You are on page 1of 3

1.B.3.

Show that if f : is a strictly increasing function and u : X is a utility function


representing preference relation , then the function v : X defined by v(x) = f(u(x)) is also a
utility function representing preference relation .
Since u() represents , then by Definition 1.B.2, for all x, y e X, x y if and only if u(x) > u(y).
Since f is a strictly increasing function, u(x) > u(y) if and only if f(u(x)) > f(u(y)).
Therefore, x y if and only if v(x) > v(y). That is, v() represents .
1.B.4. Consider a rational preference relation . Show that if u(x) = u(y) implies x ~ y and if u(x) >
u(y) implies x y, then u() is a utility function representing .
Suppose that x y. If, at the same time, y x, then (x y and y x) x ~ y, which in turn implies
that u(x) = u(y).
On the other hand, if not y x, then (x y but not y x) x y, which in turn implies that u(x) >
u(y).
Therefore, if x y, then u(x) > u(y).
Conversely, assume that u(x) > u(y). If u(x) = u(y), then x ~ y, which implies that x y (and y x).
On the other hand, if u(x) > u(y), then x y, which implies that x y (but not y x).
Therefore, u(x) > u(y) x y.
Since x y u(x) > u(y), then by Definition 1.B.2, u() represents .
1.B.5. Show that if X is finite and is a rational preference relation on X, then there is a utility
function u : X that represents . [Hint: Consider first the case in which the individuals ranking
between any two elements of X is strict (i.e., there is never any indifference and construct a utility
function representing these preferences, then extend your argument to the general case.]
As suggested by the hint, first consider the strict preference relation . Assume that is rational,
i.e., complete and transitive, on X. We use the proof by mathematical induction to show that there
exists u : X that represents .
Basis (or base case)
Trivial: Suppose X is a singleton, i.e., a set with only one element (N = 1). Then let u = c, where c is
some real number.
Alternatively and perhaps more instructively, suppose that X is a doubleton, i.e., a set with only two
elements (N = 2). Specifically, let X = {x1, x2}. Since is assumed to be complete, then either x1 x2
or x2 x1. Accordingly, let u(xi) = 2 and u(xj) = 1 if and only if xi xj, where i, j = 1, 2 and i = j.
Inductive step (Assume that the proposition is true for some N and show that it holds for N + 1)
Let X = {x1, x2, . , xN, xN+1}. The induction hypothesis claims that there is a function u that represents
. Without loss of generality, let u(x1) > u(x2) > > u(xN) (i.e., X is arranged such that x1 x2
xN).
There are three exhaustive and mutually exclusive cases on how xN+1 is related to the elements of
{x1, x2, . , xN}:
(i) xi xN+1 for every i < N + 1,
(ii) xN+1 xi for every i < N + 1, and
(iii) xi xN+1 xj for some i < N + 1 and some j < N + 1.
If case (i) applies, assign u(xN+1) < u(xN).
[Note that then we have X = {x1, x2, . , xN, xN+1} where u(x1) > u(x2) > > u(xN) > u(xN+1). In
effect, Definition 1.B.2 holds: Given X, where x1 x2 xN xN+1, there exists u : X
such that u(xi) > u(xj) if and only if xi xj.]
If case (ii) applies, assign u(xN+1) > u(x1).
Suppose case (iii) applies. Define the sets
I = {i e {1, . , N} : xi xN+1} and J = {j e {1, . , N} : xN+1 xj}.
Since is complete and the strict preference relation implies that there are no ties (or, more
precisely, that for any pair of elements in X either xk x or x xk), any n e {1, . , N} belongs to
either I or J, but not to both, and that I J = {1, . , N}.
[Suppose not.
Case (a). Suppose there is an n such that n e {1, . , N} but n e I and n e J. Since n e I, xn
xN+1. Since n e J, xN+1 xn. But (xn xN+1 and xN+1 xn) cannot be true at the same time.
Case (b). Suppose there is an n such that n e {1, . , N} but n e I and n e J, so that I J = {1,
. , N}. Then neither xn xN+1 nor xN+1 xn. But this implies that is not complete and
therefore not rational.]
Since is transitive, I and J are intervals in the sense that (a) if i e I and i' < i, then i' e I and (b) if j
e J and j' > j, then j' e J.
[Suppose i e I and i' < i. Since x1 x2 xN, it must be that xi'
xi. Since i e I, then xi
xN+1. By transitivity, (xi'
xi and xi xN+1) implies that xi'
xN+1. Therefore, i' e I.
Suppose j e J and j' > j. Since x1 x2 xN, it must be that xj xj'
. Since j e J, then xN+1
xj. By transitivity, (xN+1 xj and xj xj'
) implies that xN+1 xj'
. Therefore, j' e J.]
Let i
*
= max I and i
*
+ 1 = min J. Assign u(xN+1) a real number that lies in the open interval (u(xi*+1),
u(xi*)). Then Definition 1.B.2 holds: Given X, where x1 xi* xN+1 xi*+1 xN, there exists u
: X such that u(x1) > > u(xi*) > u(xN+1) > u(xi*+1) > > u(xN).
Now consider the weakly preferred relation (which allows the indifference relation ~) on X =
{x1, . , xN}. For each n = 1, 2, . , N, define the set Xn = {xn'
e X : xn'
~ xn}.
[This means that each element of X is assigned an index number value n and, for each xn, we define a
set Xn whose members xn'
are elements of X such that the decision maker is indifferent between xn'

and xn. There are therefore as many sets Xn as there are elements xn in X.]
Since X = {x1, . , xN} and each element xn is associated with a set Xn,
1
.
N
n
n
X X
=
=
[To see this, simply pick the xn that is the reference element of Xn for every n. Then {x1} {xN} =
X.]
Since ~ is transitive (Proposition 1.B.1(ii)), if Xn = Xn
, then Xn Xn
= C.
[Suppose not. Suppose that Xn = Xn
but Xn Xn
= C. Then there is an xk such that xk e Xn and xk e Xn
.
This implies that xk ~ xn for any xn e Xn and xk ~ xn
for any xn
e Xn
. But by transitivity, if (xn ~ xk and
xk ~ xn
) then xn ~ xn
for any xn e Xn and any xn
e Xn
. In other words, Xn = Xn
. Thus we have a
contradiction.]
Let M c {1, 2, . , N} such that
m
m M
X X
e
= and Xm = Xm'
for any m e M and m' e M with m = m'.
[Note that the elements of M consist only of the values of n such that the sets Xns are unique.
Example: Suppose the preference orderings are: x1 ~ x2, x3 ~ x5 ~ x9, x4 ~ x7, x6 ~ x8, and x1 x3 x4
x6 x10. Then X1 = X2 = {x1, x2}, X3 = X5 = X9 = {x3, x5, x9}, X4 = X7 = {x4, x7}, X6 = X8 = {x6, x8}, X10 = {x10}.
Thus,
10
1
.
n
n
X X
=
= But some Xns are duplicate sets and we can choose a selection of the Xns to get X,
e.g., X1 X3 X4 X6 X10 = X.]
Define a relation
*
on {Xm : m e M} such that Xm
*
Xm'
if and only if xm
*
xm'
.
In fact, by the definition of the set M, it cannot be the case that Xm ~ Xm'
for m = m'. Therefore, we
can use the result on above: There exists a utility function u
*
: {Xm : m e M} such that u
*
(Xm) >
u(Xm'
) if and only if Xm
*
Xm'
.
Define u : X such that u(xn) = u
*
(Xm) if m e M and xn e Xm.
Finally, we show that u represents .
If:
Suppose xn xn'
but not xn'
xn. Then xn e Xm and xn'
e Xm'
such that Xm
*
Xm'
, which implies that
u
*
(Xm) > u
*
(Xm'
) and therefore u(xn) > u(xn'
).
Suppose xn xn'
and xn'
xn. Then xn e Xm and xn'
e Xm, which implies that u(xn) = u(xn'
) = u
*
(Xm).
Combining the two cases, we have if xn xn'
, then u(xn) > u(xn'
).
Only if:
Suppose u(xn) > u(xn'
). Then it must be that xn e Xm and xn'
e Xm'
such that u(xn) = u
*
(Xm) > u
*
(Xm'
) =
u(xn'
), which implies that Xm
*
Xm'
. In other words, xn xn'
.
In effect, we have if u(xn) > u(xn'
), then xn xn'
.
1.C.2. Show that the weak axiom (Definition 1.C.1) is equivalent to the following property holding:
Suppose that (a) B, B' e B, (b) x, y e B, and (c) x, y e B'. Then if x e C(B) and y e C(B'),
we must have {x, y} c C(B) and {x, y} c C(B').
Given (a), (b), and (c), suppose WARP is satisfied. Then (i) x e C(B) implies that x e C(B') and (ii) y
e C(B') implies that y e C(B). Therefore, {x, y} c C(B) and {x, y} c C(B').
Suppose the property described above is satisfied. Since the property yields [x e C(B) and x e C(B')]
and [y e C(B') and y e C(B)], then WARP is satisfied.
Therefore WARP PROPERTY: the two are equivalent.

You might also like