Journal of Computational and Applied Mathematics 113 (2000) 65–71

www.elsevier.nl/locate/cam

A generalized xed point theorem and equilibrium point of an

abstract economy
S.P. Singha;∗ , E. Tarafdarb , B. Watsona
a
Department of Mathematics, Memorial University, St. John’s, Nf, Canada A1C 5S7
b
Department of Mathematics, The University of Queensland, Brisbane, 4072, Australia
Received 29 June 1998

Abstract
A generalization of an earlier xed point theorem to a noncompact case is proved and has been applied to prove the
c 2000 Elsevier Science B.V. All rights reserved.
existence of equilibrium points of abstract economies.
MSC: 47H10; 49A10; 90A14; 90D13

1. Introduction

In this paper all topological vector spaces will be assumed to be Hausdor. With the aid of our
xed theorem in [11] we have proved the following xed point theorem in [12].

Theorem 1.1. Let X be a nonempty convex subset of a topological vector space. Let T : X → 2X
be a setvalued mapping such that
(i) for each x ∈ X; T (x) is a nonempty convex subset of X ;
(ii) for each y ∈ X; T −1 (y) = {x ∈ X : y ∈ T (x)} contains a relatively open set Oy of X (Oy
may
S
be empty for some y);
(iii) x∈X Ox = X ; and
(iv) there exists a nonempty set X ⊂ X such that X0 is contained in a compact convex subset
T 0 c
X1 of X and the set D = x∈X0 Ox is empty or compact; where Oxc denotes the complement
of Ox in X. Then there exists a point x0 ∈ X such that x0 ∈ T (x0 ).

On the other hand, we have proved in [13] the following xed point theorem which is a gener-
alization of our xed point theorem in [11] and is useful in mathematical economics, game theory
and for problems of social sciences:
∗
Corresponding author.

0377-0427/00/$ - see front matter c 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 7 - 0 4 2 7 ( 9 9 ) 0 0 2 4 4 - 7
66 S.P. Singh et al. / Journal of Computational and Applied Mathematics 113 (2000) 65–71

Theorem 1.2. Let {X :  ∈ I } be a family of nonemptyQcompact convex sets; each in a topological

vector space E ; where I is an indexing set. Let X = ∈I X . For each  ∈ I; let T : X → 2X be
a set valued mapping such that
(i) for each x ∈ X; T (x) is a nonempty convex subset of X ;
(ii) for each x ∈ X ; T−1 (x ) = {y ∈ X : x ∈ T (y)} contains a relatively open subset Ox of X
such that
[
Ox = X (Ox may be empty for some x ):
x ∈X
Q
Then there exists a point x ∈ T (x) = ∈I T (x); i.e.; x ∈ T (x) for each  ∈ I; where P (x) = x
and P is the projection of X onto X .

The object of this paper is to obtain a noncompact version of Theorem 1.2 which will coincide
with Theorem 1.1 when I = {1} and apply this theorem to prove the existence under conditions
similar to those in [13] of a maximal element and an equilibrium point of, respectively, a qualitative
game and an abstract economy which will be shortly described.
Since we intend to apply our xed point theorem to mathematical economics and games, a brief
introduction to the relevant part of these are now in order. Shafer and Sonnenschein [9] extended
the theorem of Debreu on the existence of equilibrium in a generalized N -person game [3] or an
abstract economy [1]. In essence, Shafer and Sonnenschein maintained the spirit of the pioneering
works of Debreu, Arrow [1], Mas-Colell [7] and Gale and Mas-Colell [5,6]. Bewley [2] proved
the existence of equilibrium of an economy with innite dimensional commodity space. In recent
years, many authors (e.g., [10,13–18], etc.) have proved the existence of equilibrium of an abstract
economy with innite dimensional commodity space and innite agents.
Now, following Debreu and Shafer and Sonnenschein we will describe an abstract economy or a
generalized game by E={X ; A ; U :  ∈ I}. Where I is nite or innite (countableQor uncountable)
set of agents or players; and for each  ∈ I; X is the choice or strategy set; A : X = ∈I X → 2X is
the constraint correspondence (setvalued mapping) and U : X → R is the utility or payo function.
For
Q
each  ∈ I; X will be a nonempty subset of a topological vector space. We denote the product
∈I; 6∈ X by X− and a generic element of X− by x− .
An abstract economy instead of being given by {X ; A ; U :  ∈ I } may as well be given by
E = {X ; A ; Q :  ∈ I}, where for each  ∈ I; Q : X → 2X is the preference correspondence. The
relationship between the preference correspondence Q and the utility function U can be expressed
by the denition
Q (x) = {y ∈ X : U ([y ; x− ]) ¿ U (x)};
where for each  ∈ I; x− is the projection of x onto X− and [y ; x− ] is the point of X whose th
coordinate is y .
In the case of the economy being given by E = {X ; A ; U :  ∈ I}, a point x ∈ X is called an
equilibrium point or a generalized Nash equilibrium point of the economy E if
U (x ) = U [x ; x − ] = sup U [z ; x − ]
z ∈A (x )

for each  ∈ I where x  and x − are, respectively, projections of x onto X and X− . In this case the
equilibrium point is the natural extension of the equilibrium point introducted by Nash [8]. Now, let
 S.P. Singh et al. / Journal of Computational and Applied Mathematics 113 (2000) 65–71 67

E = {X ; A ; U :  ∈ I} be an abstract economy and let for each  ∈ I; Q be obtained as above.

Then it can be easily checked that a point x ∈ X is an equilibrium point of E if and only if for each
 ∈ I; Q (x) ∩ A (x) =  and x  ∈ A (x). Thus given an abstract economy E = {X ; Q ; A :  ∈ I}
we can dene an equilibrium point of E as follows: A point x ∈ X is said to be an equilibrium
point of the abstract economy E = {X ; Q ; A :  ∈ I} if for each  ∈ I; Q (x) ∩ A (x) =  and
x  ∈ A (x) where x  is the projection of x onto X .
Given an abstract economy E = {X ; Q ; A :  ∈ I}, for each x ∈ X we dene I (x) = { ∈
I : Q (x) ∩ A (x) 6= }. Assume that for each x ∈ X and for each  ∈ I; x 6∈ convex hull of Q (x).
For each  ∈ I , we dene the set valued mapping T : X → 2X by
(
co Q (x) ∩ A (x) if  ∈ I (x);
T (x) =
A (x) if  6∈ I (x);
where co Q (x) denotes the convex hull of Q (x).
Then it is easy to see that x ∈ X is an equilibrium point of the economy
Q
E if and only if x is a
xed point of the set valued mapping T : X → 2X dened by T (x) = ∈I T (x).

2. Generalized xed point theorem

Theorem 2.1. Let {X :  ∈ I } be a familyQof nonempty convex sets; each in a topological vector
space E ; where I is an index set. Let X = ∈I X . For each  ∈ I; let T : X → 2X be a set valued
mapping such that
(i) for each x ∈ X; T (x) is a nonempty convex subset of X ;
and
(ii) for each x ∈ X ; T−1 (x ) = {y ∈ X : x ∈ T (y)} contains a relatively open set Ox of X (Ox
may empty for some x ).
(∗) We further assume that for each Q
 ∈ I; there is a nonempty compact convex subset X̂  of
X andSa nonempty subset X0 of X̂ = ∈I X̂  such that either; for each  ∈ I;
(a) x ∈X Ox = X ;
and T
(b) the set D = x∈X0 OPc  (x) = ; where P is the projection of X onto X .
or T S
(a)0 x∈X ∈I OPc  (x) = ; and
T S
(b)0 D = x∈X ∈I OPc  (x) is compact if nonempty; where Q
A denotes the closure of A.
Then there is a point x ∈ X such that x ∈ T (x) = ∈I T (x); i.e. x ∈ T (x) for each  ∈ I;
where P (x) = x for each  ∈ I .

Before proving the theorem we rst note that the condition (a)0 implies the condition (a). To see
this, suppose (a)0 holds. Then for each
[
 ∈ I; Oxc = OPc  (x) ⊂ OPc  (x) :
∈I
T T S
Hence x∈X OPc  (x) ⊂ x∈X ∈I OPc  (x) .
68 S.P. Singh et al. / Journal of Computational and Applied Mathematics 113 (2000) 65–71

S S T
Thus x ∈X Ox = x∈X OP (x) = [ x∈X OPc  (x) ]c ⊃ X .

Proof. We rst prove the theorem under (a) and (b). Then for each x ∈ X̂ and each  ∈ I; T (x) ∩ X̂  6=
. Indeed if T (x0 ) ∩ X̂  =  for some x0 ∈ X̂ and some  ∈ I , then T
for all x ∈ X̂ ; P (x) = x 6∈ T (x0 ),
i.e. x0 6∈ T−1 (x ) ⊃ Ox . Thus x0 ∈ [T−1 (x )]c ⊂ Oxc . Hence x0 ∈ x∈X0 OPc  (x) = D as X0 ⊂ X̂ , which
contradicts that D = . Thus for each  ∈ I , we can dene a set valued mapping T̂  : X̂ → 2Xˆ 
by T̂  (x) = T (x) ∩ X̂  ; x ∈ X̂ . It is easy to see that for each  ∈ I , and each x ∈ X̂ ; T̂  (x) is
−1
a nonempty convex subset of X̂  . Now for each x ∈ X̂  ; T̂  (x ) = {x ∈ X̂ : x ∈ T̂  (x)} = {x ∈
X̂ : x ∈ T (x)T∩ X̂  } = T−1 (x ) ∩ X̂ contains S
a relatively open set Ox S ∩ X̂ = Ôx in X̂ . Also since for
each  ∈ I , x∈X0 OPc  (x) = , we have x∈X0 OP (x) = X and hence x∈X OP (x) = X . Thus for each
S S
 ∈ I; x∈Xˆ ÔP (x) = x∈Xˆ (OP (x) ∩ X̂ ) = X̂ . Hence by Theorem 1.2, there is a point x ∈ X̂ such that
Q Q
x ∈ T̂ (x) = ∈I T̂  (x) ⊂ ∈I T (x) = T (x). Thus we have proved the theorem in this case.
Now we prove the theorem under condition (a)0 Q and (b)0 . In this case we will prove the theorem
by contradiction. S
If possible, we assume that T = ∈I T has no xed point. For each x ∈ X , we
set F(x) = ∈I OPc  (x) . Then under our assumption F(x) 6=  for each x ∈ X . For, indeed F(x) = 
for some x ∈ X implies x 6∈ F(x), i.e. x 6∈ OPc  (x) for all  ∈ I , i.e. x ∈ OP (x) ⊂ T−1 (P (x)) for  ∈ I ,
i.e. P (x) ∈ T (x) for all  ∈ I , i.e. x ∈ T (x) which contradicts our assumption. More Sn
generally, the
convex hull P of each nite S
subset {x ;
1 2x ; : : : ; x n } of X is contained in the union i=1 F(x
Pi
). To see
this let x = ni=1 i xi 6∈ ni=1 F(xi ) for nite subset {xS1 ; x2 ; : : : ; x n } of X with i ¿0 and ni=1 i = 1.
This implies that for each i = 1; 2; : : : ; n; x 6∈ F(xi ) = ∈I OPc  (xi ) , i.e. x ∈ OP (xi ) ⊂ T−1 (P (xi )) for all
 ∈ I , i.e. P (xi ) ∈ T (x)Pfor all  ∈ I . But P
since T (x) is convex, we conclude that P (x) ∈ T (x)
for all  ∈ I (note that ni=1 i P (xi ) = ni=1 P (i xi ) = P (x)). This means that x is a xed point
of T . Thus we have a contraction. T
Next, we prove that for each nite subset {x1 ; x2 ; : : : ; x n } of X; x∈K F(x) 6= , where K is the
T
convex hull of X̂ ∪ {x1 ; x2 ; : : : ; x n }. If possible, let us assume that x∈K F(x) = . Then for each
y ∈ K, the set H (y) = {x ∈ K: y 6∈ F(x)} is a nonempty subset of K. Now for each  ∈ I , we
dene a set valued mapping H : K → 2K by H (y) = {P (x) ∈ K : y 6∈ OPc  (x) }, where P (K) = K .
Since H (y) 6= , there exists x ∈ K such that y 6∈ F(x). Hence y 6∈ OPc  (x) and P (x) ∈ K .
Thus H (y) 6= . Now for each P (x) ∈ K ; H−1 (P (x)) = {y ∈ K: P (x) ∈ H (y)} = {y ∈
K: y 6∈ OP(x c
)
} = OP (x) ∩ K = O P (x) , which is a relatively open set in K. Now for each  ∈ I ,
we dene a set valued mapping J : K → 2K by J (y) = convex hull of H (y); y ∈ K. Thus for
each y ∈ K; J (y) is a nonempty convex subset of K and since J (y) ⊃ H (y), it follows that for
each
T
P (x) ∈ K ; J−1 (P (x)) ⊃ H−1 S
(x) ⊃ ÔP (x) which S
isTa relatively open set in K. Furthermore
c
F(x) =  implies that X = [F(x)] = x∈K [ ∈I OP (x) ], which, in turn, implies X =
Sx∈K S x∈K
S S S
x∈K OP (x) = P (x)∈K OP (x) . Thus K = [ P (x)∈K OP (x) ] ∩ K = P (x)∈K [OP (x) ∩ K] = P (x)∈K O P (x) .

Hence by Theorem 1.2 there exists a point x ∈ K such that P (x) ∈ J (x) for each  ∈ I . This implies
that there exist Pm
points y1 ; y2 ; : : :P; ym in K such that for each  ∈ I; P (yi ) ∈ H (x); i = 1; 2; : : : ; m,
where x = i=1 i yi ; i ¿0 and mi=1 i = 1. This S
implies that for each i P = 1; 2; : : : ; m andSeach  ∈
I; x 6∈ OPc  (yi ) , i.e. for each i = 1; 2; : : : ; m; x 6∈ ∈I OPc  (yi ) = F(yi ). Thus mi=1 i yi = x 6∈ mi=1 F(yi )
which contradicts our established fact Sthat the convex hull of each nite subset T
{y1 ; y2 ; : : : ; ym } is
m
contained
T
in the corresponding
T
union i=1 F(y i ). Thus we have proved that x∈K F(x) 6= . Hence
D ∩ ( ni=1 F(xi )) ⊃ x∈K F(x) 6=  as X0 ∪T {x1 ; x2 ; : : : ; x n } ⊂ K. What we have Tn
then established is that
n
for each nite subset {x1 ; x2 ; : : : ; x n } of X; i=1 (D ∩ F(xi )) 6=  and hence i=1 (D ∩ F(xi )) 6= . Now
 S.P. Singh et al. / Journal of Computational and Applied Mathematics 113 (2000) 65–71 69

0
T
since by (b)
T
D is compact, F(x) ∩ D is closed for each x ∈ X . Hence x∈X (D ∩ F(x)) 6=  and,
therefore, x∈X
Q
F(x) 6=  which contradicts our assumption (a)0 .
Thus T = ∈I T must have a xed point.

Remark 2.1. (i) If I = {1}, then it can be seen that the above theorem reduces to Theorem 1.1.
(ii) We also note that the above theorem under (a) and (b) contains Theorem 1.2. To see this
let
T
X = X0 in Theorem 1.2. By condition (ii) of Theorem 1.2, we obtain that for each  ∈ I; D =
c
x∈X OP (x) = .
(iii) The above theorem under (a)0 and (b)0 can also be proved by the Fan–Knaster–Kuratowski–
Mazurkiewicz theorem [4,12].

3. Fixed point theorem and existence of maximal elements and equilibrium points

In this section we apply our xed point theorem to prove the existence of maximal elements and
equilibrium points of an abstract economy. As in the previous section I will denote an indexing set,
nite or innite, i.e., I is the set of agents. Q
For each  ∈ I; X will denote a nonempty set in a topological vector space E ; Q : X = ∈I X →
2X a set valued mapping (preference correspondence) and A : X → 2X a set valued mapping
(constraint correspondence).
{X ; Q :  ∈ I } will be called a qualitative game and {X ; Q ; A :  ∈ I } and abstract economy.
A point x ∈ X is called a maximal element of the game {X ; Q :  ∈ I } if Q (x)= for each  ∈ I
and a point x ∈ {x } ∈ X is called an equilibrium point of the abstract economy {X ; Q ; A :  ∈ I }
if, for each  ∈ I; x ∈ A (x) and Q (x) ∩ A (x) = .

Theorem 3.1. Let E = {X ; Q ; A :  ∈ I} be an abstract economy such that for each  ∈ I; the
following conditions hold:
(i) X is convex;
(ii) for each x ∈ X; A (x) is nonempty and convex valued;
(iii) for each x ∈ X ; {Qa−1 (x ) ∪ F } ∩ A−1
 (x ) contains a relatively open set Ox of X; (Ox may
be empty for some x ); where F = {x ∈ X : Q (x) ∩ A (x) 6= }.
(iv) for each x = {x } ∈ X; x 6∈ co Q (x);
(v) the condition (∗) of Theorem 2:1 holds with these Ox .
Then E has an equilibrium point.

Proof. For each  ∈ I , let G = {x ∈ X : Q (x) ∩ A (x) 6= } and for each x ∈ X , let I (x) = { ∈
I : Q (x) ∩ A (x) 6= }.
Now for each  ∈ I , we dene the set valued mapping T : X → 2X by
(
co P (x) ∩ A (x) if  ∈ I (x) i:e:; if x ∈ G ;
T (x) = A (x) if ;  6∈ I (x):
70 S.P. Singh et al. / Journal of Computational and Applied Mathematics 113 (2000) 65–71

Then for each x ∈ X; T (x) is nonempty and convex valued. Also for each y ∈ X , it can be easily
checked that
T−1 (y ) = [{(co Q )−1 (y ) ∩ A−1 −1
 (y )} ∩ G ] ∪ [A (y ) ∩ F ]
⊃ [{Q−1 (y ) ∩ A−1 −1
 (y )} ∩ G ] ∪ [A (y ) ∩ F ]
= [Q−1 (y ) ∩ A−1 −1
 (y )] ∪ [F ∩ A (y )]
= [Q−1 (y ) ∪ F ] ∩ A−1
 (y ):

The rst inclusion follows from the fact that as Q (x) ⊂ co Q (x) for each x ∈ X; Q−1 (y ) ⊂
(co Q )−1 (y ) for each y ∈ X . Hence by virtue of condition (iii) for each y ∈ X ; T−1 (y )
contains an open subset Oy of X . Thus all the conditions of Theorem 2.1 are satised. Hence by
Theorem 2.1 there exists a point x = {x } ∈ X such that x ∈ T (x) for each  ∈ I . It follows from
(iv) and denition of T that x is an equilibrium point.

Theorem 3.2. Let = {X ; Q :  ∈ I } be a qualitative game such that for each  ∈ I; the following
conditions hold:
(i) X is convex;
(ii) for each x ∈ X ; {P−1 (x ) ∪ F } contains a relatively open subset Ox of X where
F = {x ∈ X : Q (x) = };
(iii) for each x = {x } ∈ X; x 6∈ co Q (x);
(iv) the condition (∗) of Theorem 2:1 holds with these Ox .
Then there is a maximal element of the game .

Proof. For each  ∈ I , if we dene the set valued mapping A : X → 2X by A (x) = X ; x ∈ X ,

then the theorem will follow from Theorem 3.1.

References

[1] K. Arrow, G. Debreu, Existence of equilibrium for a competitive economy, Econometrica 22 (1954) 265–290.
[2] T.F. Bewley, Existence of equilibria in economics with innitely many commodities, J. Econom. Theory 4 (1972)
514–540.
[3] G. Debreu, A social equilibrium existence theorem, Proceedings of the National Academy of Sciences of the USA,
vol. 38, 1952, pp. 886 –893.
[4] K. Fan, Some properties of convex sets related to xed point theorems, Math. Ann. 266 (1984) 519–537.
[5] D. Gale, A. Mas-Colell, An equilibrium existence for a general model without ordered preferences, J. Math. Econom.
2 (1975) 9–15.
[6] D. Gale, A. Mas-Colell, Corrections to an equilibrium existence theorem for a general model without ordered
preference, J. Math. Econom. 6 (1979) 297–298.
[7] A. Mas-Colell, An equilibrium existence theorem without complete or transitive preferences, J. Math. Econom. 1
(1974) 237–246.
[8] J.F. Nash, Equilibrium points in N -person games, Proceedings of the National Academy of Sciences of the USA,
vol. 36, 1950, 48–49.
[9] W. Shafer, H. Sonnenschein, Equilibrium in abstract economies without ordered preferences, J. Math. Econom. 2
(1975) 345–348.
[10] K.K. Tan, X.Z. Yuan, A minimax inequality with applications to existence of equilibrium points, Bull. Australian
Math. Soc. 47 (1993) 483–503.
 S.P. Singh et al. / Journal of Computational and Applied Mathematics 113 (2000) 65–71 71

[11] E. Tarafdar, On nonlinear variational inequalities, Proceedings of the American Mathematical Society, vol. 67, 1977,
pp. 95 –98.
[12] E. Tarafdar, A xed point theorem equivalent to the Fan–Knaster–Kuratowski–Mazurkiewicz Theorem, J. Math.
Anal. Appl. 128 (1987) 475–479.
[13] E. Tarafdar, A xed point theorem and equilibrium point of an abstract economy, J. Math. Econom. 20 (1991)
211–218.
[14] S. Toussaint, On the existence of equilibria in economics with innitely many commodities and without ordered
preferences, J. Econom. Theory 33 (1984) 98–115.
[15] C.I. Tulcea, On the approximation of upper-semicontinuous correspondences and the equilibrium of generalized
games, J. Math. Anal. Appl. 136 (1988) 267–289.
[16] N. Yannelis, Maximal elements over noncompact subsets of linear topological spaces, Econom. Lett. 17 (1985)
133–136.
[17] N. Yannelis, N. Prabhakar, Existence of maximal elements and equilibria in linear topological spaces, J. Math.
Econom. 12 (1983) 233–245.
[18] X.Z. Yuan, E. Tarafdar, Maximal elements and equilibria of generalized games for condensing correspondences, J.
Math. Anal. Appl. 203 (1996) 13–30.