Session 6 - 7

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Consumer Choice Under Risk

EC101 – Lecture 7
• Uncertainty
– All face the same (rain / sunshine; economic slowdown)
– Information assymetries exist (CEO vs operator)
• Reaction
– Passive
– Active
• State of nature
– (rain, sun, cloudy)
– (high infl, low infl, high deficit…)
• Acts
– Carry umbrella, don’t
– Postpone buying decision, buy, don’t invest….)
• Payoffs = outcome (get wet, stay dry)
• Uncertainty
– What about
– What form?
• Know frequency distribution
• Have some idea of frequency distribution
• Know the extent of variation, but no ideas on freq.
• No clue
– What to do? – try to formulate as:
• States of nature : (1,2,..,s,..,S)
• Actions: (1,2,..,x,.., X)
• Consequence function c(x,s) = payoffs
Expected Utility
• Use probability theory to reach decisions in
uncertain situations
• Expected utility rule:
U(x) = ∑ps‫[ע‬c(x,s)]= E ‫(ע‬s)
– vNM utility fxn: ‫√=ע‬w; initial wealth = 36;
– Gamble: win 13 with prob 2/3 and lose 11 with
prob 1/3
– Will she accept?
Decision-making under Uncertainty
Example: Choice between two “random prospects”:

• Two 1-year projects X and Y with uncertain returns


– they require the same up-front investment of $1 million
– X will give p.v. returns of $2 mil. or $3 mil. with equal probability
– Y will give p.v. returns of $2.32 mil. or $2.6 mil. with equal probability
• In effect, the choice is between two “random variables”:
X = {1, 2; 0.5, 0.5}, Y = {1.32, 1.6; 0.5, 0.5}
• X has higher mean (expected value) and also higher variance than Y
• Your choice will depend upon your “attitude towards risk”
Issue: How do we model an individual's preference-rankings over
“random prospects” of the form: Z = {z1, z2, …, zN; μ1, μ2, …, μN)

where [z1, z2, …, zN] are alternative monetary returns and [μ1, μ2, …, μN]
are the associated probabilities?
The Expected Utility Hypothesis
Daniel Bernoulli; John von Neumann & Oskar Morgenstern

• Let the preferences of a person over “sure money” be represented


by a Bernoulli utility function u(z) [z = alternative ‘sure’ money
amounts]
– plausible assumption that ‘more money is preferred to less’ implies that
u(z) is strictly increasing in z
• Then expected utility of a prospect Z for that person is defined to be:
EU(Z) = μ1u(z1) + μ2u(z2) + …+ μNu(zN)
• The expected utility hypothesis states that every person has his
personal Bernoulli utility function u(.), and prefers a random
prospect X to another random prospect Y if and only if EU(X) > EU(Y)
Expected Utility Theorem:
• If a person’s preferences over “random prospects” satisfy properties
of ‘completeness’, ‘transitivity’, ‘monotonicity’, and the
‘independence property’, then the expected utility hypothesis is valid
i.e., we can construct an expected utility
• Independence property: If prospect X is preferred to prospect Y, then
for any other prospect Z and any probability p, the compound
prospect {X , Z; p, 1–p} is preferred to the compound prospect {Y , Z;
p, 1–p}
Applying the theory: back to the ‘project
choice’ example
• Four persons A, B, C, D with different Bernoulli utility
functions:
uA(z) = z2; uB(z) = z; uC(z) = (z)1/2; uD(z) = (z)1/4
– A’s u function is convex: increasing ‘marginal utility of money’
– B’s u function is linear: constant ‘marginal utility of money’
– C’s u function is concave: decreasing ‘marginal utility of money’
– D’s u function is more concave than C’s
• A and B prefer X to Y, C is indifferent between X and Y, and D
prefers Y to X
• For any person I with Bernoulli utility function uI(.), any
‘increasing linear transformation’: vI(z) = a + buI(z) with b > 0
also represents his utility and satisfies the expected utility
property
– but an increasing non-linear transformation [e.g. vI(z) = log.uI(z)]
does not represent his utility
A: Risk Loving : U(E(z)) <E(U(z))
U(z)
B: Risk Neutral

B
C: Risk Averse
URA U(E(z)) >E(U(z))

URN (A, B; 0.5, 0.5)

URL

A
z zB z
Attitudes toward Risk
• A person is risk averse if it is always true that she prefers getting
the ‘mean’ of a random prospect for sure rather than getting the
prospect ⇔ A person is risk averse if and only if her u(.) is strictly
concave ⇔ U(E(z)) >E(U(z))
• A person is risk loving if it is always true that she prefers getting a
random prospect rather than getting its ‘mean’ ⇔ A person is risk
loving if and only if her u(.) is strictly convex ⇔ U(E(z)) <E(U(z))
• A person is risk neutral if it is always true that she is indifferent
between a getting a random prospect and getting the ‘mean’ for
sure ⇔ A person is risk neutral if and only if her u(.) is linear ⇔ [a
linear u(.) can be represented by identity function u(z)=z for all z]

• Most individuals are risk averse


– individuals differ in their aversion to risk (a ‘more risk averse’ person
has a ‘more concave u function’ than a less risk averse person)
– in general, the richer one gets the less risk averse she is (the extent of
concavity of u function decreases as wealth increases)
Measuring Risk Aversion: ‘certainty
equivalent’/ ‘risk premium’
• For a risk averse u, the certainty
equivalent C(Z) is the sure amount
associated with prospect Z such that:
u(C(Z)) = EU(Z)
and risk premium ρ(Z) associated with Z is:
ρ(Z) = EZ – C(Z)
• Thus, risk premium of a prospect Z for a
person is the max amount of money that
she would pay to receive the ‘mean’ of Z
rather than Z

• Comparing risk premia across individuals


for a particular random prospect is one
way to evaluate their relative aversion to
risk
• Inverting this logic, starting from a ‘safe
asset’, a risk averse person needs a ‘risk
compensation’ to accept a ‘risky asset’
Nitin’s Utility function
Nitin’s level of utility Utility Curve
increases from 10 to 16 to
18 as his income increases
from 10K to 20K to 30K=>
marginal utility falls from
10 to 6 to 2

marginal utility diminishes


as income increases.

Say Nitin’s income is 15K –


his utility is 13.5

1. Will he take up a job with


the following prospect,
Z={30,000, 10,000; 0.5,0.5}, 30

if his current income is a


certain 20,000?
Certainty Equivalent and Risk Premium
Z={30,000, 10,000; 0.5,0.5}
Expected Utility
Nitin is risk averse because
he would prefer a certain Utility Curve
income of $20,000 (with a
utility of 16) to a gamble
with a .5 probability of
$10,000 and a .5
probability of $30,000 (and
expected utility of 14).

A utility of 14 is equivalent
to a certain income of
16000. Hence 16000 is
the certainty equivalent
of his prospect of
(10000,30000; 0.5,0.5) and
Rs 4000 is his risk
premium.
Risk Premium = Maximum amount of money that a risk-averse person will pay to
avoid taking a risk
Certainty Equivalent and Risk Premium
Certainty equivalent of a risky choice Risk Averse Individual
is the amount of payoff, which if
provided with certainty, would leave
the individual equally well off
Risk Premium = Amount an
individual would give up to leave her
indifferent between the risky choice
and the certain one = expected
payoff – certainty equivalent
For choice Z, U (16000)=14 = U(Z)
 16000 = certainty equivalent
 20000-16000 = 4000 = risk
premium
If Y={0,40000; .5,.5},
E(Y) = 20000
EU(Y) = .5*0+.5*20 = 10
 10,000 = certainty equivalent
 20K-10K = 10,000 = risk premium

Greater the variability, lower the certainty equivalent and greater the risk premium
Same set of choices, different utility functions
Expected Utility
• b- risk loving
She would prefer
the same gamble Utility Curve
(with expected
utility of 10.5) to
the certain
10.5
income (with a
utility of 8).
• c – risk neutral
She is indifferent
between certain
and uncertain
events with the
same expected
income
Robust Behaviour under Risk Aversion
• ‘actuarially fair’ insurance
Insurance premium = promised benefit * probability of
the loss occurring.
Most insurance not actuarially fair – why?
A risk averse individual will always pay a “positive
premium” to insure against risk
• Portfolio Diversification
Hedging as a special case
If there are N risky securities that are independently and
identically distributed, a risk-averse person will divide
her total investment portfolio equally among the N
securities
• Investing in getting more information
A Portfolio Choice Example
• You have Rs. 1 million to invest in two financial securities for a year
Asset X is a bond with a certain interest of 10% p.a.
Asset Y is a stock that will go up by 21% over the year or stay
unchanged with equal probability [expected return = 10.5%]
• If you invest y (fraction of 1 million) in Y and rest in X, you get a
random prospect:
{1.21y + 1.10(1-y), y + 1.10(1-y); 0.5, 0.5}
• So your expected utility: 0.5u(1.21y + 1.10(1-y)) + 0.5u(y + 1.10(1-
y))
• Your optimal choice is that y* that maximizes your expected utility
– If u(z) = log z, then y* = 0.5 million [50-50 split]
• If I did not know your u function, I would still be able to predict:
(i) If X is a ‘safe asset’ with return rX% p.a., and Y is ‘risky asset’
with ‘expected return’ rY% p.a., then you will hold some Y iff rY > rX
(ii) If X and Y are i.i.d. risky assets with ‘common expected return’
r% p.a., then you will hold equal amounts of X and Y

You might also like