You are on page 1of 19

Expected Utility Over Money And Applications

Econ 2100

Lecture 14, October 21

Outline
1

Expected Utility of Wealth

Betting and Insurance

Risk Aversion

Certainty Equivalent

Risk Premium

Relative Risk Aversion

Fall 2015

Probability Distribution On Wealth


Many applications of expected utility consider preferences over probability
distribution on wealth (a continuous variable).
A probability distribution is characterziedd by its cumulative distribution
function.
Denition
A cumulative distribution function (cdf) F : R ! [0; 1] satises:
x

y implies F (x)

F (y ) (nondecreasing);

limy #x F (y ) = F (x) (right continuous);a


limx !
a Recall

F (x) = 0 and limx !1 F (x) = 1.

that, if it exists, lim y #x f (y ) = lim n!1 f (x + n1 ).

Notation
F
x

denotes the mean (expected value) of F , i.e.

is the degenerate distribution function at x; i.e.


(
0 if z < x
:
x (z) =
1 if z x

=
x

x dF (x).

yields x with certainty:

Expected Utility Of Wealth


As usual, the space of all distribution functions is convex and one can dene
preferences over it.
The utility index v : R ! R is dened over wealth (can be negative).
The expected utility is the integral
of v withZrespect to F
Z
v (x)dF (x) =

vdF

0
If F is dierentiable, the expectation
is
Z
Z computed using the density f = F :
v dF = v (x )f (x ) dx :

von Neumann and Morgenstern Expected Utility


Under some
R axioms, there exists a utility function U on distributions dened as
U(F ) = v dF , for some continuous index v : R ! R over wealth, such that
Z
Z
F % G ()
v dF
v dG
Axioms not important from now on (need a stronger continuity assumption).
We always think of v as a weakly increasing function (more wealth cannot be
bad).

Betting
A Gamble
Suppose an individual is oered the following bet:
lose x with probability 1
win ax with probability p

The expected value of this bet is


pax + (1

p) ( x) = [pa + (1

p) ( 1)] x

Denition
A bet is actuarilly fair if it has expected value equal to zero (i.e. a = 1 p p ); it is
better than fair if the expected value is positive and worse than fair if it is negative.
How does she evaluate this bet? Use the expected utility model to nd out
If vNM index is v ( ) and initial wealth is w , expected utility is:
probability of winning

probability of losing

v (w + ax)
utility of wealh if win

(1

p)

v (w

x)

utility of wealth if lose

How much does she want of this bet? Answer by nding the optimal x.

Betting and Expected Utility


win ax with probability p

lose x with probability 1

The consumer solves


max pv (w + ax) + (1
x

p) v (w

x)

The FOC is
pav 0 (w + ax) = (1
rearranging
pa
(1

p)

p) v 0 (w

x)

v 0 (w x )
v 0 (w + x )

If the bet is fair, the left hand side is 1. Therefore, at an optimum, the right
hand side must also be 1.
If the vNM utility function is strictly increasing and strictly concave (v 0 > 0
and v 00 < 0), the only way a fair bet can satisfy this FOC is to solve
w + ax = w
which implies x = 0.
She will take no part of a fair bet.
What happens with a better than fair bet?

Insurance
An Insurance Problem
An individual faces a potential accident:
the loss is L with probability

nothing happens with probability 1

Denition
An insurance contract establishes an initial premium P and then reimburses an
amount Z if and only if the loss occurs.
Denition
Insurance is actuarlly fair when its expected cost is zero; it is less than fair when its
expected cost is positive.
The expected cost of an insurance contract is
P

( Z ) + (1
loss

) (0)
no loss

Fair insurance means


P= Z

=P

Insurance and Expected Utility


An Insurance Problem
An individual with current wealth W and utility function u( ) faces a potential
accident:
lose L with probability

or

lose zero with probability 1

If she buys insurance, her expected utility is


v

P +Z

wealth if loss

+ (1

)v

wealth if no loss

For example, if the loss is fully reimbursed (Z = L), this becomes


v (W P ) + (1
) v (W P ) = v (W P )

Will she buy any insurance? Yes if


v (W L P + Z ) + (1
) v (W
|
{z
expected utility with insurance

P)
}

v (W

L) + (1
{z

expected utility without insurance

How much coverage will she want if she buys any coverage?
Find the optimal Z .

) v (W )
}

The answer depends on the premium set by the insurance company P as well
as the curvature of the utility function v .

Risk Aversion
Given some F , in our notation
value of F for sure.

is a distribution that yields the expected

Denitions
The preference relation % is
risk averse if, for all cumulative distribution functions F ,
F

% F:

risk loving if, for all cumulative distribution functions F ,


F %

risk neutral if it is both risk averse and risk loving (

DM is risk averse if she always prefers the expected value


uncertain distribution F .

F ).

for sure to the

This behavioral denition does not depend on the expected utility


representation (or any other).
Risk attitudes are dened directly from preferences.

Risk Aversion: An example

Exercise
Let % be a preference relation on (R), the space of all cumulative distribution
functions, be represented by the following utility function:
x if F = x for some x 2 R
U(F ) =
0
otherwise
True of false: % is risk averse.
False: If

< 0, then F

F.

Certainty Equivalent
Denition
Given a strictly increasing and continuous vNM index v over wealth, the certainty
equivalent (CE) of F , denoted c(F ; v ), is dened by
Z
v (c(F ; v )) = v ( ) dF :
By denition, the certainty equivalent of F is the amount of wealth c( ) such
that that c( ) F .
DM is indierent between a distribution and the certainty equivalent of that
distribution.
The certainty equivalent is constructed to satises this indierence.

Unlike risk aversion, the certainty equivalent denition assumes a given


preference representation (needs some utility function that represents
preferences).
The certainty equivalent seems related to risk aversion.

Risk Premium

Denition
Given a strictly increasing and continuous vNM index v over wealth, the risk
premium of F , denoted r (F ; v ) is dened by
r (F ; v ) =

c(F ; v ):

This measures the dierence between the expected value of a particular


distribution and its certainty equivalent.
The denition of risk premium also assumes a given preference representation.
This also seems related to risk aversion.

Risk Aversion, Certainty Equivalent, and Risk


Premium
If preferences satisfy the vNM axioms, risk aversion is completely characterized
by concavity of the utility index and a non-negative risk-premium.
Proposition
Suppose % has an expected utility representation and v is the corresponding von
Neumann and Morgestern utility index over money.The following are equivalent:
1
2
3

% is risk averse;
v is concave;
r (F ; v ) 0;
The proof uses Jensens inequality.

Jensens Inequality
Jensens inequality
A function g is concave if and only if
Z
g (x) dF

xdF

This says
g (E(X ))

E(g (X ))

Consequences of Jensens inequality


Hence, v ( ) is concave if and only if
Z
vdF

dF

Since we also know that v is non decreasing,


Z
c(F ; v )
vdF
is equivalent to
v (c(F ; v ))
Z
Z
or
v dF v
vdF

vdF

Risk Aversion, CE, and Risk Premium


% is risk averse , v| is concave
v)
{z
} , r|(F ; {z
{z
}
|
(2)

(1)

(3)

We prove (1) ) (2) ) (3) ) (1). Start with (1) ) (2).

0
}

Proof.
% is risk averse, hence

% F for all F 2

R.

For any x; y 2 R and 2 [0; 1], let the discrete random variable X be such
that P(X = x) = and P(X = y ) = 1
. Let Fx ;y be the associated
cumulative distribution.
By risk aversion we have:
v(
v ( x + (1

F x ;y )

)y )

X
z

Thus v is concave.

v (z)dFx ;y (z)
v (z)P(X = z) = v (x) + (1

)v (y )

Risk Aversion, CE, and Risk Premium


% is risk averse , v| is concave
v)
{z
} , r|(F ; {z
{z
}
|
(2)

(1)

(3)

Now prove that (2) ) (3)

Proof.
Let v be concave, and X be a random variable with cdf F .
By Jensens inequality:
v (E(X ))
or
v(

F)

E(v (X ))

v (x)dF (x) = v (c(F ; v ))

Since v is an increasing function, we have


F

Thus
F

c(F ; v )

c(F ; v ) = r (F ; v )

0
}

Risk Aversion, CE, and Risk Premium


v)
% is risk averse , v| is concave
{z
} , r|(F ; {z
{z
}
|
(2)

(1)

(3) ) (1)

(3)

0
}

Proof.
Let r (F ; v )

0 for all cdfs F .

Then we have
F

c(F ; v )

which in turn implies that


v(
Hence

F)

% F for all F 2

v (c(F ; v )) =

v (x)dF (x)

R; therefore % is risk averse.

We have shown that (1) ) (2) ) (3) ) (1), thus the proof is complete.

Relative Risk Aversion


When can we say that one decision maker is more risk averse than another?
Relative risk aversion answers this question in a preference-based way.
Denition
Given two preference relations, %1 is more risk averse than %2 if and only if
F %1
for all F and x.

F %2

If DM1 prefers the lottery F to the sure payout x, then anyone who is less risk
averse than DM1 also prefers the lottery F to x .
Conversely, if DM2 prefers the sure payout x to the lottery F , then anyone who
is more risk averse than DM2 also prefers the sure payout x to the lottery F .
Again, this denition does not assume anything about preferences.
When both preferences satisfy expected utility, we have extra implications.

Relative Risk Aversion


Relative risk aversion is equivalent to: more concavity of the utility index, a
smaller certainty equivalent, and a larger risk premium.
Proposition
Suppose %1 and %2 are preference relations represented by the vNM indices v1 and
v2 . The following are equivalent:
1

%1 is more risk averse than %2 ;

v1 =

c(F ; v1 )

c(F ; v2 ), for all F ;

r (F ; v1 )

r (F ; v2 ), for all F .

v2 for some strictly increasing concave

Proof.
Question 5 in Problem Set 8

: R ! R;

Econ 2100

Fall 2015

Problem Set 8
Due 26 October, Monday, at the beginning of class
1. Kreps 6.2
2. Kreps 6.7
3. An individual has expected utility preferences with utility over money given by some twice dierentiable function v with v 0 > 0 and v 00 < 0. She can choose the value of t in the following bet: receive
txs dollars with probability s with s = 1; :::; S (note that some of the xs are negative). Her initial
wealth is w for all s.
(a) Show that

S
P

s xs

= 0 if and only if the optimal t is equal to zero.

s=1

(b) What (if anything) can you say about the optimal t when

S
P

s xs

> 0? How about

s=1

S
P

s xs

< 0?

s=1

(c) Interpret these results in terms of better than fair gambles. What would happen if v 00 = 0?
(d) Can you say anything about how those answers will change as the decision maker in question
becomes more risk averse?
4. An individual whose wealth is W > 0 could have an accident in which she loses L with probability
. Let u denote the individual utility for money, and assume this function is dierentiable at least
twice with u0 > 0 and u00 < 0. An insurance policy species the premium the consumer has to pay
regardless of the accident (P ), and the amount the insurance company reimburses if the accident
occurs (Z). The latter is chosen by the consumer.
(a) Show that if insurance is actuarilly fair, the consumer will insure the entire loss.
(b) Characterize the relation between the optimal insurance coverage and the premium P . Interpret
this result. (hint: draw a picture)
(c) Can you say anything about how those answers will change as the decision maker in question
becomes more risk averse?
5. Suppose %1 and %2 are preference relations represented by the vNM indices v1 and v2 . Prove that
the following are equivalent:
(a) %1 is more risk averse than %2 ;
(b) v1 =

v2 for some strictly increasing concave

(c) c(F; v1 )

c(F; v2 ), for all F ;

(d) r(F; v1 )

r(F; v2 ), for all F .

: R ! R;

You might also like