7 Lecture 7

Finally we move on to the theorem of the maximum. This is going to be a very useful theorem,
and it is designed to answer the following questions:
Example 13 Let j R

+
be a vector of prices, 1 be income, and consider an agent who choses
bundles r R

+
to maximize a utility function n : R

+
R subject to the budget constraint
1(j, 1) =

r R

+
|jr 1

Let 1(j, 1) be the demand function, so that

1(j, 1) = arg max
R

n(r)|jr 1
And (j, 1) be the derived utility, so that
(j, 1) = max
R

n(r)|jr 1
Can we say anything about the properties of 1 and ? In other words, do we know anything
about how demand and derived utility change with the parameters of the problem?
This is exactly what the theorem of the maximum tells us (under certain assumptions). In order
to dene these properties, we need to dene the concept of the graph of a correspondance :
Denition 28 The graph of a correspondance : 1 A is the set of pairs {r, j} A 1 such
that r is in the correspondance evaluated at j
Gr

= {{r, j} A 1 |r (j)}
Theorem 13 (The Theorem of the Maximum) Let
A and 1 be metric spaces (1 will be the set of things that are chosen, A the set of parameters)
: A 1 be compact valued and continuous (this is the constraint set dened by the
parameters)
35
) : A 1 R be continuous, (this is the utility function)
Now dene j

: A 1 as the set of maximizers of ) given parameters r

j

(r) = arg max

()
)(r, j)
and dene )

: A 1 as the maximized value of ) for ) given parameters r

)

(r) = max
()
)(r, j)
Then
1. j

is upper hemi-continuous and compact valued

2. )

is continuous
Translating into the language of the example
A is the set of price vectors and income
1 is the commodity space
is the budget correspondance
) is the utility function (note that we do not let utility depend directly on prices, but we
could if we wanted to)
j

is the demand function

)

is the derived utility

This is a really cool result. With relatively few assumptions, we are able to guarantee some neat
properties of things we really care about. The proof is somewhat cumbersome, so we will sketch it
here.
Proof. We will prove this as a set of claims:
Claim 1: j

has a closed graph. Let (j, r) be a closure point of Gr

. We need to show that this

is in Gr

. First, we show that j is feasible at r, then we show that it maximizes ) at r

36
Note that, if (j.r) is a closure point of Gr

, then we can construct a sequence (j

, r

) (j, r)
such that (j

, r

) Gr

. This implies that j

(r

) :. This in turn implies that

j

(r

) :. As is UHC and compact valued, then j

must have a subsequences that

converges to some j
0
(r), but as j

j, it must be that j (r), so j is feasible at r

Now assume that j , j

(r), then there must be some j (r) such that )(r, j) )(r, j). By
LHC, there must be some sequence j

j such that j

(r

) :. By the continuity of
), we know that
lim)(r

, j

) = )(r, j)
lim)(r

, j

) = )(r, j)
But, as j

(r

), this implies that

)(r

, j

) )(r

, j

) :
)(r, j) < )(r, j)
A contradiction (check)
Claim 2 j

is UHC and compact valued. As j

(r) is closed (by the above result) and j

(r) (r)
compact, it must be the case that j

(r) is compact, and so j

is compact valued.. Let (r

, j

)
be a sequence such that r

r and j

(r

) (r

) :.By the UHC and compact

valuedness of , we know that there is a subsequence j

that converges to some j (r). The

closed graph property tells us that, as (r

, j

) Gr

, then (r, j) Gr

, and so j j

(r),
implying that j

is UHC
Claim 3 )

is continuous. Let r

r A. We need to show that )

(r

) )

(r). We know
that there is a subsequence )

(r

) limsup)

(r

). Pick a sequence j

(r

), so
)

(r

) = )(r

, j

(r

))
Because j

is compact valued and UHC, there is a subsequence j

j j

(r). By the
continuity of ), the fact that r

r and j

j implies that
)

(r

)
= )(r

, j

)
)(r, j)
37
. but as j j

(r), )

(r), so )

(r) is the limsup of )

(r

). A similar argument proves that

)

(r) is also the liminf of )

(r

), so we are done.
38