Economy of Referential Preferences

A new mathematical approach for choice theory and general equilibrium

Teycir Goucha*

Abstract
In this paper we introduce basic notions of a new economic model where preference relations on commodities set are represented by a group action on Euclidean space instead of utility function. Conditions that ensure the existence of individual demand function and a general equilibrium in the setting of exchange economy are examined.

JEL C62, D50, D51 Keywords: General Equilibrium, Preference Relations, Group Theory.

INTRODUCTION
The mathematical modern conception of general economic equilibrium (GEE) is provided by Arrow-Debreu model developed from 1950 (Arrow, Debreu 1954). This model pictures the economy as a collection of m economic agents who make supply and demand decisions over a nite set of l commodities in order to further their own interests. The general equilibrium research program then studies many properties of economy, particularly the price, choices of agents, individual and aggregated demand functions (Balasko, 1998). In a pure exchange model, all agents are consumers, and each of them is provided with a preference relation represented by a utility function on Rl and an initial endowment e 2 Rl+ representing his supply o er in the market. Agents are assumed to take as given the market prices of goods. In exchange for his supply, each agent tries to choose the consumption bundle which maximizes his utility given his budget constraint. Such bundle represents the individual demand. Aggregated demand of an economy is the sum of all individual ones, and it is clearly a function of price. Equilibrium, is by de nition the vector price p 2 Rl+ which makes all markets clear (Supply = Demand). The centerpiece of the subject (GEE) deals with the existence and properties of equilibrium. To ensure an a rmative answer to that question, many conditions on preference relations, and hence on utility functions, are assumed. In summary, it is assumed that preferences are continuous, monotonic and convex, or equivalently, utility functions are di erentiable and concave. When these conditions hold for all agents, the economy is then called neoclassical, and equilibrium prices can be reached (Aliprantis & al, 1989).
. Teycir Goucha, Ariana, Tunisia, e-mail: goucha@hotmail.com

The aim of this paper is to build a new general formulation of consumers' choice where rationality involves not only maximization of preference, but also a well de ned reference of choice, hence our terminology of Economy of Referential Preference (ERP). Although it is clear that this approach can replace, in many instances, the conventional one based on utility function, it is not our main purpose in this paper. Since this paper is intended for both mathematicians and economists, the main basic notions of these two elds are presented. The rst section is devoted to basic tools of the GEE in the Arrow-Debreu model of the nite dimensional case. In section two we treat several examples that show the consistency of the group action approach and we explicitly determine the individual demand function. We end by proving our main theorem establishing the existence of an equilibrium in the ERP.

1. The general equilibrium model of an exchange economy

The purpose of this section is to present a brief introduction to the theory of general economic equilibrium in the neoclassical framework with a nite number of commodities and agents. For more details about notions and theorems of this section, see (Aliprantis & al, 1989, Ch 1).

Preference is a binary relation on Rl+, it's re exive, transitive and induce a total order on Rl+. The expression x <i y means x is as good as y for the consumer i. When both x <i y and y <i x we say that he is indi erent to x and y , and written x i y . The set of all preferences on Rl+ will be denoted by P . In the classical exchange economy there is a market for every good and only one period where all markets open once and simultaneously. As mathematical de nition of exchange economy we can take the following one (see Aliprantis & al, 1989, p 29):

1.1 Fundamental notion of exchange economy: commodity, preference and demand An economy is given by a set L = f1; 2; ; l g of commodities and m agents. In the pure exchange model, all agents are consumers. The consumer i 2 f1; ; m g is characterized by an initial endowment ei 2 Rl+, which represents his supply, and a preference relation <i on Rl+.

De nition 1.

An exchange economy E is a function from a non-empty set A (called the set of agent or consumers) into Rl+ P ; i.e., E : A Rl+ P .

If E : A Rl+ P is an economy, then the value Ei = (ei ; i ) represents the characteristics of agent i; respectively its initial endowment and his preference or taste. In this paper we restrict our interest to the case of exchange economy. 2

Same as in economic reality, price plays a fundamental role in this theoretical model. There is a price for each unit of commodity l; and we will denote by p = ( pl), the vector of prices. In GEE, all agents (consumers) are assumed to be price-takers: they believe that their actions (supply and demand for commodities) do not a ect prices. Prices serve to describe the rate at which commodities can be exchanged. In other words, two bundles x; y 2 Rl+ can be exchanged, if h p; x i = h p; y i, where h ; i is the scalar product in Rl. Given his initial endowment ei, and vector price p, each agent computes his own budget wi = h p; ei i: The consumer's problem consists then to choose among budget feasible bundles, those which are optimal with respect to their preference relation. This means n forothe consumer i, to choose x 2 Rl+, such that x <i y , 8 y 2 B i( p) = y 2 Rl+; h p; y i = wi . From a mathematical point of vue, an abstract preference relation is not very useful in computational perspective. Therefore we need a more sophisticated analytic tool. One of the basic considerations associated with several economic theories is the notion of utility.

De nition 2.

A utility function for a given preference relation < on such that x < y holds if and only if u(x) > u( y).

Rl+, is a function u : Rl+

The utility function is very useful and convenient for describing preference. In the hypothesis of a giving price p, the ith consumer's problem is equivalent to maximizing the utility, subject to budget constraint. Namely, the problem is:
Maximize u (x) Subject to budget constraint h p; x i 6 w
i i

(1)

When this problem admits a unique solution, it becomes by de nition the demand for the consumer i. To ensure the existence of such demand, several conditions on preference (utility function) like continuity or convexity must be assumed. We summarize these conditions on the following de nition:

De nition 3.
1. 2.

A continuous preference

on Rl+ is said to be a neoclassical whenever either:

is strictly monotone and strictly convex; or else is strictly monotone and strictly convex on int(Rl+), and everything in the interior is preferred to anything on the boundary.

For more details about axioms on preferences and their economic interpretations, see (Balasko, 1998, Ch 2). The consumer can be now de ned through his demand function which associates with every pair ( p; wi) 2 Rl+ R of a strictly positive price-vector p and a positive wealth (or income) the bundle of commodities f ( p; wi) 2 Rl which is the unique solution of (1). When ei is xed, and so it becomes the case of wi, we can view fi as function of price p, i.e fi = fi( p) 2 Rl. 3

One of main proprieties of demand function states that it is homogeneous of degree 0, ie, f ( :p) = fn( p) 8 2 R + : Then, it is possible to deal only with normalized price p 2 S l, or o p 2 l = p 2 Rl+; p1 + p2 + + pl = 1 . The set of all strictly positive prices will be denoted by S .

1.2 Conditions of existence of general equilibrium

We will restrict our attention to the neoclassical model of an exchange economy A neoclassical exchange economy is an exchange economy E : A 1. The set A of agent is nite. 2. Each agent i has a non-zero initial endowment ei (i.e., ei > 0) and his preference relation i is neoclassical. 3. The total endowment e =
P

De nition 4.

Rl+ P such that:

ei is strictly positive, i.e. e 0 holds.

The de nition of general equilibrium can be now announced: Let E : A Rl+ P , an exchange economy with m agents. Let ei 2 Rl+ denote the initial endowment of agent iP fi its demand function. We say that the vector p is an and P equilibrium price for E if I=1 fi( p) = I=1 ei. i i

De nition 5.

Theorem 6. Each neoclassical exchange economy has an equilibrium.

The proof of this theorem is based on the following mathematical result, which we will use in section two to prove our main result.

Theorem 7. Let S = p 2 Rl ; pi > 0 for i = 1; 2; ; l; p1 + p2 + pl = 1 the set of all strictly positive prices. For a function ( ) = ( 1( ); 2( ); ; l( )) from S into Rl assume that:
i. ii. is continuous and bounded from below; satis es Walra's Law, i.e., p ( p) = 0 holds for each p 2 S ;

iii. f pn g S ; pn p = ( p1; ; pl) and pk > 0 imply that the sequence f k( pn)g of the kth components of f ( pn)g is bounded; and iv. pn

p 2 @S with f pn g S imply limn 1 k ( pn)k1 = 1.

4

Then, there exists at least one vector p 2 S satisfying ( p) = 0:

For proofs of theorems 6 and 7 we refer the reader to (Aliprantis & al, 1989, p 32-34).

2. Referential preference and conditions of equilibrium

In this section it is shown by examples that preference relations on commodities set can be represented by a group action on Rl. This viewpoint sheds some new light on the economic rationality and conditions of equilibrium. In this work we will touch only a few aspects of group theory and knowledge of elementary matricial calculus is su cient ( see Roman, 2012, for details and many examples of group action).

We assert that indi erence sets of u are exactly the orbits for vectors on R2 . + Indeed, except the trivial case (c = 0) which is obviously a union of two orbits, x c > 0 o n 0 and the indi erence set Ic = x 2 R2 ; u x = c : Given any commodity x0 2 Ic, + ! y y y his orbit is nothing but Ic itself. Actually, for any g = a 0 2 G, it is clear that 0 1 ! ax0 x ~ x0 x0 g y0 = 1 y0 2 Ic. Conversely, any commodity y 2 Ic is in the orbit of y0 , ~ 0 ~ 1 ! 0 ~ ~ 0 since x = B 00 ~ C x0 = a 0 x0 where a = xx0 = 1 which is due to the fact @ A y0 1 ~ y ~ y 0
a a x x y y

Before introducing our general framework, we begin by a simple example where we can see that indi erence sets of utility function may be represented, or more precisely replaced by group action on Rl. Example 1 The commodity space is R2 and the utility function u is: + u: R2 R ; u x = xy: + y ! a 0 We choose the one-parameter's subgroup G of GL(2; R), G = 1 ; a 2 R+ . The 0 2 is simply the Matricial one on the Euclidean space, namely: action of G on R ! ! x = ax ; where g = a 0 for some a > 0 : 1 g y y 0 1
a a a

2.1 Motivations and examples of referential preference

that x0 y0 = c = xy . ~~ It remains to show that any orbit is an indi erence set. This can be ~deduced from the 0 1 1 y ) = xy , and for all x such that xy = xy we have @ 0 A x = x . ~ ~ ~~ fact that (ax)( a y ~ y y ~ 0 ~
x x y y

As indi erence set Ic is arbitrary, this is su cient to conclude that the description of indi erence sets of consumer with given utility function u can be e ciently made by a group actions on R2 . + This example gains in interest only if we are able to see how group action becomes useful to de ne a mathematical framework of consumer's theory and general equilibrium. In other words, we have to de ne a complete order on Rl+ and a consumer maximization problem in this new setting. Actually, let G a topological group and a continuous action of G on Rl. Here and subsequently, Ox denotes the orbit of x 2 Rl under group action. It is easy to check that any group action induces an equivalence relation on Rl. Indeed, such equivalence can be obviously de ned as following: 5

But since this is not su cient to give a totally (complete) order on Rl+. Some other conditions are needed. With the notation Rl++ = x 2 Rl/ xi > 0 ; 1 6 i 6 l , our basic assumption is the following:

x y i 9 g 2 G st

g ( x) = y .

Of course this implies that the quotient of Rl++ by the equivalence relation induced by the action of group is identi ed with R+. Clearly, we can deduce a preference relation on Rl++ from a group action which veri es axiom 1. Indeed, we say that x is more desirable than y when vx > vy, and they are equivalent if vx = vy. 9 g 2 G such that g(x) = y x y. We simply note, that vx = vy The above axiom is not only a simple mathematical hypothesis, but it has an evident economic meaning which asserts that consumer compares each bundle with a very simple 011 one which is v:Il = v:@ A. By identifying v:Il and v 2 R, further analysis may eventu1 ally lead to interpret v:Il in terms of a medium of exchange or money. But this is still just a mere eventuality. In many examples, axiom 1 is available for all Rl+ and the above preference can be extended to all commodities on Rl+. When this is not the case we assume that all x 2 Rl++ are preferred to anything on the the boundary. Taking into account this detail, we state the following de nition:
l

Axiom 1 011 l ; there is a unique v 2 R such that x 2 O l For all x 2 R++ + v I where Il = @ A 2 R : 1 We will denote by vx the unique real v such that we have x 2 O v I .
l

We say that a preference relation < on commodity set Rl+ is of reference type, or referential, if it is given by a continuous and globally invariant group action on Rl+ which satis es axiom 1. Returning to the previous example, where u(x1; x2) = x1x2, we can see that x 4 y u(x) 6 u( y) vx 6 vy . Actually, u(x) 6 u( y) x1x2 6 y1 y2, but since (vx ; vx) 2 Ox 2 and (vy ; vy) 2 O y, we have vx = x1x2 and v2 = y1 y2. Under the condition vx ; vy > 0. it foly lows that vx 6 vy. Now we will solve a simple problem of consumer's demand with no use of utility function. The group G are the same as in example 1.

De nition 8.

Example 2. Let p = 1 ; 2
x x t

the price vector and w = 200 the budget of the consumer. Maximize v To solve the consumer's problem which is subject to the constraint p x 6 w , we set that x = ! t 0 v for some t and vx 2 R+. It's not di cult to verify that t and vx exist and v 0 1 thatq are unique. Actually, if x = x1 2 R2 , then we can see that vx = px1x2 and they ++ x2 x1 t = x2 . The budget constraint becomes: ! p t 0 1 tv + 3 v = 200 t v p; 0 1 v = w vx = t2400p3 . 2 x 2t x +
2
x x x t

p3

t We then p obtain vx(t) = t2400p3 , that reaches its maximum at t = 4 3 , for which we have + 4 3 ) ? 1. vx = 200( ! 200 Finally, the solution of this maximization problem gives us x = 200p3 as the con3 sumer's demand. To treat the general case we must give necessary and/or su cient conditions on groups to ensure reliability and e ciency of axiom preference and so the existence of individual demand function. Indeed, under the axiom 1, we have the following theorem:

Theorem 9. Let a consumer with referential preference relation on Rl+ given by a group
constraint is equivalent to the minimization of a nonnegative continuous real valued function on the group G. Proof. Since referential preferences are determined by v = vx where x = vx the demand function is given by the solution of the following problem:
Maximize vx
subject to the budget constraint h p; x i = w

G acting on Rl+. Then, the maximization problem for the consumer under the budget
g(I ),

then

This maximization problem is clearly equivalent to nding the maximal value of v, such that h p; v g(I )i = w. So, we have to maximize v = v( g ) = hp; w (I )i .
g

But since g(I ) 2 R+ and p 2 R++, we have h p; g(I )i > 0. As h p; g(I )i 0 8 g 2 G, continuity of v( g ) follows directly from continuity of group action and scalar product on Rl. As w is xed, and w and h p; g(I )i are both positive, then the problem is equivalent to minimizing h p; g(I )i for g 2 G.
l l

In the remainder of this section we assume that referential preferences satisfy the following axiom which is the cornerstone of our approach:

Axiom 2

Rl+ which de nes his preference For consumer i 2 I , the group action's : Gi Rl+ relations on the commodity space Rl+, satieties: 9! gi 2 Gi ; such that hI; g I i 6 hI; g Il i; 8 g 2 Gi.
i

group Gi. Then its demand function is explicitly given by: fi : Rl++ Rl+, fi( p) = v I;w I g ?1g I
i p gi p i

Theorem 10. Let ei 2 Rl++ the initial endowment of consumer i whose preference is de ned by a
where p = vp

gp

I , and wi = h p; ei i is the budget of consumer i.

Proof. Let p 2 Rl++ the giving vector price. By theorem 9 the maximization problem is equivalent to minimize h p; gI i for g 2 Gi. But since p 2 Rl++, there is g p 2 Gi and vp > 0 ; such that p = vp( g I ). Then we have to minimize g I; gI for g 2 G. Now, g I; gI = I; (g g) I , and, by axiom 2, the minimum is given for gi = g p g , or ? equivalently for g = gp 1 gi. Finally, vmax = v I;w I and fi = vmax g ?1 g I = w g ?1 g I . v I; I
p p p p i p i gi p i p gi p i

Remark 11. Since p = vp

I; where gp 2 Gi, we can write p = Mp I where Mp is the diagonal matrix with entries mi;i = pi . In other word M p = vp g and the individual ? demand function for consumer i takes this form: fi( p) = I; w I M p 1 g I .
gp
p i gi i

Corollary 12. The demand function is homogeneous of degree 0. Proof. Let 2 R+, from the above expression of individual demand function, fi( p) =
hM I; e i M ?1 I . As Mp is a diagonal matrix form of the p vector, then M p = g p I; I M p, and M ?p1 = ?1M p. This clearly implies fi( p) = fi( p).
p i gi i

2.2 Referential preferences and equilibrium Example 3

We start with an example taken from (Aliprantis & al, 1989) to see how our groups' based approach is able to provide same results as the conventional one based on utility function.

Let an economy with two commodities and three agents and note that ( p1; p2) is the vector price. Utility functions of agents are u1(x; y) = xy; u2(x; y ) = x2 y and u3(x; y ) = xy2, and their initial endowment are e1 = 1 ; e2 = 1 and e3 = 2 . These assump3 1 2 tions are extracted from example 1.4.10 in Aliprantis and all]. For us, all preferences are given by groups and their actions on R2 . + ! t Consumer 1. The group of preference is the matricial subgroup G1 = 0 0 ; t > 0 : Its 1 v=v maximization problem Maximize the budget constraint p1x + p2y = b1 where X = (x; y), is equisubject to ! t 0 valent to nding the greatest v such that P ; v 0 1 1 = p1 + 2 p2 since for each 1 ! t X = (x; y ) there is a unique t > 0 and v > 0 with X = v 0 0 1 . 1 1 ! t Then, we have to nd Max v > 0; such that v p1 ; 1 = p1 + 2 p2 which gives: p2 i h p1 p2 v tp1 + 1 p2 = p1 + 2 p2 v = t(t2p1++2p2) . t
t X t t t

Now, v = v(t) reaches its optimum when dv = 0, and this occurs at t0 = dt q q 2 t = p2 > 0 we have d v < 0 , then we obtain vmax = v( p2 ) = (p1p+12pp22) . p1 p1 2 p dt2 An easy calculation establishes the demand for the rst consumer: 0q 1 B 2 0 C (p1 + 2 p2) 1 + + x1( p) = B 0 1 r1 C 2pp1p2 1 = ( p1 2p2 p2 ; p1 2 p2 p2 ). B C @ A 2 1 2
p p p p

qp

p1 .
2

Since for

Same argument and relatively simple calculation gives the following results: ! 2 t 1 + p2 Consumer 2 The group is G2 = 0 0 ; t > 0 ; v(t) = tp(1p3 + p2) reaches its maximum 1 t at

t=3

q 2p

2 p1

of the 2nd

p where vmax = 2p2 1/31 + p2 p1 2/3 p1 p1 + p2 2p2 consumer is x2( p) = 2 p13+12p2 ; p13+2p2 p p

. From this, we deduce that the demand .

Consumer 3 The group is G3 =

q

t2 0

p1 2 reached at t0 = 3 2pp21 , and vmax = p ( 2 )22/3 + 3p( 2 1 )1/3 . Then we nd x3( p) = + p2 2 1 2 4 p1 + 6 p2 as the demand of consumer 3. 1 3 p2
p p p p

1
t

; t > 0 ; the maximum of v(t) = t(21pt13+ 3pp22) is p +

2 p1 + 3 p2 ; 3 p1

To calculate the equilibrium price, it su ces to establish the common equilibrium condiP P tion: Z ( p) = 3=1 xi( p) ? 3=1 ei = 0. It follows immediately that 16 p26?113 p1 ; i i p 13 p1 ? 16 p2 = 0. The last equality gives, under the condition p + p = 1, the value of 1 2 6 p2 price equilibrium, peq = 16 ; 13 . All these results are exactly the same obtained by the 29 29 use of utility functions. Based on the above examples and results, we suggest to de ne a new mathematical framework of an exchange economy where the set I of agents is nite. This is to be de ned as:

De nition 13. An exchange economy is said to be of referential preferences if:

The consumption set coincides with Rl+; Each agent i has a strictly positive initial endowment, i.e., ei 2 Rl++ and; The preference relation all i 2 T .
i

is referential (de nition 8), and satis es axiom 2, for

The main result of this paper is provided below:

Theorem 14.

Every exchange economy of referential preferences has an equilibrium price.

Proof. It is based on Theorem 7, section 1. According to theorem 10 P excess demand function in ERP is given by Z : S the P P P Z ( p) = i fi( p) ? i ei = i 2I v I;w I g ?1g I ? e, where e = i ei 2 R++.
i p gi p i

Rl,

First, the continuity of Z is a consequence of continuity of group action and scalar product on Rl. And since all fi 2 Rl+, then Z is clearly bounded from below. Second, as fi is the solution of maximization problem under the budget constraint then h p; fi(p)i = h p; ei i, and hp; Z ( p)i = 0 follows from the equality: Z ( p) = Pi fi( p) ? Pi ei . p = ( p1; ; pl) and pk > 0. To see why the sequence Third, let now f pn g S ; pn n g of the k th components of fZ ( p )g is bounded, we consider remark 11 and this fZk(p n P P ? expression of demand function: i fi( p) = i hM I; eI i M p 1 ( g I ). Since pn 2 S , then I; ? the kth component of M p and M p 1 are nonnegative for all n and tend respectively to pk and ( pk)?1, which clearly implies fi( pn) is bounded for all i 2 f1; 2; ; m g, and conk sequently the same holds for fZk( pn g.
p i gi i n n

Last, it remains to prove that lim n 1 k Z ( pn) k1 = 1 if pn p 2 @S with f pn g S : Let j 2 f1; 2; ; m g such that p j = 0. Then pn 0, and ( pn)?1 + 1 which implies j j th component of individual demand tends to in nity, namely we have M ? 1 that the j p hM I; e i > 0 for all i, then ( f ( p)) + + 1 for all consumer i. Since I; I ( g I) j i j 1, and it follows immediately that lim n 1 k Z ( pn) k1 = 1.
n p i i gi

Conclusion
In his theory of value, Gerard Debreu wrote: ``A state of the economy is a speci cation of the action of each agent ... But these actions are not necessarily compatible with the total resources. Can one nd a price system which makes them compatible? "(Debreu, 1959, p 74) In this work we prove that if, all agents choose their preference in some group setting, and make their choice in compliance with a simple general rule of referential nature, then we can nd a system of price which makes all choices compatible. An in depth work using additional examples will certainly allow us to come across other properties of referential preference and to better grasp its economic interpretations.

References
Aliprantis, C.D., Brown, D.J., Burkinshaw, O., (1989), Existence and optimality of Competitive Equilibrium. Berlin, Heidelberg, New York, Springer. Arrow, K. J. and G. Debreu, (1954), Existence of an Equilibrium for a Competitive Economy, Econometrica, 22, 265-290. Balasko, Y. (1998) Foundations of the Theory of General Equilibrium. Academic Press, Boston. Debreu, G. (1959), Theory of Value, An Axiomatic Analysis of Economic Equilibrium, New York: Wiley. Roman, S. (2011), Fundamentals of Group Theory, An Advanced Approach, Boston, Birkhauser.

10