-order partials exist, and check whether they are continuous.
Answer: (
1
(x2 + y 2 ) sin x2 +y 2, x2 + y 2 =
6 0
f (x, y) = ,
0, x2 + y 2 = 0
f (x, 0) − f (0, 0) 1
fx0 (0, 0) = lim = lim x sin 2 = 0,
x−→0 x x−→0 x
f (0, y) − f (0, 0) 1
fy0 (0, 0) = lim = lim y sin 2 = 0.
y−→0 y y−→0 y
We need to show lim fx0 (x, y) and lim fy0 (x, y). However when (x, y) approach (0, 0) along
y−→0,x−→0 y−→0,x−→0
y axis⇒ fx0 (x, 0) = 2x sin x12 − 2x 1
x2 cos x2 . The limit does not exist. So x−→0,y−→0
lim fx0 (x, y) does not
exist.=⇒ although all first order partial exist they are not continuous at (0, 0).
2
4 First and second order partial derivatives
Let f : R2 → R be
xy(x2 −y 2 )
(
x2 +y 2 if (x, y) 6= (0, 0),
f (x, y) =
0 if (x, y) = (0, 0).
1. Find the first and second order partial derivatives with respect to x and y.
Answer:
∂f yx4 + 4y 3 x2 − y 5
= ,
∂x (x2 + y 2 )2
∂f −xy 4 − 4x3 y 2 + x5
= ,
∂y (x2 + y 2 )2
∂2f 12xy 5 − 4x3 y 3
= ,
∂x2 (x2 + y 2 )3
∂2f 4x3 y 3 − 12x5 y
= ,
∂y 2 (x2 + y 2 )3
∂2f ∂2f x6 − y 6 + 9x4 y 2 − 9x2 y 4
= = .
∂y∂x ∂x∂y (x2 + y 2 )3
∂f f (h, 0) − f (0, 0)
= lim = 0,
∂x h→0 h
∂f f (0, h) − f (0, 0)
= lim = 0,
∂y h→0 h
∂f ∂f
∂2f ∂x |(h,0)
− ∂x |(0,0)
= lim = 0,
∂x2 h→0 h
∂f ∂f
∂2f ∂y | (0,h) − ∂y |(0,0)
= lim =0
∂y 2 h→0 h
∂f ∂f yx4 +4y 3 x2 −y 5
∂2f |(0,h) −0
∂x |(0,h) − ∂x |(0,0) (x2 +y 2 )2
= lim = lim =
∂y∂x h→0 h h→0 h
h·04 +4h3 ·02 −h5 −h5
(02 +h2 )2 −0 −0
h4
= lim = lim = −1
h→0 h h→0 h
∂f ∂f
∂y |(h,0) − ∂y |(0,0)
2
∂ f
= lim = 1.
∂x∂y h→0 h
3
2. The total differential is given by
∂f (x, y) ∂f (x, y)
df = dx + dy =
∂x ∂y
yx4 + 4y 3 x2 − y 5 −xy 4 − 4x3 y 2 + x5
= 2 2 2
dx + dy
(x + y ) (x2 + y 2 )2
df ∂f (x, y) dx ∂f (x, y) dy
= + =
dy ∂x dy ∂y dy
yx4 + 4y 3 x2 − y 5 dx −xy 4 − 4x3 y 2 + x5
= +
(x2 + y 2 )2 dy (x2 + y 2 )2
∂2f
4. Let us see the values of the function ∂x2 along the line y = αx, α ∈ R
2. What can you say about quasi-concavity and quasi-convexity of this function (without performing
further calculations)?
3. What can you say about strict quasi-concavity and strict quasi-convexity of this function (without
performing further calculations)?
Answer:
1. In order to assign (strict) concavity and convexity, let us compute the Hessian matrix. We have:
!
−24x2 −10
H(x; y) =
−10 −60y 4 − 10
4
We can now compute leading principal minors:
r
2 5
|H1 | = −24x < 0 (as x > ),
12
|H2 | = 24x2 (60y 4 + 10) − 100 = 1440x2 y 4 + 10(24x2 − 10) ≥ 0.
q
5
We directly see that the second inequality turns to the equality when y = 0 and x = 12 , which is
different from (0, 0). Thus, the Hessian H(x; y) is neither PD, nor ND and the corresponding function
f (x; y) is neither strictly convex, nor strictly concave.
Let us thus check arbitrary principal minors:
Order 1: ∆11 = −24x2 < 0 and ∆21 = −(60y 4 + 10) < 0
Order 2: ∆2 = |H2 | ≥ 0.
As we have (−1)k ∆k ≥ 0, H(x; y) is negative semi-definite and the function f (x; y) is concave.
3. However, we cannot say anything about strict quasi-concavity before performing further computation.
6 Quasi-concavity
Let
f (x) = x3 ,
g(y) = −y.
3. Show that a pseudo-concave function can only achieve its global maximum at a zero gradient point.
−1 0 0
5
where we have (−1) H̄1 = 9x4 ≥ 0, (−1)2 H̄2 = −6x ∈ R. So it can not be quasi concave/convex.
3. Suppose f is pseudo concave and Df (x∗ ) = 0. If f (x∗ ) is not a global maximum, ⇒ ∃ x∗∗ 6= x∗ , s.t.
f (x∗∗ ) > f (x∗ ) ⇒ Df (x∗ )(x∗∗ − x∗ ) = 0. Therefore it is contradicting that f is pseudo concave.
6
Mathematics for Economics and Finance (Fall 2020)
Problem Set 9: Calculus and Probability
Professor: Roxana Mihet
TAs: Oksana, Laura, Remy
Due Nov 13
1 Integration
Compute the following integrals:
Re
1. x2 ln xdx
1
Z
2. cos (2x) e3x dx
Re√ R1 2
3. 1
ln x dx + 0
ex dx
R 3 R 2x
4. 2 x
(x + 2y) dydx
RπRπ sin x
5. 0 y x dxdy
Z √
sin x
6. √ dx;
x
Z
7. ex sin x dx.
Answer:
1. To compute
Ze
x2 ln xdx
1
let
ln x = u
1
x dx= du
2
x dx = dw
x3
3 =w
so
Ze Ze
x3 x3 1
x ln xdx = ln x |e1 −
2
dx =
3 3 x
1 1
e3 x3 e3 e3 1 2e3 1
= − |e1 = − + = +
3 9 3 9 9 9 9
1
2.
cos 2x = u
Z Z
−2 sin (2x) dx = du e3x e3x
I= cos (2x) e3x dx = = cos (2x) − − 2 sin (2x) dx
e3x dx = dw 3 3
e3x
3 =w
Z
e3x 2
= cos (2x) + sin (2x) e3x dx
3 3
sin (2x) = u
Z Z
2 cos (2x) dx = du e3x e3x
sin (2x) e3x dx = 3x
= sin (2x) − 2 cos (2x) dx
e dx = dw 3 3
e3x
3 =w
3x Z
e 2 e3x 2
= sin (2x) − cos (2x) ∗ e3x dx = sin (2x) − I
3 3 3 3
3x Z 3x
e 2 e 2 e3x 2
I = cos (2x) + sin (2x) e3x dx = cos (2x) + sin (2x) − I ⇒
3 3 3 3 3 3
3 2
I= cos (2x) + sin (2x) e3x + C
13 13
√
3. Introducing the following substitution: ln x = y, one can compute the integral:
√ 2
Z ( y = ln x ⇒ y 2 = ln x ⇒ ey = x )
e √ 2 2
ln x dx = dy = 21 √ln
1 1
xx
dx ⇒ dx = 2yey dy = dey =
1
y = 1 if x = e, y = 0 if x = 1
Z 1 1 Z 1 Z 1
2 2 2 2
(1st way) = ydey = yey − ey dy = e − ex dx
0 0 0
0
Z ( u=y du = dy )
1
nd y2
(2 way) = y ∗ 2ye dy = =
0 y2 y2
v=e dv = 2ye dy
1 Z 1 Z 1
y2 y2 2
= ye − e dy = e − ex dx
0 0
0
Z e √ Z 1
2
ln x dx + ex dx = e.
1 0
4.
Z 3 Z 2x Z 3 2x Z 3 3
2 2 4x3 76
(x + 2y) dydx = xy + y dx = 4x dx = = .
2 x 2 2 3 3
x 2
2
x π
RπRπ sin x
RπRx sin x
Rπ Rπ
0 y x dxdy = 0 0 x dydx = 0 sinxx y dx = 0 sin xdx = −cos x = 2.
0 0
We have:
√ √
Z √ f [g(x)] = sin x u = g(x) = x
sin x
√ ⇒ 1 ⇒ 1 √
x g 0 (x) = √ du = √ dx ⇒ dx = 2 xdu
2 x 2 x
Z Z
sin u √
⇒ 2u du = 2 sin u du = −2 cos u + C = −2 cos x + C
u
Z Z
0
f (x)g (x) dx = f (x)g(x) − f 0 (x)g(x) dx
We have:
Z f (x) = sin x f 0 (x) = cos x
ex sin x dx where ⇒
g 0 (x) = ex g(x) = ex
Z Z
⇒ ex sin x dx = ex sin x − ex cos x dx = ex sin x − A
Z f (x) = cos x f 0 (x) = − sin x
A= ex cos x dx where ⇒
g 0 (x) = ex g(x) = ex
Z Z
A = e cos x − −e sin x dx = e cos x + ex sin x dx
x x x
Z Z Z
x x x x x x sin x − cos x
So e sin x dx = e sin x − e cos x − e sin x dx ⇒ e sin x dx = e
2
2 Bounds of integration
Compute the following integrals and see what happens when you change the order of integration
Z 1Z 1
1. (1 − x2 − y 2 ) dx dy;
0 0
Z √
1Z x y
e
2. dy dx.
0 x y
3
Z y=1 Z x=1
1. (1 − x2 − y 2 ) dx dy.
y=0 x=0
Here, the region over which we want to integrate does not depend neither on x nor on y. In other
words, we want to integrate the function f (x, y) = 1 − x2 − y 2 over the square of area 1, given by
0 ≤ x ≤ 1 and 0 ≤ y ≤ 1. So, setting the bounds of integration is easy because, no matter what x I
take, y still goes from 0 to 1.
Every double integral can be computed in two steps:
ii) Take the result obtained in the point i) and plug it into the outer integral
Z y=1 1
2 2 2 2 y3 2 1 1
− y dy = y − = − =
y=0 3 3 3 0 3 3 3
Z Z √
1 x y
e
2. dy dx.
0 x y
Here the problem is more complicated. Indeed the bounds of integration involve x. To understand and
set the region of integration it is better to view some pictures.
i) First we draw the bounds of integration in the x − y plane. The region over which we integrate is
√
the area between the two curves, y = x and y = x. See Figure 1.
y=x
y
√
1 y= x
0 x
1
4
y=x
y
√
1 y= x
Z Z √
x=1 y= x y
e
dy dx
x=0 y=x y
0 x
x0
1
iii) We check where the slice of the inner variable y goes. See Figure 3.
y=x
y
√
1 y= x
√
...to y = x
2. Set the size of
the slice of the
inner variable, y.
from y = x...
Z Z √
x=1 y= x y
e
dy dx
x=0 y=x y
0 x
1
iv) We let the outer variable x vary from its first value, x = 0, to its last value, x = 1. In this way
we sum up all the slices built in points ii) and iii). See Figure 4.
5
y=x
This tells me which is the first y
slice, x = 0, and which is the
last slice, x = 1. √
1 y= x
Z Z √
x=1 y= x y
e
dy dx
x=0 y=x y
Once the bounds of integration are set, we integrate the function. We start with the inner integral.
Z √
y= x y
e
dy
y=x y
We notice that there is no closed form solution for this integral. Hence we must change the order of
integration. To do this, we do again the steps made before, but changing the role of x and y. See
Figure 5,6 and 7.
y=x
y
√
1 y= x
Z y=1 Z x=y y
e
dx dy
y=0 x=y 2 y
y0
0 x
1
6
y=x
y
from x = y 2 ... ...to x = y
√
1 y= x
Z y=1 Z x=y y
e
dx dy
y=0 x=y 2 y
0 x
1
Note: you may ask why we are going from x = y 2 to x = y and not the other way
around. This is because, in the previous case, we were going
√ from the smaller value
of y (that is y = x) to the bigger value of y (that is y = x). Here, we must keep
the same order. So, we go from the smaller value of x (that is x = y 2 ) to the bigger
value of x (that is x = y).
n o
1 1 1 1 1
= [ey ]0 − [yey ]0 − [ey ]0 = [ey − yey + ey ]0 = [2ey − yey ]0
= (2e − e) − (2 − 0) = e − 2.
7
y=x
y
This tells me which is the
first slice, y = 0, and √
1 y= x
which is the last slice,
y = 1.
Z y=1 Z x=y y
e
dx dy
y=0 x=y 2 y
8
3 Probability
Let A and B be two independent events.
3) Compute P(A | B)
Answer
1)
P(Ac ∩ B c ) = 1 − P(A ∪ B)
= 1 − (P(a) + P(B) − P(A ∩ B))
= 1 − P(A) − P(B) − P(A)P(B)
= (1 − P(A))(1 − P(B))
= P(Ac )P(B c )
2)
P(A ∩ B) P(A)P(B)
P(A | B) = = = P(A)
P(B) P(B)
9
Mathematics for Economics and Finance (Fall 2022)
Problem Set 10: Statistics and Probability
Professor: Roxana Mihet
TA: Madhushree
Due Nov 25
1 Random variable
Let’s assume the random variable Xt to be distributed as follows ∀t = 1...T :
(
1
2 if x = 1,
P[Xt = x] = 1
2 if x = −1.
t
X
Wt = Xi .
i=1
Answer:
E[Xt ] = 21 × 1 + 21 × (−1) = 0
V[Xt ] = E[Xt2 ] − E[Xt ]2
E[Xt2 ] = 12 × 12 + 21 × (−1)2 = 1, ⇒
V[Xt ] = 1 − 02 = 1
With this, we can compute the expectation and the variance of Wt .
Pt
E[Wt ] = E[ i=1 Xi ]
Pt
= i=1 E[Xi ] = 0
| {z }
Pt =0
V[Wt ] = V[ i=1 Xi ]
Pt
= i=1 V[Xi ] = t
| {z }
=1
2 Coin problem
There are exactly 64350 gold coins in a chest. One of them is drawn randomly and tossed 5 times. Each
time the outcome is "heads".
2. Now it appears that one of the coins was counterfeit : it had "heads" on both sides. What is the
probability that the coin that was drawn and tossed initially was actually the counterfeit ?
3. Later on you learn that there were another 29 counterfeits among those in the chest, all of which had
"tails" on both sides. Recalculate the probability that the coin tossed was the one with "heads" on
both sides.
Answer:
1
1. For each tossing the probability of the outcome “heads” is P (”heads”) = 12 . Outcomes of consecutive
tossings are independent events =⇒ probability of 5 consecutive outcomes “heads” is P (5 ”heads”) =
1
25 .
1 1
P (counterf eit) = =
N 64350
P (5 ”heads” | counterf eit) = 1.(Obviously)
1
64350 ·1 1 1
Therefore, P = = 2011.9 ≈ 2012 .
1
64350 ·1+( 1− 1
64350 )· 215
3. Bayes’s formula adjusted for part (c) is as follows:
1
0
64350 1
P = 1 29 30
1 = .
64350 ·1+ 64350 ·0+ 1− 64350 · 25
2011
2
3 Probability events
For events A, B, C, C1 , C2 such that 0 < P (C) < 1 which of the following statements are true?
3. P (A |C ) ≥ P (B |C ) , P (A |C c ) ≥ P (B |C c ) =⇒ P (A) ≥ P (B)
Answer:
4 Probability measure
Given a complete probability space (Ω, F, P). Prove that the conditional probability defined by
P (A ∩ B)
P ( A| B) = ,
P (B)
Answer: From the lecture notes we know that a probability measure is defined by the three axioms of
Kolmogorov, i.e
1. P (∅) = 0,
2. P (Ω) = 1,
S∞ P∞
3. P ( i=1 Ai ) = i=1 P (Ai ) for disjoint sets A1 , A2 , . . . ∈ F.
Thus, we have just to check whether the conditional probability measure satisfies all three conditions. In
order to show that all conditions are in fact true we fix the event B and just write Q (A) := P ( A| B) as
short-hand notation. Assume that all axioms are fulfilled for unconditional probabilities, like P (A) for any
set A ∈ F.
3
Ad 1.
P (∅ ∩ B) P (∅)
Q (∅) = = = 0,
P (B) P (B)
Ad 2.
P (Ω ∩ B) P (B)
Q (Ω) = = = 1,
P (B) P (B)
Ad 3.
∞ ∞
! S∞ S∞ P∞
[ P (( i=1Ai ) ∩ B) P ( i=1 (B ∩ Ai )) P (B ∩ Ai ) X
Q Ai = = = i=1 = Q (Ai ) ,
i=1
P (B) P (B) P (B) i=1
X
1 2 3
1 1 1
2 12 6 12
1 1
Y 3 6 0 6
1
4 0 3 0
Are X and Y dependent random variables? Explain your answer. What is the marginal distribution of X
and Y , respectively?
1 1 1
P (X = 1) = , P (X = 2) = , P (X = 3) =
4 2 4
1 1 1
P (Y = 2) = , P (Y = 3) = , P (Y = 4) =
3 3 3
1 1 1
P (X = 1, Y = 2) = ∗ =
4 3 12
1 1 1 1
P (X = 1, Y = 3) = ∗ = 6=
4 3 12 6
1 1 1
P (X = 1, Y = 4) = ∗ = 6= 0
4 3 12
So, X and Y are dependent random variables
4
Marginal distribution of X:
x < 1, FX (x) = 0
1 1 1
1 ≤ x < 2, FX (x) = + =
12 6 4
1 1 1 3
2 ≤ x < 3, FX (x) = + + =
4 6 3 4
3 1 1
x ≥ 3, FX (x) = + + =1
4 12 6
Marginal distribution of Y:
y < 2, FY (y) = 0
1 1 1 1
2 ≤ y < 3, FY (y) = + + =
12 6 12 3
1 1 1 2
3 ≤ y < 4, FY (y) = + + =
3 6 6 3
2 1
y ≥ 4, FY (y) = + = 1
3 3
5
6 Statistics for finance
Standard quantitative models of the stock market assume that stock returns follow a log-normal distribution.
If log X ∼ N (µ, σ 2 ), 0 < X < ∞, −∞ < µ < ∞, σ 2 > 0.
Answer:
For pdf:
2.
Z Z
1 (y−µ)2 1 (y−µ)2 −2σ 2 y
E(X) = E(elog X ) = E(eY ) = ey √ e− 2σ2 dy = √ e− 2σ 2 dy
2πσ 2πσ
(y−µ)2 −2σ 2 (y−µ)+σ 4 −σ 4 −2σ 2 µ
Z Z
1 − 1 (y−µ−σ 2 )2 σ2
= √ e 2σ 2 dy = √ e− 2σ2 + 2 +µ dy
2πσ 2πσ
Z Z
1 (y−µ−σ 2 )2 σ 2 σ 2 1 (y−µ−σ 2 )2
= √ e− 2σ2 e 2 +µ dy = e 2 +µ √ e− 2σ2 dy
2πσ 2πσ
σ2
= eµ+ 2 .
2 2 2 2 2 2
V ar(X) = EX 2 − (EX) = Eelog X − (EX) = Ee2Y − (EX) = e2µ+2σ − e2µ+σ .
7 MGF
2
1
u2 )
1. Verify that the MGF of a N (µ, σ 2 ) is ΦX (u) = e(uµ+ 2 σ .
Answer:
6
1.
+∞
Z +∞
Z
1 (x−µ)2 1 x2 +µ2 −2xµ−2σ 2 ux
ΦX (u) = E(euX ) = √ eux e− 2σ2 dx = √ e− 2σ 2 dx
2πσ 2πσ
−∞ −∞
+∞ 2 2 +∞ 2
1
Z ( ) (
x2 −2x µ+σ 2 u + µ+σ 2 u ) −(µ+σ2 u) +µ2
1
Z
−
(x−(µ+σ2 u)) 2 2
+ µu+ σ 2u
−
=√ e 2σ 2 dx = √ e 2σ 2 dx
2πσ 2πσ
−∞ −∞
+∞ 2
2 u2 1
Z (x−(µ+σ2 u)) σ 2 u2
µu+ σ −
=e 2 √ e 2σ 2
dx = eµu+ 2 .
2πσ
−∞
| {z }
=1(integral of pdf N (µ+σ 2 u,σ 2 ))
∂ΦX (u)
2. EX = ∂u |u=0 = ΦX (u) × (µ + σ 2 u)|u=0 = ΦX (0) ×µ = µ
| {z }
=1
∂ 2 ΦX (u)
E(X 2 ) = |u=0 = µ × ΦX (u) × (µ + σ 2 u) + σ 2 × u × ΦX (u) × (µ + σ 2 u) + σ 2 × ΦX (u)|u=0
∂u2
= µ × ΦX (0) × µ + σ 2 × ΦX (0) = µ2 + σ 2
V X = µ2 + σ 2 − µ2 = σ 2
7
Mathematics for Economics and Finance (Fall 2022)
Problem Set 11: Statistics and Probability
Professor: Roxana Mihet
TA: Madhushree
Due Dec 2nd
Answer:
Rb R0 Rx
1. a
pdf (t)dt = −∞
0dx + 0
λe−λt dt = 1 − e−λx
1 − e−λx = 0.05
ln 0.95
x=−
λ
Rx (t−µ)2
2. −∞
√ 1
2πσ 2
e− 2σ 2 dt = Φ( x−µ
σ )
x−µ
Φ(
) = 0.05
σ
In the table of standard normal distribution we find the closest corresponding argument
x−µ
≈ −1.64
σ
x ≈ −1.64σ + µ ≈ −1.64σ,
Pay attention that if you discuss a potential loss (a positive amount which you lose), then in α% worst
cases it will be greater than or equal to VaR. Thus, VaR is a minimum potential loss.
1
α%
losses
0 VaR
If you consider a potential return (a positive or, more interesting for us, negative amount which you
gain), then in α% worst cases it will be less than or equal to VaR. Thus, VaR will be a maximal
potential return.
α%
return
VaR
Rb R∞
3. a
pdf (t)dt = x
λe−λt dt = 0 + e−λx
e−λx = 0.05
ln 0.05
x=−
λ
R∞ (t−µ)2
4. x
√ 1
2πσ 2
e− 2σ 2 dt = 1 − Φ( x−µ
σ )
x−µ
1 − Φ( ) = 0.05
σ
x−µ
Φ(
) = 0.95
σ
In the table of standard normal distribution we find the closest corresponding argument
x−µ
≈ 1.64
σ
x ≈ 1.64σ + µ ≈ 1.64σ,
2 1
Answer: Note that the cdf of the random variable Z is a mixture of a continuous and a step function,
thus it is neither continuous nor discrete. Nevertheless, for the continuous part the pdf is given by
1 , 0 < x ≤ 1;
fX (x) = 2
0, otherwise.
Additionally, we have an atom at x = 0 with probability P (x = 0) = 12 . After we have found the pdf of Z
and identified the atom it is not to difficult to obtain the expectation and the variance of it. We have
1
1 1 1
Z
E (X) = · 0 + x dx = .
2 0 2 4
Z 1 2
1 1 2 1 5
V ar (X) = · 02 + x dx − = .
2 0 2 4 48
Answer:
R∞
1. fX (x) ≥ 0, 1 x23 dx = 1.
R∞
2. E(X) = 1 x x23 dx = 2, V ar(X) = ∞.
4 True or false?
You have two random variables, X and Y . Consider three statements:
A = ‘The expectation of the product of X and Y is equal to zero. That is, E[XY ] = 0.’
B = ‘The correlation between X and Y is zero.’
C = ‘X and Y are independent.’
For the following statements, state whether they are true or false and explain why:
1. ¬B =⇒ ¬A
2. A =⇒ B
3. C =⇒ B
Answer:
Cov(X;Y )
1. False. We have Cov(X; Y ) = E(XY ) − E(X)E(Y ). and Corr(X; Y ) = σX σY . If Corr(X; Y ) 6= 0,
Cov(X; Y ) 6= 0, but E(XY ) may still be equal to 0.
3
2. False. If E(XY ) = 0, it may be that E(X)E(Y ) 6= 0 and thus both covariance and correlation between
X and Y are different from 0.
3. True. If X and Y are independent, we have that E(XY ) = E(X)E(Y ) and Cov(X; Y ) = 0, leading
to Corr(X; Y ) = 0.
4. False. If X and Y are independent, we have that E(XY ) = E(X)E(Y ), but it does not mean that
E(XY ) = 0.
2y + 4x
f Y |X ( y| x) = .
1 + 4x
Answer:
1. From Definition 4.18 in the lecture notes we know that the conditional density of a random variable
Y given another random variable X = x is
fX,Y
(x,y)
fX (x) , if fX (x) > 0;
f Y |X ( y| x) =
0,
otherwise.
1+4x
∀x ∈ (0, 1) fX (x) = 3 > 0, therefore, the joint density of X, Y is
2y + 4x 1 + 4x 2y + 4x
fXY (x, y) = f Y |X ( y| x) fX (x) = · = .
1 + 4x 3 3
4
The corresponding marginal distribution of Y is then given by
y y
2b + 2 y 2 + 2y
Z Z
FY (y) = P(Y 6 y) = fY (b) db = db = .
0 0 3 3
2y+2
3. Finally, as ∀y ∈ (0, 1) fY (y) = 3 > 0, the conditional density of X given Y is
2y+4x
fX,Y (x, y) 3 y + 2x
f X|Y ( x| y) = = 2y+2 = .
fY (y) 3
y+1
6 Conditional Moments
For random variables X, Y ∼ N 0, σ 2 determine
1. E X X 2
2. E (X |XY )
3. E X + Y |X 2 + Y 2 .
Answer:
1. Since normal distribution is symmetric around zero, thus knowing X 2 = x one can immediately tell
√ √
that X = x with probability 1/2 or X = − x with probability 1/2. Thus E X X 2 = x =
√ √ √ √ √ √
P (X = x) · x + P (X = − x) · (− x) = 21 · x + 21 · (− x) = 0.
2
Another way to show is to notice that −X ∼ N 0, σ 2 =⇒ E X X 2 = E −X (−X) =
2
2
2
E −X X = −E X X =⇒ E X X = 0.
2. One may note that −X, −Y ∼ N 0, σ 2 .
because −X, −Y have the same distribution as X, Y . Thus E (X |XY ) = −E (X |XY ) ⇐⇒ E (X |XY ) =
0.
2 2
3. Applying the same idea, E X + Y |X 2 + Y 2 = −E (−X − Y | (−X) +(−Y ) ) = −E X + Y |X 2 + Y 2 =⇒
E X + Y |X 2 + Y 2 = 0.
5
Mathematics for Economics and Finance (Fall 2022)
Problem Set 12: Statistics and Static Optimization
Professor: Roxana Mihet
TA: Samy
Due Dec 9
estimator of θ: θb = X.
2. Show that it is not EFFICIENT estimator of θ. Hint: compute the Cramer-Rao lower bound.
Answer:
Pn
Xi nEX
1. Unbiased: E θb = EX = E i=1
n = n = EX = θ.
Pn
Xi V arX π2
Consistent: limn→∞ V arθb = limn→∞ V arX = limn→∞ V ar i=1
n = limn→∞ n = limn→∞ 3n =
0.
2
∂
2. We have − (∂θ) 2 ln f (x; θ) = 2f (x; θ) (obtained by direct computation!).
R∞ R∞
∂2
Thus I (θ) = E − (∂θ) 2 ln f (x; θ) = 2Ef (x; θ) = 2 −∞ f (x; θ) · f (x; θ) dx = 2 −∞ f 2 (x; θ) dx =
∞
2 ∞ ∞
e−x+θ e−x+θ 0
Z Z Z Z
−x+θ tdt tdt
=2 4 dx = 2 − 4 de =2 − 4 =2 4 =
−∞ (1 + e−x+θ ) −∞ (1 + e−x+θ ) ∞ (1 + t) 0 (1 + t)
Z ∞ −3
! Z ∞ ∞ Z ∞ ∞ Z ∞ −3
(1 + t) 2t (1 + t)
=2 td − =2 udv = 2uv −2 vdu = − 3 −2 − dt =
0 3 0 0 3(1 + t) 0 3
0 0
−2 ∞
(1 + t) 1
=0−2 = .
6 3
0
1 −1 3
The lower bound on the variance of all estimators of θ equals nI (θ) = n. Estimator θb = X is
efficient if its variance equals to the Cramer-Rao lower bound.
π2 3
However, V arθb = 3n > n, thus estimator θb is not EFFICIENT estimator of θ.
Remark: Note that the lower bound might not be achievable, i.e. the efficient estimator might not
exist. Also note that from the above analysis we cannot conclude anything on the optimality of θb in
the class of all unbiased estimators strictly speaking.
1
2 Unbiased estimator
Let X1 , ..., Xn be i.i.d N (µ, σ 2 ).
4. Is σ̂ 2 a biased estimator? What is the variance of σ̂ 2 ? (e) Show that σ̂ 2 has smaller MSE than
S 2 .Explain why.
Answer:
1.
n
1X 1
E X̄ = E( Xi ) = nEX1 = µ;
n i=1 n
n
1X
V ar(X̄) = V ar( Xi )
n i=1
n
1 X
= V ar( Xi )
n2 i=1
1
= nV ar(X1 )
n2
2
σ
= ;
n
n
1 X
E(S 2 ) = E{ (Xi − X̄)2 }
n − 1 i=1
n
1 X
= E{ (X 2 − 2Xi X̄ + X̄ 2 )}
n − 1 i=1 i
n n n
1 X X X
= { E(Xi2 ) − 2E(X̄ Xi ) + E( X̄ 2 )}
n − 1 i=1 i=1 i=1
| {z }
=nX̄
1
= nE(X12 ) − 2nE(X̄ 2 ) + nE(X̄ 2 )
n−1
1 σ2
= {n(σ 2 + µ2 ) − n( + µ2 )}
n−1 n
= σ2 .
2
2. We get
(n − 1)S 2
V ar( ) = 2(n − 1) ⇒
σ2
(n − 1)2
V ar(S 2 ) = 2(n − 1) ⇒
σ4
2σ 4
V ar(S 2 ) = .
n−1
3. Compute MLE σ̂ 2 ,
n
1 1 X (Xi − µ)2
L(µ, σ 2 | X) = n exp{− };
2
(2πσ ) 2 2 i=1 σ2
n
n n 1 X (Xi − µ)2
lnL = − ln 2π − ln σ 2 − ;
2 2 2 i=1 σ2
FOC:
n
∂ ln L 1X
= 2 (Xi − µ) = 0 (1)
∂µ σ i=1
n
∂ ln L n 1 X
= − + (Xi − µ)2 = 0 (2)
∂σ 2 2σ 2 2σ 4 i=1
n
1 n−1 2
From (2) ⇒ σ̂ 2 = (Xi − X̄)2 ⇒ σ̂ 2 =
P
n n S .
i=1
4. E σ̂ 2 = E( n−1 2
n S )=
n−1 2
n σ (clearly, it is biased. )
n−1 2
2 2
Bias: E(σ̂ − σ ) = n σ −σ 2 = − n1 σ 2 (Notes: as n → ∞, bias → 0 )
2(n−1)σ 4
V ar(σ̂ 2 ) = V ar( n−1 2 n−1 2 2
n S ) = ( n ) V ar(S ) = n2 .
5. MSE of σ̂ 2 is given as
2n − 1 4 2
σ < σ4 ⇒
n2 n−1
M SE(σ̂ 2 ) < M SE(S 2 ).
3
3 Markov Chains
Let πij = Pr(z 0 = zj |z = zi ) be the transition probability of an economic system to move to state zj if the
previous state was zi . The stochastic process z forms a Markov chain. Consider the matrix:
" #
π11 π12
Π= .
π21 π22
2. Assume the probability distribution over the states today is pt = (pt1 , pt2 ). What is the probability
distribution over the state in period t + 1 and t + 2, respectively?
4. A stationary distribution is a vector (q, 1 − q) with 0 ≤ q ≤ 1 such that the probability distribution
over the state in period t + 1 equals the probability distribution over the state in period t. Show that
q solves (2 − π22 − π11 )q = (1 − π22 ).
2. pt+k = pt Πk .
4. (q; 1 − q) = (q; 1 − q)Π = (qπ11 + (1 − q)(1 − π22 ), q(1 − π11 ) + (1 − q)π22 ) ⇔ q = qπ11 + (1 − q)(1 − π22 )
and 1 − q = q(1 − π11 ) + (1 − q)π22 ⇔ q(2 − π22 − π11 ) = (1 − π22 ).
Answer:
4
The FOC
∂ 1
lnL = (X − θ1 ) + (Z̄ − θ1 − θ2 ) = 0 ⇒ 4X̄ + Z̄ = 5θ1 + θ2
∂θ1 4
∂ 1 1
lnL = (Ȳ − θ2 ) + (Z̄ − θ1 − θ2 ) = 0 ⇒ 2Ȳ + Z̄ = θ1 + 3θ2
∂θ2 2 4
6X̄ − Ȳ + Z̄
θ̂1 = .
7
−2X̄ + 5Ȳ + 2Z̄
θ̂2 = .
7
1 1 1
2 2 2
∂ c − 2 (X − θ 1 ) + 2 (Y − θ 2 ) + 4 (Z − θ 1 − θ 2 )
|{z}
∂ ln f (θ1 , θ2 )
constant
=
∂θ " ∂θ#
(X − θ1 ) + 14 (Z − θ1 − θ2 )
= 1 1
.
2 (Y − θ2 ) + 4 (Z − θ1 − θ2 )
5
If DD can secure financing, the expected net present value to DD, N P VDD , of the exploration project is
V minus the underwriting fees paid to BB, X. BB, in turn, earns an expected net present value, N P VBB ,
equal to the fees. BB incurs no costs.
DD and BB now negotiate over how to split the surplus. They decide to play a Nash bargaining game.
The bargaining power of DD is α ∈ [0, 1] and the bargaining power of BB is 1 − α. The equilibrium in the
game is determined by maximizing the generalized Nash product S, where
1. State the optimization problem formally. What are the natural constraints on X implied by the
participation constraints?
2. What is the equilibrium fee X ∗ charged by BB? How does it depend on α? Interpret.
3. What are the expected net present values N P VDD and, respectively, N P VBB ? How do they depend
on α? Interpret.
You are an analyst at BB in the credit risk department. One day, your supervisor asks you to analyze
the profitabilities of all the projects that BB has financed during the last year. For simplicity, assume the
bargaining power of BB last year was 50%. For the purpose of your empirical study, you are given a sample
∗
of past fees, Xsample = (X1∗ , X2∗ , X3∗ , ..., XN
∗
), on N deals. You decide to fit the fees to an exponential
distribution.
4. If X ∗ ∼ Exp(λ) has an exponential distribution with mean λ, what is the distribution of the project
values V given your assumptions above and the Nash solution?
5. Compute by using direct computations, the two first central moments of the random variables X ∗ and
V.
7. What is the expected project value, E[V ], and what is the variance of the project values, V ar[V ], given
∗
the sample Xsample ?
Answer:
1.
2. FOC:
The fees charged by the bank is increasing with the bank’s bargaining power (1 − α).
6
3. For the bank sic previous answer, for the exploration company the NPV at equilibrium = V −X ∗ = αV
and so is decreasing with the bargaining power of the bank.
∗
4. The distribution of X ∗ is an exponential distribution with parameter λ. The distribution of V = (1−α)
X
follows also an exponential distribution with the scaled parameter: P(V ≤ x) = P(X ∗ ≤ (1 − α)x) =
1 − e−λ(1−α)x , so V ∼ Exp(λ(1 − α)).
X∗
5. First for E[V ] at equilibrium: V = 1−α , so if we set du = λe−λx dx and v = x, we get dv = dx and
u = −e−λx and it follows that:
Z ∞ ∞ Z ∞
1 1
E[V ] = xλe−λx dx = −xe
−λx
+ e−λx dx
1−α 0 1−α 0
0
∞
1 1 −λx 1 1
= − e =
1−α λ 1−αλ
0
1 ∗ ∗
We look now for V ar(V ) = (1−α) 2 V ar(X ). We want to compute V ar(X ). We need to compute
E[X ∗ 2 ], this time we set du = λe−λx dx and v = x2 , get dv = 2xdx and u = −e−λx . This leads to:
∞ Z ∞
∗2 2 −λx
E[X ] = −x e + 2xe−λx dx
0
0
Z ∞
2 2 2
= xλe−λx dx = m1 = 2 ,
λ 0 λ λ
2
where m1 is the first moment of X ∗ . Now we eventually get V ar(X ∗ ) = m2 − (m1 ) = 1
λ2 and
1
V ar(V ) = ((1−α)λ) 2.
6. Here the constant (1 − α) does not play any role: The MLE is defined by
λ̂M L = arg max L(λ |X ∗ )
λ
i=1
We then compute the log-likelihood function
n !
Xi∗ n
P
−λ
ln L(λ |X ∗ ) = ln λn e Xi∗
P
i=1 = n ln λ − λ
i=1
7
∂
∂λ ln L(λ̂M L |X ∗ ) = 0
n
n
Xi∗ = 0
P
λ̂
−
ML
i=1
n 1
λ̂M L = n = X¯∗
Xi∗
P
i=1
7. Using the ML estimator for λ we can rewrite the two first central moments of X ∗ as follows:
E [X ∗ ] = X¯∗
2
V [X ∗ ] = X¯∗
X¯∗ X¯∗ 2 2
Then we find E[V ] = 1−α = 2X¯∗ and V ar(V ) = (1−α)2 = 4X¯∗