Exercises on Optimisation in Euclidean Spaces

borrowed from
Further Mathematics for Economic Analysis
Knut Sydsaeter, Peter Hammond, Atle Seierstad
and Arne Strom,
with additional ones from the lectures notes of
J.M. Bonnisseau∗
Jean-Marc Bonnisseau†

August 31, 2021

1 Introduction, basic concepts of topology of Rn

Basic concepts of Topology of Rn: distance and norm; open sets; closed sets;
neighborhood; frontier, closure, interior of a set; bounded and compact sets;
convex combination and convex sets.
Upper and lower bounds, supremum, infimum; maximum and minimum. Con-
tinuity. Differentiability; partial, directional derivatives. A representation theo-
rem.
Eigenvalues and eigenvectors. Diagonalization. Quadratic forms. Definiteness
and semidefiniteness.

Exercise 1 (SHSS 13.1, 5) Sketch the set S = {(x, y)R2 | x > 0, w ≥ 1/x} in
the plane. Is S closed?

Exercise 2 (SHSS 13.1, 6) (a) Let E be the subset in R2 consisting of the

point (0, 0) and all point of the form (1/n, 1/m) for n = 1, 2, . . . and m = 1, 2, . . .
Is E closed?
(b) Let F be the subset in R2 defined by F = {(0, 0)} ∪ {(1/n, 1/n) | n =
1, 2, . . .}. Is F closed?
∗
EMJMD QEM, First semester, first year, 2021-2022
†
Paris School of Economics, Université Paris 1 Panthéon Sorbonne, 106-112 Boulevard de
l’Hôpital, 75647 Paris Cedex 13, France, [email protected]

1
Exercise 3 (SHSS 13.1, 7) Consider the following three subsets of R2 :
A = {(x, y) : y = 1, x ∈ ∪∞ n=1 (2n, 2n + 1)}
B = {(x, y) : y ∈ (0, 1), x ∈ ∪∞ n=1 (2n, 2n + 1)}
∞
C = {(x, y) : y = 1, x ∈ ∪n=1 [2n, 2n + 1]}
For each of these sets determine whether it is open, closed, or neither.

Exercise 4 (SHSS 13.1, 10) Let S be a subset of Rn , and let U = {U ⊂

Rn | U ⊂ S and U is open} be the family of all open subset of S. Similarly, let
F = {F ⊂ Rn | S ⊂ F and F is closed} be the family of all closed supersets of S.
(a) Show that int(S) = ∪U ∈U U . Thus int(S) is the largest open subset of S.
(b) Show that cl(S) = ∩F ∈F U . Thus cl(S) is the smallest closed set containing
S.

Exercise 5 (SHSS 13.1, 11) Show by an example that the union of infinitely
manyu closed sets need not be closed.

Exercise 6 (harder) (SHSS 13.1, 14) Prove that the empty set ∅ and the whole
space Rn are the only sets in Rn that are both open and closed.

Exercise 7 (SHSS 2.2, 2) Determine which of the following sets are convex by
drawing each in the plane.
(a) {(x, y) | x2 + y 2 < 2};
(b) {(x, y) | x ≥ 0, y ≥ 0};
(c) {(x, y) | x2 + y 2 > 8};
(d) {(x, y) | x ≥ 0, y ≥ 0, xy ≥ 1};
(e) {(x, y) | xy ≤ 1};
√ √
(f) {(x, y) | x + y ≤ 2};

Exercise 8 (SHSS 2.2, 3) Let S be the set of all poins (x1 , . . . , xn ) in Rn that
satisfy all the m inequalities


 a11 x1 + a12 x2 + . . . + a1n xn ≤ b1
a21 x1 + a22 x2 + . . . + a2n xn ≤ b2


 ......
am1 x1 + am2 x2 + . . . + amn xn ≤ bm


and moreover are such that x1 ≥ 0, . . . , xn ≥ 0. Show that S is a convex set.

Exercise 9 (SHSS 2.2, 4) If S and T are two sets in Rn and a and b are scalars,
let W = aS + bT denote the set of all points of the ax + by, where x ∈ S and
y ∈ T . (Then W is called a linear combination of the two sets.) Prove that if S
and T are both convex, then so is W = aS + bT .

Exercise 10 (SHSS 2.2, 6) (a) Let S = {x ∈ Rn | kxk ≤ r} be the closed

n-dimensional bal centred at the origin and with radius r > 0. Prove that S is
convex.

2
 S

Figure (c) Figure (d)

(b) If we replace ≤ with <, =, or ≥ in the definition of S, we get three nex

sets S1 , S2 , and S3 . Which of them are convex?

Exercise 11 (harder) (SHSS 2.2, 7) (a) Let S be a set of real numbers with
the property that if x1 , x2 ∈ S, then the midpoint 12 (x1 + x2 ) also belongs to S.
Show by an example that S is not necessarily convex.
(b) Does it make any difference if S is closed?

Exercise 12 (harder) (SHSS 13.5, 2) Determine co(S) in the cases shown in

Figures (c) and (d). (In (d), S consists of the four dots.)

Exercise 13 (harder) (SHSS 13.5, 3) Suppose that N units of a commoditiy

(50 000 barrels of oil, for example) are spread out overt points represented by
a two-dimensional coordinate system so that n1 units P are to be found at the
point x1 , n1 units at x2 , . . ., nm units at xm , where mi=1 ni = N . Explain why
z = (1/N )(n1 x1 + n2 x2 + . . . + nm xm ) is a convex combination of x1 , x2 , . . . , xm .
What is a common name for the point z?

Exercise 14 (harder) (SHSS 13.3, 1) Prove that the set S = {(x, y) | 2x − y <
2 and x − 3y < 5} is open in R2 .

Exercise 15 (harder) (SHSS 13.3, 2) Prove tha the set S = {x ∈ Rn | gj (x) ≤

0, j = 1, . . . , m} is closed if the functions gj are all continuous.

Exercise 16 (harder) (SHSS 13.3, 3) Give examples of subsets S of R and

continuous function f : R → R such that
(a) S is closed , but f (S) is not closed.
(b) S is open, but f (S) is not open.
(c) S is bounded, but f (S) is not bounded.

Exercise 17 (harder) (SHSS 13.3, 4) For a fixed a in Rn , prove that the func-
tion f : Rn → R, defined by f (x) = d(x, a) is continuous.

Exercise 18 (harder) (SHSS 2.1, 2) Let f (t) be a C 1 function of t with f 0 (t) 6=

0.

3
 (a) Put F (x, y) = f (x2 + y 2 ). Find the gradient ∇F at an arbitrary point
and show that it is parallel to the straight line segment joining the point and the
origin.
(b) Put G(x, y) = f (y/x). FInd ∇G at an arbitraty point where x 6= 0, and
show that it is orthogonal to the straight line segment joining the point and the
origin.

Exercise 19 (harder) (SHSS 2.1, 3) Compute the directional derivatives of the

following functions at the given points and in the given directions.
(a) f (x, y) = 2x + y − 1 at (2, 1), in the direction given by (1, 1).
(b) g(x, y, z) = xexy − xy − z 2 at (0, 1, 1), in the direction given by (1, 1, 1).

Exercise 20 (harder) (SHSS 2.1, 6) Suppose that f (x, y) has continuous par-
tial derivatives. Suppose too that the maximum directional derivative of f at
(0, 0) is equal to 4, and that it is attained in the direction given by the vector
from the origin to the point (1, 3). Find ∇f (0, 0).

Exercise 21 (harder) (SHSS 2.1, 9) (a) Prove that if F is C 2 and F (x, y) = C

defines y as a twice differentiable function of x, then
 
 0 F10 F20 
1 1 
00
y 00 = − 0 3 [F11 (F20 )2 − 2F12
00 0 0 00
(F10 )2 ] = − 0 3 F10 F11
00

00 
F1 F2 + F22 F12
(F2 ) (F2 )  0 00

00 
F2 F21 F22
(b) Let F (x, y) = x2 y and C = 8. Use the formula in (a) to compute y 00 at
(x, y) = (2, 2). Check the result by differentiating y = 8/x2 twice.
2
Exercise 22 (SHSS 2.9, 1) (a) Let f be defined for all (x, y) by f (x, y) = x2xy+y4
and f (0, 0) = 0. Show that f10 (x, y) and f20 (x, y) exist for all (x, y).
(b) Show that f has a directional derivative in every direction at every point.
(c) Show that f is not continuous at (0, 0). (Hint: Consider the behaviour of
f along the curve x = y 2 .) Is f differentiable at (0, 0).

Exercise 23 (SHSS 1.5, 1) For the following matrices, find the eigenvaleues
and also
 thoseeigenvectors that corresdpond to the real eigenvalues:
2 −7
(a)
3 −8
 
2 4
(b)
−2 6
 
1 4
(c)
6 −1
 
2 0 0
(d) 0 3 0
0 0 4
 
2 1 −1
(e) 0 1 1 
2 0 −2

4
  
1 −1 0
(f) −1 2 −1
0 −1 1
0 2 3
Exercise
 24 (SHSS 1.5,  2) (a) Compute X AX, A , and A when A =
a a 0 x
a a 0 and X = y .
0 0 b z
(b) FInd all the eigenvalues of A.
(c) The characteristic polynomlal p(λ) of A is a cubic function of λ. Show
that if we replace λ by A, then p(A) is the zero matrix. (This is a special case
od the Cayley-Hamilton theorem.)
   
−2 −1 4 1
Exercise 25 (SHSS 1.5, 5) Let A =  2 1 −2 , x1 = 0, x2 =
 
    −1 −1 3 1
1 1
−1, x3 = 1.
0 1
(a) Verify that x1 , x2 , and x3 are eigenvectors of A, and find the associated
eigenvalues.
(b) Let B = AA. Show that Bx2 = x2 and Bx3 = x3 . Is Bx1 = x1 ?
(c) Let C be an arbitrary n × n matrix such that C3 = C2 + C. Prove that if
λ is an eigenvalue for C, then λ3 = λ2 + λ. Show that C + In has an inverse.

Exercise 26 (SHSS 1.6, 2) (a) Let the matrices Ak and P be given by

   √ √ √ 
1 k 0 1/ 10 −3/√ 35 3/√14
Ak = 3 −2 −1 P =  √ 0 5/√35 2/ √14 
0 −1 1 3/ 10 1/ 35 −1/ 14

Find the characteristic equation of Ak and determine the values of k that make
all the eigenvalues real. What are the eigenvalues if k = 3?
(b) Show that columns of P are eigenvectors of A3 , and compute the matrix
product P0 A3 P. What do you see?

Exercise 27 (SHSS 1.6, 3) (a) Prove that if A = PDP−1 , where P and D are
n × n matrices, then A2 = PD2 P−1 .
(b) Show by induction that Am = PDm P−1 for every positive integer m.

Exercise 28 (SHSS 1.7, 5) Using a result of the course, determine the defi-
niteness of
(a) Q = x21 + 8x22
(b) Q = 5x21 + 2x1 x3 + 2x22 + 2x2 x3 + 4x23
(c) Q = −(x1 − x2 )2
(d) Q = −3x21 + 2x1 x2 − x22 + 4x2 x3 − 8x23

5
Exercise 29 (SHSS 1.7, 6) Let A = (aij )n×n be symmetric and positive semidef-
inite. Prove that A is positive definite if and only if |A| =
6 0.

Exercise 30 (SHSS 1.7, 7) (a) For what values of c is the quadratic form

Q(x, y) = 3x2 − (5 + c)xy + 2cy 2

(i) positive definite, (ii) positive semidefinite, (iii) indefinite?

(b) Let B be an n × n matrix. Show that the matrix A = B0 B is positive
semidefinite. Can you find a necessary and sufficient condition on B for A to be
positive definite, not just semidefinite?

Exercise 31 (SHSS 1.7, 8) Show that if Q = x0 Ax, where A is an n×n matrix,

is positive definite, then  
 aii aij 
(a) aii > 0, i = 1, . . . n, (b)   > 0, i, j = 1, . . . , n
aji ajj 

Exercise 32 (SHSS 1.7, 9) Let A be a symmetric matrix. Write its character-

istic polynomial as

ϕ(λ) = (−1)n (λn + an−1 λn−1 + . . . + a1 λ + a0 )

Prove that A is negative definite if and only if ai > 0 for i = 0, 1, . . . , n − 1.

Exercise 33 (SHSS 13.6, 4) If S is a set in Rn and y is a boundary point √ of

S, is y necessarily a boundary point S? (Hint: The irrational number 2 is a
boundary point of the set Q of rational numbers, but what is Q? If S is convex,
then it is true that a boundary point of S is also a boundaty of point of S.)

Exercise 34 (SHSS 13.6, 5) Some books in economics have suggested the fol-
lowing generalisation of the Minkowski’s separating hyperplane Theorem: Two
convex sets in Rn with only one point in common can be separated by a hyper-
plane. Is this statement correct? What about the assertion taht two convex sets
in Rn with disjoint interiors can be separated by a hyperplane?

Exercise 35 (SHSS 2.3, 1) Which of the functions whose graphs are shown in
the figure (e) are (presumably) convex/concave, strictly concave/strictly convex?

Exercise 36 (SHSS 2.3, 2) (a) Let f be definedfor all x, y by f (x, y) = x − y −

x2 . Show that f is concave using different results of the course.
(b) Show that −e−f (x,y) is concave.

Exercise 37 (SHSS 2.3, 3) Show that f (x, y) = ax2 +2bxy +cy 2 +px+qy +r is
strictly concave if ac−b2 > 0 and a < 0, whereas it is strictly convex is ac−b2 > 0
and a > 0.
(b) Find necessary and sufficient condition for f (x, y) to be concave/convex.

6
Exercise 38 (SHSS 2.3, 4) For what values of the constant a is the following
function concave/convex?

f (x, y) = −6x2 + (2a + 4)xy − y 2 + 4ay

Exercise 39 (SHSS 2.3, 5) Examine the convexity/concavity of the following

functions:
(a) z = x + y − ex − ex+y (b) z = ex+y + ex−y − 12 y (c) w = (x + 2y + 3z)2

Exercise 40 (SHSS 2.3, 6) Supppose y = f (x) is a production function deter-

mining output y as a function of the vector x of nonnegative factor inputs, with
f (0) = 0. Show that:
(a) If f is concave, then fii00 (x) ≤ 0 (so each marginal product fi0 (x) is decreas-
ing).
(b) If f is concave, then f (λx)/λ is decreasing as a function of λ.
(c) If f is homogeneous of degree1 (constant return to scale), then f is not
strictly concave.

Exercise
p 41 (SHSS 2.3, 7) Let f be defined for all x in Rn by f (x) = kxk =
x21 + . . . + x2n . Prove that f is convex. Is f strictly convex? (Hint: Use the
triangular inequality for the norm.)

Exercise 42 (SHSS 2.3, 8) Show that the CES function f defined for v1 > 0,
v2 > 0 by
f (v1 , v2 ) = A(δ1 v1−ρ + δ2 v2−ρ )−1/ρ (A > 0, ρ 6= 0, δ1 , δ2 > 0)
is concave for ρ ≥ −1 and convex for ρ ≤ −1, and that it is strictly concave if
ρ > −1.

Exercise 43 (SHSS 2.3, 9) (a) The Cobb-Douglas function z = f (x) = xa11 xa22 . . . xann
(a1 > 0, . . ., an > 0) is defined for all x1 > 0, . . ., xn > 0. Prove that the kth
leading principal minor of the Hessian f 00 (x) is
 
 a1 − 1 a1 ... a1 

a1 . . . ak k  a2 a2 − 1 . . . a2 
Dk = z  . .. .. .. 
(x1 . . . xk )2  .. . . . 

 ak ak . . . ak − 1

(b) Prove that Dk = (−1)k−1 ( ki=1 ai − 1)z k (xa11...x

...ak
P
2 . (Hint: Add all the other
Pkk)
rows to the first row, extract the common factor i=1 ai − 1, and then subtract
the first column in the new determinant for all the other columns.)
(c) Prove that the function is strictly concave if a1 + . . . + an < 1.

Exercise 44 (SHSS 2.4, 1) Prove that f (x, y) = 1 − x2 − y 2 defined in R2 is

concave by showing that the gradient is a supergradient.

7
Exercise 45 (SHSS 2.4, 2) Use the Jensen’s inequality to f (x) = ln(x), with
λ1 = . . . = λn = 1/n to prove that
√ 1
n
x 1 x 2 . . . xn ≤ (x1 + x2 + . . . + xn ) for x1 > 0, . . . , xn > 0
n
Exercise 46 (SHSS 2.4, 6) Prove that f (x, y) = x4 +y 4 defined in R2 is strictly
convex by showing that the gradient is a subgradient.

Exercise 47 (B chap 2, 11) Let x ∈ Rn .

1) Show that if for all y ∈ Rn , x · y ≤ 0, then x = 0.
2) Show that if for all y ∈ Rn , x · y ≥ 0, then x = 0.
3) Show that there exists a real number a such that for all y ∈ Rn , x · y ≥ a,
then x = 0.
4) Show that there exists a real number a such that for all y ∈ Rn , x · y ≤ a,
then x = 0.

Exercise 48 (B chap 2, 15) Let (uν ) and (vν ) be two sequences. We assume
that (uν ) is convergent. Show that if the set {n ∈ N | uν 6= vν } is finite, then,
(vν ) is convergent and has the same limit as (uν ).
We assume that (uν ) is not convergent. Show that if the set {ν ∈ N | uν 6= vν }
is finite, then, (vν ) is not convergent.

Exercise 49 (B chap 2, 18) Give the closure and the interior of the following
subsets of Rn .

Rn ;

Rn+ ;

Rn++ ;

a linear subspace of Rn different from Rn ;

B̄(x, r) for x ∈ Rn and r > 0;

B(x, r) for x ∈ Rn and r > 0;

in R2 , ]0, 1]2 ∪ ([1, 2] × {0});

in R2 , {(x, y) ∈ R2 | x + y ≥ 0, x2 + y 2 ≤ 1};

in R2 , {(x, y) ∈ R2 | x > 0, y > 0, xy ≥ 1}.

Exercise 50 (B chap 2, 26) We consider the linear space Rn × Rp . We define

the mapping N from Rn × Rp to R+ by N (x, y) = max{kxkn , kykp }. Show
that N is aqnorm on Rn × Rp and that it is equivalent to the Euclidean norm
Pn 2
Pp 2
k(x, y)k = i=1 (xi ) + j=1 (yj ) .

8
Exercise 51 (B chap 2, 27) We consider the linear space L(Rn , Rp ) with the
norm NL and f an element of L(Rn , Rn ). We consider the mapping Φ from
L(Rn , Rn ) to itself defined by Φ(g) = g ◦ f .
1) Show that Φ is a linear mapping. Show that it is Lipschitz continuous with a
coefficient NL (f ).
Same question with Ψ defined by Ψ(g) = f ◦ g.

Exercise 52 (B chap 3, 33) Let M be a p × n matrix. Let P be the p × p

matrix defined by P = M M t .
1) Show that P is a symmetric positive semi-definite matrix.
2) Show that if the rank of M is equal to p, then P is positive definite.
Let N be a n × n symmetric positive definite matrix. Same questions with
Q = M N M t.

Exercise 53 (B chap 3, 34) Let a ∈ R and q be a quadratic function define on

R3 as:
q(x, y, z) = x2 + (1 + a)y 2 + (1 + a + a2 )z 2 + 2xy − 2ayz
1) Compute the bilinear form ϕ associated to q.
2) Give the matrix of q in the canonical basis of R3 .
3) For which values of a, ϕ is positive definite?

Exercise 54 (B chap 3, 38) Let N be a norm on Rn . Show that N is not

differentiable at 0.

Exercise 55 (B chap 3, 39) Let f be a linear mapping from Rn to R. Show

that f is differentiable on Rn and Df (x) = f for all x ∈ Rn .

Exercise 56 (B chap 3, 43) Let f be a mapping from Rn to R. We assume

that there exists c ∈ R+ and α > 0 such that for all (x, y) ∈ (Rn )2 ,

|f (y) − f (x)| ≤ cky − xk1+α

1) Show that the partial derivatives of f at each point of Rn are vanishing.

2) Deduce that f is constant.

Exercise 57 (B chap 3, 47) Let f be a differentiable mapping from an open

subset U of Rn to Rp . We assume that f is k Lipschitz continuous on U , i.e.,
∃k > 0, ∀x, y ∈ U 2 , kf (x) − f (y)kp ≤ kkx − ykn . Show that for all x in U ,
kDf (x)kL ≤ k.

2 On optimization
Optimization in Economics: examples. Existence result: the Weierstrass theo-
rem.

9
Exercise 58 (B chap 1, 3) Let f be a function defined on C. Let us suppose
that ϕ : X ⊂ R → R is an increasing function and f (c) ∈ X for all c ∈ C.
  
max f (x) max ϕ(f (x)) min −ϕ(f (x))
(P1 ) (P2 ) (P3 )
x∈C x∈C x∈C

1) Prove that the three following problems are equivalents, that is that their sets
of solutions are the same.
2) Prove that if ϕ is continuous and val(P1 ) ∈ X, then val(P2 ) = ϕ(val(P1 )).
3) Show that if there exists a solution, then val(P2 ) = ϕ(val(P1 )).
4) Let us consider f (x) = x, C = ]0, 1[ and ϕ equal to the ceiling function, that
isϕ(x) is the smallest element of Z greater or eqal to x, or ϕ(x) = min{z ∈ Z |
x ≤ z}. Compute val(P1 ), val(P2 ) and ϕ(val(P1 )).

Exercise 59 (B chap 1, 5)
1) Prove that the function x → ax2 + bx + c with a > 0 is coercive.
2) Prove that the function x → ax3 + bx2 + cx + d with a =
6 0 is not coercive.

Exercise 60 (B chap 1, 6) Let f be a coercive function from R to R. Let g

be a function from R to R. We assume that there exists r > 0 such that for all
x ∈] − ∞, −r] ∪ [r, +∞[, f (x) ≤ g(x). Show that g is coercive.

Exercise 61 (B chap 1, 7) Let f be a continuous function from R to R. We

consider the above minimisation problem (P) above and we assume that C is
closed. Show that the problem (P) has a solution if there exists c̄ ∈ C such that
the set {c ∈ C | f (c) ≤ f (c̄)} is bounded.

Exercise 62 (B chap 1, 8) We consider the following maximisation problem:


Maximise ax2 + bx + c
(P(α))
x≥α

For which values of (a, b, c) this problem has a solution? For which values of
(a, b, c) this problem has a finite value?
When a solution exists, compute the solution and give the value of the problem.

Exercise 63 (B chap 1, 10) Find the solution(s) of the following maximisation

problem when it exists and compute the value of the problem:
 √ √
Maximise x + 2 c − x
(P(α))
x ∈ [0, c]

where c is a positive real number.


Maximise x2 + 2(c − x)
(P(α))
x ∈ [0, c]
where c is a positive real number.

10
 
Maximise ax − ex
(P(α))
x∈R
where a is a positive real number.

Exercise 64 (B chap 2, 28) Let C be a closed subset of Rn and x̄ an element

of Rn . Show that the following minimisation problem has a solution:

Minimise kx − x̄k
(P)
x∈C

Exercise 65 (B chap 2, 29) Let f be a coercive function from Rn to R. Let g

be a function from Rn to R. We assume that there exists r > 0 such that for all
x satisfying kxk ≥ r, f (x) ≤ g(x). Show that g is coercive.

Exercise 66 (B chap 2, 31) Let f be a continuous function from Rn++ to R.

We assume that for all x ∈ Rn++ , the set A = {x0 ∈ Rn++ | f (x0 ) ≥ f (x)} is closed
in Rn . Show that for all closed subset C of Rn such that C ∩ Rn++ is nonempty
and bounded, the problem

Maximise f (x)
(P)
x ∈ C ∩ Rn++

has a solution.

3 Unconstrainded optimisation
Looking for unconstrained optima: FOC; SOC.

Exercise 67 (SHSS 3.1, 4) Find the functions x∗ (r) and y ∗ (r) such that x =
x∗ (r) and y = y ∗ (r) solve the problem

max f (x, y, r) = −x2 − xy − 2y 2 + 2rx + 2ry

x,y

Exercise 68 (SHSS 3.1, 5) Find the solutions x∗ (r, s) and y ∗ (r, s) of the prob-
lem
max f (x, y, r, s) = r2 x2 + 3s2 y − x2 − 8y 2
x,y

Exercise 69 (SHSS 3.2, 1) The function

f (x1 , x2 , x3 ) = x21 + x22 + 3x23 − x1 x2 + 2x1 x3 + x2 x3

defined onR3 has only one stationary point. Show that it is a local minimum
point.

11
Exercise 70 (SHSS 3.2, 2) (a) Let f be defined for all (x, y) by f (x, y) = x3 +
y 3 − 3xy. Show that (0, 0) and (1, 1) are the only stationary points, and compute
the quadratic form associated to the Hessian matrix of f at the stationary points.
(b) Check the definiteness of the quadratic form at the stationary points.
(c) Classify the stationary points, local minimum, local maximum, saddle
point.
Exercise 71 (SHSS 3.2, 3) Classify the stationary points of
(a) f (x, y, z) = x2 + x2 y + y 2 z + z 2 − 4z
(b) f (x1 , x2 , x3 , x4 ) = 20x2 + 48x3 + 6x4 + 8x1 x2 − 4x21 − 12x23 − x24 − 4x32
Exercise 72 (SHSS 3.2, 4) Suppose f (x, y) has only one stationary point
(x∗ , y ∗ ) which is a local minimum point. Is (x∗ , y ∗ ) necessarily a global mini-
mum point? It may be surprising that the answer is no. Prove this by examining
the function defined for all (x, y) by f (x, y) = (1 + y)3 x2 + y 2 . (Hint: Look at
f (x, −2) as x → ∞.)
Exercise 73 (B chap 3, 48) For the following functions, find the critical points
where the gradient vanish.
1) f (x, y) = ln(1 + xy), (x, y) ∈ {(x0 , y 0 ) ∈ R2 | xy > −1}
2) f (x, y) = xy 2 + xy − 2x − 12y
3) f (x, y, z) = −2x2 − 2xy − xz − 21 y 2 + 2xz − 2z 2 + x − 2y − z
4) f (x, y) = x2 y 2 − 4x2 − y 2
5) f (x, y) = 2x4 + 2x2 y + y 2 − 2x2 + 1
6) f (x, y) = √ 1
+√ 1
on R2 \ {(0, 0), (1, 0)}
x2 +y 2 (x−1)2 +y 2

7) f (x, y) = x(2 ln(x) − y − 1) + ey on R∗+ × R

8) f (x, y) = (x2 + (x − 1)2 )(y − 3)2 + y
9) f (x, y) = x4 − 2x2 y + x2 + 3y 2 − 2xy − 2y + 3
10) f (x, y) = x4 − x2 + y 4 + 2xy 2 − 2y 2 − 2x + 2
11) f (x, y) = 14x2 − 6xy + 6y 2
Exercise 74 (B chap 3, 50) Let f be a continuously differentiable mapping
from U an open subset of Rn to R. Let y ∈ Rn . Show that if the following
optimisation problem 
Maximise x · y − f (x)
x∈U
has a solution x̄ then y = ∇f (x̄).
Exercise 75 (B chap 3, 51) Check for the critical points of Exercise 73 if they
satisfy the second order necessary optimality condition.

12
4 Optimisation with equality constraints
Optimization problem with equality constraints. Necessary conditions for opti-
mality: Theorem of Lagrange. The Lagrangian function: interpretation of the
Lagrange multipliers.
Equality constraints: Second order conditions. Sufficient conditions for local
optimality. Sufficient conditions for global optimality.

Exercise 76 (SHSS 3.3, 1) (a) Solve the problem max −x2 − y 2 − z 2 subject
to x + 2y + z = a.
(b) Compute the optimal value function f ∗ (a) and verify that the derivative
of the value function is equal to the multiplier.

Exercise 77 (SHSS 3.3, 2) (a) Solve the problem

max x + 4y + z subject to x2 + y 2 + z 2 = 216 and x + 2y + 3z = 0
(b) Change the first constraint to x2 + y 2 + z 2 = 215 and the second to
x + 2y + 3z = 0.1. Estimate the corresponding change in the maximum value by
using that the partial derivatives of the value function are equal to the multipliers.

Exercise 78 (SHSS 3.3, 3)(a) Solve the problem

x+y+z =1
max ex + y + z subject to
x2 + y 2 + z 2 = 1
(b) Replace the constraints by x + y + z = 1.02 and x2 + y 2 + z 2 = 0.98. What
is the approximate change in optimal value of the objective function?

Exercise 79 (SHSS 3.3, 4) (a) Solve the utility maximizing problme (assuming
m ≥ 4)
max U (x1 , x2 ) = 21 ln(1 + x1 ) + 14 ln(1 + x2 ) subject to 2x1 + 3x2 = m
(b) With U ∗ (m) as indirect utility function, show that dU ∗ /dm = λ.

Exercise 80 (SHSS 3.3, 5) (a) Solve the problem max 1 − rx2 − y 2 subject to
x + y = m, with r > 0.
(b) Find the value function f ∗ (r, m) and compute ∂f ∗ /∂r and ∂f ∗ /∂m and
verify that they are equal to the partial derivative of the Lagrangian computed
at the solution.

Exercise 81 (SHSS 3.3, 6) (a) Solve the problem

max x2 + y 2 + z 2 subject to x2 + y 2 + 4z 2 = 1 and x + 3y + 2z = 0
(b) Suppose we change the first constraint to x2 + y 2 + 4z 2 = 1.05 and the
second constraint to x + 3y + 2z = 0.05. Estimate the corresponding change in
the value function.
Pn
Exercise 82 (SHSS 3.3, 7) (a) Let U (x) = j=1 αj ln(xj − aj ), where
P α j , aj ,
pj , and m are all positive constants with nj=1 αj = 1, and with m > ni=1 pi ai .
P
Show that if x∗ solves
max U (x) subject to p · x = m, x ≥ 0

13
 then the expenditure on good j is the following linear function of prices and
income
pj x∗j = αj m + pj aj − αj ni=1 pi ai , j = 1, 2, . . . , n
P

(b) Let U ∗ (p, m) = U (x∗ ) denote the indirect utility function. Verify Roy’s
identity:
∂U ∗ ∂L
= = −λx∗i , i = 1, . . . , n
∂pi ∂pi

Exercise 83 (SHSS 3.3, 8) (a) Find the solution of the following problem by
solving the constraints for x and y: √
minimize x2 + (y − 1)2 + z 2 subject to x + y = 2 and x2 + y 2 = 1
(b) Note that there are three variables and two constraints (z does not ap-
pear in the constraints). Show that the condition on the matrix of the partial
derivatives of the constraints are not satisfied, and that there are no Lagrange
multipliers for which the Lagrangian is stationary at the solution point.

Exercise 84 (SHSS 3.3, 9)

n X
X n
Q(x1 , . . . , xn ) = aij xi xj , S = {(x1 , . . . , xn ) | x21 + . . . + x2n = 1}
i=1 j=1

Assume that the coefficient matrix A = (aij ) of the quadratic form Q is symmetric
and prove that Q attains maximum and minimum values over the set S which are
equal to the largest and smallest eigenvalues of A. (Hint: Consider first the case
n = 2. Write Q(x) as Q(x = x0 Ax. The first-order conditons give Ax = λx.)

Exercise 85 (SHSS 3.3, 10)  Consider the problem

gj x, r) = 0, j = 1, . . . , m
max x, rf (x, r) subject to
ri = bm+i , i = 1, . . . , k
where f and g1 , . . . , gm are fiven fucntions and bm+1 , . . . , bm+k are fixed pa-
rameters. (We maximize f w.r.t. both x = (x1 , . . . , xn ) and r = (r1 , . . . , rk ), but
with r1 , . . . , rk completely fixed.) Define b̃ = (0, . . . , 0, bm+1 , . . . , bm+k ) (there are
m zeros). Prove that the partial derivative of the value function with respect
to ri is equal to the Lagrangian with respect to ri computed at the solution for
i = m + 1, . . . , m + k by using the fact that the multiplier is equal to the partial
derivative of the value function for right-hand perturbation and those first-order
condition for the optimisation problem that refer to the variables ri .

Exercise 86 (SHSS 3.4, 1)

(a) Find the four points that satisfy the first-order conditions for the problem
max(min)x2 + y 2 subject to 4x2 + 2y 2 = 4
(b) Compute B2 (x, y) the determinant of the bordered Hessian of order 2 at
the four points found in (a). What can you conclude?
(c) Can you give a geometric interpretation of the problem?

14
Exercise 87 (SHSS 3.4, 2) Compute the B2 and B3 the determinant of the
bordered Hessian of order 2 and 3 for the problem
max(min)x2 + y 2 + z 2 subject to x + y + z = 1
Show that the second-order conditions for a local minimum are satisfied.

Exercise 88 (SHSS 3.4, 3) Use the second order sufficient conditions to classify
the candidates for optimality in the problem
local max(min)x + y + z subject to x2 + y 2 + z 2 = 1 and x − y − z = 1

Exercise 89 (B chap 3, 54)

Let α ∈ Rn++ = {x ∈ Rn | xi > 0, ∀i = 1, . . . , n}. The function f from Rn++ to
R is defined by
Xn
f (x) = αi ln(xi )
i=1

where ln(xi ) is the standard logarithm function of xi . Let β ∈ Rn++ . We consider

the following optimisation problem:

 Maximise
Pn f (x)
(P) i=1 βi xi = 1
x ∈ Rn++


Compute the unique point satisfying the first order necessary condition. Are
the second order necessary condition satisfied at this point?

Exercise 90 (B chap 3, 55) Let us consider the following optimisation problem:


Minimise 5x2 + 4xy + y 2
(P )
3x + 2y = 5

1) First method: solve the problem by reducing it to a one dimensional optimi-

sation problem.
2) Second method: write the first order necessary condition and find the solutions
and the multipliers.

Exercise 91 (B chap 4, 56) Let U = {x ∈ Rn | ∀i = 1, . . . , n, xi > −1}. The

function f from U to R is defined by
n
X
f (x) = ln(xi + 1)
i=1

where ln(xi + 1) is the natural logarithm of xi + 1. we consider the following

optimisation problem :

 Maximise
Pn f (x)
(P) i=1 xi = 0
x∈U


15
 Show that there exists a unique point satisfying the first order necessary con-
dition.

Exercise 92 (B chap 4, 57) Let (f i ) be n differentiable functions from R to

R. Let E be the linear subspace of Rn defined by:
n
X
n
E = {x ∈ R | xi = 0}
i=1

Let x̄ be a solution of the following optimisation problem:

Minimise ni=1 f i (xi )

 P
x∈E

Show that for all i = 2, . . . , n, (f i )0 (x̄i ) = (f 1 )0 (x̄1 ).

Exercise 93 (B chap 4, 60) For the following problem, find the points satisfy-
ingthe first order necessary conditions (minimum or maximum):
Optimise 13 x − 41 y
x2 − 2x + y 2 = 0

 Optimise ln x + ln y + ln z
x2 + y 2 + z 2 = 3
x > 0, y > 0, z > 0


Optimise 4x2 + y 2
xy + 2 = 0

Optimise xy
x2 + 4y 2 − 8 = 0

Optimise 2y 4 − 2xy 2 + x2 − 4y 2 + 2x + 2
−x + y 2 − 2 = 0

Optimise x + 3y − pz
2 2 2
x + 3y + z − 2 x2 + 3y 2 − 4 = 0
Optimise x2 − 32 x + y 2 − 32 y


x2 + y 2 − 2xy − x − y = 0

 Optimise 4x + y + 2
ln x + 2 ln y = 0
x > 0, y > 0

Optimise − 32 xy + 52 y + 83 x − 11

6
x2 + y − 1 = 0

Exercise 94 (B chap 4, 61) For the above optimisation problems, write explic-
itly the associated Lagrangian mapping and check if the second order necessary
condition is satisfied or not at the points satisfying the first order necessary con-
dition.

16