You are on page 1of 5

Chapter 3

Arrow-Debreu exchange
economies

3.1 The economy


The economy is polupated by I consumers with preferences described by
U i : RL+ ! R and resources ! i 2 RL+ , i 2 I = f1; ::; Ig :1 An allocation is an
array (x1 ; ::; xI ) 2 RLI
+ , and it is feasible if it satises

X
I X
I
xi !i
i=1 i=1
In the special case in which L = 2, I = 2, feasible allocations can be graphi-
cally represented using the Edgeworth Box.
Unless otherwise noted, we shall impose the following strong (but not
outrageous) assumptions.
Assumption 2 U i : RL+ ! R is C2 in any open subset of RL+ ;strictly monotonic
in its arguments and strictly quasi-concave. Furthermore, ! 2 RLI++ :

3.2 Pareto e ciency


An allocation x 2 RLI + is Pareto e cient if it is feasible and there is no other
^ 2 RLI
feasible allocation x + which Pareto dominates it:

U i (^
xi ) U i (xi ) for all i, > for at least one i.
1
We are conscious of the notational abuse.

17
18 CHAPTER 3. ARROW-DEBREU EXCHANGE ECONOMIES

Pareto dominance denes a (social) preference relation over the set of allo-
cations RLI
+ , which is however incomplete.
PE allocations are solutions of the problem:
i i
max
P x iU (xP) i
ix i!
s.t. i0 i0 i0
U (x ) U for all i0 6= i
0
for some given U i i0 6=i :
The focs (for an interior solution) of this problem are:

rU i =
i0 0
rU i = for all i0 6= i
X X
xi = !i
i i

0
where i i0 6=i and are the Lagrange multipliers of the two sets of con-
straints above. Thus
i0 0
rU i = rU i for all i0 6= i

and utility gradients are co-linear for all agents (Marginal rates of substitu-
tion are equalized across agents).
0
Varying the values of U i for i0 6= i we obtain the set Pareto e cient
allocations, the Pareto frontier.
Let the Utility possibility set be the image of the feasible allocations in
the space of utility levels
( )
I
X X
U = U 2 RI U U i (xi ) i=1 , for some x 2 RLI
+ such that xi !i
i i

U is closed, convex, bounded above. Furthermore, the set of Pareto optimal


utilities is dened as

UP = fU 2 U j@U 0 2 U such that U 0 > U g :2


2
In our notation, U 0 > U requires U 0i U i , for any i 2 I; with at least one strict
inequality.
3.3. COMPETITIVE EQUILIBRIUM 19

Theorem 8 (Negishi) Let x 2 RLI + be a Pareto e cient allocation. Then


I
there exist a 2 R+ , =
6 0, such that
X
i
x = arg max U i (xi )
i2I
X X
i i
s.t. x !:
i i

Furthermore,
i
i0
i0
= ; for any i0 6= i 2 I:

Proof. Consider a Pareto optimal allocation x; so that U (x) = fU i (xi )gi2I 2


UP. Furthermore, U (x) 2 bdry(U), and U is closed, convex, bounded above.
An appropriate separating hyperplane theorem then implies that there exists
a 2 RI , 6= 0, such that

U (x) U; for any U 2 U;

that is,
X
i
U (x) = arg max Ui
i2I
s.t. U 2 U:

Trivially, then
X
i
x = arg max U i (xi )
i2I
X X
i i
s.t. x !:
i i

Finally, U is unbounded below, which implies that 2 RI+ :


PE allocations support points on the outer boundary of U P .

3.3 Competitive equilibrium


Agents trade in perfectly competitive markets.
20 CHAPTER 3. ARROW-DEBREU EXCHANGE ECONOMIES

A competitive equilibrium is an allocation x = (:::; xi ; ::) 2 RLI


+ and a
L
price p 2 R++ such that

xi 2 arg max U i (xi )


s.t. pxi p! i ;

for any i 2 I; and X X


xi !i:
i i

Trade is voluntary, hence equilibrium allocations will satisfy individual ra-


tionality:
U i (xi ) U i (! i ) for all i = 1; :::; I:
Letting mi = p! i we can derive agent Is Marshallian demand from his
consumer problem, as in the previous chapter. It is convenient, however,
represent demand functions in terms of endowments rather than wealth:

xi : RL++ RL++ ! RL+ :

Let z i (p; ! i ) denote agent i 2 Is excess demand: z i (p; ! i ) = xi (p; ! i ) ! i :


Finally, the aggregate excess demand z : RL++ RLI L
++ ! R+ is dened as
X
z (p; !) = z i p; ! i :
i2I

Proposition 9 For any economy ! 2 RLI ++ the aggregate excess demand


z (p; !) satises the following properties:

smoothness: z (p; !) is C 1 ;
homogeneity of degree 0: z ( p; !) = z (p; !) ;for any > 0;
Walras Law: pz (p; !) = 0; 8p >> 0;
lower boundedness: 9s such that zl (p; !) > s, 8l 2 L;
boundary property:

pn ! p 6= 0; with pl = 0 for some l; ) max fz1 (pn ; !); :::; zL (pn ; !)g ! 1:

Lets rst study the welfare properties of competitive equilibrium alloca-


tions.
3.3. COMPETITIVE EQUILIBRIUM 21

Theorem 10 (First welfare) All competitive equilibrium allocations are


Pareto e cient.

Proof. Suppose x = (::; xi ; ::) is a competitive equilibrium allocation for


some price p and it is not Pareto e cient. Then there must be another
allocation x^ = (::; x^i ; ::) which is feasible and Pareto dominates x. But since
xi is the optimal choice of consumer i at prices p and preferences are strongly
monotone, U i (^ xi ) U i (xi ) implies p x^i p xi for all i and also, p x^i p ! i ,
with all the previous inequalities being P strict forPat least some i. Summing
i i
the latter inequality over i yields p i x
^ > p i ! which contradicts the
feasibility of x^:

Theorem 11 (Second welfare) For any Pareto P ie cient allocation x 2


RLI
+ , there exist transfers t 2 R LI
, such that i t = 0 and x 2 R LI
+ is a
L
competitive equilibrium allocation (for some prices p 2 R++ ) of the economy
LI
with initial endowment ! + t 2 R+ .

Proof. Let ti = xi ! i for all i. The theorem is an implication of the


separating hyperplane theorem.
Let ! = (! i )i2I denote the vector of aggregate endowments:Competitive
equilibrium prices are solutions of:
X X
z(p; !) xi (p; p ! i ) ! i = 0;
i i

a system of L equations in L unknowns, the prices p. By Walras law


!
X X
p xi (p; p ! i ) ! i = 0; for all p
i i

and hence at most L 1 equations are independent (the market clearing


equation for one market can be o,itted without loss of generality). By ho-
mogeneity of degree 0 in p of xi (p; p ! i ); for any i 2 I; prices can always be
normalized, e.g., restricted without loss of generality to
( )
X
p2 L 1 p 2 RL+ : pl = 1 ;
l

the L-simplex, a compact ad convex set.

You might also like