Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

THE EXPECTED UTILITY THEOREM

BEN BLUMSON

1. Introduction

The conditions of the expected utility theorem are that for all A, B, C:

(1) A  B, B  A or A ∼ B completeness

(2) if A  B and B  C then A  C transitivity

(3) A  B if and only if ApC  BpC for all p independence

(4) if A % B % C then ApC ∼ B for some p solvability

And the theorem itself is:

Theorem 1. A preference relation  satisfies conditions (1)-(4) if and only

if there is a function u such that:

(5) A  B if and only if u(A) > u(B) order preservation

(6) u(ApB) = (p × u(A)) + ((1 − p) × u(B)) additivity

And for all v satisfying (5) and (6), there is r > 0 and s such that:

(7) v(A) = r × u(A) + s. uniqueness

Date: September 11, 2018.


1
2 BEN BLUMSON

Table 1

p>q q p−q 1−p q>p p q−p 1−q

OpW O O W OpW O W W

OqW O W W OqW O O W

Let O be a best outcome and W be a worst outcome. Note that O  W .

From (3) independence, OpO  OpW for any p. So O  OpW for any p.

In other words, O for sure is better than O with any chance of W .

Now let u(O) = 1 and u(W ) = 0. And suppose O  B  W . From (4)

solvability, there is p such that OpW ∼ B for some p. Suppose for reductio

there is q 6= p such that OqW ∼ B. From (2) transitivity it follows that

OpW ∼ OqW . As in Table (1), either p > q or else q > p. In the first
q
case, we can redescribe OqW as (OsW )pW , where s = . So OpW ∼
p
(OsW )pW . So from (3) independence, O ∼ OsW . But this contradicts the

last paragraph. Mutatis Mutandis for q > p. So q = p.

So for all B such that O  B  W there is a unique p such that B ∼ OpW ,

and we can stipulate that u(B) = p. To prove the left to right part of the

theorem, it suffices to show that u satisfies clauses (5)-(7).

To prove (5) order preservation, suppose A  B. From the definition of u,

it follows that Ou(A)W  Ou(B)W . Now suppose for reductio that u(A) ≤

u(B). Then either u(A) = u(B) or u(A) < u(B). But if u(A) = u(B), then

from the last paragraph Ou(A)W  Ou(A)W , which is a contradiction. On

the other hand, if u(A) < u(B), then consult Table (2):
THE EXPECTED UTILITY THEOREM 3

Table 2

u(A) < u(B) u(A) u(B) − u(A) 1 − u(B)

Ou(A)W O W W

Ou(B)W O O W

From Table (2) we can see that Ou(A)W can be represented as the same
u(A)
as (OrW )u(B)W , where r = . Then from the next to last paragraph, it
u(B)
follows that (OrW )u(B)W  Ou(B)W . From (3) independence, it follows

that OrW  O, contradicting the first paragraph on the previous page. So

u(A) > u(B). The proof works in reverse for the right to left direction.

You might also like