Journal of Mathematical Economics 71 (2017) 36–48

Contents lists available at ScienceDirect

Journal of Mathematical Economics

journal homepage: www.elsevier.com/locate/jmateco

Cones with semi-interior points and equilibrium

Achille Basile a , Maria Gabriella Graziano a,b , Maria Papadaki c , Ioannis A. Polyrakis c,∗
a
Dipartimento di Scienze Economiche e Statistiche, Università degli Studi di Napoli Federico II, Italy
b
CSEF, Italy
c
School of Applied Mathematics and Physical Sciences, Department of Mathematics, National Technical University of Athens, Greece

article info abstract

Article history: We study exchange economies in ordered normed spaces (X , ∥ · ∥) where agents have possibly different
Received 2 May 2016 consumption sets. We define the notion of semi-interior point of the positive cone X+ of X , a notion weaker
Received in revised form than the one of interior point and we study the existence of equilibrium in the case where X+ has semi-
2 February 2017
interior points. In Section 4, we study the case where X+ has interior points and we prove a second welfare
Accepted 22 March 2017
Available online 12 April 2017
theorem and the existence of equilibrium. Subsequently we apply these results in the case where X+ has
semi-interior points. In the case of semi-interior points the supporting price vectors are continuous with
Keywords:
respect to a new norm ||| · ||| on X which is strongly related with the initial norm and the ordering, and in
Cones some sense can be considered as an extension of the norm adopted in classical equilibrium models. Many
Equilibrium examples of cones in normed and Banach spaces with semi-interior points but with empty interior are
Ordered spaces provided, showing that this class of cones is a rich one. In the last section we apply our results to strongly
Second welfare theorem reflexive cones.
Strongly reflexive cones © 2017 Elsevier B.V. All rights reserved.

1. Introduction
In this article we study equilibrium with reference to a new
class of cones in normed spaces, i.e. cones with semi-interior points.
As it is known, when a normed space (X , ∥ · ∥) ordered by the
cone P, i.e. X+ = P, represents the commodity space of a pure
exchange economy, then the existence of interior points of P is
a condition with many important implications, see for example
in Bewley (1972), Mas-Colell (1986), Aliprantis et al. (1990), Mas-
Colell and Richard (1991) and Aliprantis et al. (2002).
In particular, it is well known the relevance in general equilib-
rium theory of the assumption that the total initial endowment
ω is an interior point of X+ . Many results and proofs are based
on this assumption and its weakening. Indeed, the interior point
condition is not satisfied in several circumstances, what induces
scholars to assume suitable conditions in order to guarantee wel-
fare and existence theorems (see for example Debreu, 1954, Flo-
renzano, 1983; Bewley, 1972; Mas-Colell, 1986; Mas-Colell and
Richard, 1991; Aliprantis et al., 1990; Khan et al., 1999; Podczeck
∗ interior point of P is an interior point of P with respect to the norm
Corresponding author.
E-mail addresses: basile@unina.it (A. Basile), mgrazian@unina.it
(M.G. Graziano), papadaki@math.ntua.gr (M. Papadaki), ypoly@math.ntua.gr
(I.A. Polyrakis).
http://dx.doi.org/10.1016/j.jmateco.2017.03.002
0304-4068/© 2017 Elsevier B.V. All rights reserved.
and Yannelis, 2008 and new results in this direction in Xanthos,
2014b and Xanthos, 2014a).
Here we focus on the notion of semi-interior point of a cone P in
a normed space X , a notion that is weaker than the one of interior
point of P. By definition, for an interior point x0 of P, there exists
a real number ρ > 0 so that x0 + ρ U ⊆ P, where U is the unit
ball of X . Therefore, we have x0 + ρ U+ ⊆ P and x0 − ρ U+ ⊆ P,
where U+ = U ∩ P is the positive part of the unit ball of X defined
by the cone P. The first of the previous two inclusions is satisfied
for any vector x0 ∈ P, because the elements of ρ U+ are positive
but the second one is not always true. If a point x0 of P satisfies this
condition, i.e. if x0 − ρ U+ ⊆ P for some real number ρ > 0, then
we say that x0 is a semi-interior point of P.
When the commodity space is an ordered normed space
(X , ∥ · ∥) whose positive cone has semi-interior points, we prove
a second welfare theorem and the existence of equilibria. In both
cases, the supporting prices are continuous with respect to a new
norm of the commodity space, denoted by |||·|||. The latter is inspired
by the definition of semi-interior point of P, and defined by means
of the initial norm and the positive cone X+ . Indeed, ||| · ||| is the
Minkowski functional defined by the convex hull, co(U+ ∪ (−U+ )),
of the union of the sets U+ and −U+ . It comes out that any semi-
||| · ||| of X .
We give several examples of cones in normed spaces with semi-
interior points and empty interior.
 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48
Concerning the norm ||| · |||, it is worth to remark that if X is a
Banach space and the cone P is closed and generating, the initial
norm of X and the new norm ||| · ||| are equivalent, Proposition 3.3.
So, in the classical equilibrium models, as for example in the
finite dimensional Arrow–Debreu model and in the Banach lattice
equilibrium models these two norms are equivalent, therefore
both of them can be considered as norm of the commodity
space, i.e. as norm of these general equilibrium models. If X is
a normed space, these norms coincide on the positive cone but
the equivalence is no longer true. However in this case, by the
definition of the new norm ||| · ||| and from the above remarks, the
norm ||| · ||| of X can be considered as a natural norm for the study
of equilibrium.
In Section 4 we study equilibrium in ordered normed spaces X
whose positive cone X+ has nonempty interior, assuming that the
consumption sets of the consumers are different subcones Pi of X+ .
Following standard methods we prove a second welfare theorem,
Theorem 4.4 and the existence of quasi-valuation equilibrium,
Theorem 4.10. For the proof of equilibrium we follow the
method of Mas-Colell, Mas-Colell (1986), adopted in our case of
different consumption sets but without any lattice property of the
commodity space. We use conditions like the standard ones but we
have weakened the continuity of the utility functions by assuming
only radial or linear continuity.1
In Section 5, we extend the results of Section 4 to the case where
the cone X+ has semi-interior points. So we prove a second welfare
theorem and the existence of quasi-valuation equilibria, in the case
where X+ has semi-interior points. As we have noted above, in both
cases, the supporting price is continuous with respect to the new
norm ||| · ||| on X which is strongly related with the initial norm and
the ordering, and in some sense can be considered as an extension
of the norm adopted in classical equilibrium models.
Finally in Section 6 we apply our results in the case where the
positive cone X+ of X is strongly reflexive. Examples 6.3 and 6.4
are examples of equilibrium in economies with strongly reflexive
consumption cones.
In the Appendix we give the essential notions and results from
the theory of partially ordered linear spaces which are needed in
this paper.
2. Cones with semi-interior points
In this section we define the notion of semi-interior point of
a cone. This notion, the pioneering Example 2.5 as well as the
notion of the new norm of the next section, have been defined by
I. Polyrakis and have been announced in Paris 2014, during the XXII
European Workshop on General Equilibrium Theory, in his talk
‘‘Cones with semi-interior points and a second welfare theorem’’.
Let X be a normed space and let P be a cone of X . Suppose that X
is ordered by the cone P. We shall denote by U = {x ∈ X | ∥x∥ ≤ 1}
the unit ball of X and by U+ the set
U+ = U ∩ P .
We shall call U+ the positive part of the unit ball of X or the positive
part of the unit ball of X defined by P, if this clarification is needed.
Definition 2.1 (I. Polyrakis). The vector x0 ∈ P is a semi-interior
point of P if there exists a real number ρ > 0 so that x0 −ρ U+ ⊆ P.
Clearly, any interior point of P is a semi-interior point of P. The
next characterization will be useful in the sequel.
1 A similar form of weak continuity for preferences is adopted in the classical
paper (Debreu, 1954) of G. Debreu, see in Remark 4.12.
37
Proposition 2.2. Suppose that P is a cone of a normed space X . The
vector x0 ∈ P is a semi-interior point of P if and only if there exists a
real number k > 0 so that kx0 is an upper bound of the positive part
U+ of the unit ball of X .
Proof. Suppose that x0 is a semi-interior point of P. Then x0 −
ρ U+ ⊆ P, which implies that x0 is an upper bound of ρ U+ and
1
x is an upper bound of U+ .
ρ 0
For the converse, if we suppose that kx0 is an upper bound of
U+ then we can show that x0 − 1k U+ ⊆ P and x0 is a semi-interior
point of P. 
In the next example, X is a normed space ordered by a closed
and generating cone X+ without semi-interior points. Note that in
this example X is not complete.
Example 2.3. Let X = c00 be the space of finite real sequences
i.e. the set of sequences a = (a(i)) with a(i) ̸= 0 for at most a finite
number of i and suppose that X is ordered by the pointwise order-
ing. Then X+ = {a ∈ X | a(i) ≥ 0 for any i} is the positive cone of
X . X is a normed space with norm ∥a∥ = max{|a(i)| | i ∈ N}.
X+ does not have semi-interior points. Indeed, if we suppose
that a = (a(i)) is a semi-interior point of X+ , there exists a real
number k > 0 so that ka is an upper bound of U+ . Since ei ∈ U+
for any i, where ei (j) = 0 for any j ̸= i and ei (i) = 1, we have that
a(i) ≥ 1k ei (i) = 1k > 0 for any i, a contradiction. Hence X+ does
not have semi-interior points.
For the Krein–Smulian theorem below, see for example in
Jameson (1970, Theorem 3.5.2).
Theorem (Krein–Smulian). If the positive cone P of an ordered Ba-
nach space X is closed and generating, then P gives an open decom-
position of X , i.e. there exists a real number a > 0 so that aU ⊆
U+ − U+ .
By this theorem we obtain the next proposition:
Proposition 2.4. If X is a Banach space ordered by the closed and
generating cone P, then any semi-interior point of P is an interior
point of P.
Proof. Let x0 be a semi-interior point of P. Then x0 − ρ U+ ⊆ P, for
some ρ > 0, therefore
(x0 − ρ U+ ) + ρ U+ = x0 + ρ(U+ − U+ ) ⊆ P .
By the Krein–Smulian theorem, there exists a real number a > 0
so that aU ⊆ U+ − U+ , therefore we have x0 + aρ U ⊆ x0 + ρ(U+ −
U+ ) ⊆ P and x0 is an interior point of P. 
The next example is crucial for the development of the theory
of semi-interior points. It shows the shape of the unit ball of an
ordered normed space whose positive cone does not have interior
points but has semi-interior points. In this example we can see
that the norm takes much lower values in some directions of non
positive and non negative vectors than in the direction of positive
and negative ones. So the unit ball is ‘‘compressed’’ to the direction
of positive or negative vectors and ‘‘flattened’’ to some direction of
non positive and non negative vectors. This shape of the unit ball
gives a geometrical intuition and explains (in some sense) how the
positive cone of the space fails to have interior points but has semi-
interior points.
Example 2.5. Let Xn be the space R2 ordered by the pointwise
ordering and with norm ∥ · ∥n having, as unit ball, the polygon Dn
of R2 with vertices
(1, 0), (0, 1), (−n, n), (−1, 0), (0, −1), (n, −n).
38 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48

It is easy to show, by taking the Minkowski’s functional of Dn , that we have ∥x∥ ≤ 1 and 0 ≤ x1 e1 ≤ x and by taking norms we have
this norm is given by the formula: 0 ≤ x1 ≤ 1, therefore 0 ≤ x ≤ y. So we have that UK+ ⊆ [0, y]
which implies that UK+ is compact because each order interval of
∥(x, y)∥n = |x| + |y|, if xy ≥ 0,
a Banach lattice with a positive basis is compact, see in Singer
and (1970, Theorem 16.3). Hence the cone K is strongly reflexive. The
n−1 set UK+ ∪ (−UK+ ) is compact, therefore its convex hull
∥(x, y)∥n = max{|x|, |y|} − min{|x|, |y|}, if xy < 0.
n V = co(UK+ ∪ (−UK+ )),
Suppose that E is the space of sequences x = (xn )n∈N so that is also compact. So the set Ω = 3y + V is compact and we consider
xn = (xn1 , xn2 ) ∈ Xn and ∥xn ∥n ≤ mx for any n (the real number the cone
mx > 0 is depending on x). 
Assume that E is ordered by the cone P = {x = (xn ) ∈ E | xn ∈ P = tΩ
R2+ for any n} and that it is equipped with the norm t ≥0

∥x∥∞ = supn∈N ∥xn ∥n . generated by Ω . It is easy to show that the cone P is strongly
reflexive. Indeed, for any z = 3y + x ∈ Ω , where x ∈ V we have
Suppose also that X = P − P is the subspace of E generated by the ∥x∥ ≤ 1 therefore 2 ≤ ∥z ∥ ≤ 3∥y∥ + 1. So the set
cone P and suppose that X is ordered by the cone X+ = P. 
Let 1 = (xn ) with xn = (1, 1) for any n. Then 1 is not an interior R= tΩ
point of X+ . Indeed, if for any natural number m we take the vector 0≤t ≤1
y = (yn ) of X with ym = (−2, 2) and yn = (0, 0) for any n ̸= m,
2 is compact. For any w ∈ UP+ = P ∩ U we have that w = tz where
then it is easy to show that ∥y∥∞ = m and 1 + y ̸∈ X+ . Therefore ∥w∥ ∥w∥
1 + ρ U ̸⊆ X+ , for any ρ > 0, and 1 is not an interior point of X+ . z ∈ Ω therefore ∥w∥ = t ∥z ∥ hence t = ∥z ∥ ≤ 2 ≤ 12 . Hence
Similarly, we can show that any vector x ∈ X+ is not an interior any vector of UP+ is of the form z = 3ty + tx, where x ∈ V and
point of X+ , therefore X+ has empty interior. The positive part of 0 ≤ t ≤ 12 . This implies that UP+ ⊆ R and the set UP+ as a closed
the closed unit ball U of X is the set subset of R is compact. To show that 3y is a semi-interior point of P
U+ = {x ∈ X+ | ∥x∥∞ ≤ 1} it is enough to show that 3y is an upper bound of UP+ in the ordering
of P or equivalently that for any z ∈ UP+ we have that 3y − z ∈ P.
and it is easy to see that for any x = (xn ) ∈ U+ we have ∥xn ∥n = Suppose that z ∈ UP+ . As we have shown above it is of the form
x1n + x2n ≤ 1. So,
z = 3ty + tx, where x ∈ V and 0 ≤ t ≤ 21 , hence
1 − U+ ⊆ X+  
t
and 1 is a semi-interior point of X+ . 3y − z = 3(1 − t )y + t (−x) = (1 − t ) 3y + (−x) ∈ P ,
1−t
In particular, by Proposition 2.4, the space X is not complete.
because 1−t (−x) ∈ V for any x ∈ V and any 0 ≤ t ≤ 12 . Note that
t
In the next two examples P is a strongly reflexive cone of a t
normed space X , with semi-interior points but with empty interior. 0 ≤ 1− t
≤ 1 for any 0 ≤ t ≤ 21 . Therefore 3y is a semi-interior
In both examples the cone P generates a dense subspace of X , point of P. To show that P − P = X we remark that e1 ∈ UK+ ⊆ V ,
i.e. P − P = X . Recall that a cone P of X is strongly reflexive if therefore e1 = (3y + 0) − (3y − e1 ) ∈ P − P. Also
the positive part U+ = U ∩ P of U defined by P is ∥ · ∥-compact.   
∥e1 + ae2 ∥ e1 e1 + ae2
e2 = 3y − − 3y − ,
Example 2.6. Let E be an infinite dimensional Banach lattice with α ∥e1 + ae2 ∥ ∥e1 + ae2 ∥
a positive Schauder basis {ei }. Without loss of generality, we may therefore
suppose that this basis is normalized, i.e. ∥ei ∥ = 1, for any i.
∥e1 + ae2 ∥
For example, E can be one of the spaces c0 , ℓ1 or one of the e2 ∈ (P − P ) = P − P .
reflexive spaces ℓp with 1 < p < ∞. In Theorem 5.7 of Casini α
et al. (2013), it is proved that E+ contains a strongly reflexive cone By continuing this process we can show that ei ∈ P − P for any
which generates a dense subspace of E. In the present example, by i, therefore P − P is dense in E. If we suppose that x is an interior
a similar way we construct a strongly reflexive cone P ⊆ E+ with point of P we get a contradiction as follows: x + δ U is contained in
semi-interior points and empty interior with P − P = X . P for some real number δ > 0, therefore x + δ U is contained in a
By definition of Schauder basis, any vector x ∈ E has a unique positive multiple of UP+ , therefore the unit ball U of E is compact,
∞
expansion x = i=1 xi ei , where the real numbers xi are the contradiction. Therefore the cone P does not have interior points.
coordinates of x in the basis {ei }. Moreover, since the basis {ei } is
positive, for any vector x of E we have: x ≥ 0 if and only if xi ≥ 0 Example 2.7. In Casini et al. (2013, Example 5.9) a strongly
for any i. We start with a fixed real number α ∈ (0, 1) and we reflexive cone P of L+ 1 [0, 1] is determined which generates a dense
consider the vector subspace in L+ 1 [ 0, 1] , i.e. P − P = L+ 1 [0, 1]. Also the positive part
∞
 U+ of the unit ball of L+ 1 [ 0, 1] defined by P is dominated by the
y= α i−1 ei , vector y = 2T (η) in the ordering of L+ [ 0, 1], where T (η) is a vector
1
i=1
of P which is defined in Example 5.9 of Casini et al. (2013).
of E+ . Then we have ∥y∥ ≥ 1 because y ≥ e1 and E is a Banach Here we change the notations. So we denote the cone P of Casini
lattice. We consider the closed subcone et al. (2013, Example 5.9) by K and the set U+ by UK+ . Also we will
denote by V the convex hull of the set UK+ ∪ (−UK+ ), and by P the
K = {x ∈ X | 0 ≤ x ≤ x1 y}
cone of L+1 [0, 1] which is generated by the set Ω = 3y + V . As
i=1 α ei ∈ K , for any n. Moreover, for any x ∈ K
n i−1
of E+ . Then in the previous example we can show that P is strongly reflexive
we have that 0 ≤ x ≤ x1 y, therefore for any x ∈ K we have and the vector 3y is an upper bound of the positive part UP+ of the
∥x∥ ≤ x1 ∥y∥ because E is a Banach lattice. For any x ∈ UK+ = U ∩ K unit ball of L1 [0, 1] defined by P, in the ordering of P. Hence 3y is
 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48
a semi-interior point of P. Moreover, the space P − P generated
by P and the space K − K generated by K , coincide. Indeed, any
vector of P is of the form 3ty + tx with t ≥ 0 and x ∈ V . Also
any vector of K − K is of the form t1 x1 − t2 x2 with t1 , t2 ≥ 0 and
x1 , x2 ∈ UK+ or equivalently of the form tx, with t ≥ 0 and x ∈ V
because t1 x1 − t2 x2 = (t1 + t2 )( t +1 t x1 + t +2 t (−x2 )).
t t
1 2 1 2
So for any vector z = 3t1 y + t1 x1 −(3t2 y + t2 x2 ) of P − P we have
z = (3t1 y − 3t2 y)+(t1 x1 − t2 x2 ) ∈ (K − K )+(K − K ) = K − K . Also
for any vector w = tx of K − K we have w = (3y + tx)− 3y ∈ P − P.
Therefore P − P = K − K and P − P = K − K = L+ 1 [0, 1].
Theorem 2.8. If X is a normed space ordered by the generating cone
P, then any semi-interior point of P is an order unit of X .
Proof. Suppose that x0 is a semi-interior point of P. Then by
Proposition 2.2, there exists a real number k > 0 so that kx0 is
an upper bound of the positive part U+ of the unit ball of X . Let
x ∈ X . Then x = x1 − x2 , where x1 , x2 ∈ P and suppose that
a = max{∥x1 ∥, ∥x2 ∥}. We have
x1
x = x1 − x2 ≤ x1 = ∥x1 ∥ ≤ ∥x1 ∥kx0 ≤ akx0
∥x 1 ∥
and similarly
−x 2
x = x1 − x2 ≥ −x2 = ∥x2 ∥ ≥ −∥x2 ∥kx0 ≥ −akx0 ,
∥x 2 ∥
therefore x ∈ [−akx0 , akx0 ] and x0 is an order unit of X .
 x = (∥x1 ∥ + ∥x2 ∥)
We recall the next result from the theory of ordered spaces, see
for example Aliprantis and Tourky (2007, Theorem 2.8).
Theorem 2.9. If X is a Banach space ordered by the closed and
generating cone P, then any order unit of X is an interior point of P.
Remark 2.10. Theorem 2.8 combined with Proposition 2.2, says
that the set of semi-interior points of a cone P of a normed space
X is exactly the set of order units of X each of which is also an
upper bound of some positive multiple of the positive part U+ of
the unit ball U of X . So the set of semi-interior points of a cone P is
a subclass of the class of the order units of X . From Theorem 2.9 we
have that in Banach spaces these two classes coincide. However,
the next example shows that in normed spaces, the set of semi-
interior points of P is a proper subset of the set of the order units
of X , in general.
Example 2.11. This is an example of a normed space with
generating positive cone P with an order unit but without semi-
interior points. So by this example we get that the converse of
Theorem 2.8 is not true. Analogously to Example 2.5, suppose that
Xn is the space R2 ordered by the pointwise ordering and with norm
∥ · ∥n , having as unit ball the polygon Dn of R2 with vertices
(1, 0), (n, n), (0, 1), (−1, 0), (−n, −n), (0, −1).
As in Example 2.5, E is the space of sequences x = (xn )n∈N with
xn = (xn1 , xn2 ) ∈ Xn so that there exists a real number mx > 0
depending on x with ∥xn ∥n ≤ mx for any n. Then it is easy that
∥(n, n)∥n = 1 for any n.
Suppose that E is ordered by the cone P = {x = (xn ) ∈ E | xn ∈
R2+ for any n} and that E is equipped with the norm
∥x∥∞ = supn∈N ∥xn ∥n .
Suppose also that X = P − P is the subspace of E generated by the
cone P and suppose that X is ordered by the cone X+ = P.
It is easy to see that the constant sequence 1 i.e. the sequence
x = (xn ) with xn = (1, 1) for any n is an order unit of X . If we
suppose that z = (zn ) is a semi-interior point of X , then kz is
39
an upper bound of U+ for some k > 0. Since 1 is an order unit,
we have h1 ≥ kz for some real number h > 0, therefore h1 is
an upper bound of U+ . But for any m the vector xm = (xn ) with
xm = (m, m) and xn = (0, 0) for any n ̸= m belongs to U+
therefore (m, m) ≤ h(1, 1) for any m, which is impossible. Hence
P does not have semi-interior points.
3. The topology generated by the positive cone
Suppose that X is a normed space ordered by the closed cone
P, i.e. X+ = P, and suppose also that X+ is generating. This can
be considered as the general case because if the cone P is not
generating, we shall restrict our analysis to the normed space Y =
P − P. In this section we pass to a new norm of X . This norm is
related with the initial norm.
Let us denote by V the convex hull of the union of the positive
and the negative part of U, i.e.
V = co(U+ ∪ (−U+ )) ⊆ U .
The vectors of V are the convex combinations between positive
and negative vectors of X with norm lower than or equal to one.
In other words, the vectors of V are linear combinations of ‘‘gains’’
and ‘‘losses’’ of ‘‘size’’ at most one. In order to show that the set V
generates a norm we need to show that it is absorbing. Indeed, for
any x ∈ X we have that x = x1 − x2 where x1 , x2 ∈ P and we have
 
∥x 1 ∥ x1 ∥x 2 ∥ −x 2
+ ,
∥x1 ∥ + ∥x2 ∥ ∥x1 ∥ ∥x1 ∥ + ∥x2 ∥ ∥x2 ∥
therefore x ∈ (∥x1 ∥ + ∥x2 ∥)V and x ∈ tV for any t ≥ ∥x1 ∥ + ∥x2 ∥.
Therefore the Minkowski’s functional
q(x) = inf {t ∈ R+ | x ∈ tV },
is a norm of X and we shall denote this norm by ||| · |||.
Definition 3.1 (I. Polyrakis). We shall refer to the norm ||| · ||| of X
introduced above as the norm of X generated by the positive part
of the unit ball of X or by the positive cone X+ of X .
It is easy to see that |||x||| ≥ ∥x∥ for any x ∈ X , and |||x||| = ∥x∥ for
any x ∈ P ∪(−P ). Therefore, the |||·|||-topology of X is finer than the
initial topology of X defined by the norm ∥ · ∥ and the topological
dual of (X , ∥ · ∥) is contained in the topological dual of (X , ||| · |||).
We denote by W the unit ball of ||| · |||. Then W is the ||| · |||-closure
of V , and W ⊆ U. Also we have
W+ = W ∩ P = V ∩ P = U ∩ P = U+ .
Proposition 3.2. Suppose that X is a normed space ordered by the
closed cone P. If x0 is a semi-interior point of P, then x0 is an interior
point of P with respect to the norm ||| · ||| of X defined by X+ .
Proof. By the definition of semi-interior point, there exists ρ > 0
so that x0 − ρ U+ ⊆ P. But x0 + ρ U+ ⊆ P, therefore the convex hull
of the union of these sets is contained in P i.e.
co((x0 + ρ U+ ) ∪ (x0 − ρ U+ )) = x0 + ρ co(U+ ∪ (−U+ ))
= x0 + ρ V ⊆ P .
But the cone P is also |||·|||-closed, therefore we have that x0 +ρ W ⊆
P, hence the proposition is true. 
One more appeal to the Krein–Smulian Theorem gives the next
equivalence theorem of the two norms.
Proposition 3.3. If X is an ordered Banach space with closed and
generating positive cone X+ , then the initial norm ∥ · ∥ of X and the
norm |||·||| of X generated by the positive cone X+ of X , are equivalent.
40 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48
Remark 3.4. In the above proposition, X is a Banach space ordered
by a closed cone P. If the cone P is not generating, then the subspace
Y = P − P of X generated by P is not a Banach space in general.
Therefore, we cannot say that in Y the initial norm and the norm of
Y generated by its positive cone Y+ = P are equivalent. Indeed, in
Examples 2.6 and 2.7, the subspace Y = P − P is not a Banach space
because in both cases P has semi-interior but not interior points. In
these examples Y is dense in X .
Remark 3.5. Suppose that in an exchange economy the commod-
ity space is a normed space X , the consumption set is a closed and
generating cone P of X and suppose that X is ordered by the cone
P, i.e. X+ = P. From Proposition 3.3 we have that if X is complete,
then the norm ∥ · ∥ of X and the norm ||| · ||| of X defined by the
positive cone X+ of X are equivalent. In particular if X is a Banach
lattice these two norms are equivalent. So, in the classical equilib-
rium models as for example the finite dimensional Arrow–Debreu
model, the Banach lattices equilibrium models and in the Banach
spaces equilibrium models where the consumption set P is a closed
and generating cone of X , these two norms ∥ · ∥ and ||| · ||| of the
commodity space are equivalent, therefore both of them can be
considered as the equilibrium norm of X . If X is a normed space,
these two norms are not equivalent in general. But according to
above remarks, the study of equilibrium with respect to the norm
||| · ||| can be also considered as a natural way to extend the study of
classical equilibrium models in ordered normed spaces.
We continue below Example 2.5, where we determine the new
norm ||| · ||| of X and we compare this norm with the initial norm
∥ · ∥ of X .
Example 3.6. Suppose that X = P − P is the subspace of E of
Example 2.5. We have X+ = P and
U+ = {x = (xn ) ∈ X+ | xn = (xn1 , xn2 ) ∈ R2+ : ∥xn ∥n = xn1 + xn2 ≤ 1}.
For any x = (xn ), y = (yn ) ∈ U+ and any t ∈ [0, 1] we have
tx + (1 − t )(−y) = (t (xn1 , xn2 ) + (1 − t )(−yn1 , −yn2 )).
Since x, y ∈ Un+ and Un+ = Uℓ+1 where ∥(a, b)∥ℓ1 = |a| + |b| for
any (a, b) ∈ R2 is the ℓ1 -norm of R2 and Uℓ+1 is the ℓ1 -unit ball of
R2 . It is easy to show that Uℓ1 is the convex hull of Uℓ+1 ∪ (−Uℓ+1 ),
therefore we have that the unit ball W of the ||| · |||-norm of X is the
Cartesian product
=1 Di ,
W = Πi∞
where Di = Uℓ1 for any i. By the above remarks it is easy to see that
|||x||| = supn∈N ∥xn ∥ℓ1 ,
for any x = (xn ) ∈ X .
So if we have two not comparable commodities x = (xn ), y =
(yn ) ∈ X+ (i.e. x ̸≥ y and y ̸≥ x) and if we suppose that xi = yi
for any i ̸= n and xn = yn + (a, b) where a, b are constant real
numbers with ab < 0 and |a| = |b|. Then, by the definition of the
|a|
norm ∥ · ∥∞ we have that ∥x − y∥∞ = n but |||x − y||| = 2|a|. So,
in this case, the initial norm ∥ · ∥∞ of X destroys the differences
between the commodities x and y as n increases but the norm ||| · |||
of X perceives more precisely the differences. This shows a kind of
myopia of the norm ∥ · ∥∞ of X but only in the case of the example.
This property disappears if |a| ̸= |b|, if ab > 0 or if we pass in the
new norm ||| · ||| of X .
4. Cones with nonempty interior and equilibrium
In this section we study equilibrium in an ordered normed
space X whose positive cone X+ has nonempty interior and the
consumption sets of the consumers are different subcones Pi of X+ .
Following standard methods we prove a second welfare theorem,
Theorem 4.4 and the existence of quasi-valuation equilibrium,
Theorem 4.10. In this section we adopt the method of Mas-
Colell, Mas-Colell (1986), for the proof of equilibrium in the case
of different consumption sets. In Theorem 4.10, instead of the
continuity of the utility functions we use radial continuity and in
Theorem 4.11 linear continuity.
4.1. The model
We consider an exchange economy with l consumers. We
suppose that the commodity space X is an ordered normed space
with closed positive cone X+ . For each i = 1, 2, . . . , l, suppose that
consumer i has:
• as consumption set, a closed subcone Pi of X+ ,
• a preference relation %i which is defined on Pi , and
• as initial endowment, a nonzero vector ωi ∈ Pi .
Let us denote the economy by
E = ⟨X , (Pi , %i , ωi )i=1,2,...,l ⟩.
The total endowment is ω = i=1 ωi . Also, suppose that the
l
preferences are reflexive, complete and transitive. By monotone
preference relations we mean that for any i and x, y ∈ Pi , then
x ≥ y implies x %i y. We denote by
 
l

A= x ≡ (x1 , x2 , . . . , xl ) ∈ P1 × P2 × · · · × Pl | xi = ω ,
i =1
the set of NFD-allocations (non-free disposal allocations) or
simply the set of allocations and by
 
l

K= x ≡ (x1 , x2 , . . . , xl ) ∈ P1 × P2 × · · · × Pl | xi ≤ ω ,
i=1
the larger set of FD-allocations (free disposal allocations).
For any x, y ∈ K , we write x ≽ y if xi %i yi for any i, x ≻ y
if xi %i yi for any i and xi ≻i yi for at least one i and we shall write
x ≻≻ y if xi ≻i yi for any i.
Let us introduce the following properties repeatedly used
throughout the sequel.
(A1) for each x ∈ X+ , there exist vectors xi ∈ Pi so that x =
l
i =1 x i .
(A1b) For any i = 1, 2, . . . , l, for any x ∈ Pi and any real number
t > 0, there exists yi,x,t ∈ Pi , depending on i, x, t so that
0 ≤ yi,x,t ≤ t ω and x + yi,x,t ≻i x.
(A1c) The property (A1b) is satisfied; moreover, for any i, and any
x ∈ Pi the family (yi,x,t )t >0 of (A1b) can be chosen in such a
way that it satisfies the property: limt −→0 yi,x,t = 0 (in this
family i and x are constant).
The property A(1b) is a generalization of the classical assump-
tion that ω is extremely desirable for any consumer, in the case
where Pi = X+ for any i. Recall that a consumption vector v ∈ X+
is extremely desirable for the i-consumer or for the preference re-
lation ≽i , if v ∈ Pi and for any x ∈ Pi we have x + λv ≻i x, for any
real number λ > 0. For this definition, see for example in Aliprantis
et al. (1990).
In our model we have different consumption sets and we do
not assume that ω ∈ Pi for any i. So we cannot assume that
ω is extremely desirable for any consumer but we replace this
assumption by the properties (A1b) and (A1c).
Proposition 4.1. If ω ∈ Pi for any i and ω is an extremely desirable
for any consumer, then (A1c) is satisfied.
 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48 41

3ρ ρ
Proof. Let x ∈ Pi and t > 0. By our assumption we have that ω ∈ Pi that z2 = ω(1 − 4∥ω∥ ) is an interior point of the set K = z1 − 2 U+ ,
for any i and x + t ω ≻i x, for any t > 0, therefore (A1b) is satisfied therefore z2 + ρ2 U ⊆ K for some ρ2 > 0. Note that ∥ω − z1 ∥ = 2 ,
ρ
with yi,x,t = t ω. Also we have that limt −→0 yi,x,t = limt −→0 t ω = 0 3ρ ρω ρ
∥ω − z2 ∥ = 4
, z1 − z2 = 4∥ω∥
and ∥z1 − z2 ∥ = 4
with z1 > z2 . So
and (A1c) is true. 
we have
Proposition 4.2. If the positive cone X+ of X is normal, then (A1b) z2 + ρ2 U ⊆ K ⊆ W ⊆ F .
implies (A1c).
Since G is dense in F , there exists a sequence of G which converges
Proof. By (A1b), for any x ∈ Pi and any real number t > 0, there to z2 therefore, except of a finite number, the members of the
exists yi,x,t ∈ Pi , depending on i, x, t so that 0 ≤ yi,x,t ≤ t ω and sequence are contained in z2 + ρ2 U. So K ∩ G ̸= ∅ and suppose
x + yi,x,t ≻i x. Since the cone X+ is normal and limt −→0 t ω = 0, we ρ
that y ∈ K ∩ G. Since y ∈ K = z1 − 2 U+ ⊆ z1 − X+ we have
have that limt −→0 yi,x,t = 0, see in Aliprantis and Tourky (2007, that y ≤ z1 < ω, therefore w = ω − y > 0. Suppose that
Theorem 2.23).  y = y1 + y2 + · · · + yl where yi ∈ Pi with yi %i xi for each i. We have
ρ
Proposition 4.3. If the preferences are monotone and (A1) holds, w = ω − y ≥ ω − z1 = ω.
then for any x ∈ K there exists y ∈ A so that y % x. 2∥ω∥
ρ
Then by (A1b), for any i, any yi and t = , there exists
xi < ω and suppose that z =
l 2l∥ω∥
Proof. Let x ̸∈ A. Then ρ
i=1
vi = yi,yi ,t ∈ Pi so that 0 ≤ vi ≤ ω and yi + vi ≻i yi . Then
ω− ϕi with ϕi ∈ Pi for any i,
l l 2l∥ω∥
i=1 xi . Then we have z = i=1
it is easy to see that
therefore the allocation
m = (mi = yi + vi ) ∈ K
ỹ = (x1 + ϕ1 , x2 + ϕ2 , . . . , xl + ϕl ) ,
because
is the requested one. 
l
 l
 ρ
mi = y + vi ≤ y + ω
4.2. A second welfare theorem 2∥ω∥
i =1 i =1

Let us now recall that for x = (x1 , x2 , . . . , xl ) ∈ A, one says

that:
• x is Pareto optimal if there does not exist an allocation y ∈ A
so that y ≻ x;
• x is weakly Pareto optimal if there does not exist an allocation
y ∈ A so that y ≻≻ x.
For x = (x1 , x2 , . . . , xl ) ∈ P1 × P2 × · · · × Pl , one says that x
is supported by the linear functional p of X if p is nonzero and for
any i and any z ∈ Pi we have z ≽i xi H⇒ p(z ) ≥ p(xi ).
In the next theorem we shall prove the supportability of any
weakly Pareto optimal allocation in cones with nonempty interior,
without the assumption that the total endowment ω is itself an
interior point of the cone. The supporting price vector p is nonzero
on X , but we do not ensure that p(ω) > 0. If moreover ω is an
almost order unit of X+ , then the condition that p(ω) > 0 is
ensured.
Theorem 4.4. Suppose that in the economy E the conditions (A1) and
(A1b) are satisfied and that the preferences are monotone and convex.
If x0 is an interior point of X+ , then any weakly Pareto optimal allo-
cation is supported by a positive, continuous linear functional p of X
with p(x0 ) = 1.
If moreover the total endowment ω is an almost order unit of X ,
then we have that p(ω) > 0.
Proof. Suppose that
x = ( x1 , . . . , xl )
is a weakly Pareto optimal allocation and
= y + (ω − z1 ) ≤ y + (ω − y) = ω.
Also we have
m ≻≻ y < x.
So, by Proposition 4.3, there exists u ∈ A so that u % m ≻≻ x. This
is a contradiction because x is weakly Pareto optimal. Therefore, ω
cannot be an interior point of F .
Because of this, there exists a sequence {wn } of X which
converges to ω with wn ̸∈ F , for any n. Therefore, by the separation
theorem, for any n, there exists pn ∈ X ∗ \ {0} which separates
wn and F , i.e. pn (wn ) < pn (w) for any w ∈ F . We shall show
that pn is positive, i.e. that pn (x) ≥ 0 for any x ∈ X+ . So we
take a vector x ∈ X+ . Then, by (A1), there exist vectors φi ∈ Pi
i=1 φi . So for any λ ≥ 0 we have ω + λx =
l
so that x =
x1 + λφ1 + x2 + λφ2 . . . + xl + λφl ∈ F , because the preferences
are monotone. Hence pn (ω + λx) ≥ pn (wn ). Therefore
λpn (x) ≥ pn (wn − ω), for any λ ≥ 0.
This shows that pn (x) ≥ 0. Indeed, if we suppose that pn (x) < 0,
then for λ −→ +∞ we have λpn (x) −→ −∞, so pn (wn −
ω) = −∞, a contradiction. Hence pn (x) ≥ 0 for any x ∈ X+ ,
therefore pn ≥ 0 for any n and pn > 0 for any n, because pn ̸=
0. By our assumptions, x0 is an interior point of P therefore, by
Proposition A.2 in the Appendix, the set
∗
B = {f ∈ X+ | f (x0 ) = 1},
∗
is a weak-star compact base for the dual cone X+ = {f ∈ X ∗ |
f (x) ≥ 0, for any x ∈ X+ } of X+ .
Note that we can also suppose that pn ∈ B for any n, because
the linear functionals p (nx ) separate wn and F and belong to B.
p

G = {y = y1 + y2 + · · · + yl | yi ∈ Pi , yi %i xi for any i}.

If F is the closure of G, then F is a closed and convex subset of X+
containing the total initial endowment ω. We claim that ω is not
an interior point of F . Indeed, if we suppose that ω is an interior
point of F , then we have that ω + ρ U ⊆ F , where U is the unit ball
of X and ρ > 0. From this, we obtain a contradiction as follows:
Since ω is also an interior point of X+ , then t ω is an interior point
ρ ρ
of ρ U+ for any real number 0 < t < ∥ω∥ . Hence z1 = ω(1 − 2∥ω∥ )
n 0
Since the base B is weak star compact, the sequence {pn } has an
accumulation point p ∈ B, therefore for any V ∈ I, where I is the set
of the neighborhoods of 0 in the weak-star topology, there exists
nV so that pnV ∈ V and the net (pnV )V ∈I converges to p. So we have
p(x0 ) = 1 and also the subnet (wnV )V ∈I converges to ω. Hence, for
any x ∈ F we have
p(ω) = lim pnV (wnV ) ≤ lim pnV (x) = p(x),
V V

is an interior point of the set W = ω − ρ U+ and similarly we have therefore p supports F at ω.

42 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48

To show that p supports the allocation x, suppose that x ∈ Pi is Proposition 4.7. The condition (A1b), and therefore also (A1c),
such that x %i xi for some i. Then we have implies (A3).
ω ω ω
l
 Proof. Suppose that (A1b) is true. Let w = ( 2l1 , 2l2 , . . . 2ll ) ∈ K .
x+ xj ∈ F , ω ω
Then by (A1b) for any i there exists yi ∈ Pi so that 2li + yi ≻i 2li and
j=1,j̸=i
0 < yi ≤ t ω with t = 1
2l
. If
therefore ω
1 ω2 ωl 

l
 
l
 x= + y1 , + y2 , . . . , + yl ,
  2l 2l 2l
p x+ xj ≥ p(ω) = p xj , ω ω
j=1,j̸=i j =1 then we have that x ∈ K with ui (xi ) = ui ( 2li + yi ) > ui ( 2li ) ≥ 0
for any i therefore (A3) is true. By definition of (A1c), we have also
therefore p(x) ≥ p (xi ) and p supports the allocation x. that (A1c) H⇒ (A3). 
If ω is an almost order unit of X then by Proposition A.1 in the
Appendix, we have that p(ω) > 0.  We say that the function ui is radially continuous if for any x ∈
Pi , the restriction of ui on the half-line {tx|t ∈ R+ } is continuous,
Remark 4.5. In Aliprantis et al. (2004, Lemma 3.2), it is supposed see for example in Polyrakis (2008, Definition 6), where the notion
that the commodity space is the space X = R3 , ordered by the ice- of the radially upper and radially lower semicontinuity of the
1 utility function is used for the study of the continuity of the demand
cream cone P = {(x, y, z ) ∈ R3 | z ≥ (x2 + y2 ) 2 } and X+ = P is
correspondence. It is easy to show that ui is radially continuous
the consumption set. It is supposed that in the economy there are
on Pi if for any x ∈ Pi and for any sequence {tn } of positive real
two consumers with total endowment ω = (0, 1, 1) and utility
numbers with tn −→ t we have ui (tn x) −→ ui (tx). A stronger
functions u1 (x, y, z ) = 2z and u2 (x, y, z ) = x + 2z. It is easy to see
notion which we shall also use in this article is the one of linear
that in this example, any allocation (x1 , x2 ) is of the form x1 = aω
continuity. The function ui is linearly continuous2 if the restriction
and x2 = (1 − a)ω with a ∈ [0, 1] and also that any allocation is
of ui on any line of Pi is continuous. This is equivalent to the
Pareto optimal. Moreover, as it is shown in Aliprantis et al. (2004),
property: for any x, y ∈ Pi with y ̸= 0 and any sequence of real
any Pareto optimal allocation cannot be supported by a price vector numbers tn −→ t ∈ R we have that ui (x + tn y) −→ ui (x + ty),
p with p(ω) > 0. This is not in conflict with our Theorem 4.4 whenever x + tn y ∈ Pi for any n and x + ty ∈ Pi .
because this theorem does not imply that the supporting liner Let us introduce the total utility function
functionals p satisfy the condition p(ω) > 0. This is true if ω is
an almost order unit of P. Note that in Aliprantis et al. (2004), P is a u(x) = (u1 (x1 ), u2 (x2 ), . . . , ul (xl )),
cone with nonempty interior but ω is not an interior point and also x = (x1 , x2 , . . . , xl ) ∈ P1 × P2 × · · · × Pl ,
ω is not an almost order unit of P. For this economy all the other
assumptions of Theorem 4.4 are satisfied, therefore by our theorem and the utility space
any allocation must be supported by a nonzero price vector. It is U = {u(x) | x ∈ K }.
easy one to see that the price vector p = (0, −1, 1) is positive
with respect to the cone P. Indeed for any (x, y, z ) ∈ P we have that
z ≥ |y|, therefore (0, −1, 1)·(x, y, z ) = z − y ≥ 0. Also p is nonzero Proposition 4.8. If (A2) is satisfied, the utility space U has the
on X and it is easy to see that p supports any allocation (x1 , x2 ). properties:
Indeed, if (x, y, z ) ∈ P with u1 (x, y, z ) ≥ u1 (x1 ) or equivalently (i) U is bounded.
z ≥ a we have that p(x, y, z ) = z − y ≥ 0 = p(0, a, a) = p(x1 ) and (ii) If the preferences are radially continuous, then U is solid, in the
similarly we have that u2 (x, y, z ) ≥ u2 (x2 ) H⇒ p(x, y, z ) ≥ p(x2 ). sense that for any α ∈ U the order interval [0, α] of Rl is
But we have that p(ω) = 0. contained in U, i.e. [0, α] ⊆ U.
(iii) If the preferences are radially continuous and (A3) is true, then
4.3. Equilibrium the set U contains the positive part of a ball of Rl with center
zero and radius r.
We introduce two more conditions, (A2) and (A3), in order to Proof. (i) By (A2) each ui is bounded on the set Ωi . So for any
continue our study for the existence of equilibria. x = (x1 , x2 , . . . , xl ) ∈ K we have xi ∈ Pi , 0 ≤ xi ≤ ω, therefore
(A2) The preferences %i are defined by the utility functions ui : xi ∈ Ωi , hence 0 ≤ ui (xi ) ≤ ai , where ai is an upper bound of ui on
Pi −→ [0, +∞) with ui (0) = 0 for each i = 1, 2, . . . , l, and Ωi . Therefore the set U is bounded.
each ui is bounded on the subset (ii) Let α ∈ U. Then there exists x ∈ K with α = u(x). Suppose
that β ∈ Rl+ with 0 ≤ β ≤ α . We claim that β ∈ U. We have
Ωi = {x ∈ Pi | 0 ≤ x ≤ ω}
0 ≤ βi ≤ ui (xi ) for each i.
of Pi .
Since the restriction of ui on the line segment
Remark 4.6. The condition ui (0) = 0 does not introduce any loss {txi | t ∈ [0, 1]},
of generality because if we suppose that %i is defined by the utility
function ui and ui is monotone, then the utility function  ui (x) = is continuous, by the mean value theorem, there exists ti ∈ [0, 1]
ui (x) − ui (0) is monotone, defines %i and  ui (0) = 0. so that ui (ti xi ) = βi . For these ti we have
If ω ∈ Pi and ui is monotone, then ui is bounded on Ωi . If ui is the u(t1 x1 , t2 x2 , . . . , tl xl ) = β,
restriction of a monotone utility function on X+ , then it is bounded
on the set Ωi . therefore β ∈ U.

Assumption (A3) says what follows.

(A3) There exists x = (x1 , x2 , . . . , xl ) ∈ K so that ui (xi ) > 0, for 2 This notion of linear continuity is very similar with a notion of weak continuity
any i. of preferences, used by G. Debreu in Debreu (1954), see Remark 4.12.
 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48 43

(iii) By (A3), there exists x = (x1 , x2 , . . . , xl ) ∈ K with ui (xi ) >
0 for any i. Then u(x) ∈ U. We put r = mini {ui (xi )}. Then r > 0
because ui (xi ) > 0 for any i. If we suppose that z ∈ Rl+ with
∥z ∥ ≤ r, we have that 0 ≤ zi ≤ r for each i, therefore 0 ≤ z ≤ u(x)
and by (ii) we have that z ∈ U. Hence (iii) is true. 
intersection of U with the semiline {ts|t ∈ R+ }, is the line segment
defined by 0 and ρ(s)s.
(iii) Let {sn } be a sequence of ∆ which converges to s. Suppose
that the sequence {ρ(sn )} does not converge to ρ(s). Then there
exists a subsequence {ρ(skn )} of {ρ(sn )} so that

We will denote by ∆ the simplex of Rl+ , i.e. |ρ(skn ) − ρ(s)| ≥ a > 0, for any n.

∆ = {s = (s1 , s2 , . . . , sl ) ∈ Rl+ | s1 + s2 + · · · + sl = 1}. Since the function ρ is bounded, {ρ(skn )} has a convergent
subsequence which we denote again by {ρ(skn )} and suppose that
Under the assumption (A2), by Proposition 4.8, the utility space is
bounded and suppose that M is a ∥ · ∥1 bound of U, i.e. ∥x∥1 ≤ M
ρ(skn ) −→ λ
for any x ∈ U. For any s ∈ ∆, the set {α ∈ R+ | α s ∈ U} is and we have λ ≥ r > 0, because ρ(sn ) ≥ r > 0 for any n, where r
bounded because for any α in it, we have α = ∥α s∥1 ≤ M because is the radius of a ball whose the positive part Dr is contained in U.
∥s∥1 = 1 for any s ∈ ∆, therefore 0 ≤ α ≤ M. So the function ρ Then we have λ ̸= ρ(s) and ρ(skn )skn −→ λs ∈ U. By statement
defined by (ii), {ρ(skn )skn } is a sequence of the boundary ϑ U of U, therefore
ρ(s) = sup{α ∈ R+ | α s ∈ U}, s∈∆
λs ∈ ϑ U because ϑ U is closed. By the definition of ρ(s) we have
that λ ≤ ρ(s), and λ < ρ(s) because λ ̸= ρ(s). Therefore we have
is bounded with
λs < ρ(s)s.
0 ≤ ρ(s) ≤ M , for any s ∈ ∆.
Also we have
In the next proposition and its consequences, notably Theo-
[0, ρ(s)s] ⊆ U
rem 4.10, it is assumed that the utility space is closed. This is
the closedness condition introduced by Mas-Colell in Mas-Colell because U is solid. Therefore λs is an interior point of the order
(1986). The main idea in the proof of existence given by Mas-Colell interval [0, ρ(s)s] because 0 < λ < ρ(s), hence λs is an interior
in Mas-Colell (1986), is to support Pareto optimal allocations with point of U. This is a contradiction because λs ∈ ϑ U. Therefore
prices. This approach requires of course the existence of weakly ρ(sn ) −→ ρ(s) and the function ρ is continuous on s.
Pareto optimal allocations. Under the assumptions of Mas-Colell (iv) Assume that x ∈ K and u(x) ≥ ρ(s)s.
(1986), the compactness of the order interval [0, ω] guarantees If we suppose that ui (xi ) > ρ(s)si for any i, and if we put
the compactness of K and therefore also the compactness of the 
ui (xi )

utility space U, because it is the image of K under a continuous t ′ = min | for any i with ρ(s)si > 0 ,
function. As it is shown in Mas-Colell (1986), the closedness condi-
ρ(s)si
tion guarantees the existence of weakly Pareto optimal allocations, then we have that t ′ > 1 and it is easy to show that u(x) ≥ t ′ ρ(s)s,
so this condition, which is weaker than the compactness of [0, ω], therefore t ′ ρ(s)s ∈ U. Consequently we have t ′ ρ(s)s ∈ U with
is used by Mas-Colell in Mas-Colell (1986) as one of the basic as- t ′ > 1, a contradiction, by (ii). So, we conclude that ui (xi ) = ρ(s)si
sumptions for the existence of an equilibrium. for at least one i.
(v) Suppose that x ∈ A with u(x) ≥ ρ(s)s. If we suppose
Proposition 4.9. If (A1), (A2), (A3) are satisfied, the utility that an allocation z exists so that zi ≻i xi for any i, we have that
functions are radially continuous and the utility space U is closed, then ui (zi ) > ρ(s)si , for any i, a contradiction by (iv). Therefore x is
for each s ∈ ∆, we have: weakly Pareto optimal.
(vi) Suppose that x ∈ K with u(x) = ρ(s)s. If we suppose that
(i) ρ(s) > 0 and ρ(s)s ∈ U; x ∈ A, then x is weakly Pareto optimal by (v) and (vi) is true. So we
(ii) the intersection of U with the half-line {t ρ(s)s | t ≥ 0} is the suppose that
line segment defined by 0 and ρ(s)s;
(iii) the function ρ is continuous on s;
l

(iv) if x ∈ K with u(x) ≥ ρ(s)s, then we have that u(xi ) = ρ(s)si z =ω− xi > 0
i=1
for at least one i, where si is the i-coordinate of s;
(v) if x ∈ A with u(x) ≥ ρ(s)s, then x is weakly Pareto optimal; and we take the allocation f with fi = xi + zl for any i. Since the
(vi) if Pi = X+ for any i and the utility functions ui are monotone utility functions are monotone we have that u(f ) ≥ u(x) = ρ(s)s,
and linearly continuous, there exists a weakly Pareto optimal therefore, by (iv), there exists j so that uj (xj + zl ) = uj (xj ) = ρ(s)sj
allocation y so that u(y) = ρ(s)s. and we fix such a j. We define the vector y1 ∈ K so that
Proof. (i) By Proposition 4.8, U contains the positive part Dr of z
y1i = xi , for any i ̸= j and y1j = xj + ,
a ball of Rl with center zero and radius r > 0. So we have that l
r
∥s∥
s ∈ Dr , therefore ∥rs∥ s ∈ U. Hence, by the definition of ρ , we where j is the above fixed index. Then we have
have that ρ(s) ≥ ∥rs∥ ≥ r because 1 = sups∈∆ ∥s∥, for the Euclidean
l
l−1
norm ∥ · ∥ of Rl .

u(y1 ) = u(x) and ω− y1i = z .
Also, by the definition of ρ(s), there exists an increasing i =1
l
sequence of real numbers {αn } which converges to ρ(s) so that
αn s ∈ U. Then αn s −→ ρ(s)s, therefore ρ(s)s ∈ U because U As before, starting by y1 we consider the allocation g so that gi =
is closed. y1i + z l−l 1 1l for any i. Then as we have noted before, we have that
(ii) Let us show now that the intersection of U with the half-line uj (gj ) = uj (y1j ) = uj (xj ) for at least one j and we fix such a j. We
{t ρ(s)s} is the line segment defined by 0 and ρ(s)s. define the vector y2 ∈ K so that
Since the order interval [0, u(x)] of Rl is contained in U for any  
l−1 1
x ∈ K , we have that for each 0 ≤ t ≤ 1, t ρ(s)s ∈ U. So the line y2i = y1i , for any i ̸= j and y2j = y1j + z ,
segment defined by 0 and ρ(s)s is contained in U. By the definition l l
of ρ(s) we have that for each t > 1, t ρ(s)s ̸∈ U, therefore the for the fixed j.
44 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48

Then we have Theorem 4.10. Suppose that in the economy E the conditions (A1),
l
 
l−1
2 (A1c), (A2), are satisfied, the preferences are monotone and convex,
u(y2 ) = u(y1 ) = u(x) and ω− y2i = z . the utility functions are radially continuous and the utility space U
i =1
l is closed. If the interior of X+ is non-empty and (A4) is satisfied,
then a quasi equilibrium allocation x exists supported by a continuous,
By continuing this process we get a sequence {yn } of K so that
positive linear functional p of X .
l
 
l−1
n If moreover ω is an almost order unit of X , then we have p(ω) > 0
u(yn ) = u(x) for any n and ω− yni = z . (1) and x is a quasi-valuation equilibrium.
i =1
l
Proof. Suppose that x0 is an interior point of P. By Proposition A.2
By this process we have that for any fixed i, the sequence {yni } is of in the Appendix, the set
the form
    2  n− 
∗
B = {f ∈ X+ | f (x0 ) = 1},
z l−1 l−1 l−1
n
yi = xi + a1 + a2 + a3 + · · · + an , is a weak-star compact base for the dual cone
l l l l
∗
X+ = {f ∈ X ∗ | f (x) ≥ 0, for any x ∈ X+ } of X+ .
where ak = 1 if during the k step i = j and ak = 0 if i ̸= j. So we
have that {yni } is an increasing sequence of the line segment of X+ By (A4), for each s ∈ ∆, we select a weakly Pareto optimal
defined by the vectors xi and xi + az where allocation xs = (x1 s , x2 s , . . . , xl s ) so that u(xs ) = ρ(s)s.
∞  i By Theorem 4.4, there exists a price vector p ∈ B which supports
1 l−1 xs and we denote the set of these supporting prices by L(s), i.e.
a= = 1.
l i=0 l
L(s) = {p ∈ B | p supports xs }.
yni
Therefore the sequence { }, as an increasing sequence of a line
We define the sets
segment, is convergent to a vector yi of this line segment of X+ .
Therefore we have ui (yi ) = limn ui (yni ) = ui (xi ) for any i, because Φ (s) = {Gp (w − xs ) = (p(ω1 − xs1 ), p(ω2 − xs2 ), . . . ,
we have assumed that the functions ui are linearly continuous. p(ωl − xsl )) | p ∈ L(s)}
l
By taking limits in the second part of (1), we have that i=1 yi =
ω, hence and

y = (y1 , y2 , . . . , yl ) ∈ A, Ψ (s) = s + Φ (s).

therefore the allocation y is weakly Pareto optimal by (v). Hence If we suppose that s ∈ ∆ is a fixed point of Ψ , i.e. s ∈ Ψ (s), then
y = (y1 , y2 , . . . , yl ) is a weakly Pareto optimal allocation with we get a quasi equilibrium allocation as follows:
u(y) = u(x). 
s = s + h, where h ∈ Φ (s), therefore h = 0 ∈ Φ (s).
Proposition 4.9 naturally leads to consider a condition that we
So there exists p ∈ L(s) so that Gp (w − xs ) = 0, hence p · ωi = p · xsi
set below as condition (A4). It says that any point ρ(s)s of the
utility space is the image of a weakly Pareto optimal allocation. for each i.
The role played by this condition in the existence of equilibria Since p supports the allocation xs we have
is made clear by the next existence Theorem 4.10. By statement x ∈ Pi , x %i xi ⇒ p · x ≥ p · xsi = p · ωi ,
(vi) of Proposition 4.9, condition (A4) is fulfilled in the case where
Pi = X+ for any i, the utility functions ui are monotone and linearly hence (xs1 , xs2 , . . . , xs l ) is a quasi equilibrium allocation supported
continuous and, of course, the basic assumptions of Proposition 4.9 by p.
are also satisfied. To show that Ψ has a fixed point, we shall prove first that the
set Ψ (s) is convex for each s and that the graph of Ψ is closed. To
(A4) For any s ∈ ∆, there exists a weakly Pareto optimal allocation
show that Ψ has these two last properties it is enough to show that
x so that u(x) = ρ(s)s.
Φ has these properties. The first is obvious. To show that the graph
For each price vector p we denote by Gp the function of Φ is closed, we suppose that
Gp (x) = (p(x1 ), p(x2 ), . . . , p(xl )), x = (x1 , x2 , . . . , xl ) ∈ X l . sn −→ s, γn ∈ Φ (sn ) and γn −→ γ
Also we shall denote by and we have to show that γ ∈ Φ (s). Let
w = (ω1 , ω2 , . . . , ωl ), γn = Gpn (w − xsn ), where pn ∈ L(sn ).
the initial allocation.
Since the base B is weak star compact, the sequence {pm } has an
As it is standard, we say that a NFD-allocation x ∈ A is:
accumulation point p ∈ B, therefore for any V ∈ I, where I is the set
• a quasi equilibrium, supported by the linear functional p, if for of the neighborhoods of 0 in the weak-star topology, there exists
any i and any z ∈ Pi we have z ≽i xi H⇒ p(z ) ≥ p(ωi ); mV so that pmV ∈ V and the net (pmV )V ∈I converges to p. We shall
• a quasi-valuation equilibrium, supported by the linear show that p supports the allocation xs . Consider x ∈ Pi such that
functional p if it is a quasi equilibrium supported by the linear x %i xsi .
functional p and p(ω) > 0; Then by (A1c), for any t > 0 there exists yi,x,t ∈ Pi so that
• an equilibrium, supported by the linear functional p, if for any i yi,x,t ≤ t ω with
and any z ∈ Pi we have z ≻i xi H⇒ p(z ) > p(ωi ) and p(ω) > 0.
lim yi,x,t = 0 and x + yi,x,t ≻i x ≽i xsi ,
In the literature, in the above definitions the continuity of the t −→0
supporting functional p is also required. In our terminology, we do
therefore ui (x + yi,x,t ) > ui (xsi ) and we keep such a t > 0 constant,
not require the continuity of p in order to specify in our results the
for the moment.
topology of the continuity of p.
Since γn −→ γ and the function ρ is continuous we have that
For the conditions of the next theorem note that, by Proposi-
tion 4.7, (A1c) implies (A3). u(xsn ) = ρ(sn )sn −→ ρ(s)s = u(xs ),
 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48
therefore
ui (xi n ) −→ ui (xsi ).
s
But, as we have seen above, ui (x + yi,x,t ) > ui (xsi ), therefore
ui (x + yi,x,t ) > ( ), for any n ≥ n0 ,
s
ui xi n
for some n0 ∈ N. Since pn supports the allocation xsn we have
pn (x + yi,x,t ) = pn (x) + pn (yi,x,t ) ≥ pn (xi n )
s
for each n ≥ n0
and by taking limits as t −→ 0 we have that pn (x) ≥ pn (xi n ), for
s
each n ≥ n0 . Therefore we have
pn (x) ≥ pn (xi n ) = pn (ωi ) − pn (ωi − xi n ) = pn (ωi ) − γni ,
s s
where γ is the i coordinate of γn .
i
n
Since pnV −→ p in the weak star topology of X ∗ , by the previous
relation and the hypothesis that γn −→ γ we have that γnV −→ γ ,
and also that
p(x) ≥ p(ωi ) − γ i ,
where γ i is the i coordinate of γ . But
p(ωi ) − γ i = p(ωi ) − p(ωi − xsi ) = p(xsi ),
therefore p(x) ≥ p(xsi ) and p supports the allocation xs . Also by the
above relation we have that γ i = p · (ωi − xsi ), hence
γ = Gp (w − xs ),
therefore γ ∈ Φ (s) and the function Φ has closed graph.
We shall show now that Φ takes values in a compact subset of
Rl . For any i we have
ωi − xsi ≤ ωi ≤ ω and ωi − xsi ≥ ωi − ω ≥ −ω.
Therefore for any p ∈ L(s) and any i we have
0 ≤ |p(ωi − xsi )| ≤ p(ω) ≤ R = max{p(ω) | p ∈ B},
because the linear functional p is positive. Note that the maximum
in the above relation exists because the linear functional  ω of X ∗
so that  ω(f ) = f (ω), for any f ∈ X ∗ is weak-star continuous,
hence it takes maximum on the weak-star compact base B of
X+∗
. Therefore our function Φ takes values in the closed ball of
l
R of radius R, i.e. in a compact metric space. So, by the closed
graph theorem for multivalued functions, see Aliprantis and Border
(2006, Theorem 17.11), we have that Φ and therefore also Ψ , are
upper hemicontinuous. Let us prove now that Ψ is inward pointing,
i.e. for each s ∈ ∆ there exists y ∈ Ψ (s) and λ > 0 such that
s + λ(y − s) ∈ ∆.
Precisely, we shall show that for any s ∈ ∆ and any z ∈ Φ (s) i.e. for
any z = Gp (w − xs ), where p ∈ L(s), the vector
y=s+z
of Ψ (s) satisfies this condition. So we have to show that the vector
b = s + λ(y − s) = s + λz
belongs to ∆, where λ is a suitable real number which will be
defined below.
Let s ∈ ∆ be fixed. For this s we have selected an element xs of
A so that u(xs ) = ρ(s)s. To simplify our notations below, we shall
denote xs by x. For any i with si > 0, there exists a real number
λi > 0 such that si + λi zi > 0. Therefore we can select a real
number λ = λ0 > 0 so that bi = si + λ0 zi > 0, for each i with
si > 0.
For any i with si = 0 we have ui (xsi ) = ρ(s)si = 0 therefore
ωi ≽i xsi because
ui (ωi ) ≥ 0 = ui (xsi ).
45
Hence p(ωi ) ≥ p(xsi ), because p supports the allocation xs .
Therefore for the real number λ0 defined above and for any i
with si = 0 we have
zi = p(ωi − xsi ) ≥ 0 hence bi = si + λ0 zi ≥ 0.
So, for the real number λ0 , we have bi = si + λ0 zi ≥ 0, for any i,
hence b ≥ 0. l
Also we have that b ∈ ∆ because s ∈ ∆, and i=1 zi = 0. The
last equality is true because
z = Gp (w − xs ) = (p(ω1 − xs1 ), p(ω2 − xs2 ), . . . ,
p(ωl − xsl )) and xs ∈ A.
Therefore Ψ is inward pointing. By the theorem of Halpern–
Bergman, see Aliprantis and Border (2006, Theorem 17.54), we
have that Ψ has a fixed point and the proof has been completed. If
ω is an almost order unit of X+ , by Proposition A.1 in the Appendix,
we have p(ω) > 0. 
4.3.1. The case Pi = X+ for any i
In this subsection we consider the case where all the consumers
have as consumption set the positive cone X+ of X , i.e. Pi = X+ for
any i = 1, 2, . . . , l. First we study the form of our conditions, under
the assumption that Pi = X+ for any i. The condition (A1) is trivially
true and (A1b) and (A1c) are equivalent to the hypothesis that ω is
an extremely desirable bundle for any consumer. Note that if the
preferences are strictly monotone, then ω is an extremely desirable
bundle for any consumer. Also, if the preferences are monotone
and strictly convex, are also strictly monotone. To see this suppose
that x, y ∈ Pi with x > y. Then x = y + (x − y) and by the strict
convexity of ≽i , for any t ∈ (0, 1), we have t (y + (x − y)) + (1 −
t )y ≻i y, therefore t (x − y) + y ≻i y. But x > t (x − y) + y, and by the
monotonicity of ≻i we have x ≽i t (x − y) + y ≻i y and x ≻i y, hence
≻i is strictly monotone.
Hence, if the preferences are monotone and strictly convex,
then ω is also an extremely desirable bundle for any consumer.
If the preferences are defined by monotone utility functions
ui , then (A2) is fulfilled. Indeed, as we have observed after the
definition of (A2), we may suppose that ui (0) = 0 for any i. Also,
by the definition of Ωi and our assumption that Pi = X+ for any i,
we have that Ωi = [0, ω] for any i, therefore, for any x ∈ Ωi , we
have that 0 ≤ ui (x) ≤ ui (ω) hence ui is bounded on Ωi and (A2) is
true.
After the above remarks, by Theorem 4.10, we obtain directly
the next theorem. Note that, by statement (vi) of Proposition 4.9,
and the assumptions of the theorem, (A4) is also fulfilled.
Theorem 4.11. Suppose that in the economy E we have Pi = X+
for any i. Suppose also that the preferences ≽i are convex and they
are defined by monotone and linearly continuous utility functions.
If ω is an extremely desirable bundle for any consumer, the utility
space U is closed and X+ has nonempty interior, then a quasi
equilibrium allocation x exists supported by a continuous, positive
linear functional p of X .
If moreover ω is an almost order unit of X , then we have p(ω) > 0,
therefore x is a quasi-valuation equilibrium.
Remark 4.12. We briefly recall some main equilibrium existence
results given in frameworks close to that of our paper. Among the
papers present in the literature dealing with commodity space of
infinite dimension, Debreu (1954) is pioneering. In Debreu (1954),
the hypothesis of free disposal is assumed and also a weak axiom of
continuity for preferences (the assumption III of page 590) is used.
This assumption is the following:
For any i, any xi , x′i , x′′i ∈ Pi and any real sequence tn ∈ [0, 1] with
tn −→ t, we have:
tn x′i + (1 − tn )x′′i ≽i xi for each n H⇒ tx′i + (1 − t )x′′i ≽i xi ,
46 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48
and
tn xi + (1 − tn )xi ≼i xi
′ ′′
for each n H⇒ txi + (1 − t )xi ≼i xi .
′ ′′
It is easy to see that our assumption of linear continuity of
the utility functions implies the assumption III of Debreu and if
in our economy E we add the extra assumption (∗) below, then
assumption III implies the linear continuity of the utility functions3
and these two conditions are equivalent. For this equivalence,
the closedness of the cones Pi , the monotonicity of the utility
functions and property (∗) are only needed. The condition is a
kind of completeness of the preferences and is the following: We
say that assumption (∗) is satisfied if for any i, any xi , yi ∈ Pi with
ui (xi ) < ui (yi ) and any real number ai ∈ (ui (xi ), ui (yi )) there exist
zi , wi ∈ Pi with ui (xi ) < ui (zi ) < ai < ui (wi ) < ui (yi ). The radial
continuity of the utility functions which is used in Theorem 4.10 is
weaker than the linear continuity.
In the present article, for the proof of equilibrium, Theo-
rems 4.10 and 4.11 we follow the method of Mas-Colell (Mas-
Colell, 1986), adopted in our case of different consumption sets.
The basic assumptions in Mas-Colell (1986) are: the closedness
condition, the uniform properness of the utility functions and that
the commodity space X is a locally solid vector lattice. In this ar-
ticle we prove equilibrium by combining the closedness condition
with the existence of interior points in X+ but without any lattice
property on X . Note also that in Mas-Colell and Richard (1991), the
assumptions of Mas-Colell (1986) are weakened but the lattice as-
sumption for the commodity space is maintained. In fact it is sup-
posed in Mas-Colell and Richard (1991), that the commodity space
is a lattice, the set A of allocations is compact and the utility func-
tions are ω-uniformly proper.
Equilibrium existence theorems with infinite dimensional
normed spaces have been also proved, among the others, in Bewley
(1972) and Florenzano (1983). What is made clear in these papers
is that the main steps to ensure the existence are represented
by the possibility of supporting a weak optimum by means of
a price p lying in a weak star compact set of the price space
(whose existence follows, for example, from the assumption that
the commodity space has non empty interior), as well as by the
compactness of attainable sets in some weak compatible topology
on the commodity space (this assumption can be guaranteed
by means of suitable continuity requirements for preferences or
imposing directly the closedness of the utility set).
In particular, in Bewley (1972), in which the existence of
equilibrium is proved for an economy whose commodity space is
L∞ , to obtain the existence of equilibrium, preferences are assumed
to be continuous with respect to the Mackey topology τ (L∞ , L1 ).4
The role played by this continuity requirement is emphasized in
Araujo (1985) (see also Mas-Colell, 1986 and Mas-Colell and Zame,
1991) with reference to the commodity space X = ℓ∞ . In Araujo
(1985), the space ℓ∞ is equipped with a locally convex topology τ
finer than the weak topology σ (ℓ∞ , ℓ1 ) and coarser than the (sup)
norm topology. It is proved that if the economy satisfies a set of
natural conditions and the topology τ is coarser than the Mackey
topology τ (ℓ∞ , ℓ1 ), then the economy has equilibria. Relevantly,
3 The proof is the following: Suppose that III is satisfied, x , y ∈ P , z =
i i i n
tn xi + (1 − tn )yi , tn −→ t and z = txi + (1 − t )yi . If there exists a subsequence of
ui (zn ) that we denote again by ui (zn ) and a real number a > ui (z ) with ui (zn ) ≥ a,
for any n, then there exists w ∈ Pi so that ui (z1 ) ≥ a > ui (w) > ui (z ). Then
ui (zn ) ≥ a > ui (w) > ui (z ) for each n, hence zn ≽ w for any n, therefore
z ≽ w by III. So we have z ≽ w ≻ z, a contradiction. Similarly, if we suppose
that ui (zn ) ≤ a < ui (z ) for a subsequence of ui (zn ), there exists w ∈ Pi with
ui (z1 ) ≤ a < ui (w) < ui (z ) and we have the contradiction z ≼ w ≺ z, by III.
Therefore u(zn ) −→ u(z ) and ui is linearly continuous.
4 Further, under additional assumptions and with the consumption set equal to
the positive cone, it is proved that the supporting price can actually be chosen in L1 .
it is also proved that in the above existence theorem one cannot
dispense with the assumption that τ ⊆ τ (ℓ∞ , ℓ1 ). Indeed an
example of a two agents economy is exhibited with the following
characteristics: the commodity space is ℓ∞ endowed with τ equal
to the (sup) norm topology, all mentioned natural assumptions
are satisfied, there are no individually rational, Pareto optimal,
allocations (and therefore no equilibria). In this economy all
hypothesis of our Theorem 4.10 are fulfilled except the closedness
condition of the utility space.
The relation between the Mackey continuity and the closedness
of utility space for infinite dimensional commodity spaces is
discussed for example in Mas-Colell and Zame (1991) and Mas-
Colell (1986), where it is underlined that it is the continuity with
respect to this topology to guarantee the closedness of utility space.
In Back (2004), the commodity spaces is a vector lattice and the
consumption sets may be larger than the positive cone. Roughly
speaking, each consumption set must be extendable to a larger
consumption set with non-empty interior. In this paper the lat-
tice structure plays a fundamental role in the results. Moreover,
consumption sets in Back (2004) are supposed to be comprehen-
sive, in the sense that Pi + X+ ⊆ Pi . An approach slightly differ-
ent, based on a separating argument in the space of allocations, is
presented in Khan et al. (1999), again under the assumption that
the consumption sets coincide with the positive cone. Finally, the
asymmetric information model considered in Podczeck and Yan-
nelis (2008), produces an existence theorem which generalizes re-
sults by Bewley (1972) and Florenzano (1983) at several instances.
Notice that in this paper, the non-free disposal assumption entails
that the linear supporting price is not necessarily positive.
5. Cones with nonempty semi-interior and equilibrium
In this section we still consider the exchange economy E with
the conditions (A1)–(A4) as have been defined in Section 4. We
apply the results of the previous sections in the case where X+
has semi-interior points and we obtain a second welfare theorem
and the existence of equilibrium which are summarized in the next
theorem. Recall that ∥ · ∥ is the initial norm of X and ||| · ||| the norm
of X defined by the positive cone X+ and that the ∥ · ∥-topology is
coarser than the ||| · |||-topology of X .
Theorem 5.1. Suppose that in the economy E the preferences are
monotone and convex and x0 is a semi-interior point of X+ . Then the
following statements are true:
(i) If the conditions (A1) and (A1b) are satisfied, then any weakly
Pareto optimal allocation is supported by a positive, ||| · |||-
continuous linear functional p of X with p(x0 ) = 1.
(ii) If the conditions (A1), (A1c), (A2), (A4) are satisfied, the utility
functions are radially continuous and the utility space U is closed,
then a quasi equilibrium allocation x exists which is supported by
a positive, ||| · |||-continuous linear functional p of X .
(iii) If Pi = X+ for any i, ω is an extremely desirable bundle for
any consumer, the preferences are defined by linearly continuous
utility functions and the utility space U is closed, then a quasi
equilibrium allocation x exists, supported by a ||| · |||-continuous,
positive, linear functional p of X .
If in anyone of the above statements we suppose moreover that ω is a
||| · |||-almost order unit of X , then, in this statement, we have also that
p(ω) > 0.
Proof. Consider the space X equipped with the ||| · |||-topology.
Since the cone X+ and the subcones Pi of X+ are ∥ · ∥-closed, we
have that the cone X+ and the subcones Pi are also ||| · |||-closed
because the ∥ · ∥-topology is coarser than the ||| · |||-topology of X .
Since x0 is a semi-interior point of X+ , by Proposition 3.2, x0 is an
 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48
||| · |||-interior point of X+ . So we consider as commodity space the
ordered normed space (X , ||| · |||).
Then all the assumptions of Theorem 4.4 are satisfied, hence any
weakly Pareto optimal allocation is supported by a positive, linear
functional p of X , which is ||| · |||-continuous and p(x0 ) = 1. If ω is a
|||·|||-almost order unit of X , then by Proposition A.1 in the Appendix,
we have that p(ω) > 0.
For the proof of (ii) we remark that the utility functions are
||| · |||-radially continuous because the ||| · |||-topology of X is finer
than the ∥ · ∥-topology. So for the commodity space (X , ||| · |||),
the assumptions of Theorem 4.10 are satisfied, hence a quasi
equilibrium allocation x exists which is supported by a positive,
||| · |||-continuous, linear functional p of X . If ω is a ||| · |||-almost order
unit of X , we have p(ω) > 0.
Statement (iii) follows similarly, by Theorem 4.11. 
Example 5.2. Suppose that in the economy E the commodity
space is the space X of Example 2.5 equipped with the norm ∥ · ∥
as it has been defined in this example. In this example, we have
shown that X+ does not have interior points but X+ has semi-
interior points. If the economy satisfies statement (ii) or statement
(iii) of Theorem 5.1, then an equilibrium allocation x exists and if
moreover ω is a ||| · |||-almost order unit of X , for example if ω is
the constant vector 1 = (xn ) with xn = (1, 1) for any n, then x is a
quasi-valuation equilibrium.
6. Strongly reflexive cones
In this section we suppose that X is an ordered normed space
with strongly reflexive positive cone X+ . According to Casini et al.
(2013), the cone X+ is called strongly reflexive if the positive part
U+ = U ∩ X+ of the unit ball U of X is compact. Strongly reflexive
cones have been studied in Casini et al. (2013), where it is shown
that this class of cones is a rich class of cones in Banach spaces. Our
Examples 2.6 and 2.7 are examples of strongly reflexive cones with
semi-interior points. We start with the next easy result.
Proposition 6.1. Suppose that in the economy E the commodity
space X is a normed space ordered by the normal and strongly reflexive
cone X+ . If the preferences are defined by continuous utility functions,
then the utility space U is compact.
Proof. The cone X+ is normal, therefore 0 ≤ x ≤ y H⇒ ∥x∥ ≤
c ∥y∥, for any x, y ∈ X , where c is the constant of the normal cone
X+ . For any x = (x1 , x2 , . . . , xl ) ∈ K we have xi ∈ [0, ω] for any i,
therefore ∥xi ∥ ≤ c ∥ω∥ and
xi ∈ W = c ∥ω∥U+ .
The set W is a compact subset of X because U+ is compact.
Therefore the set K as a closed subset of the compact set W l of
X l is also compact. Hence U is compact as the image of K via the
total utility function. 
The next theorem is a reformulation of statements (ii) and (iii) of
Theorem 5.1 in the case where the cone X+ is strongly reflexive. In
this theorem, we avoid the closedness condition but we need the
full continuity of the utility functions.
Theorem 6.2. Suppose that in the economy E the preferences are
monotone and convex, the preferences are defined by continuous
utility functions and the positive cone X+ of X is normal and strongly
reflexive with semi-interior points.
(i) If the conditions (A1), (A1c), (A2) and (A4) are satisfied, a quasi
equilibrium allocation x exists which is supported by a positive,
||| · |||-continuous linear functional p of X .
47
(ii) If Pi = X+ for any i, ω is an extremely desirable bundle for any
consumer, then a quasi equilibrium allocation x exists supported
by a ||| · |||-continuous, positive linear functional p of X .
If in anyone of the above statements, we suppose moreover that ω is a
|||·|||-almost order unit of X , then in this statement we have p(ω) > 0,
therefore x is a ||| · |||-quasi valuation equilibrium.
Example 6.3. We continue Example 2.6. Let E be an infinite
dimensional Banach lattice with a normalized, positive Schauder
basis {ei }. We may suppose for example that E is one of the spaces
c0 or ℓp with 1 ≤ p < +∞. Suppose that P is the cone of
Example 2.6. Suppose that in the economy E , the commodity space
is the subspace X = P − P of E and suppose that X is ordered by the
cone P, i.e. X+ = P. As we have noted in Example 2.6, X is dense in
E and also the cone P is normal and strongly reflexive with semi-
interior points.
If the economy satisfies statement (i) or statement (ii) of
Theorem 6.2, then an equilibrium allocation x exists and if
moreover ω is a ||| · |||-almost order unit of X , then x is a quasi-
valuation equilibrium.
Example 6.4. We continue Example 2.7. Let E be the space L1 [0, 1]
and suppose that P is the subcone of E+ of this example. Suppose
that in the economy E , the commodity space is the subspace X =
P − P of E ordered by the cone P, i.e. X+ = P. As we have noted
in this example, X is dense in L1 [0, 1] and the cone P is normal
and strongly reflexive with semi-interior points. As in the previous
example, if the economy satisfies statement (i) or statement (ii) of
Theorem 6.2, then an equilibrium allocation x exists.
Acknowledgments
The Authors thank very much two anonymous referees for
their valuable comments which improved the presentation of this
article.
Appendix. Ordered normed spaces
In this section we give some essential notions and results from
the theory of (partially) ordered normed spaces which are needed
in this paper. For more information see in Aliprantis and Tourky
(2007), Jameson (1970) and Aliprantis and Border (2006).
An ordered normed space is a linear space X equipped with
a reflexive, antisymmetric and transitive relation ≥ with the
property: x ≥ y H⇒ x + z ≥ y + z and λx ≥ λy, for any x, y, z ∈ X
and any real number λ ≥ 0. Then we say that ≥ is an order relation
of X and X+ = {x ∈ X | x ≥ 0} is the positive cone of X . Note that
the order relation of X is not necessarily complete.
Let X be normed space and let P be a cone of X (i.e. P is a
nonempty, convex subset of X so that λP ⊆ P for every real number
λ ≥ 0 and P ∩ (−P ) = {0}). The cone P ⊆ X induces the order
relation ≥ in X so that x ≥ y if and only if x − y ∈ P, for any
x, y ∈ X . Then X is an ordered normed space with positive cone
the cone P, i.e. X+ = P.
If P − P = X the cone P is generating or reproducing. For any
x, y ∈ X with x ≤ y, the set [x, y] = {z ∈ X | x ≤ z ≤ y} is the
order interval defined by x, y. A linear functional f of X is positive
if f (x) ≥ 0 for each x ∈ P and strictly positive if f (x) > 0 for each
x ∈ P , x ̸= 0.
The topological dual X ∗ of X is ordered by the dual cone P 0 =
{f ∈ X ∗ | f (x) ≥ 0, for any x ∈ P }. This cone is also denoted by
∗ ∗
X+ and is also referred as the dual cone of X+ , i.e. X+ = {f ∈ X ∗ |
f (x) ≥ 0, for any x ∈ P }.
The cone P is normal if there exists a real number a > 0 so that
for any x, y ∈ X , 0 ≤ x ≤ y H⇒ ∥x∥ ≤ a∥y∥. A subset B of P is
a base for the cone P if a strictly positive linear functional f of X
48 A. Basile et al. / Journal of Mathematical Economics 71 (2017) 36–48
exists so that B = {x ∈ P | f (x) = 1}. Then we say that the base B
is defined by the functional f .
If for any x, y ∈ E the supremum x ∨ y and the infimum x ∧ y of
the set {x, y} exist in X , then X is a vector lattice and we denote by
x+ = x ∨ 0, x− = (−x) ∨ 0 and |x| = x ∨ (−x) the positive part,
the negative part and the absolute value ∞ of x.
Suppose that x0 ∈ P. Then Ix0 = n=1 [−nx0 , nx0 ] is the solid
subspace of X generated by x0 . If Ix0 = X then x0 is an order unit of
X and if Ix0 is dense in X i.e. I x0 = X then x0 is an almost order unit
or a quasi interior point of P. We underline here the difference
between the notions of semi-interior point and the notion of the
quasi interior point of P. Recall also that any interior point of P is
an order unit of X , Aliprantis and Tourky (2007, Lemma 2.5), but
the converse is not always true. For the converse the completeness
of X is needed. We have: If X is a Banach space and P is closed, then
every order unit of P is an interior point of P (Aliprantis and Tourky,
2007, Theorem 2.8). Recall that a vector x ∈ P is said strictly
positive if p(x) > 0 for any nonzero, positive, continuous, linear
functional p of X and recall also that in the case where int (P ) ̸= ∅
we have: a vector x ∈ P is strictly positive if and only if x is an
interior point of P, Aliprantis and Tourky (2007, Lemma 2.17).
Although the next property is easy, it is important for this
article.
Proposition A.1. Any almost order unit x0 of X is strictly positive.
Proof. Suppose that p(x0 ) = 0, where p is a nonzero, positive,
continuous, linear functional of X . Then for any x ∈ Ix0 , there exists
a natural number nx > 0 so that −nx x0 ≤ x ≤ nn x0 . Since p is
positive we have
−nx p(x0 ) ≤ p(x) ≤ nx p(x0 ),
therefore p(x) = 0, for any x ∈ Ix0 . Hence p = 0 on Ix0 and therefore
p = 0 on X because Ix0 is dense in X and p is continuous. This is
a contradiction because we have supposed that p ̸= 0, therefore
p(x0 ) ̸= 0, hence p(x0 ) > 0 because p is positive. 
We present the next property with his proof (see also in
Jameson, 1970 Proposition 3.8.6) because for its importance in our
article but also in economics.
Proposition A.2. If x0 is an interior point of P, then the set
∗
B = {f ∈ X+ | f (x0 ) = 1},
∗
is a weak-star compact base for the dual cone X+ = {f ∈ X ∗ | f (x) ≥
0, for any x ∈ X+ } of X+ .
Proof. Consider the continuous linear functional x0 of X ∗ so that
x0 (f ) = f (x0 ), for any f ∈ X ∗ . Then by Proposition A.1, x0 is strictly
∗ Polyrakis, I.A., 2008. Demand functions and reflexivity. J. Math. Anal. Appl. 338,
positive on X+ . Suppose that x0 + rU ⊆ X+ , where U is the closed
unit ball of X and r > 0. By the definition of B we have that B is
∗
the base of X+ defined by the linear functional x0 . For any f ∈ B we
have f (x0 + rU ) ≥ 0 H⇒ 1 + f (rU ) ≥ 0. But for any x ∈ U we have
−x ∈ U, therefore |f (x)| ≤ 1r and ∥f ∥ ≤ 1r for any f ∈ B. So the set
B is bounded. Also B is weak-star closed because (x0 ) is weak-star
∗
continuous and the cone X+ is weak-star closed, therefore the base
B is weak-star compact. 
Recall that for any A ⊆ X we denote by A, the closure of A, by
int (A) the set of interior points of A by co(A) the convex hull of A
and by co(A) the closed convex hull of A.
If X is an ordered Banach space, the sequence {xn } is a positive
basis of X if it is a Schauder basis of X and the positive cone X+ of
X and the positive cone of the basis {xn } (i.e. the set of elements
of X with positive coordinates) coincide. Note that the usual bases
of the spaces ℓp , 1 ≤ p < ∞ and the space c0 are the simplest
examples of positive bases.
Note that a Banach lattice is an ordered Banach space X which
is a lattice, with the property: |x| ≥ |y| H⇒ ∥x∥ ≥ ∥y∥, for
any x, y ∈ X . Also an ordered normed space X is a locally solid
lattice if it is a lattice and there exists a real number a > 0 so that
|x| ≥ |y| H⇒ a∥x∥ ≥ ∥y∥, for any x, y ∈ X .
References
Aliprantis, C.D., Border, K.C., 2006. Infinite Dimensional Analysis, A Hitchhiker’s
Guide, third ed. Springer.
Aliprantis, C.D., Brown, D.J., Burkinshaw, O., 1990. Existence and Optimality of
Competitive Equilibria. Springer-Verlag.
Aliprantis, C.D., Cornet, B., Tourky, R., 2002. Economic equilibrium: optimality and
price decentralization. Positivity 3, 205–241. Special Issue on Mathematical
Economics.
Aliprantis, C.D., Monteiro, P.K., Tourky, R., 2004. Non-marketed options, non-
existence of equilibria, and non-linear prices. J. Econom. Theory 114, 345–357.
Aliprantis, C.D., Tourky, R., 2007. Cones and Duality. In: Graduate Studies in
Mathematics, vol. 84. American Mathematical Society.
Araujo, A., 1985. Lack of Pareto optimal allocations in economies with infinitely
many commodities: The need for impatience. Econometrica 53 (2), 455–461.
Back, H., 2004. Edgeworth and Walras equilibria of an arbitrage-free exchange
economy. Econom. Theory 23, 353–370.
Bewley, T.F., 1972. Existence of equilibria in economies with infinitely many
commodities. J. Econom. Theory 4, 514–540.
Casini, E., Miglierina, E., Polyrakis, I.A., Xanthos, F., 2013. Reflexive Cones. Positivity
17, 911–933.
Debreu, G., 1954. Valuation equilibrium and Pareto optimum. Proc. Natl. Acad. Sci.
USA 40, 588–592.
Florenzano, M., 1983. On the existence of equilibria in economies with an infinite
dimensional commodity space. J. Math. Econom. 12, 207–219.
Jameson, G.J.O., 1970. Ordered Linear Spaces. Springer-Verlag, Berlin.
Khan, A.M., Tourky, R., Vohra, R., 1999. The supremum argument in the new
approach to the existence of equilibrium in vector lattices. Econom. Lett. 63,
61–65.
Mas-Colell, , 1986. Price equilibrium existence problem in topological vector
lattices. Econometrica 54 (5), 1039–1053.
Mas-Colell, A., Richard, S.F., 1991. A new approach to the existence of equilibria in
vector lattices. J. Econom. Theory 53, 1–11.
Mas-Colell, A., Zame, W.R., 1991. Equilibrium theory in infinite dimensional spaces.
In: Hildenbrand, W., Sonnenschein, H. (Eds.), Handbook of Mathematical
Economics, Vol. 4. pp. 1835–1898.
Podczeck, K., Yannelis, N.C., 2008. Equilibrium theory with asymmetric information
and with infinitely many commodities. J. Econom. Theory 141, 152–183.
695–704.
Xanthos, F., 2014a. A note on the equilibrium theory of economies with asymmetric
information. J. Math. Econom. 55, 1–3.
Xanthos, F., 2014b. Non-existence of weakly Pareto optimal allocations. Econ.
Theory Bull. 2, 137–146.
Singer, I., 1970. Bases in Banach Spaces I. Springer, Heidelberg.