Arrow’s Pricing Formula for Securities

Let S be a finite set of states of the world and let N be the index set for a finite set of consumers.
Generic elements of S and N are denoted by s ∈ S and i ∈ N . In a slight abuse of notation,
we also use S and N to denote the number of elements in S and N . We assume that there is
only one good; each consumer i ∈ N is endowed with x̊i0 units of the good today and with x̊is
units in state s tomorrow. It’s convenient to think of the good as dollars. Let x̊i = (x̊i1 , . . . , x̊iS ).
A consumption plan for consumer i is a (1 + S)-tuple (xi0 , xi ) ∈ R1+S
+ , and an allocation is an

N -tuple of plans, (xi0 , xi )i∈N . Each consumer evaluates consumption plans according to a utility
function ui : R1+S
+ → R. The economy is fully described by the set S of states and by the N -tuple
(ui , x̊i0 ,x̊i )i∈N of consumers.

A set of securities for this economy is an S × K matrix D. Each column of D is the S × 1 returns
vector or dividends vector of one of the securities: the element dsk specifies how many dollars one
unit of security k will return tomorrow if state s occurs. Note that dsk may be positive, zero, or
negative. The K columns of D are thus the K securities. Consumers purchase or sell units of the
securities today and hold them until tomorrow, when one of the states s ∈ S is realized and each
security k returns dsk dollars for every unit of the security a consumer owns. We denote by yki
the number of units of security k purchased by consumer i; yki may be positive, zero, or negative.
Consumer i’s portfolio is the K-tuple (y1i , . . . , yK
i
), which we denote by yi . Note that if consumer
i purchases the portfolio yi , then his vector of state-contingent returns will be the S-tuple Dyi .
(It’s most convenient here to write yi and Dyi as K × 1 and S × 1 column vectors.) We denote
the price of security k by qk , and we write q = (q1 , . . . , qK ).

Definition: An equilibrium of the securities markets defined by the matrix D is a (K +

N (1+S)
N K + N (1 + S))-tuple (q, (yi )i∈N , (xi0 , xi )i∈N ) ∈ RK
+ ×R
NK
× R+ that satisfies the utility-
maximization and market-clearing conditions:

(U-M) ∀i ∈ N : (yi , xi0 , xi ) maximizes ui (xi0 , xi ) subject to the constraints

xi0 + q · yi 5 x̊i0 and
xis 5 x̊is + K i
i.e., xi 5 x̊i + Dyi
P
k=1 dsk yk , ∀s ∈ S,
PN i PN i PN i
(M-C) i=1 x 0 = i=1 x̊ 0 and i=1 yk = 0, k = 1, . . . , K.

Examples: Our “Extended Example of Equilibrium Under Uncertainty” contains several exam-
ples of securities markets using this model. Part 3 of the example is a market with a single security,
a credit instrument such as a saving account or a bond. Part 4 adds a second security, an insurance
contract.

1
Example: Suppose there are only two states, s = H and s = L, and one security, which returns
a in state H and b in state L. By choosing y, the number of units of the security he will buy at
today’s security price q, a consumer can vary xH and xL , but not independently:
" # " #
xH − x̊H a
= y and x0 = x̊0 − qy.
xL − x̊L b

Thus, giving up y units of consumption today will only allow him to augment his consumption
tomorrow by multiples of (a, b) across the two states.

Now suppose there’s a second security, which returns c in state H and d in state L. If (c, d) is
a multiple of (a, b), then nothing is gained by the introduction of the second security: choosing
amounts y1 and y2 of the two securities still augments one’s consumption tomorrow only by mul-
tiples of (a, b). But if (a, b) and (c, d) are not multiples of one another — i.e., if they’re linearly
independent — then for any state-contingent consumptions xH and xL tomorrow, the equation
" # " #" #
xH − x̊H a c y1
=
xL − x̊L b d y2

has a solution (y1 , y2 ). Thus, in this case, state-contingent consumption tomorrow can be aug-
mented by any amounts xH − x̊H and xL − x̊L by giving up some amount of consumption today
in order to purchase some amounts y1 and y2 of the two securities. More securities would not add
anything, but would not hurt either: as long as the securities returns matrix has two linearly inde-
pendent columns (securities), any state-contingent consumptions can be achieved. More generally,
with S states, the securities returns matrix D must have S linearly independent columns — i.e.,
we must have rank D = S. We could equivalently say that the securities must span the space RS .

It seems intuitive that this spanning condition will be necessary and sufficient to ensure that the
securities markets achieve the same outcome as with complete Arrow-Debreu contingent claims
markets — that an equilibrium allocation attained via securities markets will coincide with an
Arrow-Debreu allocation. We now verify this intuition.

To simplify notation, let’s temporarily substitute z0 for x0 − x̊0 and zs for each xs − x̊s . The
key to establishing the equivalence of equilibrium outcomes is the individual consumer’s budget
constraints: we show that if the securities span RS , then both market structures present the
consumer with exactly the same budget sets at their respective equilibrium prices. In our z-
notation, the consumer’s Arrow-Debreu budget constraint is z0 + p · z = 0. We wish to be able
to show that at some security prices q the constraint z0 + q · y = 0, together with the fact that
z = Dy, makes exactly the same set of (z0 , z)’s available as the constraint z0 + p · z = 0 does.

2
The following proposition establishes that this is so if the securities span RS and if their prices are
related to the contingent claims prices p according to q = pD. The proposition then leads to the
subsequent theorem which establishes the equivalence between the securities markets equilibrium
and the Arrow-Debreu equilibrium.

Proposition: Let p ∈ RS ; let D be an S × K matrix; let q = pD ∈ RK ; and let

A = {(z0 , z) ∈ R1+S | z0 + p · z = 0} and
B = {(z0 , z) ∈ R1+S | ∃y ∈ RK : z0 + q · y = 0 and z = Dy}.
If rank D = S, then A = B.

Proof:
Note that if z = Dy then p · z = p · (Dy) = (pD) · y = q · y. We show that A ⊆ B and B ⊆ A.

(i) Let (z0 , z) ∈ A. Since rank D = S, there is a y ∈ RK that satisfies z = Dy. Since z0 + p · z = 0
(because (z0 , z) ∈ A) and p · z = q · y (because z = Dy), we have z0 + q · y = 0, and therefore
(z0 , z) ∈ B.

(ii) Let (z0 , z) ∈ B. Then, according to the definition of B, there is a y ∈ RK that satisfies both
z0 + q · y = 0 and z = Dy. Therefore p · z = q · y, and it follows that z0 + p · z = 0, and therefore
(z0 , z) ∈ A. k

Theorem: Let D be an S ×K securities returns matrix that satisfies rank D = S, and let q = pD.
If (p, (xi0 , xi )i∈N ) is an Arrow-Debreu equilibrium for the economy E = (S, (ui , (x̊i0 ,x̊i ))i∈N ), then
there is a profile (yi )i∈N of portfolios for which (q, (yi )i∈N , (xi0 , xi )i∈N ) is an equilibrium of the
securities markets defined by D for the economy E. Conversely, if (q, (yi )i∈N , (xi0 , xi )i∈N ) is a
securities-markets equilibrium, then (p, (xi0 , xi )i∈N ) is an Arrow-Debreu equilibrium for E.

Remark: Note that the allocation (xi0 , xi )i∈N is the same in both equilibria — i.e., everyone’s
state-contingent consumption is the same in both equilibria.

Proof of the Theorem: This is a simple corollary of the preceding proposition. For each i ∈ N ,
we let xi0 −x̊i0 and xi −x̊i play the roles of z0 and z in the proposition. The set A in the proposition
is therefore the set of plans (xi0 , xi ) available to consumer i — consumer i’s budget constraint —
at the equilibrium price-list p in the Arrow-Debreu equilibrium, and the set B is the set of plans
available to him at the securities prices q = pD in the corresponding securities markets. If
rank D = S, then the two sets of available plans (xi0 , xi ) are identical, and the consumer will
therefore choose the same plan when facing either price-list. Therefore the utility-maximization
and market-clearing conditions are satisfied in one case if and only if they are satisfied in the other
case. k