DC-1

SEM-II

Lesson: CONVEXITY AND CONCAVITY OF

FUNCTIONS

Lesson Developer: S.K. TANEJA

College: Ram Lal Anand College

University of Delhi

Institute of Life Long Learning, University of Delhi 1

 CONVEXITY AND CONCAVITY OF FUNCTIONS
Table of Contents

1. Learning outcomes
2. Introduction
3. Derivative Test for Concavity and Convexity
4. Second Derivative and Concavity and Convexity
4.1 Total Differential Method
4.2 Definitions
4.3 Use of Hession Matrix for the determination of Convexity and Concavity

4.4 Graphical Representation

4.5 Assumption
4.6 Theorem
5. Second derivative test for concavity and convexity.
6. Quasi-concave and Quasi-convex Functions
7. Quasi-convex function
8. Properties of Quasi-concave and Quasi-convex functions
9. Functions of multiple variables
10. Differentiable function
11. Exercises
12. References

Institute of Life Long Learning, University of Delhi 2

1. Learning outcomes

After you have read this chapter, you should be able to:-

1. Explain concept of concavity and convexity.

2. Explain concept of quasi- concavity and quasi-convexity.
3. Second Derivative Test for Concavity and Convexity.
4. Use of Bordered Hessian Determinant

2. Introduction

A function f is concave if and only if any pair of distinct point p and R in the domain of f
and 0   1

f ( p  (1 –  ) R )  f ( p)  (1 –  ) f ( R )

0 0 1 1
Where p = ( x1 , x2 ) and R = ( x1 , x2 )

The definition can be extended to strict concavity by changing the weak inequality ≥ to
the strict inequality >.

A function f is convex if and only if any pair of distinct points p and R in domain of f and
for 0<θ<1

f ( p  (1 –  ) R )   f ( p)  (1 –  ) f ( R )

The right hand side is the height of line segment and the left hand side is the height of
the arc AB.

Figure 1

Institute of Life Long Learning, University of Delhi 3

Till now we have been discussing concavity and convexity of functions of one variable
only. The conditions for concavity and convexity, strict and non-strict can be defined for
functions of many variables. We shall discuss the concept of concavity and convexity for
a two variable function.

ƶ = f(x1, x2)

The function f(x,y) is concave (convex) if and only if for any pair of distinct points A and
B on its graph (a-surface) the line segment lies either on or below (above) the surface
except at point A and B. Strict concavity requires the line segment AB lies below the arc
AB. Imagine a dome-shaped surface. The surface of convex function typically be bowl-
shaped. For non-strictly concave and convex function the line segment AB is allowed to
lie on the surface itself, some portion of the surface, or even the entire surface may be
flat rather than curved

Figure 2

3. Derivative Test for Concavity and Convexity

In the case of functions of two or more than two variable, it becomes difficult to use
diagrammatic method or algebraic method to determine the concavity or convexity of
function. The functions are such that they require a lot of algebraic manupulation to use
the algebraic formula. A way out is to use the derivatives if the functions is
differentiable.

A differentiable function f(x) = f(x1, x2,..., xn) is concave if and only if for any given point
0 2 0 1 1 1
p = ( x1 , x2 ,....., xn ) and any other point R = ( x1 , x2 ,...., x n ) in convex domain

n
f ( R )  f ( p)   f i ( f ) ( R – p)
L 1

In the case of function of two variables this can be written as :

Institute of Life Long Learning, University of Delhi 4

 0 0 1 1
given P = ( x1 , x2 ) and R = ( x1 , x2 )

f ( R )  f (p)  f1 (x i ) (x 02 – x1 )  f 2 (x 20 ) (x12 – x11 )

In the case of convex function the inequality will be reversed.

Geometrically it means that for a concave function the tangent plane on point p on the
graph of the function lies initially above the graph of the function.

In the case of a convex function graph of the function lies strictly above all the tangent
planes or the hyper planes, except the point of tangency.

Example

2 2
ƶ = x1  x2

The function is convex if for all X = (x1, x2) and Y = (y1, y2)

f(Y) – f(X) ≥  f (X) (y

i i – x i )  (y12  y22 ) – ( x12  x22 ) ≥ 2x1 (y1–x1) + 2x2 (y2–x2)

f f
(where  2 x1 and  2 x2 ) = 2 x1 y1 – 2 x12  2 x2 y2 – 2 x22
x1 x2

shifting the right hand side to left hand side we get.

y12  y 2 – x12 – x22 – 2 x1 y1  2 x12  2 x2 y2 – 2 x 22 ≥ 0

 y12  x12  2 x1 y1  y22  x22 – 2 x2 y2 ≥ 0

 ( y1 – x1 ) 2  ( y2 – x2 ) 2 ≥ 0

The expression in the brackets will remain positive whatever the value of (x1, x2) and
(y1, y2). This proves that the function is convex

4. Second Derivative and Concavity and Convexity

Till now we have discussed about curvature properties of the function by using algebra
or first derivative concavity and convexity of a function is usually discussed using the
second derivative. The second derivative shows how the function represented by the first
derivative changes. In the case of function of one variable we saw that if f''>0 is convex
which means that for f'>0 the function increases more rapidly as x increases while for
f'<0 the function values full less quickly. For f''<0 the function is concave which means
that for f'>0 the function value increases less quickly as x increases while for f'<0 the
function value falls more quickly.

We cannot use the method of determining concavity and convexity for function of two
variables (or n variables). Second partial derivatives cannot be used directly because
there are infinite number of paths that one can take from same point.

Institute of Life Long Learning, University of Delhi 5

Example

2 2
ƶ = x1  x2 – 20 x1 x2

f1 = 2x1 – 20x2 and f11 = 2

f2 = 2x2 – 20x2 and f22 = 2 and f12 = –20

f11 and f22 are positive, the function is not strictly convex in all directions. Cross partial
derivative also plays a role in determining the curvature of the function.

4.1 Total Differential Method

In order to determine the concavity (convexity) of the functions of two variables (this
approach can be extended to n-variables also) we shall use the method of total
differential.

Let y = f(x)

the first order differential at point x = x0 is dy = f (x0) dx

dy is a function of both x and dx. Let us regard dx as a given constant. dx is infinitely

small change in x. Now we find the total differential of dy which we can write

d (dy )
d2y = d(dy) = dx
dx

d [ f '( x) dx]
 dx
dx

d (dx )
 f ''( x ) dx dx  f '( x ) dx
dx

= f''(x) dx2 to (since dx is constant)

= f''(x) dx2

This is called the second total differential of f(x). Since the term dx2 = (dx)2, it is strictly
positive for any value of dx ≠ 0. It follows that d2y has some sign as f''(x). Therefore the
determination of convexity and concavity which relies on the sign of f''(x) can be
presented using the sign of d2y. A function is convex if f''(x) ≥ 0 and concave if f''(x) ≤ 0
then

d2y = f''(x) dx2 ≥ 0 for convex function

d2y = f''(x) dx2 ≤ 0 for concave functions

The same conditions relating to the sign of d2y to concavity/convexity apply to functions
of n-variables. Here we shall explain this method for two variables.

y = f(x1, x2)

Institute of Life Long Learning, University of Delhi 6

Total differential

f f
dy  dx1  dx2
x1 x2

= f1 dx1 + f2 dx2

The second order total differential is the total differential of dy

(dy) (dy )
d (dy )  d 2 y  dx1  dx2
x1 x2

 ( f1dx1  f 2 dx2 )  ( f1dx1  f 2 dx2 )

 dx1  dx2
x1 x2

2 2
= f11 dx1  f 21 dx1 dx2  f12 dx1 dx2  f 22 dx2

2 2
= f11 dx  2 f12 dx1 dx2  f 22 dx2 ` (f12 = f21)

The expression makes it clear that d2y depends on cross partial derivative f12 as well as
f11 and f22.

A function y = f(x1, x2) is twice continuously differentiable. If d2y>0 whenever at least

one of the d x1 or dx2 is non-zero is convex. If d2y<0 then the function is concave

4.2 Definations

Def: A twice continuously differentiable function y = f(x1, x2) is concave if and only if,
d2y is everywhere negative semi definite

f11 ≤ 0, f12 ≤ 0, f11f22 – f12 ≥ 0

Def : A twice continuously differentiable function y = f(x1, x2) is convex if and only if d2y
is everywhere positive semi definite.

If the second order total differential is satisfies the condition d2y ≶ 0 then the function is
strictly concave/convex.

The method of determining the sign of d2y directly can involve a lot of algebraic
manipulation even when the function is function of two variables. In an earlier topic
dealing with maxima and minima we have used quadratic forms and their properties to
determine maxima-minima. Here also we can use the same method to determine the
sign of d2y.

d 2 y  f11 dx12  2 f12 dx1 dx2  f 222 dx22

Institute of Life Long Learning, University of Delhi 7

We can write the right hand side as

f f12   dx1 
[dx1 dx2 ]  11  ...............
 f 21 f 22   dx2 

We known that f12 = f21 from young's theorem. It follows that 2×2 matrix is symmetric.
This matrix whose elements are second order partial derivatives and cross partial
derivatives is called the Hession matrix and is denoted by H. Hession matrix can be used
to determine the concavity and convexity of the function.

4.3 Use of Hession Matrix for the determination of Convexity and Concavity

Def: For any function y = f(x1, x2, ...., xn) = f(X) where X............. which is twice
diffeerntiable with Hession H, the function f is strictly concave on Rn iff H is negative
definite for all X in Rn, that is

Let dx = [dx1 dx2]

Then

d2y = dxT H dX

d2y = dxT H dx < 0

The function f is strictly convex on Rn if and only if H is positive definite for all x € Rn,
that is

d2y = dXT H dx > 0.

The Hession H is positive definite if and only if all the leading principal minors of Matrix H
are positive.

 f11 f12 f13 

 
H = f 21 f 22 f 23 

f f 32 f33 
 31

For example if y = f(x1, x2, x3) the leading principal minors are

f11 f12
|H1| = |f1|, |H2| = |H3| = |H| = |f11 f12 f13|
f 21 f 22

If |H1|<0, |H2|>0 |H3|>0 the f(x1, x2, x3) is strictly convex (the Hession
is positive definite.

If |H1|<0, |H2|>0, |H3|<0

then H is negative definite, and the f(x1, x2, x3) is strictly concave.

We now revise the condition for concavity and convexity.

Institute of Life Long Learning, University of Delhi 8

Suppose H is the Hession matrix associated with twice differentiable function y = f(X) =
f(x1, x2, ...., xn) x ∈ Rn then

H is positive definite on Rn if and only if its leading principle minors are positive |H1|>0,
|H2|>0, .... |Hn| = |H|>0 for X ∈ Rn

This means that d2y>0 and so f is strictly concave.

H is negative definite on Rn if and only if its leading principle minors alternate in sign
begining with a negative value for |H1|.
>0 n is even
|H1|<0 |H2|>0, ... |Hn| = |H|
<0 n is odd

This means that d2y<0 and f is strictly concave.

Note : A leading principle minor of order r of Hession matrix is found by suppressing the
last n–r rows and columns.

Example. In the case of a 3×3 matrix |H1| is found by suppressing 2nd and 3rd rows and
columns.

|H1| = f11

A leading principle minor of order 2 |H2| is found by suppressing the third row and third
column.

So far we have given the conditions for strict concavity and strict convexity. There are
functions which are not strictly concave/convex. They are concave or convex.

Example

A twice differentiable function y = f(x1, x2...xn) is concave if and only if d2y is everywhere
negative semidefinite.

In terms of Hession matrix this means H is negative semi definite on Rn if and only if all
its principle minors alternate in sign begining with negative or zero value for k=1 (HK).

|Hx1| ≤ 0, |HX2| ≥ 0 ... |HXn| = |H| { ≥ n is even

≤ n is odd

A twice differentiable function y = f(x1, x2, ... xn) is convex if and only if d2y is
everywhere positive semi-definite.

In terms of H this means all its principal minors are positive or zero

|HX1| ≥ 0 |HX2| ≥ 0..........., |HXn| = |H| ≥ 0

Note : Principal minors of order are found by suppressing n-k rows and columns of H.

Institute of Life Long Learning, University of Delhi 9

  f11 f12 f13 
 
Example H = f 21 f 22 f 23 

f f 32 f33 
 31

|H1| = f11, H22, H33

f11 f12 f13

|HX2| = f 21 f 22 f 23
f31 f 32 f33

For a 3×3 Hession matrix there are seven principal minor.

For an n×n matrix there are 2n–1 principal minor.

In the case of 2×2 matrix there are three principal minors

f11 f12
|HX1| f11, f12 |HX2| =
f 21 f 22

For concave function

f11 f12
f11 ≤ 0, f22 ≤ 0, ≥0
f 21 f 22

4.4 Graphical Representation

Let us draw a straight line

If we choose any two points b and d on the line and connect then by a line. The line
connecting the points b and d also lies on the straight line which we drew in the
beginning. All the point on the line connecting b and d lie on the original line.

Figure 3(a) Figure 3(b)

Now look at the circle choose any two points on or in the circle and connect them by a
straight line x1x2. This straight line also lies within the circle.

Institute of Life Long Learning, University of Delhi 10

Now consider the Fig. 3(b). In this case if we connect the two points x1 and x2 by a
straight line we find that there are many points on the x1x2 which do not lie within the
figure shown above.

We are now in a position to define the property of straight line and the circle shown
above.

A set is convex if the line joining any two points of the set lies entirely within the set.
The straight line and the circle which includes the area within the circle is an example of
convex set.

Figure 4

In fig. 4, we have drawn the first quadrant of the eucledium space. If we take two points
a1 and a2 in the figure and connect them by a straight line then the entire line lies within
1 2
the quadrant. For example if we look at the point ƶ = a1  a2 the point ƶ lies on the
3 3
straight line connecting the two points a1 and a2. Any point on the straight line
connecting points a1 and point a2 can be expressed as :

a1 + (1–)a2 where 0 ≤ ≤ 1 [or ∈ [0, 1]]

Such a point is called a convex combination of pair of points a1 and a2

Def: A set S is convex if for every pair of points x1 ∈ S and x2 ∈ S, point x̄ = x1 + (1–
)x2is also an element of , for every value of when 0 ≤  ≤ 1.
A set containing only one point is a convex set. Null set is also considered as a convex
set.

Figure 5

Institute of Life Long Learning, University of Delhi 11

A sold circle is a convex set.

Algeberically x2 + y2 ≤ r defines a convex set.

The equation depicts the circle plus the point in it.

The circle is not hallow. All the three figures given above are examples of convex sets.

Solid figure given below are not convex sets

Figure 6

In these figures there is a feature of reentrance (and also a hole). This is a cause for
non-convexity. To qualify as a convex set, the set of points in the figure must contain no
holes, and its boundary points must be not be reentered anywhere.

The geometric definition of convexity also applies to points in 3-space as well as n-

space. A solid cube is a convex set. A hollow cylender is non convex.

A function which gives rise to a hill over its entire domain is a concave function. A
function which gives rise to a valley over its entire domain is convex function.

If the hill (or valley) does not contain any flat surface then the function is suidth be
strictly concave (convex) function. In case a function which give rise to hill (or valley)
and contains flat surface also, is a concave function (convex function).

4.5 Assumption

The domain of the function is a convex set. This assumption is necessary because we
use the combination of x1 and x2 in the domain D to prove whether the f is a concave or
convex function.

Institute of Life Long Learning, University of Delhi 12

 Figure 7

In fig. 7, let x ≥ 0 be the domain of the function. This domain is a convex set. If we take
two values of x in the domain x1 and x2. The associated values of the function are f(x1)
and f(x2) connect these two points by a straight line AB. The graph of the function is also
given in the fig1. The graph of f is shown by are AB.

The straight line lies below the arc AB. The value of the function at x̄ between x1 and x2
is f(x̄) = C. This is higher than the point (D) on line AB immediately above the value x̄. It
is clear that f(x̄) > d. This property can be expressed as strict concavity of the function.

The value of x between x1 and x2 can be written as

x̄ = x1 + (1 – )x2 where ∈ [0, 1] or 0 ≤ ≤ 1

The point d on the line AB can be expressed as

d = f(x1) + (1 – ) f(x2) 0 ≤ ≤

The point c on the arc AB is

c = f(x1 + (1 – )x2)

The function is strictly concave if

c = f(x1 + (1 – )x2) > f(x1) + (1 – ) x2

In simple words if we take only two points on the domain where the domain is a convex
set then convex combination of these points is also in the domain of the function.

4.6 Theorem

The f is strictly concave if

c = f(x1 + (1 – )x2) > f(x1) + (1 – ) f(x2) ∈ [0, 1] and x ∈ Dan interval which is a
convex set.

The function f is concave if f(x1 + 1 – )x2) ≥ f(x1) + (1 – ) f(x2) 0 ≤≤1 f(x)

Institute of Life Long Learning, University of Delhi 13

 Figure 8

In fig. 8, line AB lies entirely above the graph of the function except at point A & point B.

the f(x) is a convex function. A convex function bends below the line joining points f(x1)
and f(x2) (AB).

A function f(x) is strictly convex if

f(x1 + (1 – )x2) < f(x1) + (1 – )x2 where ∈ [0, 1]

It is a concave function if

f(x1 + (1 – )x2) ≤ f(x1) + (1 – )x2

A linear function is a convex and concave function because it satisfies the conditions of
both convex and concave function.

Figure 9a Figure 9b

Figure9c

Institute of Life Long Learning, University of Delhi 14

Another way of defining concave and convex functions is as discussed below.

Def : A function f is concave on the convex set D

where D = [a, b] if {(y, x)| h ≤ f(x), x ∈ D} is a convex set.

For a convex function the inequality is reversed. In fig 9(a) the function is convex and in
fig 9(b) the function is concave. In fig 9(a) look at all the points above the graph of the
function and below the straight line k parallel to x-axis. The set of points satisfy the
above definition and are actually a convex set. Similarly the shaded area in fig b is a
convex set. The function depicted in fig a is a convex function on the convex domain [a,
b].

Now observe fig 9c. There is a tangent at point x0. The tangent line h on any point on a
concave function will lie above the function (except at point f(x0)). For a convex function
the tangent line at any point x0 will lie below the graph of the function.

It means if f is concave and differentiable the n

f(x) ≤ f(x0) + f'(x0) (x – x0) (1)

for any x0 and any other point X.

The right hand side is actually the equation of the tangent at x0 on the function. If we
move slightly away from point x0 on either side of x0 the tangent line at point x0 lies
above the graph of the function.

For convex function the inequality is reversed.

f(x) ≥ f(x0) + f'(x0) (x – x0) (2)

In the case of convex function (if the function is differentiable) the tangent line at (x0,
f(x0)) will lie below the graph of the function except at point (x0, f(x0))

5. Second derivative test for concavity and convexity.

If the function is differentiable twice then we can use the second derivative to test the
concavity and convexity.

Figure 10a Figure 10b

Institute of Life Long Learning, University of Delhi 15

In fig. 10(a) tangent at any point on the graph of the function lies below the graph of the
function. In fig. 10(b) the tangent line on any point on the graph of the function lies
entirely above the graph of the function. We can use the second derivative to determine
the concavity and convexity of the function. We know that the sign of the first derivative
on an interval (a, b) tells us whether the function is increasing or decreasing. We also
know that the first derivative at a point on (a, b) give us the slope of the tangent at that
point.

f'(x) > 0 on an interval (a, b) means that the function is increasing on (a, b)

f'(x) < 0 on an interval (a, b) means that the function is decreasing on (a, b)

f''(x) (the second derivative) is the derivative of f'(x)

therefore f''(x) > 0 iff f'(x) is increasing an (a, b)

f''(x) < 0 iff f'(x) is decreasing on (a, b)

If f''(x) > 0 it means the slope of the tangent is increasing as we move from left to right
on the graph. In the fig (a) the slope of the tangent is increasing when we move from x1
to x2, where x1 < x2. This happens when the function is convex.

If f''(x) < 0 on (a, b) then tangent becomes flatter when we move to x0 from the left. It
means the slope of the tangent is decreasing as we move from left to right on the graph.
In other words when we move from x1 to x2 where x1 < x2 the slope of the tangent is
decreasing.

We conclude f is strictly concave on interval I if and only if f''(x) < 0 for all x in the
interior of I.

Function f is strictly convex on interval I if and only if f''(x) > 0 for all x on interior of I.

Example

Show that x + x is convex on R2

The function is convex if

Y X
(y + y ) – (x + x ) ≥ (2x1 2x2)
Y X

= 2x1y1 – 2x + 2x2y2 – 2x

y + y – x – x2 + 2x + 2x – 2x1y1 – 2x2y2 ≥ 0

y + y + x + x – 2x1y1 – 2x2y2 ≥ 0

= (y1 – x1)2 + (y2 – x2)2 ≥ 0

This is true for all (x1 x2) and (y1, y2) in R2

Institute of Life Long Learning, University of Delhi 16

We can also use the second derivative test to check the convexity

f1 = 2x1, f2 = 2x2, f11=2, f12 = 0, f22 = 2

f11 < 0 f22 > 0 f11 f22 – (f12)2 > 0 = 4  0 > 0 proved

6. Quasi-concave and Quasi-convex Functions

While discussing concave and convex functions we saw that if the function is concave
(convex) there is no need to check the second order condition to determine whether the
function achiever maxima (minima) or not. When we are dealing with problem of
constrained optimization, it is again possible to dispense with the second order condition.
In the case of constrained optimization quasi-concavity of the function obviates need for
second order condition for determining the maxima. In a similar manner quasi-convex
function removes the need for second order condition when we are trying to find out
minimum of the function.

In the beginning we shall discuss the concept of quasi-concavity and quasi-convexity in

the case of function of single variable. We are using fig. 11(a).

1) Let f be a function of x [y = f(x)]

2) x and y are non-negative [x, y ≥ 0]

This means that we are in the first quadrant.

3) The domain of the function is convex set.

4) Choose two distinct point xi and xj (or x1 and x2) such that xi < xj in the convex
domain of the function.

5) The function f(x) forms an arc between xi and sj such that f (xi) = A and f (xj) =B

Fig 11(a) Fig 11(b)

In fig. 11(a) point B is higher in height than A. In other words f(xj) > f(xi). The function
is strictly quasi-concave if all other points on are AB are higher in height than point A.

Institute of Life Long Learning, University of Delhi 17

In fig. 11 (b).

1) The domain of the is Ist quadrant (xi, xj ≥ 0)

Domain is convex set.

2) Two distinct points are xi and xj such that xi < xj.

3) The function f(x) forms an arc CD between xi and xj such that f(xi) = c and f(xj) =
D. In fig(b) f(xj) < f(xi). The function is strictly quasi-concave if all other points on
the arc are lower in height than f(xi)

Now we shall give an algebric definition of quasi-concavity and quasi-convexity.

Let f be a function of x. Then for any two distinct points xi and xj in the convex domain of
the function such that xi < xj and 0 < θ < 1, the function is strictly quasi-concave
function if the following inequality is satisfied.

f(xj) ≥ f(xi)  f[θxi + (1–θ)xj] > f(xi)

If we replace the strict inequality with weak inequality then the function is quasi-
concave.
The weak inequality implies that there is some horizontal straight line segment also on
the arc AB.

7. Quasi-convex function

Suppose f is a function of x. then for any two points xi and xj and for 0 < θ < 1 the
function is strictly quasi-convex if the following inequality is satisfied.

f(xi) > f(xj) f(xi) > f[θxi + (1 – θ)xj]

If we replace the strict inequality with weak inequality the function satisfy the condition
is quasi-concave.

f(xi) ≥ f(xj) f(xi)≥ f[θxi + (1 – θ)xj]

Differentiable functions

If f is a function of x and is differentiable then the function is strictly quasi-concave if for

the two points xi and xj on their convex domain such that xi < xj, the following inequality
is satisfied.
f(xj) > f(xi) f'(xi) (xj – xi) ≥

where f'(xi) = /x=xi

The function is quasi-concave if

f(xj) ≥ f(xi) f'(xi) (xj – xi) ≥ 0

The function is strictly quasi-convex if

f(xi) > f(xj) f'(xi) (xj – xi) < 0

Institute of Life Long Learning, University of Delhi 18

Quasi-convex if

f(xi) ≥ f(xj) f'(xi) (xj – xi) ≤ 0

8. Properties of Quasi-concave and Quasi-convex functions

1) A linear function is both quasi-convex and quasi-concave.

2) All concave (convex) functions (strict or non strict) arc quasi-concave (quasi-
convex). But the opposite is not true.

3) If f(x) quasi convex (strict or non-strict) then –f(x) is quasi-concave.

9. Functions of multiple variables

Let a function z = f(x, y) of two variables

The function is defined on R2++ (x, y ≥ 0)

Let u = (x1, y1) and v = (x2, y2) and v > u

then f(x, y) is quasi-concave if

f(v) > f(x)  f[θx + (1 – θ)v] > f(x) strictly quasi-concave

f(v) ≥ f(x)  f[θx + (1 – θ)v] ≥ f(x) quasi-concave

f(v) < f(x)  f[θx + (1 – θ)v] < f(x) strictly quasi-convex

f(v) ≤ f(x)  f[θx + (1 – θ)v] ≤ f(x) quasi-convex

10. Differentiable function

Suppose a function z = f(x1, x2, ........ xn) is twice continuously differentiable. The quasi-
concavity and quasi-convexity of the function can be checked with the help of first and
second partial derivatives of the function arranged as a bordered determinant.

0 ……….
………….
|B| = …………
.. .. ..
. …………

The bordered determinant appear to be similar to bordered Hessiam |H| determinant

which we use when we deal with constrained maxima (minima). In the case of of
bordered Hessian |H| the first row (column) of |H| consists of 0 and fire derivatives of
constraint. In this case first row (column) of |B| consists of 0 and first derivatives of the
function f (x1, x2,...., fxn) It is because the quasi-concavity (quasi-convexity) depends
exclusively on the partial derivatives of function of itself. We use B along with its leading
principal minor

Institute of Life Long Learning, University of Delhi 19

 0
0
|B1| = |B2| = |Bn| = |B| =
0

The domain of the function is non-negative or thant (that is x1, x, x3 .......... xn ≥ 0)

z = f(x1, x2,.......xn) is quasi concave on the .

|B1| ≤ 0, |B2| ≥ 0, |Bn| = |B| ≤ 0 if n is odd

= |B| ≥ 0 if n is even
The function is strictly quasiconcave on the
If
|B1| < 0, |B2| > 0,..... |Bn| = |B| < if n is odd
= |B| > if n is even

Quasi-convex.

|B1| ≤ 0, |B2| ≤ 0, |B3| ≤ 0,...., |Bn| = |B| ≤ 0

Strictly quasiconvex if

|B1| < 0, |B2| < 0,...... |Bn| = |B| < 0.

11. Exercises

1) Are the following function quasiconcave ? Which of them are also concave

a) f(x, y) = x1/2 + y1/4 fx all x, y≥ 0

b) f(x, y) = x1/4 + y3/4 x, y ≥ 0

c) f(x, y) = x2 y3

d) f(x, y) = x y2

2) Which of these function defined on are quasiconvex which are also convex

a) +

b) 3 +4

c) 2x1 + 3x2 –

3) Are the following functions concave or convex?

a) z = – (x + x )
is it a concave function ?

b) z = xα yβ show that it is concave

if 0 α < , β < 1 and 1 – (α + β) > 0

Institute of Life Long Learning, University of Delhi 20

 c) –3x2 + 2xy – y2 + 3x – 4y + 1 Ans (Concave)

d) x4 + x2 + y2 + y4 – 3x – 8y (Convex)

e) f(x, y) = – xy Neither concave nor convex.

f) x – y – x2

g) –6x2 + (2a + 4)xy – y2 + 4ay

What is the value of a which will make the function concave

h) f(x, y) = –2x2 – y2 + 4x + 4y – 3 for all (x, y)

has a maximum at (x, y) = 1, 2

4) Is this function concave or convex?

a) (x1 + x2) /
defined on R2++

Show that is concave but not strictly concave

b) 3x1 + x defined on all R2

Show that it is neither strictly concave nor strictly convex.

12. References

Allen, R.G,D, Mathematical Analysis for Economists, London: Macmillan and Co. Ltd

Knut Sydsaeter and Peter J. Hammond, Mathematics for Economic Analysis, Prentice Hall

Carl P. Simon and Lawrence Blume, Mathematics for Economists, London: W .W. Norton & Co.

Institute of Life Long Learning, University of Delhi 21