State Prices, Arbitrage, and Contingent Claims

Prof. Sven Rady, Ph.D.

Microeconomics I BGSE, University of Bonn, Winter 2022/23

For investors, capital markets serve two purposes: they allow investors to transfer resources
over time (saving/borrowing), and they allow investors to shift resources in response to the
risks they face (insurance, risk sharing). The aim of this chapter is to investigate the allocative
functions of capital markets in the simplest possible framework, and to derive fundamental
properties of asset prices. This will then provide the foundations for option pricing.
We shall need the following notation for vectors x:
• x ≥ 0 if all components are ≥ 0;
• x > 0 if all components are ≥ 0, and at least one is > 0;
• x  0 if all components are > 0.
` for the set of all x ∈ IR` which are ≥ 0, and IR`
We shall write IR+ ++ for the set of all x  0.
Finally, given two vectors x and y of the same length, we say x ≥ y if x − y ≥ 0, etc.

1 The One-Period State Preference Model

Consider a simple economy with two dates, t = 0 and 1, a single physical good, and S different,
mutually exclusive, possible events at date 1, labelled s = 1, . . . , S. Each of these states is
assumed to occur with strictly positive probability. Thus, the economy has the ‘event tree’
shown in Figure 1.
An asset is a set of contractual obligations to deliver or receive a specified number of units of
the good in each of the S states that can occur at date 1. Suppose there are N assets, labelled
n = 1, . . . , N . Let the dividend of asset n in state s be dsn units of the good. An investor who
holds these assets in quantities xn obtains
N
X
ys = dsn xn
n=1

units of the good in state s. In vector and matrix notation, this means

y = Dx

where
y1 x1 d11 · · · d1N
     
 ..   ..   .. ..  ∈ IRS×N .
y =  .  ∈ IRS , x =  .  ∈ IRN , D= . . 
yS xN dS1 · · · dSN
State Prices & Arbitrage 2 Prof. S. Rady

t=0 t=1

w
......... s=1
..........
..........
..................
.
..
..........
..........
.........
..........

..........
..........
.....................
.............
w
............ s=2
.......... ..............
.......... ..............
.................. .. ...
. .....................
.. ....
.......... ..............
.......... ..............
.......... ..............
..........
........................
......... .........................
..............
..........................
w
.... s=3
.......... .............. ..........................
.......... .............. ..........................
......
............. ........................... ............. ........... ...........................
..........
... ................. ..........
......... ..............
. ..........................
..........................
.......... .............. ...................................................
........................
w q
.
.. . ...
..
..............................................................................
qq
. .
..................
.........................
........................
.......... ...............
.......... ..............
.......... ............... q
.......... .
.......... ...........................
.......... .............. qq
q
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
.......... ..............
..........
..........
..........
..........
..............
w
..............
s=S−1
..........
..........
..........
..........
..........
..........
..........
..........
w
.........
s=S

Figure 1: The event tree of the one-period state-preference model

The price of asset n at date 0 will be denoted by qn , so we have the asset price vector
q1
 
 .. 
q =  .  ∈ IRN .
qN
This implies that the portfolio1 x costs
N
qn xn = q> x.
X

n=1
The vector " # " #
−q> x −q>
= x ∈ IRS+1
Dx D
represents the dividend stream associated with portfolio x.2
A consumption plan specifies a level c0 of consumption at date 0, and a level of consumption
cs in each state at date 1. Using the notation
c1
 
 ..  S
c =  .  ∈ IR+
cS
1
For our current purposes, it is most convenient to identify portfolios in terms of the number of assets which
are held rather than the amount or the fraction of wealth invested.
2
Note that the top entry refers to date 0, while the remaining S entries refer to the S different states at date
1; this will be our notational convention whenever we collect date 0 and date 1 quantities into a single vector of
length S + 1.
State Prices & Arbitrage 3 Prof. S. Rady

for the date 1 part, we can write a consumption plan as

" #
c0 S+1
∈ IR+ .
c

An agent in this economy is characterized by an endowment

" #
e0 S+1
∈ IR+
e

of the consumption good, and by preferences over consumption plans. We describe these
preferences by a utility function U (c0 , c).
Preferences may, but need not, conform to the expected utility hypothesis. If they do, then
the overall utility U can be written as
S
X
U (c0 , c) = ps u(c0 , cs )
s=1

with probabilities ps and a von Neumann–Morgenstern utility function u(c0 , c1 ). Often, utility
is assumed to be additively separable, which means that u(c0 , c1 ) = u0 (c0 ) + u1 (c1 ), hence
S
X
U (c0 , c) = u0 (c0 ) + ps u1 (cs ).
s=1

The probabilities can be subjective, i.e., different across agents. We only require that each
agent assign positive probability ps > 0 to each state s.
" # " #
c0 e0
A consumption plan is feasible for an individual with endowment if it satisfies
c e
the S + 1 budget constraints
" # " # " #
c0 e0 −q>
≤ + x
c e D

for some portfolio x. The individual aims to maximize the utility of consumption over all
feasible plans.

2 State Prices
A vector ψ  0 of length S is called a vector of state prices for the given assets if

q> = ψ > D,

that is,
S
X
qn = ψs dsn
s=1
for all n. The intuition behind this concept is the following. If we interpret ψs as the price of
one unit of consumption in state s, then the last equation says that the given asset prices are
consistent with these prices for state contingent consumption: the asset price equals the sum
of the prices of dividends.
State Prices & Arbitrage 4 Prof. S. Rady

State prices play a central role in financial economics. As we shall see below, their existence
is closely linked to the internal consistency of a given system of dividends and prices. As a
consequence, they turn out to be of crucial importance in the theory (and practice!) of option
pricing.
Before moving on to these issues, however, let us briefly note that state prices need not be
unique. In fact, if we rewrite the linear system that defines state prices as
D> ψ = q,
we see that we have N equations for S unknowns. After eliminating redundant assets, we are
in fact left with only rank(D) ≤ S equations.3 Thus, the set of state price vectors is either
empty, or a linear manifold of dimension S − rank(D).

3 Arbitrage
An arbitrage is a portfolio x with dividend stream
" #
−q>
x > 0.
D
In other words, an arbitrage portfolio can be set up at a non-positive cost, yet it yields a
non-negative payoff in each state; moreover, either the initial cost is negative, or the payoff
is positive in at least one state. Thus, an arbitrage either gives you something today without
asking anything back tomorrow (a ‘free lunch’), or it creates something out of nothing (a ‘money
machine’).
This cannot happen in equilibrium, provided there is at least one rational investor who prefers
more to less. In fact, this investor would try to run the arbitrage strategy on an ever larger
scale, making unboundedly high profits, but never reaching an optimum. By doing so, on
the other hand, he would drive up the demand for assets that are ‘too cheap’, and decrease
the demand for assets that are ‘too expensive’. Reacting to these changes in demand, asset
prices would have to adjust until a situation without arbitrage opportunities is reached. Thus,
absence of arbitrage is one of the fundamental postulates of asset pricing. In particular, it is a
necessary condition for asset markets to be in equilibrium.
This raises an important question: How can we verify whether a given system of asset prices
and dividends is arbitrage-free? It seems impractical to check the above system of inequalities
for all possible portfolios x. Fortunately, it turns out that the existence of a state price vector
is a necessary and sufficient condition for absence of arbitrage, and you will see in an exercise
that this condition is easy to verify.

4 The Fundamental Theorem of Asset Pricing

The following theorem summarizes the links between absence of arbitrage, existence of state
prices, and existence of optimal portfolios for non-satiable investors. It is one of the most
important results of financial economics. While we present this theorem only in the simplest
possible setting, one can derive extensions to more complex (e.g. multi-period) models of asset
markets along the same lines.
3
Recall that the rank of a matrix is the maximum number of linearly independent rows or linearly independent
columns, and that D and D> have the same rank. As before, we call an asset redundant if its dividend can be
replicated as a linear combination of the dividends of other assets.
State Prices & Arbitrage 5 Prof. S. Rady

Theorem 1 (Fundamental Theorem of Asset Pricing) Given a matrix of dividends and

a vector of asset prices, the following three statements are equivalent:

(i) There is no arbitrage.

(ii) There exists a vector of state prices.

(iii) There is an optimal portfolio for some investor who prefers more to less.

We shall first prove the easy parts of this theorem.

Proof of (ii) ⇒ (i): Let ψ be a state price vector. Then
" #
−q>
[1 ψ> ] x = (−q> + ψ > D)x = 0
D

for all portfolios x. Now, if x were an arbitrage, we would have

" #
−q>
x > 0,
D

hence " #
−q>
[1 ψ> ] x>0
D
– a contradiction.
This proof formalizes the intuition that, if asset prices are consistent with a set of prices for
state-contingent consumption, it should be impossible to generate a free lunch or a money
machine. If, for example, a portfolio gives us positive (in the > 0 sense) dividends tomorrow,
it will increase our consumption tomorrow. Evaluated at the given state prices, this additional
consumption has a positive price. So the portfolio itself must trade at a positive price – it
cannot be an arbitrage.
Proof of (iii) ⇒ (i): We show the contraposition of this statement, i.e., ¬(i) ⇒ ¬(iii): if
there is an arbitrage, then the portfolio
" problem
# has no solution for a non-satiable investor.
e0
Consider an agent with endowment and utility function U which is strictly increasing
e
in all its S + 1 arguments. If x is an arbitrage and α ≥ 0, then the individual’s utility from
holding α units of portfolio x,
" # " # !
e0 −q>
U +α x ,
e D

is strictly increasing in α. So there is no optimal portfolio for this agent.

Note that this proof just formalizes the verbal argument presented in the previous section:
faced with an arbitrage opportunity, a non-satiable investor would like to run the arbitrage
strategy on an infinite scale.
The converse implications, (i) ⇒ (ii) and (i) ⇒ (iii), are more delicate since they require explicit
construction of a state price vector and an agent with optimal demand, respectively. In the
proof of (i) ⇒ (ii), we shall use the following version of
State Prices & Arbitrage 6 Prof. S. Rady

.... .... .... ....

.... .... .... ....
6 ... ... ... ...
.... ....... ....... ....... .......
.... .... .... .... ....
.... .... .... .... ....
.... ...... ...... ...... ...... ......
.... .... .... .... .... ....
... ... ... ... ... ...
.... ....... ....... ....... ....... ....... .......
.... .... .... .... .... .... ....
.... .... .... .... .... .... ....
.... ...... ...... ...... ...... ...... ...... ......
.... .... .... .... .... .... .... ....
... ... ... ... ... ... ... ...
.... ....... ....... ....... ....... ....... ....... ....... .......
.........
φ
.... .... .... .... .... .... .... .... ....
... ... ... ... ... ... ... ... ...
.... ....... ....... ....... ....... ....... ........ ........ ........ ........ K
.........
......... .... .... .... .... .... ..... .... .... .... ....
... ... ... ... ... ... ... ... ... ...
.........
......... .... ....... ....... ....... ....... ........ ....... ....... ....... ....... .......


......... .... .... .... .... ..... .... .... .... .... .... ....
......... ... ... ... ... ... ... ... ... ... ... ...
.........
.........
.... ....... ....... ....... ........ ....... ....... ....... ....... ....... ....... .......
......... .... .... .... .... .... .... .... .... .... .... .... ....
......... .... .... .... .... .... .... .... .... .... .... .... ....
......... . . . . . . . . . . . .
.........
...... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... .......
......... .... .... .... .... .... .... .... .... .... .... .... .... ....
......... . . . . . . . . . . . . .
.........
.........
.... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... .......
......... . . . . .
.... .... .... .... .... .... .... .... .... .... .... .... ..... ..... . . . . . . .
......... .... .... .... .... .... .... .... .... .... .... .... .... .... ....
.........
.. ... .. . .
.. .. .. .. .. .. .. .. .. .. .. .. .. ..... .....
. . . .. ... ... .. . .. . ... .
.. ..
. ..
.
.........
......... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... ....... .......

t ......... ... ... ... ... ... ... ... .... .... .... .... .... .... .... ....
.........
.........
-
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
.........
M .........
.........
........

Figure 2: Separation of a linear space M and the non-negative orthant K for ` = 2

Separation Theorem Let M be a linear subspace of IR` , and K = IR+ ` the non-negative cone
` >
in IR . If M ∩ K = {0}, then there exists a vector φ  0 such that φ z = 0 for all z ∈ M .
In other words, if the linear subspace M and the non-negative cone K have only the zero vector
in common, then there is a vector φ which is orthogonal to M and points in the interior of K.
Figure 2 illustrates this fact for ` = 2.
Proof of (i) ⇒ (ii): Let
(" # )
−q> S+1 N
M= x ∈ IR : x ∈ IR
D
S+1
be the linear space of marketed dividend streams, and K = IR+ the non-negative cone in
IRS+1 . If there is no arbitrage, then the only non-negative marketed dividend stream is the
zero dividend stream, so M ∩ K = {0}. Applying the above separation theorem for ` = S + 1,
S+1
we can conclude that there exists a vector φ ∈ IR++ such that
" #
> −q>
φ x=0
D
for all portfolios x. Normalizing the first entry in φ to 1, we can write
" #
1
φ=

S . As φ> = [ 1 ψ > ], we now have

for some ψ ∈ IR++
" #
−q>
0=[1 ψ> ] x = (−q> + >
D)x
D
State Prices & Arbitrage 7 Prof. S. Rady

for all x ∈ IRN . But this implies −q> + ψ > D = 0, so ψ is indeed a state price vector.
Due to the importance of the theorem and the simple intuition behind the above three proofs,
you should spend some time going through them.
The final part of the proof is less intuitive, and less important for our purposes. Rather than
showing (i) ⇒ (iii) directly, we shall prove (ii) ⇒ (iii). Of course, given that (i) ⇒ (ii), this
will be enough to complete the proof of the theorem.
Proof of (ii) ⇒ (iii): Let ψ be a state price vector. Consider an agent with date 0 endowment
e0 = η > 0, date 1 endowment e = 0, and utility function
S
X
U (c0 , c) = − exp(η − c0 ) − ψs exp(−cs ).
s=1

The investor wants to choose a portfolio x so as to maximize

S
U (η − q> x, Dx) = − exp(q> x) −
X
ψs exp (−(Dx)s ) .
s=1

Differentiating with respect to xn , we get the first order conditions

S
− exp(q> x)qn +
X
ψs exp (−(Dx)s ) dsn = 0
s=1

for n = 1, . . . N . As ψ is a state price vector, these conditions are satisfied for x = 0. Since
U is strictly concave, the first order conditions are sufficient for an optimum. So the trivial
portfolio x = 0 is indeed optimal for this investor.
Incidentally, this proof shows that any arbitrage-free system of asset prices can be supported
as a financial markets equilibrium. In fact, choose a vector of state prices and consider an
economy which is populated by agents with endowments and preferences of the same type as
in the proof (not necessarily all equal). Suppose that financial assets are in zero net supply. At
the given asset prices, all investors are happy just to consume their date 0 endowments and not
to hold any assets at all. Thus, all markets clear, and we have indeed an equilibrium, although
in a highly special and stylized economy.
In an exercise, you will investigate how a state price vector can be constructed directly from
the marginal utilities of an investor who has an optimal demand.

5 Pricing of Contingent Claims

Any vector y ∈ IRS defines a contingent claim, with the component ys specifying how many
units of the physical good are receivable or due at date 1 if state s occurs. A claim y is called
attainable if there is a portfolio x such that y = Dx. In this case, x is called a replicating
portfolio. Let n o
A = Dx : x ∈ IRN
denote the linear space of attainable claims.

Proposition 1 Suppose there is no arbitrage. Then a contingent claim y is attainable if and

only if the scalar product ψ > y is constant on the set of all state price vectors ψ. In this case,
ψ > y equals the date 0 investment required to attain the claim. In other words, q> x = ψ > y for
any portfolio x that replicates y.
State Prices & Arbitrage 8 Prof. S. Rady

Proof: Suppose y is attainable. Let x be a replicating strategy, and consider an arbitrary

state price vector ψ. Then,

ψ > y = ψ > (Dx) = (ψ > D)x = q> x

does not depend on ψ, which proves the necessity part of the proposition.
Conversely, suppose y is not attainable, i.e., y 6∈ A. We can decompose the claim as

y = yA + z

where yA ∈ A (this is the attainable part of the claim) and z 6= 0 is orthogonal to A, i.e.,
0 = z> (Dx) = (z> D)x for all portfolios x. This implies z> D = 0. Given a state price vector
ψ, and choosing α sufficiently small so that ψ + αz  0, we have

(ψ + αz)> D = ψ > D + αz> D = ψ > D = qT ,

so ψ + αz is also a state price vector. But now,

(ψ + αz)> y = ψ > y + αz> y = ψ > y + αz> z > ψ > y.

This settles the sufficiency part.

As for the initial investment required to replicate an attainable claim y, running the first chain
of equalities above in reverse shows that q> x = ψ > y for all replicating portfolios x. So the
initial investment does not depend on the choice of replicating portfolio.
Suppose an attainable claim is issued as a new financial asset. Clearly, this new asset must
trade at a price equal to the initial investment required to replicate the claim – any other price
would lead to arbitrage opportunities. If, for example, the price exceeded the necessary initial
investment, you could take a short position in the asset and set up the replicating portfolio at
the same time (this is one application of the cherished motto ‘sell high, buy low’...). You could
cash in the difference today, and there would be no liabilities tomorrow: a free lunch! This
surprisingly simple idea is the basis of option pricing, the most successful branch of financial
theory over the last three decades (more on this later).
To summarize, attainable claims, whether traded or not, are uniquely priced ‘by arbitrage’,
and their ‘arbitrage prices’ are given by the pricing operator

Ψ : A → IR
y 7→ q> x = ψ > y

where x is any replicating portfolio for y, and ψ any state price vector.

6 Risk-Neutral Valuation
We shall show next that the arbitrage price of an attainable claim can be calculated as the
expected value of the claim’s payoff under some artificial probabilities, discounted at a suitable
rate of return. This is the basis of the so-called ‘risk-neutral valuation’ approach to option
pricing.
We denote the (gross) rate of return on asset n by R̃n . This random variable takes the value
dsn
Rsn =
qn
State Prices & Arbitrage 9 Prof. S. Rady

in state s = 1, . . . , S. Assuming that Rsn > 0 for all s, we can define the artificial probabilities
{πsn } associated with asset n by

πsn = ψs Rsn , s = 1, . . . , S.

{πsn } are indeed probabilities since

S
X S
X
πsn = ψs Rsn
s=1 s=1
S
X dsn
= ψs
s=1
qn

= 1,

where the last step follows from the Fundamental Theorem of Asset Pricing.
Now we price an attainable claim y using asset n as numeraire:

Ψ(y) = ψ > y
S
X
= ψs ys
s=1
S
−1 n
X
= Rsn πs ys
s=1
h i
= Eπn R̃n−1 ỹ

where Eπn is the expectation operator under the probabilities {πsn }. Thus the price of the
claim y (whose payoff can be regarded as the random variable ỹ) is the expectation under the
probabilities {πsn } of the payoff of the claim discounted at the interest rate R̃n . The probabilities
{πsn } are called risk-neutral probabilities with respect to asset n. Often the most natural choice
for the numeraire asset is the riskfree asset, if this asset is available.

7 Insurable States and Arrow Securities

A contingent claim which pays one unit of the physical good in a particular state, and nothing
in all other states, is called an Arrow security. More precisely, the Arrow security for state s is
defined by the payoff vector
0
 
 .. 
 . 
 

 0 

is =  1
 

 
 0 
..
 
 
 . 
0
with the entry 1 in row s. State s is called insurable if the corresponding Arrow security is
attainable, i.e., if it can be replicated by some portfolio x. The justification for this terminology
is that investors can use such a replicating portfolio to insure themselves against the occurrence
of state s, for example by shifting consumption to this state if their endowment es is particularly
low.
State Prices & Arbitrage 10 Prof. S. Rady

In your classwork, you are asked to prove the following result:

Proposition 2 Suppose there is no arbitrage. Then a state is insurable if and only if it has a
unique state price.

8 Equilibrium
We have not explicitly defined equilibrium so far. We begin by briefly describing the Arrow-
Debreu model of general equilibrium, and then turn to equilibrium in the asset market model
set out above.
There are H agents and ` goods. The commodity space is IR` , and the consumption set is IR+ `
h ` h
for every agent. Agent h = 1, . . . , H has an endowment e ∈ IR++ and a utility function u on
` .
IR+
`
Definition 1 An (Arrow-Debreu) equilibrium is a price vector p ∈ IR++ together with an
h H
allocation {c }h=1 such that:
(a) Agents optimize given the prices p, i.e. for every h, ch solves

max uh (c) s.t. p> c ≤ p> eh .

`
c∈IR+

(b) Markets clear:

H
X H
X
ch = eh .
h=1 h=1

PH PH
An allocation {ch }H h `
h=1 is feasible if c ∈ IR+ for all h, and h=1 c
h ≤ h
h=1 e .

Definition 2 An allocation {ch }H h=1 is Pareto efficient if it is feasible and there does not exist
a feasible allocation {c }h=1 such that uh (ch ) > uh (ch ) for every agent h.
h H

In the presence of local nonsatiation (which is implied by strict monotonicity of utility func-
tions), this definition is equivalent to the usual one wherein Pareto efficiency means that it is
impossible to make someone better off without making someone else worse off.

Theorem 2 (First Welfare Theorem) An equilibrium allocation is Pareto efficient.

Proof: Let {ch }H h H

h=1 be an equilibrium allocation with equilibrium prices p. Suppose {c }h=1
is not Pareto efficient. Then there exists a feasible allocation {c }h=1 such that u (c ) > uh (ch )
h H h h

for every agent h. Since ch is utility-maximizing subject to the agent’s budget constraint, it
follows that, for every agent h, ch is not budget-feasible:

p> ch > p> eh .

Summing over agents:

H H
p> c h > p>
X X
eh .
h=1 h=1

This implies that the allocation {ch }H

h=1 is not feasible, a contradiction.

Time and uncertainty are introduced into the Arrow-Debreu model as follows. Suppose the
economy unfolds as an event-tree, each node being indexed by the date and the state of the
State Prices & Arbitrage 11 Prof. S. Rady

world. Suppose further that at the initial date all physical goods can be traded conditional on
any date-event. Then the Arrow-Debreu model can be applied wholesale by simply interpreting
goods as dated state-contingent commodities. With this interpretation, all trading takes place
at the initial date, after which markets close down, with deliveries being made as contractually
determined at the initial date.
In Section 1-7 of this chapter we have studied a one-period economy with a single physical good.
An Arrow-Debreu equilibrium for such an economy is easily defined with respect to S + 1 dated
state-contingent commodities, corresponding to the S + 1 nodes of the corresponding event
tree.
Given N assets with payoff matrix D, we define an asset market equilibrium in this one-period
economy as follows.

Definition 3 An asset market equilibrium consists of an asset price vector q ∈ IRN , an allo-
cation {ch0 , ch }H h H
h=1 and an assignment of portfolios {x }h=1 such that:
(a) Agents optimize given the asset prices q, i.e. for every h, (ch0 , ch ) solves

max uh (c0 , c)
S+1
(c0 ,c)∈IR+

under the constraint that

" # " # " #
c0 eh0 −q>
≤ + x for some portfolio x ∈ IRN .
c eh D

(b) Asset markets clear:

H
X
xh = 0.
h=1

It follows directly from the Fundamental Theorem of Asset Pricing that absence of arbitrage
is a necessary condition for equilibrium, provided there is at least one nonsatiated investor.

Proposition 3 Suppose there is at least one agent with strictly monotone preferences. Then
an equilibrium asset price vector does not permit arbitrage.

9 Complete Markets
The set of attainable claims
A = {Dx : x ∈ IRN }
is also called the marketed subspace or asset span. It is the column space of the asset payoff
matrix D, and therefore a linear subspace of IRS . The dimension of A is equal to the rank
of D. We say that asset markets are complete if dim(A) = rank(D) = S. When markets are
complete, we have A = IRS , and any distribution of payoffs can be generated by choosing an
appropriate portfolio. Markets are incomplete if rank(D) < S.
In the absence of arbitrage, completeness can be characterized in terms of state prices.

Proposition 4 Suppose there is no arbitrage. Then asset markets are complete if and only if
they possess a unique vector of state prices.
State Prices & Arbitrage 12 Prof. S. Rady

Proof: We have seen above that the number of ‘degrees of freedom’ in the selection of a state
price vector is S − rank(D). Thus, state prices are unique if and only if rank(D) = S, but this
is equivalent to completeness.
Alternatively, we can interpret this result in terms of the above notion of insurable states.
Markets are complete if and only if all states are insurable (why?). According to Proposition
2, this means that all state prices are pinned down uniquely.
If asset markets are complete, then a consumption allocation is an asset market equilibrium
allocation if and only if it is an Arrow-Debreu equilibrium allocation. The key to this result is
the observation that, in the presence of complete markets, the S + 1 budget constraints which a
feasible consumption plan has to satisfy can be collapsed into a single date 0 budget constraint.
" #
c0
Proposition 5 With complete markets, a consumption plan is feasible for an individual
c
" #
e0
with endowment if and only if
e

c0 + Ψ(c) ≤ e0 + Ψ(e).

Proof: Suppose the consumption plan is feasible, i.e.,

" # " # " #
c0 e0 −q>
≤ + x
c e D

for some portfolio x. Let ψ be the state price vector. Multiplying the above vector inequality
through by [ 1 ψ > ], we find c0 + Ψ(c) ≤ e0 + Ψ(e) + (−q> + ψ > D)x = e0 + Ψ(e).4 Conversely,
consider a consumption plan that satisfies c0 + Ψ(c) ≤ e0 + Ψ(e). The agent’s net demand at
date 1 is y = c − e. As markets are complete, y is attainable: there is a portfolio x such that
y = Dx. Now,
c0 ≤ e0 + Ψ(e − c) = e0 − Ψ(y) = e0 − q> x
and
c = e + y = e + Dx,
so the consumption plan is feasible.

Suppose we have an asset market equilibrium consisting of the asset price vector q ∈ IRN and
the allocation {ch0 , ch }H
h=1 , and asset markets are complete. Given S possible states at time 1,
we have ` = S + 1 state-contingent commodities: one at time 0, and S of them at time 1. We
want to consider the corresponding Arrow-Debreu model. Define a price vector p ∈ IRS+1 for
the state-contingent commodities by setting
" #
1
p=
ψ

where ψ ∈ IRS is the unique state price vector for the asset market equilibrium. Then the
budget constraint of household h becomes
" #> " # " #> " #
1 ch0 1 eh0
≤ ,
ψ ch ψ eh
4
Note that this part of the proposition does not rely on completeness.
State Prices & Arbitrage 13 Prof. S. Rady

that is, c0 + Ψ(c) ≤ e0 + Ψ(e). By Proposition 5, this is the same budget constraint as in
the asset market model, so the household’s optimal consumption plans coincide! Therefore, an
asset market equilibrium allocation at asset prices q constitutes an Arrow-Debreu equilibrium
allocation at commodities prices p, and vice versa.
The following result is now an immediate consequence of the First Welfare Theorem (Theo-
rem 2).

Theorem 3 Suppose markets are complete. Then an asset market equilibrium allocation is
Pareto efficient.

It was said at the beginning that capital markets have two purposes for investors: they allow
investors to transfer resources over time, and to shift resources across states. Complete markets
perform both functions perfectly because any contingent claim (and hence any re-allocation of
date 1 income) is attainable at a well-defined price in terms of date 0 consumption. Put
differently, there are well-defined relative prices for consumption today versus consumption
tomorrow, and for consumption in state s versus consumption in state s0 . In fact, these relative
prices are derived directly from the unique set of state prices. One unit of consumption today
is worth
1
ψ1 + . . . + ψS
units of sure consumption tomorrow.5 In the same way, one unit of consumption in state s is
worth
ψs
ψs0
units of consumption in state s0 .

5
Note that one unit of sure consumption corresponds to the claim i1 + . . . + iS .