Download as pdf or txt
Download as pdf or txt
You are on page 1of 95

ECON 104: Mathematics for Economics

SMU School of Economics; Fall 2021


Answer Key to Homework Assignments
Instructor: Takashi Kunimoto
November 22, 2021

Throughout the semester, you are supposed to upload your work (in pdf file)
at “Assignments” at our class website (on eLearn) by 5:00pm on the date specified
below.

1 Homework 1 (Due Date: Aug 26 (Thu), 2021)


Question 1.1 (20 points) Let A = {2, 3, 4}, B = {2, 5, 6}, C = {5, 6, 2}, and
D = {6}. Answer the following questions.

1. Determine each of the following statements is true or false: 4 ∈ C; 5 ∈ C;


A ⊆ B; D ⊆ C; B = C; and A = B.

• 4 ∈ C: false
• 5 ∈ C: true
• A ⊆ B: false
• D ⊆ C: true
• B = C: true
• A = B: false.

2. Find A ∩ B; A ∪ B; A\B; B\A; (A ∪ B)\(A ∩ B); A ∪ B ∪ C ∪ D; A ∩ B ∩ C;


and A ∩ B ∩ C ∩ D.

• A ∩ B = {2}
• A ∪ B = {2, 3, 4, 5, 6}
• A\B = {3, 4}

1
• B\A = {5, 6}
• (A ∪ B)\(A ∩ B) = {3, 4, 5, 6}
• A ∪ B ∪ C ∪ D = {2, 3, 4, 5, 6}
• A ∩ B ∩ C = {2}
• A ∩ B ∩ C ∩ D = {∅}

Question 1.2 (20 points) Consider the proposition 2x + 5 ≥ 13. Answer the
following questions.

1. Is the condition x ≥ 0 necessary or sufficient, or both necessary and sufficient


for the inequality to be satisfied?

2x + 5 ≥ 13 ⇔ x ≥ 4.

So, the condition x ≥ 0 is a necessary condition for the inequality to be


satisfied.

2. Answer the same question when x ≥ 0 is replaced by x ≥ 50.

The condition x ≥ 50 is sufficient for the inequality to be satisfied.

3. Answer the same question when x ≥ 0 is replaced by x ≥ 4.

The condition x ≥ 4 is a necessary and sufficient condition for the inequality


to be satisfied.

Question 1.3 (20 points) In each case below, find the slope of the tangent to the
graph of f at the specified point:

1. f (x) = 3x + 2, at (x, y) = (0, 2)


0
Since we have f (x) = 3, the slope of f is 3 at (0, 2).

2. f (x) = x2 − 1 at (x, y) = (1, 0)


0 0
f (x) = 2x. So, f (1) = 2.

3. f (x) = 2 + 3/x, at (x, y) = (3, 3)


0 0
f (x) = −3/x2 . So, f (3) = −1/3.

4. f (x) = x3 − 2x, at (x, y) = (0, 0)


0 0
f (x) = 3x2 − 2. So, f (0) = −2.

2
5. f (x) = x + 1/x, at (x, y) = (−1, −2)
0 0
f (x) = 1 − 1/x2 . So, f (−1) = 0.
6. f (x) = x4 , at (x, y) = (1, 1)
0 0
f (x) = 4x3 . So, f (1) = 4.
Question 1.4 (20 points) Let f (x) = ax2 + bx + c where a, b, c are nonzero
constants. Answer the following questions.
1. Show that
f (x + h) − f (x)
= 2ax + b + ah.
h
f (x + h) − f (x) 1
a(x + h)2 + b(x + h) + c − ax2 − bx − c

=
h h
1
ax2 + 2ahx + ah2 + bx + bh − ax2 − bx

=
h
1
2ahx + ah2 + bh = 2ax + b + ah.

=
h
0
2. Using the previous result, show that f (x) = 2ax + b.
0 f (x + h) − f (x)
f (x) = lim = lim 2ax + b + ah = 2ax + b.
h→0 h h→0

0
3. For what value of x is f (x) = 0?
b
x=− .
2a
Question 1.5 (20 points) Let f (x) = −x3 + 4x2 − x − 6. Examine where f (x)
is nondecreasing and/or nonincreasing.
We compute
√ ! √ !
0 2 2 4− 13 4+ 13
f (x) = −3x + 8x − 1 = −(3x − 8x + 1) = −3 x − x− .
3 3

Therefore,
 √ √
0 ≥ 0 if (4 − 13)/3√ ≤ x ≤ (4 + 13)/3

f (x)
≤ 0 if x ≥ (4 + 13)/3 or x ≤ (4 − 13)/3.
√ √
This implies f (x) is nondecreasing
√ if (4 − 13)/3
√ ≤ x ≤ (4 + 13)/3 and it is
nonincreasing if x ≥ (4 + 13)/3 or x ≤ (4 − 13)/3.

3
2 Homework 2 (Due Date: Sep 2 (Thu), 2021)
Question 2.1 (4 points) Find the limit below precisely:

x2 + 7x − 8
lim .
x→1 x−1
Since x2 + 7x − 8 = (x − 1)(x + 8), we obtain

x2 + 7x − 8 (x − 1)(x + 8)
lim = lim
x→1 x−1 x→1 x−1
= lim x + 8 = 9.
x→1

Question 2.2 (12 points) Find the derivatives of the following functions, where
a, p, q, and b are constants.

1.
1
y=
(x2 + x + 1)5

0
y = −5(x2 + x + 1)−6 · (2x + 1) = −5(2x + 1)(x2 + x + 1)−6 .

2. r

q
y= x+ x+ x

Note first that y = (x + (x + x1/2 )1/2 )1/2 . So,


  
0 1 1/2 1/2 −1/2
 1 1/2 −1/2 1 −1/2
y = x + (x + x ) · 1 + (x + x ) · 1+ x
2 2 2

3.
y = xa (px + q)b .

0
y = axa−1 (px + q)b + xa · b(px + q)b−1 · p = xa−1 (px + q)b−1 {a(px + q) + bpx}
= xa−1 (px + q)b−1 [p(a + b)x + aq]

Question 2.3 (5 points) Let

t2
g(t) = .
t−1
00
Find g (2).

4
We first compute the following:

0 2t(t − 1) − t2 · 1 t2 − 2t
g (t) = =
(t − 1)2 (t − 1)2
00 (2t − 2)(t − 1)2 − (t2 − 2t) · 2(t − 1)
g (t) =
(t − 1)4
2(t − 1)2 − 2(t2 − 2t)
=
(t − 1)3
2
=
(t − 1)3
00
Therefore, we obtain g (2) = 2.

Question 2.4 (10 points) Let U (x) = x and g(u) = u3 . Then, show that U (x)
is a concave function but f (x) = g(U (x)) is “not” a concave function.
have x ≥ 0 as the relevant domain of the functions, U and
Note that we must √
f . Otherwise, U (x) = x is not well-defined. We first compute the following:

01 −1/2
U (x) = x
2
00 1
U (x) = − x−3/2
4
00
So, as long as x > 0, we have U (x) < 0. Thus, U is a concave function as long as
x > 0. Next we consider f (x) = x3/2 . We compute the following:

0 3 1/2
f (x) = x
2
00 3 −1/2
f (x) = x .
4
00
So, as long as x > 0, we have f (x) > 0. This implies that f is not a concave
function.

Question 2.5 (8 points) Find the first-order derivatives of the following func-
tions.

1. y = ex + x2
0
y = ex + 2x.

2. y = 5ex − 3x3 + 8
0
y = 5ex − 9x2 .

5
3. y = x/ex
0 1 · ex − xex 1−x
y = 2x
=
e ex
4. y = (x + x2 )/(ex + 1).

0 (1 + 2x)(ex + 1) − (x + x2 )ex 1 + 2x x(1 + x)ex


y = = −
(ex + 1)2 ex + 1 (ex + 1)2

Question 2.6 (15 points) Find the intervals where the following functions are
nondecreasing:

1. y = x2 /e2x

We compute
0 2x · e2x − x2 · 2e2x −2x(x − 1)
y = 2x 2
= .
(e ) e2x
Therefore, we have 
0 ≥ 0 if 0 ≤ x ≤ 1
y
≤ 0 if x ≥ 1 or x ≤ 0
So, y is nondecreasing over [0, 1].

2. y = ex − e3x

We compute
0
y = ex − 3e3x = ex (1 − 3e2x ).
Therefore, we have 
0 ≥ 0 if x ≤ −(ln 3)/2
y
≤ 0 if x ≥ −(ln 3)/2
So, y is nondecreasing over (−∞, −(ln 3)/2].

3. y = e2x /(x + 2)

We first note that this function is not well-defined when x = −2. So, we
exclude x = −2 as part of the domain. Next we compute

0 2e2x · (x + 2) − e2x · 1 e2x (2x + 3)


y = =
(x + 2)2 (x + 2)2
Therefore, we have 
0 ≥ 0 if x ≥ −3/2
y
≤ 0 if x ≤ −3/2
Then, we conclude that y is nondecreasing over [−3/2, +∞).

6
Question 2.7 (8 points) Compute the first and second derivatives of the follow-
ing functions:

1. y = ln x + 3x − 2
0 1
y = +3
x
00 1
y = − 2.
x

2. y = x2 − 2 ln x
0 2
y = 2x −
x
00 2
y = 2+ 2
x

3. y = x3 ln x
0 1
y = 3x2 ln x + x3 · = 3x2 ln x + x2
x
00 1
y = 6x ln x + 3x2 · + 2x = 6x ln x + 5x.
x

4. y = (ln x)/x
1
0 · x − (ln x) · 1
x 1 − ln x
y = 2
=
x2 x2
1
00 − x · x − (1 − ln x) · 2x 2 ln x − 3
y = =
x4 x3

Question 2.8 (12 points) Find the equation for the tangent to the graph of the
following functions:

1. y = ln x at the three points with x-coordinates: 1, 1/2, and e


0
We first compute y = 1/x. We second note that x = 1 corresponds to y = 0;
x = 1/2 to y = − ln 2; and x = e to y = 1.

• Consider the case where (x, y) = (1, 0). At this point, we have the
equation for the tangent to the graph as y − 0 = 1(x − 1), i.e., y = x − 1.
• Consider the case where (x, y) = (1/2, − ln 2). At this point, we have
the equation for the tangent to the graph as y + ln 2 = 2(x − 1/2), i.e.,
y = 2x − (ln 2 + 1).

7
• Consider the case where (x, y) = (e, 1). At this point, we have the
equation for the tangent to the graph as y − 1 = (1/e)(x − e), i.e.,
y = e−1 x.

2. y = xex at the three points with x-coordinates: 0, 1, and −2


0
We first compute y = (x + 1)ex . Second, we note that x = 0 corresponds to
y = 0; x = 1 to y = e; and x = −2 to y = −2e−2 .

• Consider the case where (x, y) = (0, 0). At this point, we have the
equation for the tangent to the graph as y − 0 = (x − 0), i.e., y = x.
• Consider the case where (x, y) = (1, e). At this point, we have the
equation for the tangent to the graph as y − e = 2e(x − 1), i.e., y =
2ex − e.
• Consider the case where (x, y) = (−2, −2e−2 ). At this point, we have
the equation for the tangent to the graph as y + 2e−2 = −e−2 (x + 2),
i.e., y = −e−2 x − 4e−2 .

Question 2.9 (10 points) Suppose F is a differentiable function of one variable,


0
with F (0) = 0 and F (0) 6= −1, and that y is defined implicitly as a differentiable
function of x by the equation
x3 F (xy) + exy = x.
0
Find an expression for y at the point (x, y) = (1, 0).
We differentiate this equation with respect to x:
0 0 0
3x2 F (xy) + x3 F (xy) · (y + xy ) + exy · (y + xy ) = 1.
Evaluating this equation at (x, y) = (1, 0), we obtain
0 0 0
3F (0) + F (0)y + y = 1.
0
Taking into account that F (0) = 0 and F (0) 6= −1, we get
0 1
y = .
1 + F 0 (0)
Question 2.10 (16 points) Consider the extended standard macroeconomic model
which accommodates international trade as follows:
0
(i) Y = C + I¯ + X̄ − M ; (ii) C = f (Y ), with 0 < f (Y ) < 1; (iii) M = g(Y ).
Here X̄ is an exogenous constant that denotes exports, whereas M denotes the
volume of imports. The function g in (iii) is called an “import function,” which is
0 0
assumed to satisfy 0 < g (Y ) < f (Y ). Answer the following questions.

8
¯
1. Obtain an equation that defines Y as a function of exogenous investment I.

Plugging (ii) and (iii) into (i), we obtain

Y = f (Y ) + I¯ + X̄ − g(Y ).

2. Find an expression for dY /dI¯ by implicit differentiation.

Differentiating Y = f (Y ) + I¯ + X̄ − g(Y ) with respect to I,


¯ we have

dY 0 dY 0 dY
¯ = f (Y ) ¯ + 1 − g (Y ) ¯ .
dI dI dI
We therefore obtain
dY 1
¯ = .
dI 1 − f (Y ) + g 0 (Y )
0

¯
3. Discuss the sign of dY /dI.

It is reasonable to assume that the quantity of imports increases as the income


0
increases. That is, g (Y ) > 0. Then, we conclude that dY /dI¯ > 0 because
0
1 − f (Y ) > 0.

4. Find an expression for d2 Y /dI¯2 .

We duplicate the equation we had in part 2 of the same question:


dY 0 dY 0 dY
¯ = f (Y ) ¯ + 1 − g (Y ) ¯ .
dI dI dI
Differentiating this equation with respect to I, ¯ we obtain
2 2
d2 Y d2 Y d2 Y
 
00 dY 0 00 dY 0
= f (Y ) + f (Y ) − g (Y ) − g (Y ) .
dI¯2 dI¯ dI¯2 dI¯ dI¯2

Summarizing the terms above, we further obtain


2
d2 Y

0 0
 00 00
 dY
(1 − f (Y ) + g (Y )) ¯2 = f (Y ) − g (Y ) .
dI dI¯

Therefore,
00 00 2
d2 Y

f (Y ) − g (Y ) dY
¯ = .
dI 2 1 − f 0 (Y ) + g 0 (Y ) dI¯

9
3 Homework 3 (Due Date: Sep 9 (Thu), 2021)
Question 3.1 (12 points) Find dx/dy of the following functions:

1. y = e−x−5
dx 1 1
= = −x−5
= −ex+5
dy dy/dx −e

2. y = ln(e−x + 3)

dx 1 1 e−x + 3
= = 1 = − = −1 − 3ex .
dy dy/dx e−x +3
· (e−x · −1) e−x

3. xy 3 − x3 y = 2x.

Differentiating the equation with respect to y, we obtain


dx 3 dx dx
y + x · 3y 2 − 3x2 · · y − x3 = 2 .
dy dy dy

Rearranging the terms for dx/dy, we have

dx x3 − 3xy 2
= 3 .
dy y − 3x2 y − 2

Question 3.2 (9 points) Find the linear approximations to the following func-
tions about x = 0:

1. f (x) = (1 + x)−1

We first compute
0
f (x) = −(1 + x)−2 .
So,
0
f (x) ≈ f (0) + f (0)(x − 0) = 1 − x.

2. f (x) = (1 + x)5

We first compute
0
f (x) = 5(1 + x)4 .
So,
0
f (x) = f (0) + f (0)(x − 0) = 1 + 5x.

10
3. f (x) = (1 − x)1/4

We first compute
0 1 1
f (x) = (1 − x)−3/4 · (−1) = − (1 − x)−3/4 .
4 4
So,
0 1
f (x) = f (0) + f (0)(x − 0) = 1 − x
4
Question 3.3 (9 points) Let f (x) = 10 − 10e−x/2 be defined for all x ∈ R. An-
swer the following questions.

1. Show that f (x) has an inverse function g(x).


0
The derivative of f is given as f (x) = −10e−x/2 · (−1/2) = 5e−x/2 > 0. So,
f (x) is strictly increasing so that it is a one-to-one mapping. Thus, f has an
inverse function.

2. What is the domain of g?

Since e−x/2 → ∞ as x → −∞, we have limx→−∞ f (x) = −∞. Since e−x/2 →


0 as x → ∞, we have limx→∞ f (x) = 10. Taking into account that f (·) is
strictly increasing, we obtain (−∞, 10) as the domain of g.
0
3. Note that f (0) = 0. Compute g (0).

By the inverse function theorem, we have


0 1 1
g (0) = 0 = .
f (0) 5

Question 3.4 (10 points) Let f be defined for all x by


 2
x − 1 if x ≤ 0
f (x) =
−x2 if x > 0

Answer the following questions.

1. Draw a graph of the function f .

We omit this.

2. Is f continuous at x = 0?

It is “not” continuous at x = 0 because limx→0+ f (x) = 0 6= −1 = limx→0− f (x).

11
Question 3.5 (12 points) Evaluate the following limits:

1.
lim (x2 − 3x + 2)
x→3−

lim (x2 − 3x + 2) = lim− (x − 2)(x − 1) = lim− (x − 2) lim− (x − 1) = 1 · 2 = 2.


x→3− x→3 x→3 x→3

2.
x2 − 3x + 14
lim +
x→−2 x+2
Since we have x2 − 3x + 14 = (x + 2)(x − 5) + 24, we obtain

x2 − 3x + 14 24
lim + = lim + (x − 5) + lim + = −7 + ∞ = ∞.
x→−2 x+2 x→−2 x→−2 x + 2

3. √
3− x + 17
lim
x→−1 x+1
We first obtain √
3− x + 17
lim = −∞
x→−1+ x+1
and second we obtain

3− x + 17
lim − = +∞
x→−1 x+1
Since the limit from below do not coincide with that from above, we have no
such a limit in this case.

4.  
1 5
lim − 2
x→3 x − 3 (x − x − 6)
We observe
1 5 1 5 x+2−5 x−3 1
− 2 = − = = = .
x − 3 (x − x − 6) x − 3 (x − 3)(x + 2) (x − 3)(x + 2) (x − 3)(x + 2) x+2

So,  
1 5 1 1
lim − 2 = lim = .
x→3 x − 3 (x − x − 6) x→3 x + 2 5

Question 3.6 (12 points) Let f1 (x) = x, f2 (x) = x, f3 (x) = x2 , and f4 (x) =
1/x. Compute the following:

12
1. limx→∞ (f1 (x) + f2 (x)) = 2 limx→∞ x = +∞

2. limx→∞ (f1 (x) − f3 (x))

lim (f1 (x) − f3 (x)) = lim (x − x2 ) = lim x(1 − x)


x→∞ x→∞ x→∞
= ( lim x) · ( lim (1 − x)) = −∞
x→∞ x→∞

3. limx→∞ f1 (x)/f3 (x)


x 1
= lim 2
= lim = 0.
x→∞ x x→∞ x
4. limx→∞ f1 (x)f2 (x)
= lim x2 = ∞.
x→∞

5. limx→∞ f1 (x)f4 (x)


1
= lim x · = lim 1 = 1.
x→∞ x x→∞
6. limx→∞ f3 (x)f4 (x)
1
= lim x2 · = lim x = +∞.
x→∞ x x→∞

Question 3.7 (6 points) Compute the differential of f (x) = 1 + x3 . What is
the approximate change in f (x) when x changes from x = 2 to x = 2 + dx, where
dx = 0.2.
We first compute

0 1 3x2 1
f (x) = (1 + x3 )−1/2 · 3x2 = √ .
2 2 1 + x3
0
Since dy = f (x)dx, we obtain

3x2 1
dy = √ dx.
2 1 + x3
Evaluating this at x = 2 and dx = 0.2, we obtain
3 1
dy ≈ · 4 · · 0.2 = 0.4.
2 3

x
Question 3.8 (6 points) Prove that f (x) = e − 3 has a unique zero in the
interval (1, 4).

13
We first compute
0 √ 1
f (x) = e · x−1/2
x
2
Since f (·) is differentiable, it is continuous as well. We also check
f (1) = e − 3 < 2.8 − 3 = −0.2 < 0
f (4) = e2 − 3 > 7 − 3 = 4 > 0
Then, by intermediate value theorem, there exists x∗ ∈ (1, 4) such that f (x∗ ) = 0.
0
We also confirm that f (x) > 0 for all x > 0. Hence, f (·) is strictly increasing for
all x > 0. This implies that x∗ must be the unique such point.

Question 3.9 (9 points) Find the domain of the following functions.

1. f (x, y) = 1/(ex+y − 3)

The domain of this function is {(x, y) ∈ R2 | x + y 6= ln 3}.


2. f (x, y) = ln(x − a)2 + ln(y − b)2 , where a, b ∈ R are constants.

The domain of this function is {(x, y) ∈ R2 | x 6= a and y 6= b}.


3. f (x, y) = 2 ln(x − a) + 2 ln(y − b), where a, b ∈ R are constants.

The domain of this function is {(x, y) ∈ R2 | x > a and y > b}.

Question 3.10 (9 points) Find all the first-order and second-order derivatives
(including cross derivatives) of the following functions.

1. f (x, y) = x7 − y 7
0
f1 (x, y) = 7x6
0
f2 (x, y) = −7y 6
00
f11 (x, y) = 42x5
00
f22 (x, y) = −42y 5
00 00
f12 (x, y) = f21 (x, y) = 0

2. f (x, y) = x5 ln y
0
f1 (x, y) = 5x4 ln y
0
f2 (x, y) = x5 /y
00
f11 (x, y) = 20x3 ln y
00
f22 (x, y) = −x5 /y 2
00 00
f12 (x, y) = f21 (x, y) = 5x4 /y

14
3. f (x, y) = (x2 − 2y 2 )5 .
0
f1 (x, y) = 10x(x2 − 2y 2 )4
0
f2 (x, y) = −20y(x2 − 2y 2 )4
00
f11 (x, y) = 10(9x2 − 2y 2 )(x2 − 2y 2 )3
00
f22 (x, y) = −20(x2 − 18y 2 )(x2 − 2y 2 )3
00 00
f12 (x, y) = f21 (x, y) = −160xy(x2 − 2y 2 )3

Question 3.11 (6 points) Let


1
f (x, y) = ln(x3 − y) + 3(x − 1)y + (x + 1)y.
7
Show that the two points (x, y) = (e, 0) and (x, y) = (2, 7) are both on the same
level curve of f .
This is proved by establishing the relation f (e, 0) = f (2, 7).
1
f (e, 0) = ln(e3 − 0) + 3(e − 1)0 + (e + 1) · 0 = 3 ln e + 3 = 6
7
1
f (2, 7) = ln(8 − 7) + 3(2 − 1)7 + (2 + 1) · 7
7
= ln 1 + 3 + 3 = 6.

4 Homework 4 (Due Date: Sep 16 (Thu), 2021)


Question 4.1 (10 points) Find dz/dt for the following cases:

1. z = F (x, y) = x ln y + y ln x; x = t + 1; and y = ln t.
 
dz ∂F dx ∂F dy  y x 1
= + = ln y + ·1+ + ln x ·
dt ∂x dt ∂y dt x y t
   
ln t 1 t+1
= ln(ln t) + + + ln(t + 1)
t+1 t ln t

2. z = F (x, y) = ln x + ln y; x = Aeat ; and y = Bebt where A, B, a, b are nonzero


constants.
dz ∂F dx ∂F dy 1 1
= + = aAeat + bBebt
dt ∂x dt ∂y dt x y
= a + b.

Question 4.2 (10 points) Find ∂z/∂t and ∂z/∂s for the following cases:

15
1. z = F (x, y) = x + y 2 , where x = t − s and y = ts.

∂z ∂F ∂x ∂F ∂y
= + = 1 · 1 + 2y · s = 1 + 2ts2
∂t ∂x ∂t ∂y ∂t
∂z ∂F ∂x ∂F ∂y
= + = 1 · (−1) + 2y · t = −1 + 2t2 s.
∂s ∂x ∂s ∂y ∂s

2. z = F (x, y) = 2x2 + 3y 3 , where x = t2 − s and y = t + 2s3 .

∂z ∂F ∂x ∂F ∂y
= + = 4x · 2t + 9y 2 · 1 = 8t(t2 − s) + 9(t + 2s3 )2
∂t ∂x ∂t ∂y ∂t
∂z ∂F ∂x ∂F ∂y
= + = 4x · (−1) + 9y 2 · 6s2 = −4(t2 − s) + 54s2 (t + 2s3 )2 .
∂s ∂x ∂s ∂y ∂s

Question 4.3 (16 points) Let u = ln(x3 + y 3 + z 3 − 3xyz). Answer the following
questions.

1. Show that
∂u ∂u ∂u
x +y +z = 3.
∂x ∂y ∂z

We first set v = x3 + y 3 + z 3 − 3xyz. Next we compute the following:

∂u 1 ∂v 1
= · = (3x2 − 3yz)
∂x v ∂x v
∂u 1 ∂v 1 2
= · = (3y − 3xz)
∂y v ∂y v
∂u 1 ∂v 1
= · = (3z 2 − 3xy).
∂z v ∂z v

∂u ∂u ∂u 1
3x3 − 3xyz + 3y 3 − 3xyz + 3z 3 − 3xyz

x +y +z =
∂x ∂y ∂z v
3 3
x + y 3 + z 3 − 3xyz = 3.

=
v

2. Show that  
∂u ∂u ∂u
(x + y + z) + + = 3.
∂x ∂y ∂z

16
We compute the following:
 
∂u ∂u ∂u 3 3
x + xy 2 + xz 2 − x2 y − xyz − x2 z

x + + =
∂x ∂y ∂z v
 
∂u ∂u ∂u 3 2
x y + y 3 + yz 2 − xy 2 − y 2 z − xyz

y + + =
∂x ∂y ∂z v
 
∂u ∂u ∂u 3 2
x z + y 2 z + z 3 − xyz − yz 2 − xz 2

z + + =
∂x ∂y ∂z v
Summing over the above equations, we obtain
 
∂u ∂u ∂u 3 3
x + y 3 + z 3 − 3xyz = 3.

(x + y + z) + + =
∂x ∂y ∂z v

Question 4.4 (16 points) A curve in the xy-plane is given by the equation 2x2 +
xy + y 2 − 8 = 0. Answer the following questions.
0 00
1. Find y , y and the equation for the tangent at the point (x, y) = (2, 0).

Let F (x, y) = 2x2 + xy + y 2 − 8. First we observe that F (2, 0) = 8 − 8 = 0.


Second, we compute
∂F/∂x = 4x + y
∂F/∂y = x + 2y.
Assuming that ∂F/∂y 6= 0, by the implicit function theorem, we obtain
0 ∂F/∂x 4x + y
y =− =− .
∂F/∂y x + 2y
Since ∂F (2, 0)/∂y = 2 at (x, y) = (2, 0), we evaluate
0 8
y |(x,y)=(2,0) = − = −4.
2
The equation for the tangent at the point (x, y) = (2, 0) is given as follows:
y − 0 = −4(x − 2) ⇔ y = −4x + 8.
00
We derive y as follows:
0 0
00 (4 + y )(x + 2y) − (4x + y)(1 + 2y )
y =− ,
(x + 2y)2
where x + 2y 6= 0. Therefore, we have
00 −8(1 − 8)
y |(x,y)=(2,0) = − = −14.
4

17
2. Which points on the curve have a horizontal tangent?
0
To have a horizontal tangent, we must have y = 0. This implies that 4x+y =
0, i.e., y = −4x. Moreover, we must be on F (x, y) = 0 simultaneously. So,
plugging y = −4x into F (x, y) = 0, we have

2x2 − 4x2 + 16x2 − 8 = 0 ⇒ x2 = 4/7 ⇒ x = ±2/ 7.
√ √ √ √
Therefore, the curve has a horizontal tangent at (2/ 7, −8/ 7) and (−2/ 7, 8/ 7).

Question 4.5 (12 points) Suppose that a firm has√a single input L and a single
output Q. The production function is Q = f (L) = L. This firm maximizes its
profit π(L), given that p (output price) and w (wage for a worker)
√ are parameters
given by the market. Then, we have π(L) = pQ − wL = p L − wL. Answer the
following questions.
0
1. Find L∗ such that π (L∗ ) = 0.
0 p p2
π (L) = L−1/2 − w = 0 ⇒ L∗ = .
2 4w2
2. Express L∗ obtained in the previous question as a function p and w, that is,
L∗ (p, w). Find ∂L∗ (p, w)/∂w. Interpret the sign of ∂L∗ (p, w)/∂w.

We have L∗ (p, w) = p2 /(4w2 ). So,


∂L∗ (p, w) p2 p2
= · (−2)w−3 = − 3 < 0.
∂w 4 2w
We interpret ∂L∗ (p, w)/∂w < 0 as follows: When it is more costly to hire a
worker, the firm’s labor demand goes down.
3. Find ∂L∗ (p, w)/∂p. Interpret the sign of ∂L∗ (p, w)/∂p.
∂L∗ (p, w) p
= > 0.
∂p 2w2
We interpret ∂L∗ (P, w)/∂p > 0 as follows: When it becomes possible to sell
one unit of output at higher price, the firm has an incentive to hire more
workers.

Question 4.6 (12 points) Let f (x, y) = 3x2 + xy − y 2 . Answer the following
questions.

1. Compute f (1.02, 1.99) exactly.

f (1.02, 1.99) = 3(1 + 0.02)2 + (1 + 0.02)(2 − 0.01) − (2 − 0.01)2 = 1.1909.

18
2. Let f (1.02, 1.99) = f (1+0.02, 2−0.01) and use a linear approximation around
(1, 2) to find an approximate numerical value for f (1.02, 1.99). How large is
the approximation error?

We first compute the following:


0
f1 (x, y) = 6x + y
0
f2 (x, y) = x − 2y.

So,
0 0
f (1.02, 1.99) ≈ f (1, 2)+f1 (1, 2)·0.02+f2 (1, 2)·(−0.01) = 1+8·0.02−3·(−0.01) = 1.19.

So, the size of the approximation error is 0.0009.

Question 4.7 (12 points) Let


       
2 4 −2 4 2 3 1 1
A= ; B= ; C= ; and D = .
1 2 1 −2 6 9 1 3

Calculate 3A + 2B − 2C + D, AB, and C(AB).

 
−1 15
3A + 2B − 2C + D =
−6 −13
    
2 4 −2 4 2 · (−2) + 4 · 1 2 · 4 + 4 · (−2)
AB = =
1 2 1 −2 1 · (−2) + 2 · 1 1 · 4 + 2 · (−2)
 
0 0
= .
0 0

Clearly, we have  
0 0
C(AB) = .
0 0

Question 4.8 (12 points) Let


   2 
p q 0 p 2pq q 2
T =  p/2 1/2 q/2  and S =  p2 2pq q 2 
0 p q p2 2pq q 2

be two matrices where p, q are constants such that p + q = 1. Answer the following
questions.

19
1. Prove that T S = S.
 
p3 + p2 q 2p2 q + 2pq 2 pq 2 + q 3
T S =  p3 /2 + p2 /2 + p2 q/2 p2 q + pq + pq 2 pq 2 /2 + q 2 /2 + q 3 /2 
p3 + p2 q 2p2 q + 2pq 2 pq 2 + q 3
 
p2 (p + q) 2pq(p + q) q 2 (p + q)
=  p2 (p + q + 1)/2 pq(p + q + 1) q 2 (p + q + 1)/2 
p2 (p + q) 2pq(p + q) q 2 (p + q)
= S (∵ p + q = 1).

2. Prove that T 2 = T /2 + S/2.


 2 
p + pq/2 pq + q/2 q 2 /2
T 2 =  p2 /2 + p/4 pq/2 + 1/4 + pq/2 q/4 + q 2 /2 
p2 /2 p/2 + pq pq/2 + q 2
 2 
p /2 + p(p + q)/2 pq + q/2 q 2 /2
=  p2 /2 + p/4 pq + 1/4 q/4 + q 2 /2 
2 2
p /2 p/2 + pq q /2 + q(p + q)/2
 
p/2 + p2 /2 q/2 + pq 0 + q 2 /2
=  p/4 + p2 /2 1/4 + pq q/4 + q 2 /2  (∵ p + q = 1)
0 + p2 /2 p/2 + pq q/2 + q 2 /2
   2 
p/2 q/2 0 p /2 pq q 2 /2
=  p/4 1/4 q/4  +  p2 /2 pq q 2 /2  = T /2 + S/2.
0 p/2 q/2 p2 /2 pq q 2 /2

3. Prove that T 3 = T /4 + 3S/4.

T 3 = T (T 2 ) = T (T /2 + S/2) (∵ T 2 = T /2 + S/2)
= T 2 /2 + (T S)/2 = (T /2 + S/2)/2 + S/2 (∵ T 2 = T /2 + S/2 and T S = S)
= T /4 + 3S/4.

5 Homework 5 (Due Date: Sep 23 (Thu), 2021)


Question 5.1 (12 points) Use Cramer’s rule to solve the following system of
equations.

x1 − x2 + x3 = 2
x1 + x2 − x3 = 0
−x1 − x2 − x3 = −6.

20
We reexpress the above system of equations:
    
1 −1 1 x1 2
Ax = b ⇔  1 1 −1   x2  =  0  .
−1 −1 −1 x3 −6

Next we compute |A|:



1+1 1
−1 2+1 −1
1 + 1 · (−1)3+1 −1 1

|A| = 1 · (−1) + 1 · (−1)
−1 −1 −1 −1 1 −1
= (−1 − 1) − (1 + 1) + (1 − 1) = −4.

We compute the following:



2 −1 1
1+1 1
−1 3+1 −1
1
0
1 −1
= 2 · (−1) −1 −1 + (−6) · (−1) 1 −1


−6 −1 −1
= 2(−1 − 1) − 6(1 − 1) = −4.


1 2 1

= 2 · (−1)1+2 1 −1 + (−6) · (−1)3+2 1 1

1
0 −1 −1 −1 1 −1


−1 −6 −1
= −2(−1 − 1) + 6(−1 − 1) = −8.


1 −1 2

1 = 2 · (−1)1+3 1
1
+ (−6) · (−1)3+3 1 −1
1 0 −1 −1 1 1

−1 −1 −6
= 2(−1 + 1) − 6(1 + 1) = −12.

By Cramer’s rule, we obtain x1 = 1; x2 = 2; and x3 = 3.

Question 5.2 (12 points) Prove that each of the following determinants is zero:

1.
1 2 3 1+2 2+4 3+5 3 6 8

2 4 5 = 2 4 5 = 2 4 5 = 0.

3 6 8 3 6 8 3 6 8

2.
1 a b+c

1 b c+a ,

1 c a+b

21
where a, b, c are constants.

1
a b + c 1 a b + c − (a + b + c) 1 a −a



1
b c + a = 1 b c + a − (a + b + c) = 1 b −b

= 0.

1 c a + b 1 c a + b − (a + b + c) 1 c −c

3.
x − y x − y x2 − y 2

1 1 x+y ,

y 1 x
where x, y are constants.

x − y x − y x2 − y 2 x − y x − y (x + y)(x − y)

1 1 x + y = 1 1 x+y

y 1 x y 1 x

1 1 x+y

= (x − y) 1 1 x + y = 0
y 1 x

Question 5.3 (15 points) Calculate the following determinants. Please leave all
the details of the process of derivation for determinants.
1.
1 2 4

1 3 9

1 4 16

1 2 4
= 4 · (−1) 1 3 + 9 · (−1) 1 2 + 16 · (−1) 1 2
1+3
2+3
3+3

1 3 9
1 4 1 4 1 3
1 4 16
= 4(4 − 3) − 9(4 − 2) + 16(3 − 2) = 2.

2.

1 2 3 4

0 −1 0 11

2 −1 0 3
−2 0 −1 3

1 2 3 4
0 −1 11 1 2 4
0 −1 0 11
1+3

+ (−1) · (−1)4+3 0 −1 11
= 3 · (−1) 2 −1 3
2 −1 0 3
−2 0 3 2 −1 3
−2 0 −1 3

0 −1 11 1 2 4

= 3 2 −1 3 + 0 −1 11 ,
−2 0 3 2 −1 3

22
where

0 −1 11
2+1
−1 11 3+1
−1 11
2 −1 3 = 2 · (−1) 0 3 − 2 · (−1) −1 3


−2 0 3
= −2(−3 − 0) − 2(−3 + 11) = −10

1 2 4
1+1 −1 11 3+1 2 4

0 −1 11 = 1 · (−1) + 2 · (−1)

2 −1 3
−1 3 −1 11

= (−3 + 11) + 2(22 + 4) = 60.

Therefore, the determinant of the matrix is 30.

3.

2 1 3 3


3 2 1 6


1 3 0 9

2 4 1 12



2 1 3 3

2
3·1 3 3

2 3
3 3

= 1 3
3 2 1 6 3·2 1 6 1 3 6 1 6
= =0

1 3 0 9
3 1 3·3 0 9 3 1 9
0 9

2 4 1 12 2 3·4 1 12 2 12 1 12

Question 5.4 (12 points) Find the inverse of


 
−2 3 2
A =  6 0 3 .
4 1 −1

You are asked to leave all the details of the process of derivation for A−1 .
We first compute |A|:

3 2 + 3 · (−1)2+3 −2 3

|A| = 6 · (−1)2+1
1 −1 4 1
= −6(−3 − 2) − 3(−2 − 12) = 72.

23
Next we compute each cofactor of A:

1+1
0 3
C11 = (−1) = −3
1 −1

1+2
6 3
C12 = (−1) = 18
4 −1

1+3
6 0
C13 = (−1) =6
4 1

2+1
3 2
C21 = (−1) =5
1 −1

2+2
−2 2
C22 = (−1) = −6
4 −1

2+3
−2 3
C23 = (−1) = 14
4 1

3+1
3 2
C31 = (−1) =9
0 3

3+2
−2 2
C32 = (−1) = 18
6 3

−2 3
C33 = (−1)3+3 = −18
6 0

Thus, we obtain
     
C11 C21 C31 −3 5 9 −1/24 5/72 1/8
1  1 
A−1 = C12 C22 C32  = 18 −6 18  =  1/4 −1/12 1/4 
|A| 72
C13 C23 C33 6 14 −18 1/12 7/36 −1/4

Question 5.5 (8 points) Let A be a 2 × 2 matrix as follows:


 √ 
1 −1 − 3
A= √ .
2 3 −1

Answer the following questions.

1. Show that A3 = I.
 √  √   √ 
1 −1 − 3 −1 − 3 1 −1 3
2
A = √ √ = √
4 3 −1 3 −1 2 − 3 −1
 √  √   √ √ 
1 −1 − 3 −1 3 1 1 + 3 − 3 + 3
3
A = √ √ = √ √ = I.
4 3 −1 − 3 −1 4 − 3+ 3 3+1

24
2. Use the previous result to find A−1 .

Since A3 = I, we have A−1 = A−1 A3 = A2 . Thus,


 √ 
1 −1 3
−1
A = √ .
2 − 3 −1

Question 5.6 (5 points) Suppose that X is an m×n matrix and that |X T X| = 6 0.


Let A be defined as A = Im − X(X T X)−1 X T , where Im denotes the m × m identity
matrix. Show that A2 = A.

A2 = (Im − X(X T X)−1 X T )(Im − X(X T X)−1 X T )


= Im − Im X(X T X)−1 X T − X(X T X)−1 X T Im + X(X T X)−1 X T X(X T X)−1 X T
= Im − 2X(X T X)−1 X T + X(X T X)−1 X T
= Im − X(X T X)−1 X T = A.
Question 5.7 (10 points) Find a matrix X that satisfies AB + CX = D, where
 
  3 1    
−2 0 1 1 2 −9 3
A= ; B=  0 1 ; C=
 ; and D = .
1 −1 5 3 4 −8 17
−1 2
We first compute the following:
 
 3 1   
−2 0 1  −7 0
AB = 0 1 =
1 −1 5 −2 10
−1 2
     
−9 3 −7 0 −2 3
D − AB = − =
−8 17 −2 10 −6 7
   
−1 1 4 −2 −2 1
C = =
4 − 6 −3 1 3/2 −1/2

Since X = C −1 (D − AB), we have


      
−2 1 −2 3 4−6 −6 + 7 −2 1
X= = =
3/2 −1/2 −6 7 −3 + 3 9/2 − 7/2 0 1
Question 5.8 (12 points) Consider the following homogeneous system of equa-
tions:
ax + by + cz = 0
bx + cy + az = 0
cx + ay + bz = 0,

25
where a, b, c are constants. Prove that this system has a nontrivial solution if and
only if a3 + b3 + c3 − 3abc = 0.
We re-write the system of equations in matrix form:
    
a b c x 0
 b c a   y  =  0 .
c a b z 0
By the theorem we covered in the class, we know that the system of equations has
a nontrivial solution if and only if

a b c

b c a = 0.

c a b

So, the rest of the proof is reduced to the following computation:



a b c
c a b c
+ c(−1) b c
1+1 2+1 3+1

b c a = a(−1) + b(−1)

c a a b a b c a
b
= a(bc − a2 ) − b(b2 − ac) + c(ab − c2 ) = −a3 − b3 − c3 + 3abc = 0.
Thus, we obtain a3 + b3 + c3 − 3abc = 0.

Question 5.9 (4 points) Express the vector (4, −11) as a linear combination of
(2, −1) and (1, 4).
     
4 2 1
=3 −2 .
−11 −1 4

Question 5.10 (10 points) Find the equation for the plane going through the
points (1, 0, 2), (5, 2, 1), and (2, −1, 4).
Let a = (1, 0, 2), b = (5, 2, 1), and c = (2, −1, 4). We compute the following:
a − b = (−4, −2, 1)
b − c = (3, 3, −3)
We know that two vectors a − b and b − c are on the plane. Define p = (1, p2 , p3 )
as a vector that is orthogonal to both a − b and b − c. Then, we obtain p as follows:
p · (a − b) = −4 − 2p2 + p3 = 0
p · (b − c) = 3 + 3p2 − 3p3 = 0.
We therefore obtain p = (1, −3, −2). The equation is given as follows:
p · (x − a) = 0 ⇔ (x1 − 1) − 3(x2 − 0) − 2(x3 − 2) = 0 ⇔ x1 − 3x2 − 2x3 + 3 = 0.

26
6 Homework 6 (No Need for Submission)
Question 6.1 The two equations u2 v − u = x3 + 2y 3 and eux = vy define u and v
as differentiable functions of x and y around the point P = (x, y, u, v) = (0, 1, 2, 1).
Answer the following questions.

1. Find the differentials of u and v expressed in terms of the differentials of x


and y.

Let us denote each equation by the following implicit function:

g1 (u, v, x, y) = u2 v − u − x3 − 2y 3 = 0
g2 (u, v, x, y) = eux − vy = 0.

We consider the differentials of g1 and g2 :


∂g1 ∂g1 ∂g1 ∂g1
dg1 = du + dv + dx + dy = 0
∂u ∂v ∂x ∂y
⇔ (2uv − 1)du + u2 dv − 3x2 dx − 6y 2 dy = 0
⇔ (2uv − 1)du + u2 dv = 3x2 dx + 6y 2 dy
∂g2 ∂g2 ∂g2 ∂g2
dg2 = du + dv + dx + dy = 0
∂u ∂v ∂x ∂y
⇔ xeux du − ydv + ueux dx − vdy = 0
⇔ xeux du − ydv = −ueux dx + vdy.

2. Find ∂u/∂x and ∂v/∂x at P .

We re-write the system of equations derived in the previous question in matrix


form:     
2uv − 1 u2 du 3x2 dx + 6y 2 dy
= .
xeux −y dv −ueux dx + vdy
Setting dy = 0, we obtain
    
2uv − 1 u2 du 3x2 dx
=
xeux −y dv −ueux dx
    
2uv − 1 u2 ∂u/∂x 3x2
⇔ =
xeux −y ∂v/∂x −ueux

At P , the above representation is reduced to


    
3 4 ∂u/∂x 0
= .
0 −1 ∂v/∂x −2

27
So, we obtain
   −1    
∂u/∂x 3 4 0 −8/3
= = ,
∂v/∂x 0 −1 −2 2

where  −1  
3 4 1 1 4
= .
0 −1 3 0 −3

3. Find ∂u/∂y and ∂v/∂y at P .

We start from the matrix representation for the system of equations we de-
rived in the previous question. Setting dx = 0, we simplified the representa-
tion:
    2 
2uv − 1 u2 du 6y dy
ux =
xe −y dv vdy
 2
   2 
2uv − 1 u ∂u/∂y 6y
⇔ =
xeux −y ∂v/∂y v

At P , this representation is reduced to


    
3 4 ∂u/∂y 6
= .
0 −1 ∂v/∂y 1

So, we obtain
   −1    
∂u/∂y 3 4 6 10/3
= =
∂v/∂y 0 −1 1 −1

4. If x increases by 0.1 and y decreases by 0.2 from their values at P , what are
the approximate changes in u and v?

We denote by ∆u the approximate change in u; by ∆v the approximate


change in v; by ∆x the approximate change in x; and by ∆y the approximate
change in y. Then, we use the formula of the linear approximation:
∂u ∂u
∆u = ∆x + ∆y
∂x ∂y
∂v ∂v
∆v = ∆x + ∆y
∂x ∂y

28
We want to evaluate ∆u and ∆v at P and in so doing, we set ∆x = 0.1 and
∆y = −0.2. So,
∂u ∂u
∆u|(0,1,2,1) = · (1/10) + · (−1/5)
∂x (0,1,2,1) ∂y (0,1,2,1)

8 1 10 1 14
= − · − · =−
3 10 3 5 15
∂v ∂v
∆v|(0,1,2,1) = · (1/10) + · (−1/5)
∂x (0,1,2,1) ∂y (0,1,2,1)

1 1 2
= 2· +1· = .
10 5 5

Question 6.2 Show that f (x, y) = x3 + xy is “not” homogeneous of any degree.


Suppose on the contrary that there exists k ∈ R such that f (x, y) is homoge-
neous of degree k. If f (x, y) is homogeneous of degree k, we have

f (tx, ty) = tk f (x, y) ⇔ (tx)3 + (txty) = tk (x3 + xy) (∗)

for all t > 0 and (x, y) ∈ R2 . Since equation (∗) must hold for all (x, y), we set
(x, y) = (1, 1). In this case, we have

t3 + t2 = 2tk (∗∗)

for all t > 0. Since equation (∗∗) must hold for all t > 0, we set t = 2. Then,

23 + 22 = 2 · 2k ⇔ 2k = 6 ⇔ k = log2 6.

Similarly, since equation (∗∗) must hold for all t > 0, we set t = 4. Then,

43 + 42 = 2 · 4k ⇔ 4k = 40 ⇔ 22k = 40 ⇔ 2k = log2 40 ⇔ k = log2 40
√ √
Since 40 > 6, we have log2 40 > log2 6, which is the desired contradiction.
Hence, f (x, y) is not homogeneous of any degree.

Question 6.3 Let f (x, y) = xy/(x2 + y 2 ). Answer the following questions.

1. Show that the function f is homogeneous.

Fix t > 0. We compute the following:

(tx)(ty) t2 xy
f (tx, ty) = = = f (x, y).
(tx)2 + (ty)2 t2 (x2 + y 2 )

Therefore, f (·) is homogeneous of degree zero.

29
2. Explicitly check Euler theorem holds in this case.

We first compute the following:

0 y(x2 + y 2 ) − xy · 2x y(y 2 − x2 )
f1 (x, y) = =
(x2 + y 2 )2 (x2 + y 2 )2
2 2
0 x(x + y ) − xy · 2y x(x2 − y 2 )
f2 (x, y) = = 2
(x2 + y 2 )2 (x + y 2 )2

So,
0 0 xy(y 2 − x2 ) xy(x2 − y 2 )
xf1 (x, y) + yf2 (x, y) = + = 0.
(x2 + y 2 )2 (x2 + y 2 )2
0 0
This is equivalent to xf1 (x, y)+yf2 (x, y) = kf (x, y), where k = 0. Therefore,
Euler theorem holds in this case.

Question 6.4 The following system of equations defines x = x(u, v) and y =


y(u, v) as differentiable functions of u and v around the point P = (x, y, u, v) =
(1, 4, 1, −1).

y 2 + 2u2 + v 2 − xy = 15
2y 2 + u2 + v 2 + xy = 38.

Think of u and v as exogenous (i.e., independent) variables and x and y as en-


dogenous (i.e., dependent) variables. Answer the following questions.

1. Find the differentials of x and y expressed in terms of the differentials of u


and v.

Taking the differentials of the above system of equations, we obtain

2ydy + 4udu + 2vdv − xdy − ydx = 0


4ydy + 2udu + 2vdv + xdy + ydx = 0.

We summarize this system as follows:

ydx + (x − 2y)dy = 4udu + 2vdv


−ydx − (x + 4y)dy = 2udu + 2vdv.

We further express this system in matrix forms:


    
y x − 2y dx 4udu + 2vdv
= .
−y −(x + 4y) dy 2udu + 2vdv

30
2. Use Cramer’s rule to find ∂x/∂u and ∂y/∂u at P .

Plugging dv = 0 into the above system, we obtain


    
y x − 2y dx 4udu
= .
−y −(x + 4y) dy 2udu
This can be translated into the following in terms of partial derivatives:
    
y x − 2y ∂x/∂u 4u
= .
−y −(x + 4y) ∂y/∂u 2u

Evaluating the above at (x, y, u, v) = (1, 4, 1, −1), we simplify the system


into the following:
    
4 −7 ∂x/∂u 4
= .
−4 −17 ∂y/∂u 2
By Cramer’s rule, we obtain

4 −7

∂x 2 −17 −54 9
= = =
∂u 4 −7 −96 16

(1,4,1,−1)
−4 −17

4 4

∂y −4 2 24 1
= = =− .
∂u 4 −7 −96 4

(1,4,1,−1)
−4 −17

3. Use Cramer’s rule to find ∂x/∂v and ∂y/∂v at P .

Plugging du = 0 into the system, we obtain


    
y x − 2y dx 2vdv
= .
−y −(x + 4y) dy 2vdv
This can be translated into the following in terms of partial derivatives:
    
y x − 2y ∂x/∂u 2v
= .
−y −(x + 4y) ∂y/∂u 2v

Evaluating the above at (x, y, u, v) = (1, 4, 1, −1), we simplify the system


into the following:
    
4 −7 ∂x/∂v −2
= .
−4 −17 ∂y/∂v −2

31
By Cramer’s rule, we obtain

−2 −7

∂x −2 −17 20 5
= = = −
∂v −7 −96 24

4

(1,4,1,−1)
−4 −17

4
−2
∂y −4 −2 −16 1
= = = .
∂v −7 −96 6

(1,4,1,−1) 4
−4 −17

4. Compute the approximate values of x and y that correspond to u = 0.9 and


v = −1.1.

Using the linear approximation, we have the following formula:

∂x(u, v) ∂x(u, v)
x(u + du, v + dv) ≈ x(u, v) + du + dv
∂u ∂v
∂y(u, v) ∂y(u, v)
y(u + du, v + dv) ≈ y(u, v) + du + dv
∂u ∂v
Setting u = 1, v = −1, du = −0.1, and dv = −0.1 in the above approximation
forms, we obtain

x(0.9, −1.1) = x(1 − 0.1, −1 − 0.1)


∂x(1, −1) ∂x(1, −1)
≈ x(1, −1) + · (−0.1) + · (−0.1)
 ∂u
   ∂v
9 1 5 1
= 1+ · − − · −
16 10 24 10
9 1
= 1− + ≈ 1 − 0.05625 + 0.02083 = 0.96458
160 48
y(0.9, −1.1) = y(1 − 0.1, −1 − 0.1)
∂y(1, −1) ∂y(1, −1)
≈ y(1, −1) + · (−0.1) + · (−0.1)
 ∂u   ∂v
1 1 1 1
= 4− · − + · −
4 10 6 10
1
= 4+ ≈ 4.0083.
120

Question 6.5 We consider the function f (x, y) = y 3 +3x2 y. Answer the following
questions.

32
1. Determine the degree of homogeneity of f .

f (tx, ty) = (ty)3 + 3(tx)2 (ty) = t3 y 3 + 3t3 x2 y = t3 (y 3 + 3x2 y) = t3 f (x, y),

for all (x, y) ∈ R2 and t > 0. So, the function f is homogenous of degree 3.

2. Find the slope of the tangent line to the level curve f (x, y) = −13 at an
arbitrary point on the curve, and find in particular the equation of the tangent
line at the point (2, −1).

By the implicit function theorem, we have

0 dy ∂f (x, y)/∂x 6xy 2xy


y = =− =− 2 2
=− 2 ,
dx ∂f (x, y)/∂y 3y + 3x y + x2

where (x, y) 6= (0, 0). We can easily confirm that f (2, −1) = −13. When
(x, y) = (2, −1), we have

dy 2 · 2 · (−1) 4
=− = .
dx (x,y)=(2,−1) 1+4 5

So, the equation of the tangent line at (2, −1) is


4 4 13
y + 1 = (x − 2) ⇔ y = x − .
5 5 5

3. Examine whether the level curve f (x, y) = −13 is convex or concave around
the point (2, −1).
00
We compute y = d2 y/dx2 :
0 0
d2 y
00 (2y + 2xy )(x2 + y 2 ) − 2xy(2x + 2yy )
y = 2 =− .
dx (x2 + y 2 )2

At the point (2, −1), we have

00 (2(−1) + 2 · 2 · (4/5))5 − 2 · 2 · (−1)(2 · 2 + 2(−1) · (4/5))


y |(2,−1) = −
(4 + 1)2
−10 + 16 + 16 − 32/5
= −
25
78
= −
125
Hence, the level curve f (x, y) = −13 is concave around the point (2, −1).

33
4. Show that no point on the level curve f (x, y) = −13 lies above the x-axis.

We have
f (x, y) = y 3 + 3x2 y = −13.
Then, we manipulate the above equation as follows:
y(y 2 + 3x2 ) = −13.
We observe that f (0, 0) = 0 6= −13 so that (x, y) = (0, 0) is not on the
level curve f (x, y) = −13. So, we always have y 2 + 3x2 > 0. This further
implies have y < 0 in order to satisfy y(y 2 + 3x2 ) = −13. This completes the
argument.
Question 6.6 Consider the following maximization problem:
1
max f (x) = − x3 + 2x2 − 2x − 72,
x∈D 3
where D = {x| x ≥ 0} is the constraint set. Then, obtain the solution to this
maximization problem. You are required to obtain this answer analytically without
resorting to any graphical method.
0
We first check the first-order condition f (x) = 0:
0 √ √ √
f (x) = −x2 + 4x − 2 = 0 ⇒ −(x − (2 − 2))(x − (2 + 2)) = 0 ⇒ x = 2 ± 2.
√ √ √ √
Since 1 < 2 < 2, we have 2 − 2 > 0 so that both√ 2 − 2,√2 + 2 belong to the
interior of D. So, the two stationary points x = 2 − 2, 2 + 2 are the candidates
for the solution. We also know the following:
 √

 < 0 if 0 ≤ x < √
2− 2
 = 0 if x = √ 2− 2



0
f (x) > 0 if 2 − 2 < √ x < 2 + 2
= 0 if x = 2 + 2




 √
< 0 if x > 2 + 2
√ √
This implies that f (2 − 2) < f (2 + 2). Therefore,√if a maximum point of this
problem is an interior point of D, it must be x = 2 + 2. We compute
√ 4 √
f (2 + 2) = (1 + 2) − 72.
3
Since we only consider interior points of D thus far, the only other possible solution
candidate is x = 0. In this case, we have
f (0) = −72.
√ √
Since f (2 + 2) > f (0), x = 2 + 2 is the solution to the maximization problem.

34
Question 6.7 Let  
a11 a12 a13
A =  a21 a22 a23 
a31 a32 a33
be a 3 × 3 matrix where aij is a constant for each i, j = 1, 2, 3. Let |A| denote
the determinant of a matrix A and Aij denote the cofactor corresponding to (i, j)
component of the matrix A. Assume |A| 6= 0. A linear system of 3 equations and
3 unknowns is given as follows:

a11 x1 + a12 x2 + a13 x3 = b1


a21 x1 + a22 x2 + a23 x3 = b2 (∗)
a31 x1 + a32 x2 + a33 x3 = b3 ,

where bj is a constant for each j = 1, 2, 3. We begin by assuming that (c1 , c2 , c3 ) is


a solution to (∗) so that

a11 c1 + a12 c2 + a13 c3 = b1


a21 c1 + a22 c2 + a23 c3 = b2 (∗∗)
a31 c1 + a32 c2 + a33 c3 = b3 ,

Answer the following questions

1. Show that for each j, k = 1, 2, 3 with j 6= k,

a1k A1j + a2k A2j + a3k A3j = 0.

We prove this by dividing our argument into the following cases:

Case 1: (j, k) = (1, 2)

What we want to show is

a12 A11 + a22 A21 + a32 A31 = 0.

The left hand side of the above equation is obtained by the determinant of a
matrix which is the same as A except that we replace the first column of A
with its second column:
a12 a12 a13

a22 a22 a23 .

a32 a32 a33
Since this determinant contains two identical columns in it, it is 0. This
completes the argument for this case.

35
Case 2: (j, k) = (1, 3)

What we want to show is

a13 A11 + a23 A21 + a33 A31 = 0.

The left hand side of the above equation is obtained by the determinant of a
matrix which is the same as A except that we replace the first column of A
with its third column column:

a13 a12 a13

a23 a22 a23 .

a33 a32 a33

Since this determinant contains two identical columns in it, it is 0. This


completes the argument for this case.

Case 3: (j, k) = (2, 1)

What we want to show is

a11 A12 + a21 A22 + a31 A32 = 0.

The left hand side of the above equation is obtained by the determinant of a
matrix which is the same as A except that we replace the second column of
A with its first column column:

a11 a11 a13

a21 a21 a23 .

a31 a31 a33

Since this determinant contains two identical columns in it, it is 0. This


completes the argument for this case.

Case 4: (j, k) = (2, 3)

What we want to show is

a13 A12 + a23 A22 + a33 A32 = 0.

The left hand side of the above equation is obtained by the determinant of a
matrix which is the same as A except that we replace the second column of
A with its third column column:

a11 a13 a13

a21 a23 a23 .

a31 a33 a33

36
Since this determinant contains two identical columns in it, it is 0. This
completes the argument for this case.

Case 5: (j, k) = (3, 1)

What we want to show is

a11 A13 + a21 A23 + a31 A33 = 0.

The left hand side of the above equation is obtained by the determinant of a
matrix which is the same as A except that we replace the third column of A
with its first column column:

a11 a12 a11

a21 a22 a21 .

a31 a32 a31

Since this determinant contains two identical columns in it, it is 0. This


completes the argument for this case.

Case 6: (j, k) = (3, 2)

What we want to show is

a12 A13 + a22 A23 + a32 A33 = 0.

The left hand side of the above equation is obtained by the determinant of a
matrix which is the same as A except that we replace the third column of A
with its second column column:

a11 a12 a12

a21 a22 a22 .

a31 a32 a32

Since this determinant contains two identical columns in it, it is 0. This


completes the argument for this case.

The analysis in all the cases above completes the argument.

2. We multiply the first equation in (∗∗) by the cofactor A11 ; multiply the second
equation in (∗∗) by A21 ; and finally multiply the third equation in (∗∗) by
A31 . Then, we add all the equations so obtained. The result is

(a11 A11 + a21 A21 + a31 A31 )c1 + (a12 A11 + a22 A21 + a32 A31 )c2
+ (a13 A11 + a23 A21 + a33 A31 )c3 = b1 A11 + b2 A21 + b3 A31 .

37
Show that c1 = A1 /|A|, where

b1 a12 a13

A1 = b2 a22 a23 .

b3 a32 a33

We can simplify the equation obtained above into the following:


(a11 A11 + a21 A21 + a31 A31 )c1 = b1 A11 + b2 A21 + b3 A31 ,
because a12 A11 + a22 A21 + a32 A31 = 0 and a31 A11 + a32 A21 + a33 A31 = 0 by
the previous question. Thus, we obtain

b1 a12 a13

b2 a22 a23

b1 A11 + b2 A21 + b3 A31 b3 a32 a33 A1
c1 = = = .
a11 A11 + a21 A21 + a31 A31 |A| |A|
3. Show that c2 = A2 /|A| and c3 = A3 /|A|, where

a11 b1 a13 a11 a12 b1

A2 = a21 b2 a23 and A3 = a21 a22 b2 .
a31 b3 a33 a31 a32 b3
In doing so, you are required to provide the whole derivation process which
justify your answer.

We consider each case in turn:

Case I: c2 = A2 /|A|

We multiply the first equation in (∗∗) by the cofactor A12 ; multiply the second
equation in (∗∗) by A22 ; and finally multiply the third equation in (∗∗) by
A32 . Then, we add all the equations so obtained. The result is
(a11 A12 + a21 A22 + a31 A32 )c1 + (a12 A12 + a22 A22 + a32 A32 )c2
+ (a13 A12 + a23 A22 + a33 A32 )c3 = b1 A12 + b2 A22 + b3 A32 .
We can simplify the above equation into the following:
(a12 A12 + a22 A22 + a32 A32 )c2 = b1 A12 + b2 A22 + b3 A32 ,
because a11 A12 + a21 A22 + a31 A32 = 0 and a31 A12 + a32 A22 + a33 A32 = 0 by
the first question. Thus, we obtain

a11 b1 a13

a21 b2 a23

b1 A12 + b2 A22 + b3 A32 a31 b3 a33 A2
c2 = = = .
a12 A12 + a22 A22 + a32 A32 |A| |A|

38
Case II: c3 = A3 /|A|

We multiply the first equation in (∗∗) by the cofactor A13 ; multiply the second
equation in (∗∗) by A23 ; and finally multiply the third equation in (∗∗) by
A33 . Then, we add all the equations so obtained. The result is

(a11 A13 + a21 A23 + a31 A33 )c1 + (a12 A13 + a22 A23 + a32 A33 )c2
+ (a13 A13 + a23 A23 + a33 A33 )c3 = b1 A13 + b2 A23 + b3 A33 .

We can simplify the above equation into the following:

(a13 A13 + a23 A23 + a33 A33 )c3 = b1 A13 + b2 A23 + b3 A33 ,

because a11 A13 + a21 A23 + a31 A33 = 0 and a12 A13 + a22 A23 + a32 A33 = 0 by
the first question. Thus, we obtain

a11 a12 b1

a21 a22 b2

b1 A13 + b2 A23 + b3 A33 a31 a32 b3 A3
c3 = = = .
a13 A13 + a23 A23 + a33 A33 |A| |A|

Question 6.8 Let a and b be positive constants and α be a constant such that
0 < α < 1. Let f (x) = ln(αax + (1 − α)bx ). Answer the following questions.
0
1. Using the notion of limits, provide the precise definition of f (x).

0 f (x + h) − f (x) ln(αax+h + (1 − α)bx+h ) − ln(αax +(1 − α)bx )


f (x) = lim = lim
h→0 h h→0 h
0
2. Compute f (x).
0 αax ln a + (1 − α)bx ln b
f (x) =
αax + (1 − α)bx
0
3. Compute f (0).
0
f (0) = α ln a + (1 − α) ln b = ln aα b(1−α) .

4. Compute the following:


1
lim αah + (1 − α)bh h .

h→0

HINT: Use the following relation: X = eln X where X = (αah + (1 − α)bh )1/h .

39
We execute the following series of computation:
 1  1
lim ln αah + (1 − α)bh h = lim ln αah + (1 − α)bh

h→0 h→0 h
1
ln αah + (1 − α)bh − ln 1 (∵ ln 1 = 0)

= lim
h→0 h
1
ln αah + (1 − α)bh − ln αa0 + (1 − α)b0
 
= lim
h→0 h
0
= f (0)
= α ln a + (1 − α) ln b
= ln aα b(1−α) .
Since X = eln X , we obtain
1
  
1 limh→0 ln {αah +(1−α)bh } h
lim αah + (1 − α)b h h

= e
h→0
0
= ef (0)
α (1−α) )
= eln(a b = aα b(1−α) .

Question 6.9 Find local maximum/minimum points for the following functions:

1. y = (x + 2)e−x

We find the stationary points of this function:


0
y = e−x − (x + 2)e−x = −(x + 1)e−x = 0.
There is the unique stationary point x = −1. We next compute the second
derivative of y:
00
y = −e−x + (x + 1)e−x = xe−x .
00
At x = −1, we have y |x=−1 = −e < 0. Therefore, x = −1 is a local
maximum point.
2. y = ln x + 1/x

For this function to be well-defined, we must have x > 0. We find the


stationary points of this function:
0 1 1 1
y = − 2 = 2 (x − 1).
x x x
So, there is the unique stationary point x = 1. We next compute the second
derivative of y:
00 1 2 1
y = − 2 + 3 = 3 (2 − x).
x x x

40
00
At x = 1, we have y |x=1 = 1 > 0. Therefore, x = 1 is a local minimum
point.

3. y = x3 e−x .

We find the stationary points of this function:


0
y = 3x2 e−x − x3 e−x = x2 e−x (3 − x) = 0.

So, x = 0 and x = 3 are the stationary points. We next compute the second
derivative of y:
00
y = 2xe−x (3 − x) − x2 e−x (3 − x) − x2 e−x = x(x2 − 6x + 6)e−x .
00
At x = 3, we have y |x=3 = −9e−3 < 0. So, x = 3 is a local maximum point.
00
However, at x = 0, we have y |x=0 = 0. So, in the domain of x such that
x > 0, x = 0 is considered a local minimum point. On the other hand, in the
domain of x such that x < 0, x = 0 is considered a local maximum point.
Therefore, x = 0 is neither a local maximum nor minimum point.

Question 6.10 Find the values of x that maximizes/minimize the following func-
tions:

1. f (x) = ex + e−2x

We first compute

 < 0 if x < ln 2/3
0
f (x) = ex − 2e−2x −2x 3x
= e (e − 2) = 0 if x = ln 2/3
> 0 if x > ln 2/3

because e−2x > 0 for any x ∈ R. So, x = ln 2/3 is the minimum point of f .

2. f (x) = 9 − (x − a)2 − 2(x − b)2 , where a, b are constants.

We compute

 > 0 if x < (2a + 4b)/6
0
f (x) = −2(x − a) − 4(x − b) = −(6x − 2a − 4b) = 0 if x = (2a + 4b)/6
< 0 if x > (2a + 4b)/6

So, x = (2a + 4b)/6 is the maximum point of f .

41
3. f (x) = ln x − 5x, where x > 0.

We compute 
1  > 0 if 0 < x < 1/5
0
f (x) = − 5 = 0 if x = 1/5
x
< 0 if x > 1/5

So, x = 1/5 is the maximum point of f .

Question 6.11 Find the maximum and minimum of each function over the indi-
cated interval.

1. f (x) = −2x − 1 over [0, 3]

Since f (·) is strictly decreasing in x, we obtain x = 0 is the maximum point


of f over [0, 3] and x = 3 is the minimum point. Accordingly, f (0) = −1 is
the maximum of f and f (3) = −7 is the minimum over [0, 3].

2. f (x) = x3 − 3x + 8 over [−1, 2]

Since f (x) is continuous (because it is differentiable) and [−1, 2] is a closed,


bounded interval of the real line, by the extreme-value theorem, there always
exist the maximum and minimum value of f over [−1, 2]. We first find the
stationary points of f :
0
f (x) = 3x2 − 3 = 3(x2 − 1) = 3(x + 1)(x − 1) = 0.

Thus, x = 1, −1 are the stationary points. We then have f (−1) = −1+3+8 =


10 and f (1) = 1 − 3 + 8 = 6. Note that f (2) = 8 − 6 + 8 = 10. Therefore,
the maximum of f over [−1, 2] is 10 and the minimum of f over it is 6.

3. f (x) = (x2 + 1)/x over [1/2, 2]

First, f (x) is well-defined over [1/2, 2]. Second, [1/2, 2] is a closed, bounded
interval of the real line and f (x) is differentiable over (0, ∞), which implies
it is continuous over [1/2, 2]. By the extreme-value theorem, there always
exist the maximum and minimum value of f over [1/2, 2]. We first find the
stationary points of f :

0 2x · x − (x2 + 1) · 1 x2 − 1 (x + 1)(x − 1)
f (x) = = = = 0.
x2 x2 x2
So, x = −1, 1 are the stationary points but x = −1 is excluded from the
domain [1/2, 2]. We then have f (1) = 2. Note that f (1/2) = 5/2 and
f (2) = 5/2. So, the maximum of f is 5/2 and the minimum of f is 2.

42
Question 6.12 Consider a firm that produces a single commodity to maximize its
profit. The total revenue generated in a certain period by producing and selling
Q units is R(Q) = 10Q − Q2 /1000 dollars for Q ∈ [0, 10000], whereas C(Q) =
5000 + 2Q for Q ≥ 0 denotes the associated total dollar cost. Find the value of Q
that maximizes profit.
We first define the profit function π(Q) of the firm:

Q2
π(Q) = R(Q) − C(Q) = 10Q − − 2Q − 5000.
1000
Since [0, 10000] is a closed, bounded interval of the real line and π(Q) is a continuous
function (because it is differentiable), by the extreme-value theorem, there always
exist the maximum and minimum values of π(Q) over [0, 10000]. Second, we find
the stationary points of π:
0 Q
π (Q) = 10 − − 2 = 0 ⇔ Q = 4000
500
We then compute π(4000) = 10, 000. We also compute π(0) = −5000 and π(10000) =
−25000. Therefore, Q = 4000 is the profit maximizing quantitity.

Question 6.13 Find local maximum/minimum points and inflection points for the
following functions.

1. f (x) = (ln x)/x2

First, we find the stationary points of f :


1
0
x
· x2 − 2x ln x 1
f (x) = = (1 − 2 ln x) = 0.
x4 x3
00
So, x = e1/2 is the unique stationary point. Second, we compute f (x):

00 − x2 · x3 − 3x2 · (1 − 2 ln x) 1
f (x) = 6
= − 4 (5 − 6 ln x).
x x
Therefore, x = e5/6 is an inflection point of f . Moreover, we can check
 
00 1/2 1 1 2
f (e ) = − 2 5 − 6 · = − 2 < 0.
e 2 e

So, x = e1/2 is a local maximum point of f .

43
2. f (x) = e2x − 2ex

We first find the stationary points of f :


0
f (x) = 2e2x − 2ex = 2ex (ex − 1) = 0.

Because ex > 0 for any x, x = 0 is the unique stationary point of f . Second,


00
we compute f (x):
00
f (x) = 4e2x − 2ex = 2ex (2ex − 1).

Then, x = − ln 2 is an inflection point of f because ex > 0 for any x. Third,


we check
00
f (0) = 4 − 2 = 2 > 0.
So, x = 0 is a local minimum point of f .

3. f (x) = (x2 + 2x)e−x

We first compute the stationary points of f :


0 √ √
f (x) = (2x+2)e−x −(x2 +2x)e−x = −e−x (x2 −2) = −e−x (x+ 2)(x− 2) = 0.

Since e−x > 0 for any x, x = ± 2 are the stationary points. Second, we
00
compute f (x):
00 √ √
f (x) = e−x (x2 −2)−2xe−x = e−x (x2 −2x−2) = e−x (x−1− 3)(x−1+ 3).
−x
Once again,
√ since e > 0 for any x, we have two inflection points of f :
x = 1 ± 3. Third, we check the following:
00 √ √ √ √

f ( 2) = e− 2 (2 − 2 2 − 2) = −2 2e− 2 < 0
00 √ √ √ √ √
f (− 2) = e 2 (2 + 2 2 − 2) = 2 2e− 2 > 0.
√ √
Thus, x = 2 is a local maximum point of f and x = − 2 is a local minimum
point of f .

Question 6.14 Let f (x) = x2 /(x2 + 2). Answer the following questions.
0
1. Compute f (x) and determine where f (x) is increasing/decreasing.

2 2
2x(x + 2) − x · 2x 4x  > 0 if x > 0
0
f (x) = = = 0 if x = 0
(x2 + 2)2 (x2 + 2)2 
< 0 if x < 0

44
2. Find possible inflection points.
00
We compute f (x):

00 4(x2 + 2)2 − 4x · 2(x2 + 2) · 2x 4(x2 + 2)(2 − 3x2 )


f (x) = =
(x2 + 2)4 (x2 + 2)4
p p
−12(x + 2/3)(x − 2/3)
=
(x2 + 2)3
p
So, there are two inflection points of f : x = ± 2/3.

3. Determine the limit of f (x) as x → ±∞.


1
lim = lim =1
x→∞ x→∞ 1 + 2/x2
1
lim = lim =1
x→−∞ x→−∞ 1 + 2/x2

7 Homework 7 (Due Date: Oct 21 (Thu), 2021)


Question 7.1 (10 points) Investigate the definiteness of the following quadratic
forms.

1. Q(x1 , x2 ) = −x21 + 2x1 x2 − 6x22

We re-express Q(x1 , x2 ):

Q(x1 , x2 ) = −(x1 − x2 )2 − 5x22 .

It is easy to see that Q(x1 , x2 ) ≤ 0 for any (x1 , x2 ) and it never be the case
that Q(x1 , x2 ) = 0 unless (x1 , x2 ) = (0, 0). Therefore, Q(x1 , x2 ) is negative
definite.

2. Q(x1 , x2 ) = 4x21 + 2x1 x2 + 25x22

We re-express Q(x1 , x2 ):
 x2 2 99 2
Q(x1 , x2 ) = 4 x1 + + x2
4 4
It is easy to see that Q(x1 , x2 ) ≥ 0 for any (x1 , x2 ). Moreover, it never be
the case that Q(x1 , x2 ) = 0 unless (x1 , x2 ) = (0, 0). Therefore, Q(x1 , x2 ) is
positive definite.

45
Question 7.2 (10 points) Investigate the definiteness of the following symmetric
matrices.

1.  
2 3
A=
3 7

We compute the following:

a11 = 2 > 0
|A| = 2 · 7 − 3 · 3 = 14 − 9 = 5 > 0,

where    
a11 a12 2 3
A= = .
a21 a22 3 7
So, A is positive definite.

2.  
2 4
B=
4 7

We compute the following:

b11 = 2 > 0
|B| = 2 · 7 − 4 · 4 = −2 < 0,

where    
b11 b12 2 4
B= =
b21 b22 4 7
So, B is indefinite.

Question 7.3 (12 points) Annual profits for a firm are given by

π(x, y) = −x2 − y 2 + 22x + 18y − 102,

where x is the amount spend on research, and y is the amount spent on advertising.
Answer the following questions.

1. Find the profit when (x, y) = (10, 8).

π(10, 8) = −100 − 64 + 220 + 144 − 102 = 98

2. Find the profit when (x, y) = (12, 10).

π(12, 10) = −144 − 100 + 264 + 180 − 102 = 98

46
3. Find the only possible values of x and y that can maximize profits, and the
corresponding profit.

We rewrite the expression of π(x, y):

π(x, y) = −(x − 11)2 − (y − 9)2 + 100.

This clearly shows that (x, y) = (11, 9) maximizes its profit.

Question 7.4 (12 points) A firm produces two different kinds, A and B, of a
commodity. The daily cost of producing x units of A and y units of B is

C(x, y) = 2x2 − 4xy + 4y 2 − 40x − 20y + 514.

Suppose that the firm sells all its output at a price per unit of $24 for A and
$12 for B. In what follows, you can assume that there is a solution to the profit
maximization problem. Answer the following questions.

1. Find the daily production levels x and y that maximize profit.

Define π(x, y) as the profit of the firm:

π(x, y) = 24x + 12y − C(x, y) = −2x2 + 64x − 4y 2 + 32y + 4xy − 514.

We first find the stationary points of π:


0
π1 (x, y) = −4x + 64 + 4y = 0
0
π2 (x, y) = −8y + 32 + 4x = 0.

Solving the above system of equations, we obtain (x, y) = (40, 24) as the
unique stationary point. Second, we compute the profit achieved by (x, y) =
(40, 24):

π(40, 24) = −2(40)2 + 64 · 40 − 4 · (24)2 + 32 · 24 + 4 · 40 · 24 − 514 = 1150.

Third, we consider the maximized profit the firm can achieve when x = 0 or
y = 0, each of which corresponds to the boundary points.
Case 1: x = 0

In this case, we have

π(0, y) = −4y 2 + 32y − 514 = −4(y − 4)2 − 450.

Thus, the maximized profit in this case is -450.

47
Case 2: y = 0

In this case, we have

π(x, 0) = −2x2 + 64x − 514 = −2(x − 16)2 − 258.

Thus, the maximized profit in this case is -258.

The domain of the firm’s profit maximization problem is {(x, y) ∈ R2 | x ≥


0, y ≥ 0}. This domain is equivalent to

{(x, y) ∈ R2 | x > 0, y > 0} ∪ {(x, y) ∈ R2 | x ≥ 0, y = 0} ∪ {(x, y) ∈ R2 | x = 0, y ≥ 0} .


| {z } | {z }
interior points boundary points

Since we assume that there is a solution to the profit maximization prob-


lem, the unique stationary point (x, y) = (40, 24) is the profit maximizing
production plan.

2. The firm is required to produce exactly 54 units per day of the two kinds
combined. What is the optimal production plan now?

We now must satisfy the following equation: x + y = 54. Plugging y = 54 − x


into π(x, y), we obtain

π(x, 54 − x) = −2x2 + 64x − 4(54 − x)2 + 32(54 − x) + 4x(54 − x) − 514


= −10x2 + 680x − 10450
= −10(x − 34)2 + 1110.

This clearly shows that (x, y) = (34, 20) maximizes its profit, given the con-
straint that x + y = 54.

Question 7.5 (12 points) Consider the function f defined for all (x, y) by f (x, y) =
5 − x2 + 6x − 2y 2 + 8y. Answer the following questions.

1. Find its all partial derivatives of first and second order.


0
f1 (x, y) = −2x + 6
0
f2 (x, y) = −4y + 8
00
f11 (x, y) = −2
00
f22 (x, y) = −4
00 00
f12 (x, y) = f21 (x, y) = 0.

48
2. Find the only stationary point and classify it by using the second-derivative
test.
0 0
The stationary point can be computed by solving f1 (x, y) = 0 and f2 (x, y) =
0. So, (x, y) = (3, 2) is the unique stationary point of f . Next, we form the
associated Hessian matrix:
 
−2 0
H(x, y) = .
0 −4

We compute the following objects of H(x, y):


00
f11 (x, y) = −2 < 0
|H(x, y)| = 8 > 0.

This means that H(x, y) is negative definite for all (x, y). Therefore, the
stationary point (x, y) = (3, 2) is a local maximum point of f .

Question 7.6 (12 points) Consider a firm that sells a product in two isolated
geographical areas. If it wants to, it can charge different prices in the two different
areas because what is sold in one area cannot easily be resold in the other. Suppose
that this monopolist faces two inverse demand functions P1 (Q1 ) = 200 − 2Q1 and
P2 (Q2 ) = 180 − 4Q2 , and the cost function is C(Q1 , Q2 ) = 20(Q1 + Q2 ). Answer
the following questions.

1. How much should be sold in two markets to maximize total profit? What
are the corresponding prices?

Let π(Q1 , Q2 ) be the profit of the firm:

π(Q1 , Q2 ) = (200 − 2Q1 )Q1 + (180 − 4Q2 )Q2 − 20(Q1 + Q2 )


= −2Q21 + 180Q1 − 4Q22 + 160Q2
= −2(Q1 − 45)2 − 4(Q2 − 20)2 + 5650.

This clearly shows that (Q1 , Q2 ) = (45, 20) maximizes its total profit. The
maximized profit is 5650.

2. How much profit is lost if it becomes illegal to discriminate prices?

In this case, we must satisfy P (Q1 ) = P (Q2 ). That is,

200 − 2Q1 = 180 − 4Q2 ⇔ Q1 = 10 + 2Q2

49
Plugging Q1 = 10 + 2Q2 into π(Q1 , Q2 ), we have

π(10 + 2Q2 , Q2 ) = −2(10 + 2Q2 )(10 + 2Q2 − 90) − 4Q22 + 160Q2


= −12Q22 + 440Q2 + 1600
2
552

55
= −12 Q2 − + 1600 + 12 · 2 .
3 3
 2
55
= −12 Q2 − + 5633 + 1/3
3

So, (Q1 , Q2 ) = (140/3, 55/3) maximizes its total profit, given the constraint
that P1 = P2 . Besides, the lost payoff is

5650 − 5633 − 1/3 = 16 + 2/3.

Question 7.7 (16 points) Find the maximum and minimum points for the fol-
lowing problems.

1. f (x, y) = x3 + y 3 − 9xy + 27 subject to 0 ≤ x ≤ 4 and 0 ≤ y ≤ 4.

Define D as the domain of f :

D = (x, y) ∈ R2 | 0 ≤ x ≤ 4 and 0 ≤ y ≤ 4 .


We first claim that D is a bounded set. Let B = {(x, y) ∈ R2 | x2 + y 2 ≤ 36}.


Fix (x, y) ∈ D. Since 0 ≤ x ≤ 4 and 0 ≤ y ≤ 4, we have

x2 + y 2 ≤ 42 + 42 = 32 < 36

Thus, D ⊆ B so that D is a bounded set. We next claim D is a closed set.


Define

D1 = {(x, y) ∈ R2 | x ≥ 0}
D2 = {(x, y) ∈ R2 | x ≤ 4}
D3 = {(x, y) ∈ R2 | y ≥ 0}
D4 = {(x, y) ∈ R2 | y ≤ 4}

Since g 1 (x, y) = x and g 2 (x, y) = y are clearly continuous functions, we know


that D1 , D2 , D3 , and D4 are all closed. We can define D = D1 ∩D2 ∩D3 ∩D4 .
Since any intersection of finitely many closes is closed, we conclude that D
is closed. Besides, since f (x, y) is the sum of polynomials so that it is indeed

50
continuous. Therefore, by the extreme-value theorem, there exist both a
maximum point and a minimum point of f over D.
We compute the stationary points of f as the points satisfying the following
system of equations:
0
f1 (x, y) = 3x2 − 9y = 0
0
f2 (x, y) = 3y 2 − 9x = 0

Solving this system, we obtain


( 2 )
3 27
x(x3 −27) = 0 ⇔ x(x−3)(x2 +3x+9) = 0 ⇔ x(x−3) x+ + =0
2 4

So, we have x = 0, 3. Since y = x2 /3, we obtain (x, y) = (0, 0), (3, 3) as the
stationary points of f . We then compute

f (0, 0) = 27
f (3, 3) = 27 + 27 − 81 + 27 = 0.

In what follows, we compute the value of f over the boundary of D. Suppose


x = 0. Then, we have
f (0, y) = y 3 + 27,
which achieves the maximum value of 91 when y = 4 and the minimum value
of 27 when y = 0. Suppose y = 0. Then, we have

f (x, 0) = x3 + 27,

which achieves the maximum value of 91 when x = 4 and the minimum value
of 27 when x = 0.
Suppose y = 4. Then, we have

f (x, 4) = x3 − 36x + 91.

Since we have the following



√ √

0 2 ≤ 0 if 0 ∈ [0,√
2 3]
f1 (x, 4) = 3(x − 12) = 3(x + 2 3)(x − 2 3)
≥ 0 If x ∈ [2 3, 4]

This shows that f (x, 4) achieves the minimum value at x = 2 3 over [0, 4].
So, √ √
f (2 3, 4) = 91 − 48 3 > 7.

51
Suppose x = 4. Here the role of x and y is symmetric. Therefore, we also
obtain √ √
f (4, 2 3) = 91 − 48 3 > 7.
In sum, (0, 4) and (4, 0) achieves the maximum value of 91 and (3, 3) achieves
the minimum value of 0.
2. f (x, y) = x2 + 2y 2 − x subject to x2 + y 2 ≤ 1.

Define D as the domain of f :


D = {(x, y) ∈ R2 | x2 + y 2 ≤ 1}.
Clearly, D is closed and bounded so that it is a compact set. Since f (x, y)
is the sum of polynomials, it is also a continuous function. Then, by the
extreme-value theorem, f achieves the maximum and minimum values over
D.
First, we find the stationary points of f by solving the following system of
equations:
0
f1 (x, y) = 2x − 1 = 0
0
f2 (x, y) = 4y = 0.
So, we have (x, y) = (1/2, 0) as the unique stationary point of f . We also
compute f (1/2, 0) = −1/4.
Second, we consider the boundary of D. So, assume that x2 + y 2 = 1.
Plugging y 2 = 1 − x2 into f (x, y), we obtain
2


2 1 9
f (x, ± 1 − x2 ) = −x − x + 2 = − x + +
2 4
√ √ √
x = −1/2 and x2 +y 2 = 1 imply that y = ± 3/2. So, (x, y) = (−1/2, 3/2), (−1/2, − 3/2)
achieve the maximum over the boundary of D. Its maximum value is 9/4.
Besides, x = 1 achieves the minimum value of f over the boundary of D. So,
we compute f (1, 0) = 0.

In sum, (−1/2, ± 3/2) achieves the maximum value of 9/4 and (1/2, 0)
achieves the minimum value of −1/4.
Question 7.8 (16 points) A firm uses capital K, labor L, and land T to produce
Q units of a commodity, where
Q = K 2/3 + L1/2 + T 1/3 .
Suppose that the firm is paid a positive price p for each unit it produces, and
the positive prices it pays per unit of capital, labor, and land are r, w, and q,
respectively. Answer the following questions.

52
1. Express the firm’s profit as a function π of (K, L, T ).

π(K, L, T ) = p(K 2/3 + L1/2 + T 1/3 ) − rK − wL − qT.

2. Find the values of K, L, and T , as functions of the four prices, that maximizes
the firm’s profit. Here we assume that a maximum exists.

To maximize the firm’s profit, we must satisfy the following first-order con-
ditions:
 3
0 2 −1/3 ∗ 2p
πK = pK −r =0⇔K =
3 3r
0 1 −1/2 ∗
 p 2
πL = pL −w =0⇔L =
2 2w
 3/2
0 1 −2/3 p
πT = pT − q = 0 ⇔ T∗ = .
3 3q

3. Let Q∗ denote the optimal number of units produced and K ∗ the optimal
capital stock. Show that ∂Q∗ /∂r = −∂K ∗ /∂p.

We first compute Q∗ :
 2  1/2
∗ ∗ 2/3 ∗ 1/2 ∗ 1/3 2p p p
Q = (K ) + (L ) + (T ) = + +
3r 2w 3q

Second, we compute the following:

∂Q∗ ∂ 4p2 8p2


 
= = −
∂r ∂r 9r2 9r3
∂K ∗ 8p3 8p2
 

= = .
∂p ∂p 27r3 9r3

So, we confirm ∂Q∗ /∂r = −∂K ∗ /∂p.

8 Homework 8 (Due Date: Oct 28 (Thu), 2021)


Question 8.1 (25 points) Consider the following problem:

max xy subject to x + 3y = 24.


(x,y)∈R2

Answer the following questions.

53
1. Use Lagrangian method to find the unique solution candidate.

We first set up the Lagrangian:

L(x, y, λ) = xy − λ(x + 3y − 24).

The first-order conditions are


0
Lx = y − λ = 0
0
Ly = x − 3λ = 0
0
Lλ = x + 3y − 24 = 0

We then obtain (x, y, λ) = (12, 4, 4) as the unique stationary point.

2. Show whether the solution candidate found in previous question is indeed a


solution or not.

Since x + 3y = 24, we plug x = 24 − 3y into f (x, y) = xy, we obtain

f (24 − 3y, y) = y(24 − 3y) = −3y 2 + 24y = −3(y − 4)2 + 48

This clearly shows that f (24 − 3y, y) is maximized at (x, y) = (12, 4). There-
fore, the Lagrangian method indeed allows us to obtain the right solution.

Question 8.2 (25 points) Consider the following minimization problem with an
equality constraint:

min 2 x2 + y 2 subject to x + 2y = a,
(x,y)∈R

where a is a constant. Answer the following questions.

1. Solve this constrained minimization problem. Do not use the Lagrangian


method for this question.

Let f (x, y) = x2 + y 2 . Plugging x = a − 2y into f (x, y), we obtain


2
a2

2 2 2 2 2a
f (a − 2y, y) = (a − 2y) + y = 5y − 4ay + a = 5 y − + .
5 5

This implies that f (a−2y, y) is minimized at y = 2a/5. Therefore, (x∗ , y ∗ ) =


(a/5, 2a/5) is the solution to the problem.

54
2. Show that the Lagrangian method leads to the same solution you obtained
in the previous question.

We set up the Lagrangian:


L(x, y, λ) = x2 + y 2 − λ(x + 2y − a).
The FOCs are
0
Lx = 2x − λ = 0,
0
Ly = 2y − 2λ = 0,
0
Lλ = x + 2y − a = 0.
Solving the above system of equations, we obtain (x, y, λ) = (a/5, 2a/5, 2a/5).
This corresponds to the solution we obtained in the previous question.
3. Verify that the derivative of the optimal value of the objective function with
respect to a is equal to the Lagrange multiplier.

Using the first result, we first compute


f (a/5, 2a/5) = a2 /5.
So,
df (a/5, 2a/5) 2a
= ,
da 5
which is the same as the Lagrange multiplier which we obtained in the second
question.
4. Does the corresponding maximization problem with the same equality con-
straint have a solution? If so, solve it. If not, argue why.

We reproduce the following:


2
a2

2a
f (a − 2y, y) = 5 y − + .
5 5
Note that this expression already takes the equality constraint into account.
This implies that f (a−2y, y) → ∞ as y → ±∞. Therefore, the corresponding
maximization problem has no solution.

Question 8.3 (25 points) Consider the problem


max xy subject to x + y = 2.
(x,y)∈R2

Answer the following questions.

55
1. Solve the problem by transforming it into an unconstrained optmization prob-
lem with one variable.

Plugging y = 2 − x into xy, we obtain


xy = x(2 − x) = −x2 + 2x = −(x − 1)2 + 1.
Therefore, (x, y) = (1, 1) is the solution.
2. Show that the Lagrangian method leads to the same solution.

We set up the Lagrangian:


L(x, y, λ) = xy − λ(x + y − 2).
The FOCs are
0
Lx = y − λ = 0 ⇒ y = λ
0
Ly = x − λ = 0 ⇒ x = λ
0
Lλ = x + y − 2 = 0
From the first two conditions, we obtain x = y. Plugging this into the
third condition, we obtain (x, y, λ) = (1, 1, 1). Indeed, this is the answer we
obtained in the previous question.
3. Show that the unique solution you obtain does “not” maximize the La-
grangian with the associated Lagrangian multiplier. Does this matter? Pro-
vide your argument.

Let us compute the optimal value of the Lagrangian:


L(1, 1, 1) = 1 − 1 − 1 + 2 = 1.
Let us consider the following:
L(2, 2, 1) = 4 − 2 − 2 + 2 = 2 > 1 = L(1, 1, 1).
So, (1, 1) does not maximize the value of the Lagrangian when λ = 1. How-
ever, we claim that this entails no problem. This is because (x, y) = (2, 2)
does not satisfy the equality constraint. Specifically, x + y = 4 > 2, which is
simply impossible. This implies that (2, 2) is not a feasible point.

Question 8.4 (25 points) We revisit Question 8.2.


min x2 + y 2 subject to x + 2y = a,
(x,y)∈R2

where a is a constant. Answer the following questions.

56
1. Define the bordered Hessian matrix of this problem.

We first compute the second-order derivatives of L(x, y, 2a/5):


00
Lxx = 2
00
Lyy = 2
00 00
Lxy = Lyx = 0

Second, we compute the first derivatives of the constraint function g(x, y) =


x + 2y = a:
0 0
gx (x, y) = 1 and gy (x, y) = 2.
Finally, we form the bordered Hessian matrix:
 0 0   
0 gx gy 0 1 2
0 00 00
H̄ =  gx Lxx Lxy  =  1 2 0 
0 00 00
gy Lyx Lyy 2 0 2

2. Compute the determinant of bordered Hessian matrix defined in the previous


question.

2+1 1 2 3+1 1 2

|H̄| = 1·(−1) +2·(−1) 2 0 = −(2−0)+2(0−4) = −10 < 0.
0 2

3. Is each stationary point you obtained using the Lagrangian method in Ques-
tion 8.2 a local maximum or minimum point?

Since |H̄| < 0, we conclude that the unique stationary point we obtained in
Question 8.2 is a local minimum point.

9 Homework 9 (Due Date: Nov 4 (Thu), 2021)


Question 9.1 (25 points) Let f (x, y) = 3 + x3 − x2 − y 2 and D be the domain
of f which is given by points in the x − y-plane satisfying x2 + y 2 ≤ 1 and x ≥ 0.
We consider the problem:

max or (min(x,y)∈R2 ) f (x, y) subject to (x, y) ∈ D.


(x,y)∈R2

Answer the following questions.

57
1. Show that the domain D is compact.

Define

D1 = {(x, y) ∈ R2 | x2 + y 2 ≤ 1},
D2 = {(x, y) ∈ R2 | x ≥ 0}.

Then, we can define D = D1 ∩ D2 . Let g 1 (x, y) = x2 + y 2 and g 2 (x, y) = x.


Since g 1 (x, y) = x2 + y 2 is the sum of two polynomials, it is a continuous
function. We observe that g 2 (x, y) = x is a linear function of x so that it is
continuous. Therefore, D1 and D2 are closed. Since the intersection of two
closed sets is closed, D is also closed.
We now claim that D is bounded. Since D is a subset of D1 , our proof
reduces to showing that D1 is a bounded set. However, D1 is a closed ball
itself. So, we can easily conclude that D1 is bounded. Hence, D is both
closed and bounded so that it is compact.

2. Define the set of all interior points of D (we simply call it the interior of D)
and the set of all boundary points of D (we simply call it the boundary of
D).

The interior of D is defined as

Int(D) = (x, y) ∈ R2 | x2 + y 2 < 1 and x > 0 .




The boundary of D is defined as

∂D = D\Int(D)= {(x, y) ∈ R2 | x2 + y 2 = 1, x ≥ 0} ∪ {(x, y) ∈ R2 | x = 0, −1 ≤ y ≤ 1}.

3. Find the stationary points of the function f , and classify them.

We check the first-order conditions:


0
fx (x, y) = 3x2 − 2x = 0 ⇔ x(3x − 2) = 0
0
fy (x, y) = −2y = 0 ⇔ y = 0.

So, we have two stationary points (x, y) = (0, 0), (2/3, 0). We next consider
the second-order derivatives of f :
00
fxx (x, y) = 6x − 2
00 00
fxy (x, y) = fyx (x, y) = 0
00
fyy (x, y) = −2.

58
Then, we have the Hessian matrix H(x, y):
 
6x − 2 0
H(x, y) = .
0 −2
At (x, y) = (0, 0), we have
 
−2 0
H(0, 0) =
0 −2
Now, we can check that h11 (0, 0) = −2 < 0 and |H(0, 0)| = 4 > 0. Thus,
H(0, 0) is negative definite so that (0, 0) is a local maximum point. At
(x, y) = (2/3, 0), we have
 
2 0
H(2/3, 0) =
0 −2
Here we have |H2 (2/3, 0)| = −4 < 0. This implies that H(3/2, 0) is indefinite
so that (3/2, 0) is a saddle point.
4. Find the global extreme points of f in D.

Since f is differentiable, it is also continuous. We can easily check that the


constraint set D is a compact set. By the extreme value theorem, f achieves
both maximum and minimum in D. The only stationary point in the interior
of D is (3/2, 0). Thus, it only remains to check the extreme points in the
boundary of D.

Assume x = 0 and −1 ≤ y ≤ 1. Then, f (0, y) = 3 − y 2 , which achieves


the maximum at y = 0 and achieves the minimum at y = 1, −1. So, there
are three candidates for the extreme points along x = 0 and −1 ≤ y ≤ 1:
(x, y) = (0, 0), (0, 1), (0, −1).

Assume x2 + y 2 = 1 and 0 ≤ x ≤ 1. In this case, we have f (x, y) =


3 + x3 − x2 − 1 + x2 = 2 + x3 , which achieves the minimum of 2 at x = 0 and
the maximum of 3 at x = 1. Thus, along x2 + y 2 = 1 and 0 ≤ x ≤ 1, there
are three candidates for the extreme points: (x, y) = (1, 0), (0, −1), (0, 1).

Therefore, the only possible candidates for the extreme points are (x, y) =
(0, 0), (0, 1), (0, −1), (1, 0). We check the following:
f (0, 0) = 3
f (0, 1) = f (0, −1) = 2
f (1, 0) = 3
Hence, (0, 0) and (1, 0) are the global maximum points in D and (0, 1) and
(0, −1) are the global minimum points.

59
Question 9.2 (20 points) Let f (x, y) = 24x − x2 + 16y − 2y 2 and g(x, y) =
x2 + 2y 2 . Consider the problem
max f (x, y) subject to g(x, y) = 44.
(x,y)∈R2

Answer the following questions.


1. Solve this problem.

We set up the Lagrangian:


L(x, y, λ) = 24x − x2 + 16y − 2y 2 − λ(x2 + 2y 2 − 44).
The FOCs are:
0 12
Lx = 24 − 2x − 2λx = 0 ⇒ x =
λ+1
0 4
Ly = 16 − 4y − 4λy = 0 ⇒ y =
λ+1
0 2 2
Lλ = x + 2y − 44 = 0.
Form the first two conditions, we obtain x = 3y. Plugging this into the third
condition, we obtain y = ±2. Therefore, we have two solution candidates:
(x, y, λ) = (6, 2, 1) and (−6, −2, 1). We compute the following:
f (6, 2) = 24 · 6 − 62 + 16 · 2 − 2 · 22 = 132
f (−6, −2) = 24 · (−6) − (−6)2 + 16 · (−2) − 2 · (−2)2
= −144 − 36 − 32 − 8 = −220.
So, we narrow down the solution candidates into (x, y, λ) = (6, 2, 1). We
compute the second-order derivatives of the Lagrangian.
00
Lxx = −2 − 2λ
00
Lyy = −4 − 4λ
00 00
Lxy = Lyx = 0.
We form the Hessian matrix H associated with the Lagrangian at λ = 1:
 
−4 0
H= .
0 −8
We compute the following from the Hessian matrix H:
00
L11 = −4 < 0
|H| = 32 > 0.
Therefore, H is negative definite, i.e., negative semidefinite for all (x, y) when
λ = 1. Hence, (6, 2) is the solution to the corresponding maximization prob-
lem.

60
2. What is the approximate change in the optimal value of f (x, y) if 44 is
changed into 45?

We know that the Lagrangian multiplier corresponds to the rate of change in


the optimal value of the objective function as the constraint is relaxed. So,
the approximate change in the optimal value is λ = 1.

Question 9.3 (15 points) Each week an individual consumes quantities x and y
of two goods, and works for l hours. These three quantities are chosen to maximize
the utility function
U (x, y, l) = α ln x + β ln y + (1 − α − β) ln(L − l),
which is defined for 0 ≤ l < L and for x, y > 0. Here α and β are positive
parameters satisfying α + β < 1. The individual faces the budget constraint px +
qy = wl, where p is the unit price of good x; q is the unit price of good y; and w is
the wage per hour. Define γ = (α + β)/(1 − α − β). Find the individual’s demands
x∗ , y ∗ , and labour supply l∗ as functions of p, q, and w. Assume that the solution
to this constrained optimization problem exists.
We set up the Lagrangian:
L(x, y, l, λ) = α ln x + β ln y + (1 − α − β) ln(L − l) − λ(px + qy − wl).
The FOCs are:
0 α
Lx = − λp = 0
x
0 β
Ly = − λq = 0
y
0 1−α−β
Ll = − + λw = 0
L−l
0
Lλ = px + qy − wl = 0
From the first two conditions, we have
βp
y= x.
αq
From the first condition, we have
α
λ= .
px
From the third condition, we have
1−α−β
l =L−
λw

61
Plugging λ = α/(px) into the above equation, we obtain
(1 − α − β)px
l =L−
αw
So far, we express y and l in terms of functions of x. So, plugging these expressions
into the fourth condition, we have
βp (1 − α − β)px
px + q x = wL − .
αq α
Solving the above equation, we have
αwL
x =
p
βwL
y =
q
l = (α + β)L.

Let g(x, y, l) = px + qy − wl. We compute

∇g(x, y, l) = (p, q, −w) .

This implies that for any feasible point (x, y, l) satisfying px + qy = wl, we have
∇g(x, y, l) 6= (0, 0, 0). Thus, the constraint qualification holds. Furthermore, since
the solution to the given constrained optimization problem is assumed to exist,
the unique stationary point we found is indeed the solution to the constrained
maximization problem. That is, the solutions are:
αwL
x∗ (p, q, w) =
p
βwL
y ∗ (p, q, w) =
q

l (p, q, w) = (α + β)L.

Question 9.4 (25 points) Consider the problem:

max f (x, y, z) = x + 2y + ln(1 + z) subject to x2 + y 2 − az = 0,


(x,y,z)∈R3

where a is a constant. Answer the following questions.

1. Write down the necessary conditions for this problem.

We set up the Lagrangian function:

L(x, y, z) = x + 2y + ln(1 + z) − λ(x2 + y 2 − az).

62
The necessary conditions for this problem are summarized as:
0
(1) Lx = 1 − 2λx = 0
0
(2) Ly = 2 − 2λy = 0
0 1
(3) Lz = + aλ = 0
1+z
(4) x2 + y 2 − az = 0.

2. Show that this problem has no solution when a = 0.

From (4), we have x2 + y 2 = 0. This implies that (x, y) = (0, 0). In this
case, the objective function becomes f (x, y, z) = ln(1 + z). Since f is strictly
increasing, f cannot achieve the maximum.
3. Show that this problem has no solution when a = 1.

Plugging a = 1 into (3), we obtain λ = −1/(1 + z) 6= 0. Since λ 6= 0,


we have x = 1/2λ and y = 1/λ from (1) and (2), respectively. Besides,
λ = 1/(1 + z) implies z = −(1 + λ)/λ. When a = 1, plugging (x, y, z) =
(1/2λ, 1/λ, (1 − λ)/λ) into (4), we obtain
 2
1 1 1+λ 2 1
+ + = 0 ⇔ 4λ + 4λ + 5 ⇔ 4 λ + + 4 = 0.
4λ2 λ2 λ 2
Clearly, this equation has no solutions in the real number. Therefore, the
solution does not exist for a = 1.
4. Assume that the necessary conditions for this problem (which you are sup-
posed to derive in part 1 of this question) suffice to guarantee the existence
of solutions to this problem. Then, find the range of values of a for which
there is a solution to this problem.

In the previous questions, we can exclude the case of a = 0 for which there
is no solution to the problem. From (3), we have λ = −1/a(1 + z) 6= 0. This
is equivalent to z = −(1 + λa)/(λa). From (1) and (2), we have x = 1/(2λ)
and y = 1/λ. Plugging (x, y, z) = (1/(2λ), 1/λ, −(1 + λa)/(λa)) into (4), we
obtain
1 1 1 + λa
2
+ 2 +a = 0 ⇔ 4aλ2 + 4λ + 5 = 0.
4λ λ λa
In order for this equation to have real roots, we must have 4 − 20a ≥ 0, which
implies that a ≤ 1/5. Indeed, this explains why we have no solution when
a = 1. Let g(x, y, z) = x2 + y 2 − az. Define

∇g(x, y, z) = (2x, 2y, −a).

63
This means that the constraint qualification is satisfied as long as a 6= 0.
Therefore, the relevant range of values of a is {a ∈ R| a < 0 or 0 < a ≤ 1/5}.

Question 9.5 (15 points) Consider the following constrained maximization prob-
lem:
max −x2 − y 2 subject to 3y − x ≥ 10
x,y

Solve this problem.


We set up the Lagrangian.

L(x, y, λ) = −x2 − y 2 − λ(x − 3y + 10).

The Kuhn-Tucker conditions are

Lx = −2x − λ = 0
Ly = −2y + 3λ = 0
λ ≥ 0; x − 3y + 10 ≤ 0; λ[x − 3y + 10] = 0.

From the first two conditions above, we obtain x = −λ/2 and y = 3λ/2. We divide
our argument into the following two cases.
Case 1: λ > 0
From the complementary slackness condition, we must have x − 3y + 10 = 0.
Plugging x = −λ/2 and y = 3λ/2 into this, we have λ = 2. Therefore, (x, y, λ) =
(−1, 3, 2) is a solution candidate.
Case 2: λ = 0
In this case, we have x = y = 0. In this case, however, we have x−3y +10 = 10,
which contradicts the inequality constraint x − 3y + 10 ≤ 0. Thus, we ignore this
case.
We compute the second-order derivatives of the Lagrangian:
00
Lxx = −2
00
Lyy = −2
00 00
Lxy = Lyx = 0.

Hence, the associated Hessian matrix is


 
−2 0
H= .
0 −2

64
We compute the following objects of H.

h11 = −2 < 0
|H| = 4 > 0.

Thus, H is negative definite so that the Lagrangian is concave. Therefore, the


unique solution candidate (x, y) = (−1, 3) is indeed a solution to the problem.

10 Homework 10 (Due Date: Nov 11, 2021)


Question 10.1 (25 points) Consider the following maximization problem with
inequality constraints:

max x + y − ex − ex+y subject to x + y ≥ 4, x ≥ −1, and y ≥ 1


x,y

Answer the following questions.

1. Formally define D as the set of all points (x, y) satisfying all the inequality
constraints given.

Define the following:

D1 = (x, y) ∈ R2 | y ≥ −x + 4


D2 = (x, y) ∈ R2 | x ≥ −1


D3 = (x, y) ∈ R2 | y ≥ 1 .


We can thus define D = D1 ∩ D2 ∩ D3 .

2. Find all points (x, y) that satisfy the Kuhn-Tucker conditions.

We set up the Lagrangian.

L(x, y, λ1 , λ2 , λ3 ) = x+y−ex −ex+y −λ1 (−x−y+4)−λ2 (−x−1)−λ3 (−y+1).

The Kuhn-Tucker conditions are


0
(1) Lx = 1 − ex − ex+y + λ1 + λ2 = 0
0
(2) Ly = 1 − ex+y + λ1 + λ3 = 0
(3) λ1 ≥ 0; x + y ≥ 4; λ1 (−x − y + 4) = 0
(4) λ2 ≥ 0; x ≥ −1; λ2 (−x − 1) = 0
(5) λ3 ≥ 0; y ≥ 1; λ3 (−y + 1) = 0.

We divide our argument into the following cases.

65
Case 1: λ2 > 0 and λ3 > 0

From (4) and (5), we obtain x = −1 and y = 1. In this case, however, we


have x + y = 0, which contradicts the inequality x + y ≥ 4 in (3). So, we
ignore this case.
Case 2: λ2 > 0 and λ3 = 0

From (4), we have x = −1. Plugging x = −1 and λ3 = 0 into (1) and (2),
we obtain
0
(1 ) 1 − e−1 − ey−1 + λ1 + λ2 = 0
0
(2 ) 1 − ey−1 + λ1 = 0
0 0
From (1 ) and (2 ), we obtain λ2 = e−1 and λ1 = ey−1 − 1. If λ1 > 0, by (3),
we must have y = 5. So, (x, y, λ1 , λ2 , λ3 ) = (−1, 5, e4 − 1, e−1 , 0) is a solution
candidate. If λ1 = 0, we must have y = 1. In this case, however, we have
x + y = −1 + 1 = 0, which contradicts the inequality x + y ≥ 4 in (3). So,
we ignore the case for λ1 = 0.
Case 3: λ2 = 0 and λ3 > 0

From (5), we obtain y = 1. Plugging λ2 = 0 and y = 1 into (1) and (2), we


obtain
00
(1 ) 1 − ex − ex+1 + λ1 = 0
00
(2 ) 1 − ex+1 + λ1 + λ3 = 0
00 00
From (1 ) and (2 ), we obtain λ3 = −ex < 0, which contradicts our hypoth-
esis that λ3 > 0. So, we ignore this case.

Case 4: λ2 = 0 and λ3 = 0

Plugging λ2 = λ = 0 into (1) and (2), we obtain


000
(1 ) 1 − ex − ex+1 + λ1 = 0
000
(2 ) 1 − ex+1 + λ1 = 0
000 000
From (1 ) and (2 ), we obtain −ex = 0, which is impossible. So, we ignore
this case.

In sum, we have the unique solution candidate (x, y, λ1 , λ2 , λ3 ) = (−1, 5, e4 −


1, e−1 , 0).

66
3. Find the solution to this constrained maximization problem.

We compute the second-order derivatives of the Lagrangian.


00
Lxx = −ex − ex+y
00
Lyy = −ex+y
00 00
Lxy = Lyx = −ex+y .

Second, we form the Hessian matrix associated with the Lagrangian:


   
h11 (x, y) h12 (x, y) −ex − ex+y −ex+y
H(x, y) = = .
h21 (x, y) h22 (x, y) −ex+y −ex+y

Third, we compute the leading principal minors of H(x, y):

h11 (x, y) = −ex − ex+y < 0


|H(x, y)| = ex+y (ex + ex+y ) − ex+y ex+y = ex (ex+y ) > 0.

Therefore, the Hessian matrix H(x, y) is negative definite. This implies that
the Lagrangian function is concave. So, we conclude that (x, y, λ1 , λ2 , λ3 ) =
(−1, 5, e4 − 1, e−1 , 0) is the solution to the problem.

Question 10.2 (25 points) Consider the following maximization problem with
inequality constraints:

max x + ay subject to x2 + y 2 ≤ 1 and x + y ≥ 0,


x,y

where a is a constant. Answer the following questions.

1. Define D as the set of points (x, y) satisfying the inequality constraints.

We define the following:

D1 = (x, y) ∈ R2 | x2 + y 2 ≤ 1


D2 = (x, y) ∈ R2 | y ≥ −x .


Then, we can define D = D1 ∩ D2 .

2. Write down the Kuhn-Tucker conditions for this problem.

We set up the Lagrangian:

L(x, y, λ1 , λ2 ) = x + ay − λ1 (x2 + y 2 − 1) − λ2 (−x − y).

67
The Kuhn-Tucker conditions are
0
(1) Lx = 1 − 2λ1 x + λ2 = 0
0
(2) Ly = a − 2λ1 y + λ2 = 0
(3) λ1 ≥ 0; x2 + y 2 ≤ 1; λ1 (x2 + y 2 − 1) = 0
(4) λ2 ≥ 0; x + y ≥ 0; λ2 (−x − y) = 0.

3. Find the solution candidates satisfying the Kuhn-Tucker conditions for all
possible values of the constant a.

First we exhaust all the possible solution candidates by solving the necessary
conditions derived in the previous question.

Case 1: λ1 > 0 and λ2 > 0

From (3) and (4), we must satisfy x2 + 2


√y = 1√ and y = −x.
√ This
√ implies that
there are two √ (x, y) = (1/ 2, −1/ 2), (−1/ 2, 1/ 2). Plugging
√ solutions:
(x, y) = (1/ 2, −1/ 2) into (1) and (2), we obtain
1−a
λ1 = √
2 2
−(a + 1)
λ2 =
2
In order that λ1 > 0 and λ2 > 0, we must satisfy a < −1.
√ √
Plugging (x, y) = (−1/ 2, 1/ 2) into (1) and (2), we obtain
a−1
λ1 = √
2 2
−(a + 1)
λ2 =
2
It is easy to see that there is no value
√ of a √ such that λ1 > 0 and λ2 > 0. So,
when a < −1, we have (x, y) = (1/ 2, −1/ 2) as a solution candidate.
Case 2: λ1 > 0 and λ2 = 0

Plugging λ2 = 0 into (1) and (2), we obtain


0
(1 ) 1 − 2λ1 x = 0
0
(2 ) a − 2λ1 y = 0.
0 0
From (1 ) and (2 ), we have
1−a
λ1 =
2(x − y)

68
In order for λ1 to be well-defined, we must have x = 6 y and a 6= 1. So,
plugging λ1 = (1 − a)/2(x − y)) into (1), we obtain y = ax.
From (3), we have x2 + y 2 = 1. So, plugging y = ax into x2 + y 2 = 1, we have
1
x 2 + a2 x 2 = 1 ⇒ x = ± √ .
1 + a2
Then, there are two possible solutions (x, y):
 
1 a
(x, y) = √ ,√
1 + a2 1 + a2
 
1 a
(x, y) = −√ , −√
1 + a2 1 + a2
√ √
Plugging (x, y) = (1/ 1 + a2 , a/ 1 + a2 ) into the expression for λ1 , we ob-
tain √
1 + a2
λ1 = > 0,
2
which is consistent with our hypothesis. We also need to satisfy x + y ≥ 0
from (4). So, we √ √ that√a ≥ −1. Therefore, when a ≥ −1,
need to assume
(x, y, λ1 , λ2 ) = (1/ 1 + a , a/ 1 + a2 , 1 + a2 /2, 0) is a solution candidate.
2
√ √
Plugging (x, y) = (−1/ 1 + a2 , −a/ 1 + a2 ) into the expression for λ1 , we
obtain √
1 + a2
λ1 = − < 0,
2
which contradicts our hypothesis that λ1 > 0. So, we ignore this as a solution.
Case 3: λ1 = 0 and λ2 > 0

Plugging λ1 = 0 into (1) and (2), we obtain


00
(1 ) 1 + λ2 = 0
00
(2 ) a + λ2 = 0.

These equations are consistent with each other only when a = 1. However,
in this case, we have λ2 = −1, which contradicts our hypothesis that λ2 > 0.
So, we ignore this case.
Case 4: λ1 = 0 and λ2 = 0

Plugging λ1 = λ2 = 0 into (1), we obtain 1 = 0, which is impossible. So, we


ignore this case.

69
In sum, we have the unique solution candidate which depend on the value of
a:  √ √
√(1/ 2, −1/√ 2) if a < −1
(x, y) = 2 2
(1/ 1 + a , a/ 1 + a ) if 1 > a ≥ −1 or a > 1

4. Find the solution to this problem.

Let g 1 (x, y) = x2 + y 2 − 1 and g 2 (x, y) = −x − y. We also write ∇g 1 (x, y) =


(2x, 2y) and ∇g 2 (x, y) = (−1, −1) as the gradient vector of g 1 and g 2 , respec-
tively. We shall show that the constraint qualification holds by considering
the following four cases:

(a) Both g 1 and g 2 are not binding.


In this case, the constraint qualification holds trivially.
(b) g 1 is binding, while g 2 is not binding.

In this case, we have ∇g 1 (x, y) = (2x, 2y) = (2x, ±2 1 − x2 ), which
cannot be equal to (0, 0) regardless of the value of x. So, the constraint
qualification holds.
(c) g 2 is binding, while g 1 is not binding.
In this case, we have ∇g 2 (x, y) = (−1, −1) 6= (0, 0). So, the constraint
qualification holds.
(d) Both g 1 and g 2 are binding.
Suppose, by way of contradiction, that the constrain qualification fails,
i.e., ∇g 1 (x, y) and ∇g 2 (x, y) are parallel to each other. This implies that
(2x, 2y) = (−α, −α) for some α 6= 0 so that x = y = −α/2. However,
x + y = −α 6= 0 because α 6= 0. This contradicts the hypothesis that g 2
is binding. Therefore, in this case, the constraint qualification holds.

By the necessity of the Kuhn-Tucker conditions, it only remains to make


sure that the constrained optimization problem has a solution so that the
solution candidate we identify is indeed the solution. First, the objective
function f (x, y) = x + ay is a linear function so that it is clearly continuous.
Since g 1 (x, y) = x2 +y 2 −1 and g 2 (x, y) = −x−y are the sums of polynomials,
D1 and D2 are closed. Further, since D = D1 ∩ D2 and any intersection of
finitely many closed sets is closed, D is closed. We also claim that D is
bounded. This suffices to show that D1 is bounded, as D ⊆ D1 . Define
B = {(x, y)| x2 + y 2 ≤ 1}. Fix (x, y) ∈ D1 . Then,

x2 + y 2 ≤ 1.

70
Thus, we obtain (x, y) ∈ B. So, we have D1 ⊆ B so that D1 is bounded.
This implies that D is a compact set. By the extreme-value theorem, the
constrained optimization problem has a solution.
Question 10.3 (25 points) Consider the following problem:
max f (x, y) = x2 + 2y subject to x2 + y 2 ≤ m, y ≥ 0,
(x,y)∈R2

where m is a positive constant. Answer the following questions.


1. Write down the Kuhn-Tucker conditions.

We set up the Lagrangian:


L(x, y) = x2 + 2y − λ1 (x2 + y 2 − m) − λ2 (−y).
The Kuhn-Tucker conditions are:
0
(1) L1 = 2x − 2λ1 x = 0
0
(2) L2 = 2 − 2λ1 y + λ2 = 0
(3) λ1 ≥ 0; x2 + y 2 ≤ m; and λ1 (x2 + y 2 − m) = 0
(4) λ2 ≥ 0; y ≥ 0; and λ2 (−y) = 0.

2. Solve this constrained maximization problem “for each possible value of m.”

We divide our argument into the following cases:

Case 1: x2 + y 2 = m and y = 0

Plugging y = 0 into (2), we obtain λ2 = −2, which contradicts (4). So, we


ignore this case.
Case 2: x2 + y 2 = m and y > 0

Due to the complementary slackness condition of (4), we have λ2 = 0. Plug-


ging λ2 = 0 into (2), we obtain λ1 = 1/y. Furthermore, plugging λ1 = 1/y
into (1), we obtain x(y − 1) = 0. This implies that there are three cases:
x = 0, y = 1, or both.

= 0, then x2 + y 2 =√m yields y = ± m. Since y > 0, we must have
If x √
y =√ m. Then,
√ λ1 = 1/ m and λ2 = 0. So, we have found (x, y, λ1 , λ2 ) =
(0, m, 1/ m, 0) as a candidate for the solution.

If y = 1, λ1 = 1. Plugging y = 1 into x2 + y 2 = m, we obtain√x = ± m − 1
provided√ m ≥ 1. So, when m ≥ 1, we have (x, y, λ1 , λ2 ) = ( m − 1, 1, 1, 0)
and (− m − 1, 1, 1, 0) as the candidates for the solution.

71
Case 3: x2 + y 2 < m and y = 0

Plugging y = 0 into (2), we obtain λ2 = −2, which contradicts λ2 ≥ 0 as


part of (4). So, we ignore this case.
Case 4: x2 + y 2 < m and y > 0

By (3) and (4), we have λ1 = λ2 = 0. Plugging these into (2), we obtain


2 = 0, which is clearly a contradiction. So, we ignore this case.

When
√ 0 < m < 1, we have√the only√ one candidate for the solution: (x, y) =
(0, m). In this case, f (0, √ m) = 2√ m. When√m ≥ 1, we have three
√ candi-
√ f ( m − 1, 1) = f√
dates for the solution: f (0, m) = 2 m and (− m − 1, 1) =
m − 1 + 2 = m + 1. Note that m + 1 ≥ 2 m if and only if ( m − 1)2 ≥ 0,
which is obviously satisfied.

Consider the case where 0 < m < 1. In this case, we have λ1 = 1/ m and
λ2 = 0. So, the corresponding Lagrangian function is given as

y2
 
2 1 2 2 1 5
L(x, y) = x +2y− √ (x +y −5)−0·(−y) = 1 − √ x2 − √ +2y+ √ .
m m m m
We derive the Hessian matrix H of the Lagrangian function:
!
1 − √1m
 
h11 (x, y) h12 (x, y) 0
H(x, y) = =
h21 (x, y) h22 (x, y) 0 − √1m

Since 0 < m < 1, we have 1 − 1/ m < 0, which implies that |h11 (x, y)| < 0
and |H(x, y)|
√ > 0. Therefore, the Lagrangian function is concave so that
(x, y) = (0, m) is the solution to the problem when 0 < m < 1.

Now consider the case where m ≥ 1. In this case, λ1 = 1 and λ2 = 0. So,


the Lagrangian function is given as

L(x, y) = x2 + 2y − (x2 + y 2 − m) − 0 · (−y) = −y 2 + 2y + m.

Then, the Lagrangian function becomes a function of y only and it is clearly


concave. Therefore,
√ when m ≥ 1, we have two solutions to the problem:
(x, y) = (± m − 1, 1).

Question 10.4 (25 points) Consider the problem:

max f (x, y) = xy + x2 subject to g 1 (x, y) = x2 + y ≤ 2, g 2 (x, y) = −y ≤ −1.

Answer the following questions.

72
1. Write down the Kuhn-Tucker Conditions for this problem.

We set up the Lagrangian function:


L(x, y) = xy + x2 − λ1 (x2 + y − 2) − λ2 (−y + 1).
The Kuhn-Tucker conditions are
0
(1) L1 = y + 2x − 2λ1 x = 0
0
(2) L2 = x − λ1 + λ2 = 0
(3) λ1 ≥ 0; x2 + y ≤ 2; and λ1 (x2 + y − 2) = 0
(4) λ2 ≥ 0; y ≥ 1; and λ2 (−y + 1) = 0

2. Solve this problem.

We divide our argument into the following cases:

Case 1: x2 + y = 2 and y = 1

In this case, we have x = ±1 and y = 1. Consider the case where (x, y) =


(1, 1). Plugging (x, y) = (1, 1) into (1) and (2), we obtain λ1 = 3/2 and
λ2 = 1/2. So, (x, y) = (1, 1) is a candidate for the solution.

Consider the case where (x, y) = (−1, 1). Plugging (x, y) = (−1, 1) into (1)
and (2), we obtain λ1 = 1/2 and λ2 = 3/2. Thus, (x, y) = (−1, 1) is a
candidate for the solution.

Case 2: x2 + y = 2 and y > 1

By (4), we have λ2 = 0. Plugging this into (2), we obtain λ1 = x. Plugging


this into (1), we have y + 2x − 2x2 = 0. Since x2 + y = 2, we also have
y = 2 − x2 . Combining these two equalities,
√ we obtain 3x2 − 2x − 2 = 0. The
to this equation are x = (1 ± 7)/3. But we √
solutions √ must have λ1 = x ≥ 0,
x = (1 + 7)/3 is only admissible. Plugging x = (1 + 7)/3 into y = 2 − x2 ,
we have
√ √
1 √ 2 18 − (8 + 2 7) 2(5 − 7) 2(5 − 2) 2
y = 2 − (1 + 7) = = < = < 1.
9 9 9 9 3
This contradicts our hypothesis that y > 1. So, we ignore this case.

Case 3: x2 + y < 2 and y = 1

From (1), we have λ1 = 0. Plugging λ1 = 0 and y = 1 into (1), we have


x = −1/2. Furthermore, plugging x = −1/2 and λ1 = 0 into (2), we have
λ2 = 1/2. Thus, (x, y) = (−1/2, 1) is a candidate for the solution.

73
Case 4: x2 + y < 2 and y > 1

From (3) and (4), we have λ1 = λ2 = 0. Plugging this into (2), we have
x = 0. Plugging x = 0 and λ1 = 0 into (1), we have y = 0, which contradicts
the hypothesis that y > 1. So, we ignore this case.

In sum, we have three candidates for the solution: (x, y) = (1, 1), (−1, 1),
and (−1/2, 1). At these values, we have

f (1, 1) = 2
f (−1, 1) = 0
f (−1/2, 1) = −1/4.

Among these, the objective function is highest at (1, 1).


We next check the constraint qualification. Recall that g 1 (x, y) = x2 + y and
g 2 (x, y) = −y. We need to check all the four cases below where the solution
can possibly be found.

Case 1: no constraints are binding.

In this case, the constraint qualification holds automatically.

Case 2: g 1 is binding, while g 2 is not.

Clearly, ∇g 1 (x, y) = (2x, 1) 6= (0, 0). So, the constraint qualification holds in
this case.

Case 3: g 1 is not binding, while g 2 is binding.

Clearly, ∇g 2 (x, y) = (0, −1) 6= (0, 0). So, the constraint qualification holds
in this case.

Case 4: g 1 and g 2 are binding.

Suppose, on the contrary, that there exist α 6= 0 and a feasible point (x, y)
such that ∇g 1 (x, y) = αg 2 (x, y). This implies
   
2x 0
= .
1 −α

This is clearly impossible. So, ∇g 1 (x, y) and ∇g 2 (x, y) are not parallel to
each other so that the constraint qualification holds in this case.

74
Finally, we will establish the existence of solutions to the constrained opti-
mization problem. Define

D1 = {(x, y) ∈ R2 | g 1 (x, y) ≤ 2}
D2 = {(x, y) ∈ R2 | g 2 (x, y) ≤ 1}.

We are now able to define D = D1 ∩ D2 as the set of feasible points for this
constrained optimization problem. Since x2 + y is the sum of polynomials
and −y is a linear function of y, D1 and D2 are closed. In addition, D is
closed, as the intersection of two closed sets is closed. We shall claim that
D is bounded. Define B = {(x, y) ∈ R2 | x2 + y 2 ≤ 5}. Fix (x, y) ∈ D
arbitrarily. Since (x, y) ∈ D, we satisfy x2 + y ≤ 2, which further implies
that x2 ≤ 2 − y. Taking into account that x2 ≥ 0, we thus have y ≤ 2. We
check the following:

x2 + y 2 = x2 + y + y 2 − y ≤ 2 + y 2 − y ≤ 1 + y 2 ≤ 5.
|{z} |{z} |{z}
x2 +y≤2 y≥1 y≤2

Hence, (x, y) ∈ B so that D ⊆ B. Therefore, D is a bounded set. We


know that the objective function is continuous and the set of feasible points
defined by the constraints is compact (both closed and bounded). Due to the
extreme value theorem, the solution is guaranteed to exist.
By the necessity of the Kuhn-Tucker conditions, we conclude that (x∗ , y ∗ ) =
(1, 1) is the solution.

11 Homework 11 (Due Date: Nov 18 (Thu), 2021


Question 11.1 (8 points) Show the following equalities:

1. Z
1 1
x2 ln xdx = x3 ln x − x3 + C
3 9

d 1 1 1
(RHS) = x2 ln x + x3 · − x2 = x2 ln x,
dx 3 x 3
where RHS denotes the right hand side of the above equation.

2. Z √
2
1 √ 2 1  √
2

x + 1dx = x x + 1 + ln x + x + 1 + C
2 2

75
!
1 2 −1/2
d 1 2 1 1 1 + (x + 1) · 2x
(RHS) = (x + 1)1/2 + x(x2 + 1)−1/2 · 2x + 2

dx 2 4 2 x + x2 + 1
1 1 + x(x2 + 1)−1/2
 
1 2 −1/2
 2 2
= (x + 1) x +1+x +
2 2 x + (x2 + 1)1/2
1 (1 + x(x2 + 1)−1/2 )(x − (x2 + 1)1/2 )
 
1 2 −1/2 2
= (x + 1) (2x + 1) +
2 2 (x + (x2 + 1)1/2 )(x − (x2 + 1)1/2 )
1 2 1
(x + 1)−1/2 (2x2 + 1) + (x2 + 1)1/2 − x2 (x2 + 1)−1/2

=
2 2
1 2 −1/2
= (x + 1) (2x + 1 + (x2 + 1) − x2 ) = (x2 + 1)1/2 .
2
2

where C is an arbitrary constant.

Question 11.2 (12 points) Evaluate the following integrals:

1. Z 1
xdx.
0
1 1
x2
Z 
1
xdx = = .
0 2 0 2
2. Z 2
(2x + x2 )dx
1
2 2 2  2  2 x3 2
Z Z Z  
2 2 16
(2x+x )dx = 2xdx+ x dx = x 1 + = (4−1)+(8/3−1/3) = .
1 1 1 3 1 3
3. Z 3  
1 2 1 3
x − x dx
−2 2 3
Z 3  
1 2 1 3 1  3 3 1  4 3 1 1 5
x − x dx = x −2 − x −2 = (27+8)− (81−16) = .
−2 2 3 6 12 6 12 12
0
Question 11.3 (8 points) Find the function f (x) if f (x) = ax2 + bx, where a, b
are constants and f (x) satisfies the following all three equations:
0
1. f (1) = 6
00
2. f (1) = 18
R2
3. 0 f (x)dx = 18.

76
First, we have the following expression for f (x):
a b
f (x) = x3 + x2 + c,
3 2
0 00
where c is a constant. Second, since f (x) = ax2 +bx, it follows that f (x) = 2ax+b.
The first two conditions gives us the following equations:
0
f (1) = a + b = 6
00
f (1) = 2a + b = 18.
Hence, we obtain a = 12 and b = −6. Plugging these values into the expression
for f (x), we obtain
f (x) = 4x3 − 3x2 + c.
The last condition gives us the following:
Z 2
2
f (x)dx = x4 − x3 + cx 0 = (16 − 8 + 2c) − 0 = 18.

0

Thus, we have c = 5 so that f (x) = 4x3 − 3x2 + 5.

Question 11.4 (6 points) Given ρ 6= −1, show that


xρ+1 xρ+1
Z
xρ ln xdx = ln x − + C,
ρ+1 (ρ + 1)2
where C is an arbitrary constant.

d xρ xρ
(RHS) = xρ ln x + − = xρ ln x.
dx ρ+1 ρ+1
Question 11.5 (12 points) Calculate the following integrals:
1. √
4
e x
Z
√ √ dx
1 x(1 + e x )

Set u = x. Then, we have du = x−1/2 /2dx, which implies dx = 2udu.
Note also that x = 1 corresponds to u = 1 and x = 4 corresponds to u = 2.
Therefore,
Z 4 √ Z 2
e x eu
√ √ dx =
u
· 2udu
1 x(1 + e x ) 1 u(1 + e )
Z 2
eu
= 2 u
du
1 1+e
= 2 [ln(1 + eu )]21 = 2 ln(1 + e2 ) − 2 ln(1 + e).

77
2. Z 1/3
1
dx
0 ex +1
−x
(Hint: substitute t = e )
Set t = e−x . Then, we have dt = −e−x dx, which implies dx = −dt/t. Note
that x = 0 corresponds to t = 1 and x = 1/3 corresponds to t = e−1/3 .
Therefore,
Z 1/3 Z e−1/3  
1 1 1
x
dx = 1 · − dt
0 e +1 1 t
+1 t
Z 1
1
= dt = [ln(1 + t)]1e−1/3 = ln 2 − ln(1 + e−1/3 ).
e−1/3 1 + t

3. Z 41
1
√ √ dx
8.5 2x − 1 − 4 2x − 1
4
(Hint: substitute z = 2x − 1)
Set z 4 = 2x − 1. Then, we have 4z 3 dz = 2dx, which implies dx = 2z 3 dz.
Note that x = 8.5 corresponds to z = 2 and x = 41 corresponds to z = 3.
Therefore,
Z 41 Z 3
1 2z 3
√ √ dx = 2
dz
8.5 2x − 1 − 4 2x − 1 2 z −z
Z 3 2
z
= 2 dz
2 z −1
Z 3 Z 3 Z 3 
1
= 2 zdz + 1dz + dz
2 2 2 z −1
 
5
= 2 + 1 + ln 2 = 7 + 2 ln 2.
2

Question 11.6 (15 points) In statistics, the normal (or Gaussian) density func-
tion with mean µ and variance σ 2 is defined by
(x − µ)2
 
1
f (x) = √ exp −
σ 2π 2σ 2
in the interval (−∞, ∞). Prove the following results.

1. Z +∞
f (x)dx = 1
−∞

78
√ √
Set
√ u = (x − µ)/ 2σ. Then, we have du = dx/( 2σ), which implies dx =
σ 2du. Note that x = −∞ corresponds to u = −∞ and x = ∞ corresponds
to u = ∞. Therefore,
Z +∞ Z ∞
1 2 √
f (x)dx = √ e−u · σ 2du
−∞ σ 2π
Z−∞

1 2
= √ e−u du = 1.
−∞ π

2. Z +∞
xf (x)dx = µ
−∞
√ √
If u = (x − µ)/ 2σ, x = µ + 2σu. Therefore,
Z +∞
1
Z ∞ √ 2
xf (x)dx = √ (µ + 2σu)e−u du
−∞ π −∞
µ
Z ∞
−u2
√ Z ∞ −u2
= √ e du + 2σ ue du = µ.
π −∞ −∞

3. Z +∞
(x − µ)2 f (x)dx = σ 2 .
−∞
√ √
If u = (x − µ)/ 2σ, then x − µ = 2σu, which implies (x − µ)2 = 2σ 2 u2 .
Therefore,
Z +∞
2
Z ∞
2 2 1 −u2
√ 2σ 2 ∞ 2 −u2
Z
(x − µ) f (x)dx = 2σ u √ e σ 2du = √ u e du
−∞ −∞ σ 2π π −∞

By integration by parts, we then compute


Z ∞ Z ∞
2 −u2 u  −u2 0
u e du = − (e ) du
−∞ −∞ 2
1 ∞ −u2
h u i∞ Z
−u2
= − e + ue du
2 −∞ 2 −∞

π
=
2
Therefore, we obtain
Z +∞
(x − µ)2 f (x)dx = σ 2 .
−∞

79
You√should take into account the following hint: use the substitution u = (x −
µ)/ 2σ together with the following results (without proof):
Z +∞
2 √
e−x dx = π
−∞
Z +∞
2
xe−cx dx = 0
−∞

where c is a constant.

Question 11.7 (12 points) An amount $1,000 earns interest at 5% per year.
Answer the following questions.

1. What will this amount have grown to after ten years when interest is com-
pounded monthly?
1000(1 + 0.05/12)120 ≈ 1647

2. What will this amount have grown to after ten years when interest is com-
pounded continuously?

1000e0.05·10 = 1000e0.5 ≈ 1648.

3. What will this amount have grown to after 50 years when interest is com-
pounded monthly?
1000(1 + 0.05/12)600 ≈ 12119

4. What will this amount have grown to after 50 years when interest is com-
pounded continuously?

1000e0.05·50 = 1000e2.5 ≈ 12182

Question 11.8 (12 points) Consider a tree that is planted at time t = 0, and
let P (t) be its current market value at time t, where P (t) = (t + 5)2 for all t ≥ 0.
Assume that the interest rate is 5% per year, and assume continuous compounding.
Answer the following questions.

1. Let f (t) denote the present value of the tree at time t. What is f (t)?

f (t) = (t + 5)2 e−0.05t

2. At what time t∗ should this tree be cut down in order to maximize its present
value?

80
0
We compute f (t):
0
f (t) = 2(t + 5)e−0.05t − 0.05(t + 5)2 e−0.05t

 > 0 if t ∈ [0, 35)
= 0.05(t + 5)(35 − t)e−0.05t = 0 if t = 35
< 0 if t > 35

Therefore, t∗ = 35.

3. Find limt→∞ f (t).


0
Due to the sign variation of f (t), we have limt→∞ f (t) = 0.

Question 11.9 (15 points) Examine the convergence of the following geometric
series, and find their sums when they exist:

1.
1 1 1
+ 2 + 3 + ···
p p p
The generic k-th term of the sequence is denoted ak = (1/p)k . When |p| > 1,
we have
lim ak = 0.
k→∞

When |p| > 1, we also obtain



X 1/p 1
(1/p)k = = .
k=1
1 − 1/p p−1

2.
√ 1
x+ x + 1 + √ + ···
x
The generic k-th term of the sequence is denoted
 k−1
1
ak = x √ .
x

When x > 1, i.e., x > 1, we obtain limk→∞ ak = 0. When x > 1, we also
obtain ∞ √
X x x x
ak = =√ .
k=1
1 − √1x x−1

81
3. ∞
X
x2n
n=1

First, we observe

X
x2n = x2 + x4 + x6 + x8 + · · ·
n=1

The generic k-th term of the sequence is denoted ak = x2 (x2 )k−1 . When
x2 < 1, i.e., −1 < x < 1, we obtain

X x2
ak = .
k=1
1 − x2

12 Homework 12 (No Need for Submission)


Question 12.1 Consider the following maximization problem with two equality
constraints:

max (xy + yz + zx) subject to x + y + 2z = 3, x + 3y = 1.


(x,y,z)∈R3

Answer the following questions.

1. Find the solution candidate for this problem using the Lagrangian method.

We set up the Lagrangian:

L(x, y, z, λ1 , λ2 ) = xy + yz + zx − λ1 (x + y + 2z − 3) − λ2 (x + 3y − 1).

The first-order conditions are


0
(1) Lx = y + z − λ1 − λ2 = 0,
0
(2) Ly = x + z − λ1 − 3λ2 = 0,
0
(3) Lz = y + x − 2λ1 = 0.

We also have two equalities:

(4) x + y + 2z = 3,
(5) x + 3y = 1.

(1) - (2) yields


x−y
y − x + 2λ2 = 0 ⇒ λ2 = (∗)
2

82
From (5), we have
0
(5 ) x = 1 − 3y.
From (3), we have
0 x+y
(3 ) λ1 = .
2
0
Plugging (5 ) into (∗), we have
1 − 4y
λ2 = . (∗∗)
2
0 0
Plugging (5 ) into (3 ), we obtain
00 1 − 2y
(3 ) λ1 = .
2
0
Plugging (5 ) into (4), we obtain
0
(4 ) z = 1 + y.
0 00 0
Finally, plugging (5 ), (∗∗), (3 ), and (4 ) into (1), we have
1 − 2y 1 − 4y
y + (1 + y) − − = 0 ⇒ y = 0.
2 2
0 00 0
Plugging y = 0 into (5 ), (∗∗), (3 ), and (4 ), respectively, we have (x, y, z, λ1 , λ2 ) =
(1, 0, 1, 1/2, 1/2). This is the unique solution candidate we obtained via the
Lagrangian method.
2. Check whether or not the solution candidate you identified in the previous
part of the question is indeed a solution to the constrained optimization
problem. Justify your answer.
0 0
Plugging (5 ) and (4 ) into the objective function, we obtain

xy + yz + zx = (1 − 3y)y + y(1 + y) + (1 − 3y)(1 + y)


= y − 3y 2 + y + y 2 + 1 + y − 3y − 3y 2
= −5y 2 + 1.

Obviously, this is maximized at y = 0. Then, we also obtain

x = 1 − 3y = 1
z = 1 + y = 1.

Thus, we obtain (x, y, z) = (1, 0, 1) as the maximum point of this constrained


optimization problem and it is exactly the one we obtained in the previous
question.

83
Question 12.2 Consider the following optimization (both maximization and min-
imization) problem with an equality constraint:

max(min)x2 + y 2 subject to 4x2 + 2y 2 = 4.

Answer the following questions.

1. Find all the solution candidates using the Lagrangian method.

We set up the Lagrangian:

L(x, y, λ) = x2 + y 2 − λ(4x2 + 2y 2 − 4).

The first-order conditions are


0
(1) Lx = 2x − 8λx = 2x(1 − 4λ) = 0,
0
(2) Ly = 2y − 4λy = 2y(1 − 2λ) = 0.

We also have the following equality:

(3) 4x2 + 2y 2 = 4.

We continue our investigation by the following four cases:


Case 1: x = y = 0

This case is invalid because (3) does not hold at (0, 0).

Case 2: x = 0 and y 6= 0

From (2), we have λ = 1/2. Plugging λ = 1/2 into (1), we must have x = 0.
So, this is consistent with our assumption. Plugging x = 0 into (3), we have

2y 2 = 4 ⇒ y 2 = 2 ⇒ y = ± 2.
√ √
Hence, we have two solution candidates: (x, y, λ) = (0, 2, 1/2), (0, − 2, 1/2).
Case 3: x 6= 0 and y = 0

From (1), we have λ = 1/4. Plugging λ = 1/4 into (2), we must have y = 0.
So, this is consistent with our assumption. Plugging y = 0 into (3), we obtain

4x2 = 4 ⇒ x = ±1.

Hence, we have two solution candidates: (x, y, λ) = (1, 0, 1/4), (−1, 0, 1/4).
Case 4: x 6= 0 and y 6= 0

This case is invalid because we have λ = 2 6= 4 = λ, which is a contradiction.

84
2. Find all local maximum points and local minimum points for this constrained
optimization problem.

Let g(x, y) = 4x2 + 2y 2 . So, we have


0 0
∇g(x, y) = (g1 (x, y), g2 (x, y)) = (8x, 4y).

We compute the following:


00
L11 (x, y, λ) = 2(1 − 4λ)
L”22 (x, y, λ) = 2(1 − 2λ)
00 00
L12 (x, y, λ) = L21 (x, y, λ) = 0.

We define the bordered Hessian matrix B̄(x, y, λ):


 0 0   
0 g1 (x, y) g2 (x, y) 0 8x 4y
0 00 00
B̄(x, y, λ) =  g1 (x, y) L11 (x, y, λ) L12 (x, y, λ)  =  8x 2(1 − 4λ) 0 .
0 00 00
g2 (x, y) L21 (x, y, λ) L22 (x, y, λ) 4y 0 2(1 − 2λ)

We continue our discussion by considering the following cases.



Case 1: (x, y, λ) = (0, 2, 1/2)

The bordered Hessian matrix becomes


 √ 
√ 0 0 4 2
B̄(0, 2, 1/2) =  √0 −2 0  .
4 2 0 0

Hence,

0 0 4 2
√ √

0 −2
= 4 2(−1) √
1+3
|B̄(0, 2, 1/2)| = √0 −2 0 4 2 0 = 64 > 0.

4 2 0 0

Thus, (x, y) = (0, 2) is a local maximum point.

Case 2: (x, y, λ) = (0, − 2, 1/2)

The bordered Hessian matrix becomes


 √ 
√ 0 0 −4 2
B̄(0, − 2, 1/2) =  0√ −2 0 .
−4 2 0 0

85
Hence,


0 0 −4 2 √

0
√ −2
|B̄(0, − 2, 1/2)| = 0√ −2 0 = −4 2(−1)1+3 = 64 > 0.
−4 2 0 −4 2 0
0

Thus, (x, y) = (0, 2) is a local maximum point.

Case 3: (x, y, λ) = (1, 0, 1/4)

The bordered Hessian matrix becomes


 
0 8 0
B̄(1, 0, 1/4) =  8 0 0  .
0 0 1

Hence,

0 8 0
= 8(−1) 8 0
1+2

|B̄(1, 0, 1/4)| = 8 0 0 0 1
= −64 < 0.

0 0 1

Thus, (x, y) = (1, 0) is a local minimum point.

Case 4: (x, y, λ) = (−1, 0, 1/4)

The bordered Hessian matrix becomes


 
0 −8 0
B̄(−1, 0, 1/4) =  −8 0 0  .
0 0 1

Hence,

0 −8 0
−8 0
|B̄(−1, 0, 1/4)| = −8 0 0 = (−8) · (−1)1+2 = −64 < 0.
0 0 1
0 1

Thus, (x, y) = (1, 0) is a local minimum point.

3. Find the global maximum point(s) and global minimum point(s) for this
constrained optimization problem.

We investigate this problem by considering the following two cases.

Case I: λ = 1/2

86
We define the Hessian matrix associated with the Lagrangian:
   00 00 
h11 h12 L11 (x, y, 1/2) L12 (x, y, 1/2)
H(x, y, λ = 1/2) = = 00 00
h21 h22 L21 (x, y, 1/2) L22 (x, y, 1/2)
 
−2 0
= .
0 0

Since h11 = −2 ≤ 0, h22 = 0 ≤ 0, and |H(x, y, 1/2)| = 0 for all √ (x, y),
H(x,√y, 1/2) is negative semidefinite for all (x, y). Therefore, (0, 2) and
(0, − 2) are the global maximum points.

Case II: λ = 1/4

We define the Hessian matrix associated with the Lagrangian:


   00 00 
h11 h12 L11 (x, y, 1/2) L12 (x, y, 1/2)
H(x, y, λ = 1/4) = = 00 00
h21 h22 L21 (x, y, 1/2) L22 (x, y, 1/2)
 
0 0
= .
0 1

Since h11 = 0 ≥ 0, h22 = 1 ≤ 0, and |H(x, y, 1/4)| = 0 for all (x, y),
H(x, y, 1/2) is positive semidefinite for all (x, y). Therefore, (1, 0) and (−1, 0)
are the global minimum points.

Question 12.3 Let f (x, y) = ax + y where a is a constant and g(x, y) = x2 + 3y 2 .


Consider the following maximization problem with inequality constraints:

max f (x, y) subject to g(x, y) ≤ 2, x ≥ 0, y ≥ 0.

Answer the following questions.

1. Solve this constrained maximization problem for all values of the constant a.

We set up the Lagrangian:

L(x, y, λ1 , λ2 , λ3 ) = ax + y − λ1 (x2 + 3y 2 − 2) − λ2 (−x) − λ3 (−y).

The Kuhn-Tucker conditions are


0
(1) Lx = a − 2λ1 x + λ2 = 0,
0
(2) Ly = 1 − 6λ1 y + λ3 = 0,
(3) λ1 ≥ 0, x2 + 3y 2 ≤ 2, λ1 (x2 + 3y 2 − 2) = 0,
(4) λ2 ≥ 0, x ≥ 0, λ2 (−x) = 0,
(5) λ3 ≥ 0, y ≥ 0, λ3 (−y) = 0.

87
We investigate the K-T conditions by considering the following four cases.

Case 1: λ2 > 0 and λ3 > 0

From (4) and (5), we have x = y = 0. Plugging y = 0 into (2), we obtain

1 + λ3 = 0 ⇒ λ3 = −1,

which contradicts the hypothesis that λ3 > 0. So, this case is invalid.

Case 2: λ2 > 0 and λ3 = 0

Since λ2 > 0, we know from (4) that x = 0. Plugging λ3 = 0 into (2), we


obtain
1 − 6λ1 y = 0 ⇒ λ1 y = 1/6.
Taking into account that λ1 ≥ 0 from (3), we know that λ1 y = 1/6 implies
that λ1 > 0. Since λ1 > 0, p have x2 + 3y 2 = 2. From x = 0, this
by (3), we √
√ y = ± 2/3 =
equality implies that √ ± 6/3. Since we have y ≥ 0 from (5),
we must have y = 6/3. So, λ1 = 6/12.
Plugging x = 0 into (1), we have

a + λ2 = 0 ⇒ λ2 = −a.

Since we assume that λ2 > 0, we must have a < 0. Thus, for any a < 0, we
have √ √
(x, y, λ1 , λ2 , λ3 ) = (0, 6/3, 6/12, −a, 0)
as the solution candidate.
Case 3: λ2 = 0 and λ3 > 0

Since λ3 > 0, from (5), we have y = 0. Plugging y = 0 into (2), we have

1 + λ3 = 0 ⇒ λ3 = −1,

which contradicts the hypothesis that λ3 > 0. So, this case is invalid.

Case 4: λ2 = 0 and λ3 = 0

Plugging λ3 = 0 into (2), we have


1
1 − 6λ1 y = 0 ⇒ λ1 y = .
6
This further implies that λ1 > 0 and y > 0 because λ1 ≥ 0 from (3) and
y ≥ 0 from (5). Since λ1 > 0, by (3), we have x2 + 3y 2 = 2.

88
Plugging λ2 = 0 into (1), we have
a
a − 2λ1 x = 0 ⇒ λ1 x = .
2
Since λ1 > 0 and x ≥ 0 from (4), we must have a ≥ 0. We consider the
following subcases:

Case 4-1: a = 0

Since we know that λ1 > 0, we must have x = 0. Then, taking√ into account
x = 0 and y > 0, we know that x2 +3y 2 = 2 implies y = 6/3. Plugging
that √
y = 6/3 into λ1 y = 1/6, we have
√ √
6 1 6
λ1 = ⇒ λ1 = .
3 6 12
Hence, for a = 0, we have
√ √
(x, y, λ1 , λ2 , λ3 ) = (0, 6/3, 6/12, 0, 0)

as the solution candidate.

Case 4-2: a > 0

Since λ1 > 0, we have x = a/(2λ1 ) from λ1 x = a/2. Similarly, since λ1 > 0,


we have y = 1/(6λ1 ) from λ1 y = 1/6. Plugging x = a/(2λ1 ) and y = 1/(6λ1 )
into x2 + 3y 2 = 2, we have
a2 3
2
+ =2
4(λ1 ) 36(λ1 )2
⇔ 9a2 + 3 = 72(λ1 )2
p
3(3a2 + 1)
⇔ λ1 = √ (∵ λ1 > 0)
72
p
3(3a2 + 1)
⇔ λ1 = √
6 2
p
6(3a2 + 1)
⇔ λ1 =
12
Therefore, we obtain
6a
x = p
6(3a2 + 1)
2
y = p
6(3a2 + 1)

89
So, for any a > 0, we have
p !
6a 2 6(3a2 + 1)
(x, y, λ1 , λ2 , λ3 ) = p ,p , , 0, 0
6(3a2 + 1) 6(3a2 + 1) 12

as the solution candidate. Furthermore, it is easy to see that the solution


above is reduced to the one for a = 0 when we set a = 0 in it. To summarize,
we obtain the unique solution candidate: for any value of a,
  √ 2 
 √ 6a 6(3a +1)
,√ 2 ,
2
12
, 0, 0 if a ≥ 0
(x, y, λ1 , λ2 , λ3 ) = 6(3a2 +1)
√ 6(3a +1)

(0, 6/3, 6/12, −a, 0) if a < 0.

We now check if the solution candidate we found is a solution to the orig-


inal constrained optimization problem. We analyze this by considering the
following two cases.

Case A: a ≥ 0

We recall the Lagrangian:

L(x, y, λ∗1 , 0, 0) = ax + y − λ∗1 (x2 + 3y 2 − 2),

where p
6(3a2 + 1)
λ∗1 = .
12
We compute the following:
00
Lxx = −2λ∗1
00
Lyy = −6λ∗1
00 00
Lxy = Lyx = 0.

We now form the Hessian matrix associated with the Lagrangian:


 √ 
   ∗
 6(3a2 +1)
h11 h12 −2λ1 0 − √0 2
H(x, y, λ∗1 ) = = ∗ = 6 
h21 h22 0 −6λ1 6(3a +1)
0 − 2
p
Since h11 = − 6(3a2 + 1)/6 < 0 and |H(x, y, λ∗1 )| = (3a2 + 1)/2 > 0 for all
(x, y), H(x, y, λ∗1 ) is negative definite for all (x, y).

Case B: a < 0

90
We recall the Lagrangian:
L(x, y, λ∗∗ ∗∗ 2 2
1 , 0, 0) = ax + y − λ1 (x + 3y − 2),

where
λ∗∗
1 = −a.

We now form the Hessian matrix associated with the Lagrangian:


−2λ∗1
     
∗ h11 h12 0 2a 0
H(x, y, λ1 ) = = =
h21 h22 0 −6λ∗1 0 6a
Since h11 = a < 0 and |H(x, y, λ∗1 )| = 12a2 > 0 for all (x, y), H(x, y, λ∗∗
1 ) is
negative definite for all (x, y).

Thus, for any value of a, the associated Hessian matrix is negative definite
for all (x, y). Hence, the solution candidate we found is a solution to the
constrained optimization problem.
2. Let (x∗ (a), y ∗ (a)) be the solution to the constrained maximization problem
for each value of a. Then, obtain the expression for f ∗ (a) = f (x∗ (a), y ∗ (a)).

From the previous question, we have that for any value of a,


  
 √ 6a

, √ 22 if a ≥ 0
(x∗ (a), y ∗ (a)) = 6(3a2 +1)
 √ 6(3a +1)
0, 36 if a < 0

Then, for any value of a, we have


( √
6(3a2 +1)
∗ ∗ ∗
f (a) = f (x (a), y (a)) = √3
if a ≥ 0
6
3
if a < 0

3. Show that f ∗ (a) is a continuous function of a.

For any a > 0, we have


df ∗ (a) 1 1 −1/2 6a
= · 6(3a2 + 1) · 36a = p .
da 3 2 6(3a2 + 1)
Since f ∗ (a) is differentiable at any√positive value of a, f ∗ (a) is continuous
on (0, ∞). For any a < 0, f ∗ (a) = 6/3 so that it is trivially continuous on
(−∞, 0). So, f ∗ (a) is continuous on (−∞, 0). What remains to be shown is
the continuity of f ∗ (·) at a = 0. We can show this as follows:
p √ √
∗ 6(3a2 + 1) 6 6
lim+ f (a) = lim+ = = lim− = lim− f ∗ (a).
a→0 a→0 3 3 a→0 3 a→0

91
Question 12.4 Compute the following:

1. Z 2
(x − 1)4 (2 − x)dx
1
We set

f (x) = 2 − x,
1
g(x) = (x − 1)5 .
5
Then, by integration by parts,
Z 2 Z 2 Z 2
4 0 2 0
(x − 1) (2 − x)dx = f (x)g (x)dx = [f (x)g(x)]1 − f (x)g(x)dx
1 1 1
 2 Z 2
1 1
= (x − 1)5 (2 − x) − (−1) (x − 1)5 dx
5 1 1 5
1 2 1
(x − 1)6 1 = .
 
=
30 30

2.
1 √
Z
1 − xdx
0

1
1 √ 1
Z Z 
1/2 2 3/2 2
1 − xdx = (1 − x) dx = − (1 − x) = .
0 0 3 0 3

3. Z 1
(3x + 1)5/2 dx
0
Set
u = 3x + 1.
Then, we have du/dx = 3; x = 0 corresponds to u = 1; and x = 1 corresponds
to u = 4. By integration by substitution, we have
Z 1 4
1 4 5/2
Z 
5/2 1 2 7/2
(3x + 1) dx = u du = u
0 3 1 3 7 1
2 7/2 2 7 2
= (4 − 1) = (2 − 1) = (128 − 1)
21 21 21
254
= .
21

92
4. Z 1
x2 ex dx
0
Set

f (x) = x2
g(x) = ex .

Then, by integration by parts, we have


Z 1 Z 1 Z 1
2 x 0 1 0
x e dx = f (x)g (x)dx = [f (x)g(x)]0 − f (x)g(x)dx
0 0 0
Z 1
 2 x 1
= x e 0−2 xex dx
0
Z 1
= e−2 xex dx.
0

Set

h(x) = x,
k(x) = ex .

Once again, by integration by parts, we have


Z 1 Z 1 Z 1
x 0 1 0
xe dx = h(x)k (x)dx = [h(x)k(x)]0 − h (x)k(x)dx
0 0 0
Z 1
= [xex ]10 − ex dx = e − [ex ]10 = e − (e − 1) = 1.
0

Thus, Z 1
x2 ex dx = e − 2.
0

5. Z 3
x2 ln xdx
1
Set

f (x) = ln x
x3
g(x) = .
3

93
Then, by integration by parts, we have
Z 3 Z 3 Z 3
2 0 3 0
x ln xdx = f (x)g (x)dx = [f (x)g(x)]1 − f (x)g(x)dx
1 1 1
 3 3 Z 3
x 1
= ln x − x2 dx
3 1 3 1
1  3 3 1
= 9 ln 3 − x 1 = 9 ln 3 − (33 − 1)
9 9
26
= 9 ln 3 − .
9

Question 12.5 Consider a tree that is planted at time t = 0, and let P (t) be its
current market value at time t, where P (t) is twice-continuously differentiable such
that P (t) > 0 for all t ≥ 0. Assume that the interest rate is 100r% per year, and
assume continuous compounding. Answer the following questions.

1. Let f (t) be the present discounted value of the tree at time t. Find the
expression for f (t).
f (t) = P (t)e−rt .

2. Find the condition under which f (·) is concave. In the rest of the question,
we assume that f (·) is concave.
0
We first compute f (t):
0 0 0
f (t) = P (t)e−rt − rP (t)e−rt = e−rt (P (t) − rP (t)).
00
We next compute f (t):
00 0 00 0
f (t) = −re−rt (P (t) − rP (t)) + e−rt (P (t) − rP (t)
h 0 00 0
i
= e−rt −rP (t) + r2 P (t) + P (t) − rP (t)
h 00 0
i
−rt 2
= e P (t) − 2rP (t) + r P (t) .

00
Since f (·) is concave if and only if f (t) ≤ 0 for any t ≥ 0, the condition for
the concavity of f (·) is reduced to
00 0
P (t) − 2rP (t) + r2 P (t) ≤ 0

for any t ≥ 0.

94
3. At what time t∗ should this tree be cut down in order to maximize its present
discounted value? You can assume that t∗ > 0.

Since f (·) is assumed to be concave, t∗ can be obtained by the first-order


0
condition f (t) = 0. That is,
0 ∗ 0 0
f (t∗ ) = e−rt (P (t∗ ) − rP (t∗ )) = 0 ⇒ P (t∗ ) − rP (t∗ ) = 0.
0
So, t∗ is the solution to P (t∗ ) − rP (t∗ ) = 0.

4. Find dt∗ /dr. What is the sign of dt∗ /dr?


0
Let g(t∗ , r) = P (t∗ ) − rP (t∗ ) = 0. By the implicit function theorem, we have

dt∗ ∂g/∂r −P (t∗ ) P (t∗ )


=− = − = ,
dr ∂g/∂t∗ P 00 (t∗ ) − rP 0 (t∗ ) P 00 (t∗ ) − rP 0 (t∗ )
00 0
where we claim that P (t∗ ) − rP (t∗ ) < 0. We check this below:
00 0 00 0 0
P (t∗ ) − rP (t∗ ) = P (t∗ ) − 2rP (t∗ ) + r2 P (t∗ ) + r(P (t∗ ) − rP (t∗ ))
00 0
= P (t∗ ) − 2rP (t∗ ) + r2 P (t∗ ) < 0,
0
where the second equality follows because P (t∗ ) − rP (t∗ ) = 0 from the first-
order condition and the last strict inequality follows because the concavity of
00 0
f (·) and t∗ > 0, both of which imply P (t∗ ) − rP (t∗ ) < 0. Thus, dt∗ /dr < 0
because P (t∗ ) > 0.

95

You might also like