Chiang/Wainwright: Fundamental Methods of Mathematical Economics

CHAPTER 11
EXERCISE 11.2
Use Table 11.1 to find the extreme value(s) of each of the following four
functions, and determine whether they are maxima or minima:
1. z = x 2 + xy + 2 y 2 + 3
2. z = − x 2 − y 2 + 6 x + 2 y
3. z = ax 2 + by 2 + c ; consider each of the three subcases:
(a) a>0, b>0 (b) a<0, b<0 (c) a and b opposite in sign
4. z = e − 2x + 2 y + 3
2x 2

Ans:

1. The derivatives are: f x = 2 x + y , f y = x + 4 y , f xx = 2 , f yy = 4 , and

f xy = 1 . Then first-order condition requires that 2x + y = 0 and x + 4 y = 0 .

Thus we have x * = y * = 0 implying z * = 3 (which is a minimum)

2. The derivatives are: f x = −2 x + 6 , f y = −2 y + 2 , f xx = −2 , f yy = −2 , and

f xy = 0 . Then first-order condition requires that − 2x = 6 and − 2 y = −2 .

Thus we have x * = 3 y * = 1 so that z * = 10 (which is a maximum)

3. f x = 2ax , f y = 2by , f xx = 2a , f yy = 2b , and f xy = 0 . The first-order

condition requires that 2ax = 0 and 2by = 0 . Thus x * = y * = 0 so that

z * = c . The second derivatives give us f xx f yy = 4ab , and f xy2 = 0 . Thus:

(a) z * is a minimum if a,b>0.

(b) z * is a maximum if a,b<0.
(c) z * gives a saddle point if a and b have opposite signs.

4. f x = 2(e 2 x − 1) , f y = 4 y , f xx = 4e 2 x , f yy = 4 , and f xy = 0 . The first-order

condition requires that e 2 x = 1 and 4 y = 0 . Thus x * = y * = 0 so that

z * = 4 . Since f xx f yy = 4(4) exceeds f xy2 = 0 , z * = 4 is minimum.

1
 5. Consider the function z = ( x − 2) 4 + ( y − 3) 4 .
(a) Establish by intuitive reasoning that z attains a minimum (z * = 0) at
x * = 2 and y * = 3 .
(b) Is the first-order necessary condition in Table 11.1 satisfied?
(c) Is the second-order sufficient condition in Table 11.1 satisfied?
(d) Find the value of d 2 z . Does it satisfy the second-order necessary
condition for a minimum in (11.9)?

Ans:
(a) And pair ( x, y) other than (2,3) yields a positive z value.
(b) Yes. At x* = 2 and y* = 3 , we find f x = 4( x − 2) 3 and

f y = 4( y − 3) 3 = 0 .

(c) No. At x * = 2 and y * = 3 , we have f xx = f yy = f xy = f yx = 0 .

(d) By (11.6), d 2 z = 0 . Thus (11.9) is satisfied.

EXERCISE 11.3
1. By direct matrix multiplication, express each of the following matrix
products as a quadratic form:
⎡ 4 2⎤ ⎡ u ⎤ ⎡5 2⎤ ⎡ x ⎤
(a) [u v]⎢ ⎥ ⎢ ⎥ (c) [x y]⎢ ⎥⎢ ⎥
⎣ 2 3⎦ ⎣ v ⎦ ⎣ 4 0⎦ ⎣ y ⎦

⎡− 2 3 ⎤ ⎡u ⎤ ⎡f xx f xy ⎤ ⎡dx ⎤
(b) [u v]⎢ (d) [dx dy]⎢
⎥⎢ ⎥
⎣ 1 − 4⎦ ⎣ v ⎦ ⎣f yx f yy ⎥⎦ ⎢⎣dy ⎥⎦

Ans:
(a) q = 4u 2 + 4uv + 3v 2
(b) q = −2u 2 + 4uv − 4 v 2
(c) q = 5x 2 + 6 xy

(d) q = f xx dx 2 + 2f xy dxdy + f yy dy 2

2. In Prob.1b and c, the coefficient matrices are not symmetric with respect to
the principal diagonal. Verify that by averaging the off-diagonal elements and
⎡− 2 2 ⎤ ⎡5 3⎤
thus converting them, respectively, into ⎢ ⎥ and ⎢ ⎥ we will get
⎣ 2 − 4⎦ ⎣3 0 ⎦
the same quadratic forms as before.

2
Ans: For (b): q = −2u 2 + 4uv − 4 v 2 . For (c): q = 5x 2 + 6xy . Both are the same
as before.

3. On the basis of their coefficient matrices (the symmetric versions), determine

by the determinantal test whether the quadratic forms in Prob. 1a, b, and c are
either positive definite or negative definite.

Ans:
⎡ 4 2⎤
(a) ⎢ ⎥ : 4 > 0,4(3) > 2 2 -- positive definite
⎣ 2 3⎦
⎡− 2 2 ⎤
(b) ⎢ ⎥ : −2 < 0,−2(−4) > 2 2 -- negative definite
⎣ 2 − 4⎦
⎡5 3⎤
(c) ⎢ ⎥ : 5 > 0,5(0) < 3 2 -- neither
⎣3 0 ⎦

4. Express each of the following quadratic forms as a matrix product involving

a symmetric coefficient matrix:
(a) q = 3u 2 − 4uv + 7 v 2 (d) q = 6xy − 5 y 2 − 2 x 2

(b) q = u 2 + 7uv + 3v 2 (e) q = 3u 12 − 2u 1 u 2 + 4u 1 u 3 + 5u 22 + 4u 32 − 2u 2 u 3

(c) q = 8uv − u 2 − 31v 2 (f) q = − u 2 + 4uv − 6uw − 4 v 2 − 7 w 2

Ans:
⎡3 − 2⎤ ⎡ u ⎤
(a) q = [u v]⎢
⎣− 2 7 ⎥⎦ ⎢⎣ v ⎥⎦
⎡1 3.5⎤ ⎡u ⎤
(b) q = [u v]⎢
⎣3.5 3 ⎥⎦ ⎢⎣ v ⎥⎦
⎡− 1 4 ⎤ ⎡u ⎤
(c) q = [u v]⎢
⎣4 − 31⎥⎦ ⎢⎣ v ⎥⎦
⎡− 2 3 ⎤ ⎡x ⎤
(d) q = [x y]⎢
⎣3 − 5⎥⎦ ⎢⎣ y ⎥⎦

⎡ 3 − 1 2 ⎤ ⎡ u1 ⎤
(e) q = [u 1 u2 u 3 ]⎢⎢− 1 5 − 1⎥⎥ ⎢⎢u 2 ⎥⎥
⎢⎣ 2 − 1 4 ⎥⎦ ⎢⎣ u 3 ⎥⎦

⎡ − 1 2 − 3⎤ ⎡ u ⎤
(f) q = [u v w ]⎢⎢ 2 − 4 0 ⎥⎥ ⎢⎢ v ⎥⎥
⎢⎣− 3 0 − 7 ⎥⎦ ⎢⎣ w ⎥⎦

3
5. From the discriminants obtained from the symmetric coefficient matrices of
Prob. 4, ascertain by the determinantal test which of the quadratic forms are
positive definite and which are negative definite.

Ans:
(a) 3 > 0,3(7) > (−2) 2 -- positive definite
(b) 1 > 0,1(3) < (3.5) 2 -- neither
(c) − 1 < 0,−1(−31) > 4 2 -- negative definite
(d) − 2 < 0,−2(−5) > 3 2 -- negative definite

3 −1 2
3 −1
(e) 3 > 0, = 14 > 0, − 1 5 − 1 = 37 > 0 -- positive definite
−1 5
2 −1 4

−1 2
(f) − 1 < 0, =0 -- neither (no need to check D 3 )
2 −4

6. Find the characteristic roots of each of the following matrices:

⎡ 4 2⎤ ⎡− 2 2 ⎤ ⎡5 3⎤
(a) D = ⎢ ⎥ (b) E = ⎢ ⎥ (c) F = ⎢ ⎥
⎣ 2 3⎦ ⎣ 2 − 4⎦ ⎣3 0 ⎦
What can you conclude about the signs of the quadratic forms u ′Du , u ′Eu
and u ′Fu ? (Check your results against Prob.3.)

Ans:
(a) The characteristic equation is
4−r 2
= r 2 − 7r + 8 = 0
2 3− r

Its roots are r1 , r2 = 12 (7 + 17 ) . Both roots being positive, u ′Du is positive

definite.

(b) The characteristic equation is r 2 + 6r + 4 = 0 , with roots r1 , r2 = −3 ± 5 .

Both roots being negative, u ′Eu is negative definite.

(c) The characteristic equation is r 2 − 5r − 9 = 0 , with roots r1 , r2 = 12 (5 ± 61) .

Since r1 is positive, but r2 is negative, u ′Fu is indefinite.

4
 ⎡ 4 2⎤
7. Find the characteristic vectors of the matrix ⎢ ⎥.
⎣2 1⎦

4−r 2
Ans: The characteristic equation = r 2 − 5r = 0 has the roots r1 = 5
2 1− r

and r2 = 0 . (Note: This is an example where D = 0 ). Using r1 in (11.13’), we

⎡− 1 2 ⎤ ⎡ x 1 ⎤
have ⎢
− ⎥ ⎢ x ⎥ = 0 . Thus x 1 = 2 x 2 . Upon normalization, we obtain the
⎣ 2 4 ⎦⎣ 2 ⎦

⎡ 25 ⎤
first characteristic vector v1 = ⎢ 1 ⎥ . Next, using r2 in (11.13’), we have
⎣⎢ 5 ⎥⎦

⎡ 4 2⎤ ⎡ x 1 ⎤
⎢2 1 ⎥ ⎢ x ⎥ = 0 . Therefore, x 1 = − 2 x 2 . Upon normalization, we obtain
1

⎣ ⎦⎣ 2 ⎦

⎡− 15 ⎤
v 2 = ⎢ 2 ⎥ . These results happen to be identical with those in Example 5.
⎣⎢ 5 ⎦⎥

8. Given a quadratic form u ′Du , where D is 2 × 2 , the characteristic equation

of D can be written as

d 11 − r d 12
=0 (d 12 = d 21 )
d 21 d 22 − r

Expand the determinant; express the roots of this equation by use of the
quadratic formula; and deduce the following:

(a) No imaginary number (a number involving − 1 ) can occur in r1 and

r2 .
⎡ c 0⎤
(b) To have repeated roots, matrix D must be in the form of ⎢ ⎥.
⎣0 c ⎦
(c) To have either positive or negative semidefiniteness, the discriminant of

the quadratic form may vanish, that is, D = 0 is possible.

5
 Ans: The characteristic equation can be written as
r 2 − (d 11 + d 22 )r + (d 11d 22 − d 12 d 21 ) = 0

Thus r1 , r2 = 12 [(d 11 + d 22 ) ± (d 11 + d 22 ) 2 − 4(d 11d 22 − d 12 d 21 ) ]

(a) The expression under the square-root sign can be written as

E = d 11
2
+ 2d 11d 22 + d 222 − 4d 11d 22 + 4d 12 d 21
= d 11
2
− 2d 11d 22 + d 222 + 4d 12 d 21 = (d 11 − d 22 ) 2 + 4d 12
2
≥0
Thus no imaginary number can occur in r1 and r2 .
(b) To have repeated roots, E has to be zero, which can occur if and only if
d 11 = d 22 (say, =c) and at the same time d 12 = d 21 = 0 . This would mean
⎡ c 0⎤
that matrix D takes the form of ⎢ ⎥.
⎣0 c ⎦
(c) Positive or negative semidefiniteness allows a characteristic root to be zero
(r=0), which implies the possibility that the characteristic equation reduces to

d 11d 22 − d 12 d 21 = 0 , or D = 0 .

EXERCISE 11.4
Find the extreme values, if any, of the following four functions. Check whether
they are maxima or minima by the determinantal test.

1. z = x 12 + 3x 22 − 3x 1 x 2 + 4x 2 x 3 + 6x 32

2. z = 29 − ( x 12 + x 22 + x 32 )

3. z = x 1 x 3 + x 12 − x 2 + x 2 x 3 + x 22 + 3x 32

4. z = e 2 x + e − y + e w − (2 x + 2e w − y)
2

Then answer the following questions regarding Hessian matrices and their
characteristic roots.

Ans:
1. The first-order condition
f 1 = 2 x 1 − 3x 2 = 0
f 2 = −3x 1 + 6x 2 + 4x 3 = 0
f 3 = 4x 2 + 12 x 3 = 0
is a homogeneous linear-equation system in which the three equations are
independent. Thus the only solution is

6
 x 1* = x *2 = x *3 = 0 so that z* = 0

2 −3 0
The Hessian is −3 6 4 , with H1 = 2 > 0 , H 2 = 3 > 0 , and
0 4 12

H 3 = 4 > 0 . Consequently, z * = 0 is a minimum.

2. The first-order condition consists of the three equations

f 1 = −2 x 1 = 0 f 2 = −2 x 2 = 0 f 3 = −2 x 3 = 0

Thus x 1* = x *2 = x *3 = 0 so that z * = 29

−2 0 0
The Hessian is 0 −2 0 , with H 1 = −2 < 0 , H 2 = 4 > 0 , and
0 0 −2

H 3 = −8 < 0 . Consequently, z * = 29 is a maximum.

3. The three equations in the first-order conditions are

2x 1 + x 3 = 0
2x 2 + x 3 = 1
x 1 + x 2 + 6x 3 = 0

Thus x 1* = 1
20 x *2 = 11
20 x *3 = − 202 so that z * = − 11
40 . Since the Hessian is

2 0 1
0 2 1 , with H 1 = 2 > 0 , H 2 = 4 > 0 , and H 3 = 20 > 0 , the z * value
1 1 6

is a minimum.
4. By the first-order condition, we have

f x = 2e 2 x − 2 = 0 , f y = −e − y + 1 = 0 , f w = 2we w − 2e w = 0
2

Thus x * = 0 y * = 0 w * = 1 so that z * = 2 − e
Note: The value of x * and y * are found from the fact that e 0 = 1 . Finding
w * is more complicated. One way of doing it is as follows: First, rewrite the

equation f w = 0 as we w e w . Taking natural logs yield

2

or ln w + ln e w = ln e w
2

7
 or ln w + w 2 = w
or ln w = w − w 2
If we draw a curve for ln w , and another for w − w 2 , their intersection
point will give us the solution. The ln w curve is a strictly concave curve
with horizontal intercept at w = 1 . The w − w 2 is a hill-type parabola with
horizontal intercepts w = 0 and w = 1 . Thus the solution is w * = 1.

4 0 0
The Hessian is 0 1 0 when evaluated at the stationary point, with all
0 0 4e

leading principal minors positive. Thus z * is a minimum.

5. (a) Which of Prob. 1 through 4 yield diagonal Hessian matrices? In each such
case, do the diagonal elements possess a uniform sign?
(b) What can you conclude about the characteristic roots of each diagonal
Hessian matrix found? About the sign definiteness of d 2 z ?
(c) Do the results of the characteristic-root test check with those of the
determinantal test?

Ans:
(a) Problems 2 and 4 yield diagonal Hessian matrices. The diagonal elements are
all negative fro problem 2, and all positive for problems 4 and 5.
(b) According to (11.16), these diagonal elements represent the characteristic
roots. Thus the characteristic roots are all negative ( d 2 z negative definite)
for problem 2, and all positive ( d 2 z positive definite) for problem 4.
(c) Yes.

6. (a) Find the characteristic roots of the Hessian matrix for Prob. 3.
(b) What can you conclude from your results?
(c) Is your answer to (b) consistent with the result of t he determinantal test
for Prob. 3?

Ans:
(a) The characteristic equation is, by (11.14):

2−r 0 1
0 2−r 1 =0
1 1 6−r

8
 Expanding the determinant by the method of Fig. 5.1, we get
(2 − r )(2 − r )(6 − r ) − (2 − r ) − (2 − r ) = 0
or (2 − r )[(2 − r )(6 − r ) − 2] = 0 [factoring]
or (2 − r )(r − 8r + 10) = 0
2

Thus, from the (2 − r ) term, we have r1 = 2 . By the quadratic formula, we

get from the other term: r2 , r3 = 4 ± 6 .

(b) All three roots are positive. Thus d 2 z is positive definite, and z * is a
minimum.
(c) Yes.

EXERCISE 11.5
1. Use (11.20) to check whether the following functions are concave, convex,
strictly concave, strictly convex, or neither:
(a) z = x 2 (b) z = x 12 + 2x 22 (c) z = 2 x 2 − xy + y 2

Ans:
(a) Let u and v be any two distinct points in the domain. Then
f (u ) = u 2 f ( v) = v 2 f [θu + (1 − θ) v] = [θu + (1 − θ) v] 2
Substituting these into (11.20), we find the difference between the left- and
right-side expressions in (11.20) to be
θu 2 + (1 − θ) v 2 − θ 2 u 2 − 2θ(1 − θ)uv − (1 − θ) 2 v 2
= (1 − θ)u 2 − 2θ(1 − θ)uv + θ(1 − θ) v 2
= (1 − θ)(u − v) 2 > 0 [since u ≠ v ]
Thus z = x 2 is strictly convex function.
(b) Let u = (u 1 , u 2 ) and v = ( v1 , v 2 ) be any two distinct points in the domain.
Then
f (u ) = u 12 + 2u 22 f ( v) = v12 + 2 v 22
f [θu + (1 − θ) v] = [θu 1 + (1 − θ) v1 ] 2 + 2[θu 2 + (1 − θ) v 2 ] 2
The difference between the left- and right-side expressions in (11.20) is
θ(1 − θ)(u 12 − 2u 1 v1 + v12 + 2u 22 − 4u 2 v 2 + 2 v 22 ) = θ(1 − θ)[(u 1 − v1 ) 2 + 2(u 2 − v 2 ) 2 ] > 0
Thus z = x 12 + 2x 22 is a strictly convex function.
(c) Let u = (u 1 , u 2 ) and v = ( v1 , v 2 ) be any two distinct points in the domain.
Then
f (u ) = 2u 12 − u 1 u 2 + u 2 f ( v) = 2 v12 − v1 v 2 + v 22
f [θu + (1 − θ) v] = 2[θu 1 + (1 − θ) v12 ] − [θu 1 + (1 − θ) v1 ] ⋅ [θu 2 + (1 − θ) v 2 ] + [θu 2 + (1 − θ) v 2 ] 2
The difference between the left- and right-side expressions in (11.20) is
θ(1 − θ)[(2u 12 − 4u 1 v1 + 2 v12 ) − u 1 u 2 + u 1 v 2 + v1 u 2 − v1 v 2 + (u 22 − 2u 2 v 2 + v 22 )]

9
 = θ(1 − θ)[2(u 1 − v1 ) 2 − (u 1 − v1 )(u 2 − v 2 ) + (u 2 − v 2 ) 2 ] > 0
because the bracketed expression is positive, like θ(1 − θ) . [The bracketed
expression, a positive-definite quadratic form in the two variables (u 1 − v1 )
and (u 2 − v 2 ) , is positive since (u 1 − v1 ) and (u 2 − v 2 ) are not both zero
in our problem.] Thus z = 2 x 2 − xy + y 2 is strictly convex function.

2. Use (11.24) or (11.24’) to check whether the following functions are concave,
convex, strictly concave, strictly convex, or neither:
(a) z = − x 2 (b) z = ( x 1 + x 2 ) 2 (c) z = − xy

Ans:
(a) With f ′(u ) = −2u , the difference between the left- and right-side expressions
in (11.24) is
− v 2 + u 2 + 2u ( v − u ) = − v 2 + 2uv − u 2 = −( v − u ) 2 < 0
Thus z = − x 2 is strictly concave.
(b) Since f 1 (u 1 , u 2 ) = f 2 (u 1 , u 2 ) = 2(u 1 + u 2 ) , the difference between the left-
and right-side expressions in (11.24’) is
( v1 + v 2 ) 2 − (u 1 + u 2 ) 2 − 2(u 1 + u 2 )[(v1 − u 1 ) + ( v 2 − u 2 )]
= ( v1 + v 2 ) 2 − 2( v1 + v 2 )(u 1 + u 2 ) + (u 1 + u 2 ) 2
= [( v1 + v 2 ) − (u 1 + u 2 )]2 ≥ 0
A zero value cannot be ruled out because the two points may be, e.g.,
(u 1 , u 2 ) = (5,3) and ( v1 , v 2 ) = (2,6) . Thus z = ( x 1 + x 2 ) 2 is convex, but
not strictly so.
(c) Since f 1 (u 1 , u 2 ) = −u 2 , and f 2 (u 1 , u 2 ) = −u 1 , the difference between the
left- and right-side expressions in (11.24’) is
− v1 v 2 + u 1 u 2 + u 2 ( v1 − u 1 ) + u 1 ( v 2 − u 2 )
= − v1 v 2 + v1 u 2 + u 1 v 2 − u 1 u 2 = ( v1 − u 1 )( v 2 − u 2 ) <> 0
Thus z = − xy is neither convex nor concave.

3. In view of your answer to Prob. 2c, could you have made use of Theorem III
of this section to compartmentalize the task of checking the function
z = 2 x 2 − xy + y 2 in Prob. 1c? Explain your answer.

Ans: No. That theorem gives a sufficient condition which is not satisfied.

4. Do the following constitute convex sets in the 3-space?

(a) A doughnut (b) A bowling pin (c) A perfect marble

10
Ans: (a) No. (b) No. (c) Yes.

5. The equation x 2 + y 2 = 4 represents a circle with center at (0,0) and with a

radius of 2.

(a) Interpret geometrically the set {( x , y) x 2 + y 2 ≤ 4}.

(b) Is this set convex?

Ans:
(a) The circle with its interior, i.e. a disk.
(b) Yes.

6. Graph each of the following sets, and indicate whether it is convex:

(a) {( x , y) y = e x } (c) {( x , y) y ≤ 13 − x 2 }

(b) {( x , y) y ≥ e x } (d) {( x , y) xy ≥ 1; x > 0, y > 0}

Ans:
(a) The set of points on an exponential curve; not a convex set.
(b) The set of points lying on or above an exponential curve; a convex set.
(c) The set of points lying on or below an inverse U-shaped curve; a convex set.
(d) The set of points lying on or above a rectangular hyperbola in the positive
quadrant; a convex set.

⎡10⎤ ⎡ 4⎤
7. Given u = ⎢ ⎥ and v = ⎢ ⎥ , which of the following are convex
⎣6⎦ ⎣8 ⎦
combinations of u and v?
⎡7 ⎤ ⎡5.2⎤ ⎡6.2⎤
(a) ⎢ ⎥ (b) ⎢ ⎥ (c) ⎢ ⎥
⎣7 ⎦ ⎣7.6⎦ ⎣8.2⎦

Ans:
(a) This is a convex combination, with θ = 0.5 .
(b) This is again a convex combination, with θ = 0.2 .
(c) This is not a convex combination.

11
 8. Given two vectors u and v in the 2-space, find and sketch:
(a) The set of all linear combinations of u and v.
(b) The set of all nonnegative linear combinations of u and v.
(c) The set of all convex combinations of u and v.

Ans:
(a) This set is the entire 2-space.
(b) This set is a cone bounded on one side by a ray passing through point u, and
on the other side by a ray passing through point v.
(c) This set is the line segment uv.

9. (a) Rewrite (11.27) and (11.28) specifically for the cases where the f and g
functions have n independent variables.
(b) Let n = 2 , and let the function f be shaped like a (vertically held)
ice-cream cone whereas the function g is shaped like a pyramid. Describe the
sets S ≤ and S ≥ .

Ans:

(a) S ≤ ≡ {( x 1 , K , x n f ( x 1 , K , x n ) ≤ k} (f convex)

S ≥ ≡ {( x 1 , K , x n g ( x 1 , K , x n ) ≥ k} (g concave)

(b) S ≤ is a solid circle (or disk); S ≥ is a solid square.

EXERCISE 11.6
1. If the competitive firm of Example 1 has the cost function C = 2Q12 + 2Q 22
instead, then:
(a) Will the production of the two goods still be technically related?
(b) What will be the new optimal levels of Q1 and Q 2 ?
(c) What is the value of π12 ? What does this imply economically?

Ans:
(a) No, because the marginal cost of one commodity will be independent of the
output of the other.
(b) The first-order condition is
π1 = P10 − 4Q1 = 0 π 2 = P20 − 4Q 2 = 0

Thus Q1* = 14 P10 and Q *2 = 14 P20 . The profit is maximized, because the

12
 −4 0
Hessian is , with H 1 < 0 and H 2 > 0 . The signs of the
0 −4
principal minors do not depend on where they are evaluated. Thus the
maximum in this problem is a unique absolute maximum.
(c) π12 = 0 implies that the profit-maximizing output level of one commodity
is independent of the output of the other (see first-order condition). The firm
can operate as if it has two plants, each optimizing the output of a different
product.

2. A two-product firm faces the following demand and cost functions:

Q1 = 40 − 2P1 − P2 Q 2 = 35 − P1 − P2 C = Q12 + 2Q 22 + 10
(a) Find the output levels that satisfy the first-order condition for maximum
profit. (Use fractions.)
(b) Check the second-order sufficient condition. Can you conclude that this
problem possesses a unique absolute maximum?
(c) What is the maximal profit?

Ans:
(a) By the procedure used in Example 2 (taking Q1 and Q 2 as choice

variables), we can find Q1* = 3 74 , Q *2 = 4 149 , P1* = 6 141 , P2* = 24 72 .

−4 2
(b) The Hessian is , with H 1 = −4 and H 2 = 28 . Thus the
2 −8
sufficient condition for a maximum is met.
(c) Substituting the P * ’s and Q * ’s into the R and C functions, we get

R * = 134 98
43
, C * = 65 85
98 , and r = 68 7
4

3. On the basis of the equilibrium price and quantity in Example 4, calculate the

point elasticity of demand ε di (for i = 1,2 ). Which market has the highest

and the lowest demand elasticities?

dQ1 P1*
Ans: c d1 = dP1 Q1* = 1 39
4 6 = 138 . Similarly, c d 2 = 1 60
5 9 = 4
3 , and c d 3 = 1 45
6 5 = 3
2 .

The highest is c d1 ; the lowest is c d 2 .

13
4. If the cost function of Example 4 is changed to C = 20 + 15Q + Q 2
(a) Find the new marginal-cost function.
(b) Find the new equilibrium quantities. (Use fractions.)
(c) Find the new equilibrium prices.
(d) Verify that the second-order sufficient condition is met.

Ans:
(a) C′ = 15 + 2Q = 15 + 2Q1 + 2Q 2 + 2Q 3
(b) Equating each MR to the MC, we obtain the three equations:
10Q1 + 2Q 2 + 2Q 3 = 48 2Q1 + 12Q 2 + 2Q 3 = 90 and
2Q1 + 2Q 2 + 14Q 3 = 60

Thus Q1* = 2 97
88
, Q *2 = 6 97
51
, Q *3 = 2 97
91
.

(c) Substituting the above into the demand equations, we get

P1* = 51 97
36
, P2* = 72 97
36
, P3* = 57 97
36

(d) Since R 1′′ = −8 , R ′2′ = −10 , R ′3′ = −12 , and C′′ = 2 , we do find that: (1)
R 1′′ − C′′ = −10 (2) R 1′′R ′2′ − (R 1′′ + R ′2′ )C′′ = 80 + 36 = 116 > 0 , and (3)

H = −960 − (80 + 96 + 120)(2) = −1552 < 0

5. In Example 7, how would you rewrite the profit function if the following
conditions hold?
(a) Interest is compounded semiannually at an interest rate of i 0 per annum,
and the production process takes 1 year.
(b) Interest is compounded quarterly at an interest rate of i 0 per annum, and
the production process takes 9 months.

Ans:

(a) π = P0 Q(a , b)(1 + 12 i 0 ) −2 − Pa 0 a − Pb 0 b

(b) π = P0 Q(a , b)(1 + 14 i 0 ) −3 − Pa 0 a − Pb 0 b

6. Given Q = Q(a , b) , how would you express algebraically the isoquant for
the output level of, say, 260?

Ans: Q(a , b) = 260

14
EXERCISE 11.7
For Probs. 1 through 3, assume that Q ab > 0 .
1. On the basis of the model described in (11.45) through (11.48), find the

comparative-static derivatives (∂a * ∂Pa 0 ) and (∂b * ∂Pa 0 ) . Interpret the

economic meaning of the result. Then analyze the effects on a * and b * of

a change in Pb 0 .

Ans:
(a) We may take (11.49) as the point of departure. Letting Pa 0 alone vary (i.e.,
letting dP0 = dPb 0 = dr = dt = 0 ), and dividing through by dPa 0 ≠ 0 , we get
the matrix equation

⎡P0 Q aa e − rt P0 Q ab e − rt ⎤ ⎡ ( )⎤ = ⎡1⎤
∂a *
∂Pa 0
⎢
⎣P0 Q ab e
− rt ⎥⎢
P0 Q bb e − rt ⎦ ⎢⎣ ( )⎥⎥⎦ ⎢⎣0⎥⎦
∂b*
∂Pa 0

Hence, by Cramer’s Rule,

( )=
∂a *
∂Pa 0
P0 Q bb e − rt
J
< 0 and ( )= −
∂b*
∂Pa 0
P0 Q bb e − rt
J
<0

The higher the price of input a, the smaller will be the equilibrium levels of
inputs a and b.
(b) Next, letting Pb 0 alone vary in (11.49), and dividing through by dPb 0 ≠ 0 ,
we can obtain results similar to (a) above:

( )= −
∂a *
∂Pb 0
P0 Q ab e − rt
J
< 0 and ( )=
∂b*
∂Pb 0
P0 Q aa e − rt
J
<0

2. For the problem of Example 7 in Sec. 11.6:

(a) How many parameters are there? Enumerate them.
(b) Following the procedure described in (11.45) through (11.50), and
assuming that the second-order sufficient condition is satisfied, find the

comparative-static derivatives (∂a * ∂P0 ) and (∂b * ∂P0 ) . Evaluate their

signs and interpret their economic meanings.

(c) Find (∂a * ∂i 0 ) and (∂b * ∂i 0 ) , evaluate their signs, and interpret their

economic meanings.

Ans:
(a) P0 , i 0 , Pa 0 , Pb 0 .

15
(b) From the first-order condition, we can check the Jacobian
∂F1 ∂F1
P0 Q aa (1 + i 0 ) −1 P0 Q ab (1 + i 0 ) −1 −2
Q aa Q ab
J = ∂a ∂b
= = P 2
(1 + i )
P0 Q ab (1 + i 0 ) −1 P0 Q bb (1 + i 0 ) −1
∂F 2 ∂F 2 0 0
∂a ∂b
Q ab Q bb

the be positive at the initial equilibrium (optimum) since the second-order

sufficient condition is assumed to be satisfied. By the implicit-function
theorem, we can then write

a * = a * (P0 , i 0 , Pa 0 , Pb 0 ) and b * = b * (P0 , i 0 , Pa 0 ,P b 0 )

we can also write the identities

P0 Q a (a * , b * )(1 + i 0 ) −1 − Pa 0 ≡ 0

P0 Q a (a * , b * )(1 + i 0 ) −1 − Pb 0 ≡ 0

Taking the total differentials, we get (after rearrangement) the following pair
of equations corresponding to (11.49):

P0 Q aa (1 + i 0 ) −1 da * + P0 Q ab (1 + i 0 ) −1 db * = −Q a (1 + i 0 ) −1 dP0 + P0 Q a (1 + i 0 ) −2 di 0 + dPa 0

P0 Q ab (1 + i 0 ) −1 da * + P0 Q bb (1 + i 0 ) −1 db * = −Q b (1 + i 0 ) −1 dP0 + P0 Q b (1 + i 0 ) −2 di 0 + dPb 0

Letting P0 alone vary (i.e., letting di 0 = dPa 0 dPb 0 = 0 ), and dividing

through by dP0 ≠ 0 , we get

⎡P0 Q aa (1 + i 0 ) −1 P0 Q ab (1 + i 0 ) −1 ⎤ ⎡ ∂∂Pa0 ⎤ ⎡− Q a (1 + i 0 ) −1 ⎤
*

⎢ −1 ⎥⎢ * ⎥ = ⎢ ⎥
⎣ P0 Q ab (1 + i 0 ) P0 Q bb (1 + i 0 ) −1 ⎦ ⎢⎣ ∂∂bP0 ⎥⎦ ⎣− Q b (1 + i 0 ) −1 ⎦

Thus ( )=
∂a *
∂P0
( Q b Q ab − Q a Q bb ) P0 (1+ i 0 ) −2
J
>0

( )=
∂b*
∂P0
( Q a Q ab − Q b Q aa ) P0 (1+ i 0 ) −2
J
>0

(c) Letting i 0 alone vary, we can similarly obtain

⎡P0 Q aa (1 + i 0 ) −1 P0 Q ab (1 + i 0 ) −1 ⎤ ⎡ ∂∂ai 0 ⎤ ⎡ P0 Q a (1 + i 0 ) −2 ⎤
*

⎢ −1 ⎥⎢ * ⎥ = ⎢ ⎥
⎣ P0 Q ab (1 + i 0 ) P0 Q bb (1 + i 0 ) −1 ⎦ ⎢⎣ ∂∂bi 0 ⎥⎦ ⎣P0 Q b (1 + i 0 ) − 2 ⎦

Thus ( )=
∂a *
∂i 0
( Q a Q bb − Q b Q ab ) P02 (1+ i 0 ) −3
J
>0

( )=
∂b*
∂i 0
( Q b Q aa − Q a Q ab ) P0 (1+ i 0 ) −3
J
>0

16
3. Show that the results in (11.50) can be obtained alternatively by
differentiating the two identities in (11.48) totally with respect to P0 , while
holding the other exogenous variables fixed. Bear in mind that P0 can affect
a * and b * by virtue of (11.47).

Ans: Differentiating (11.49) totally respect to P0 , we get

Q a e − rt + P0 Q aa ( )e
∂a *
∂P0
− rt
+ P0 Q ab ( )e
db*
dP0
− rt
=0

Q b e − rt + P0 Q ab ( )e
∂a *
∂P0
− rt
+ P0 Q bb ( )e
db*
dP0
− rt
=0

Or, in matrix notation,

⎡P0 Q aa e − rt ( )
P0 Q ab e − rt ⎤ ⎡ ∂a * ∂P0 ⎤ ⎡− Q a e − rt ⎤
⎥=⎢
⎢
⎣P0 Q ab e
− rt ⎥⎢
( ) ⎥
P0 Q bb e − rt ⎦ ⎣ ∂b * ∂P0 ⎦ ⎣− Q b e − rt ⎦

which leads directly to the results in (11.50).

4. A Jacobian determinant, as defined in (7.27), is made up of first-order partial

derivatives. On the other hand, a Hessian determinant, as defined in Sec. 11.3
and 11.4, has as its elements second-order partial derivatives. How, then, can

it turn out that J = H , as in (11.46)?

Ans: In (11.46), the elements of the Jacobian determinant are first-order partial
derivatives of the components of the first-order condition shown in (11.45). Thus,
those elements are really the second-order partial derivatives of the (primitive)
objective function – exactly what are used to construct the Hessian determinant.

17