Vous êtes sur la page 1sur 12

MIDTERM I

Fall 2013

Rules:

Do not begin the exam until instructed to do so.

This is a closed-book, closed-notes exam. Calculators are allowed as long as they


do not store formulas.

You have 80 minutes to complete the exam.

Please be courteous to other students if you leave the exam room early.

You may not discuss this exam with anyone or transmit information about this
exam to anyone until everyone in the course has taken either this exam or the
make-up exam. Doing so is a violation of University regulations.

Instructions and Information:

Anything written on the backside of pages will not be graded. There is extra
space at the end of the exam if you run out of room in the space provided.

Points are proportional to the number of minutes attributed to each question.

On quantitative questions, you must show your work to get credit. If you think
there is ambiguity in the assumptions of the question, state your assumptions
clearly.

On qualitative questions your answers should be succinct yet complete. You will
be evaluated on both the clarity and the content of your answers. Answering
with either bullet points or full sentences is fine. Points will be deducted for
comments that are needlessly long or not relevant.

On True or False questions, remember to justify each answer. Answers without


justification will not be graded. State clearly whether your answer is True or
False. It depends is not a valid answer.

Remember to answer all parts of all questions.

Please sign this after you have finished the exam:

I have not cheated on this exam. I will comply with University regulations in all
matters related to this exam.

Signature:

Name:

Section:
MIDTERM I
Fall 2013

Part I: True or False, and Justify. (35 minutes) Determine whether each of the
following statements is true or false, and state either true or false. Then provide
a justification. Answers without justification will not be graded, and quality of your
justification matters.

Question 1: (5 minutes) Consider a risk averse agent and a gamble with positive
expected value. If the agent is risk averse enough, there will be no amount of money
that he will be willing to risk even though the gamble has positive expected value.

False. As long as the agent puts positive utility on money, he would be willing to risk
some amount of money in a gamble that has a positive expected value. If we think
about increasing a risk aversion parameter, i.e., making an agent more and more risk
averse, this only means that the amount the agent is willing to risk is decreasing.

Question 2: (5 minutes) Quadratic Utility satisfies Second Order Stochastic Domi-


nance but fails to satisfy First Order Stochastic Dominance.

True. SOSD is satisfied because among gambles that are mean-preserving spreads, the
quadratic utility function will always select the gamble with the higher expected utility.
The mean will be the same across the two gambles, but the variance will be larger
for the riskier gamble. FOSD is not satisfied because the quadratic utility function
is decreasing in money after some point. We could certainly construct a gamble (as
demonstrated in lecture) with the following properties: (1) gambles have the same mean
and dierent variances, but one is not a mean-preserving spread of the other; (2) the
quadratic utility function will select the gamble with the lower variance; and (3) given
some concave utility function, an expected utility maximizer would be better o choosing
the gamble with the higher variance.
MIDTERM I
Fall 2013

Question 3: (5 minutes) A risk averse agent will never buy insurance with a zero
load.

False. Mossins theorem tells us that if an insurance policy is actuarially fair (i.e., has
zero load), then a risk-averse agent will always choose to fully insure.

Question 4: (5 minutes) Consider a gamble with a 50% chance of halving the agents
wealth and a 50% chance of doubling the agents wealth. A risk averse agent may
accept the gamble.

True. The agents choice regarding whether to take this gamble will depend on his level
of risk aversion. Notice that the gamble is not symmetric so the expected value of the
gamble is not equal to the agents original level of wealth: the upside is twice as large as
the downside (100]% of wealth vs. 50% of wealth.). The expected value of the gamble
is (1/2)(2W ) + (1/2)(W/2) = 5W/4 > W .
MIDTERM I
Fall 2013

Question 5: (5 minutes) An agents response to a permanent shock to income will be


the same as that to a temporary shock to income only if U 000 () = 0

False. Regardless of the degree of patience, an individual will fully adjust his consump-
tion in both periods in response to a permanent and to a temporary income shock. The
amount by which which consumption changes in these two cases will be dierent and
depends on their impact on the PDV of lifetime income. The degree of patience only
tells us about how consumption is allocated between c1 and c2 .

Question 6: (5 minutes) A risk averse person with a decreasing coefficient of absolute


risk aversion necessarily has a decreasing coefficient of relative risk aversion.

False. A decreasing coefficient of absolute risk aversion means that as wealth increases,
the amount of money that one is willing to put at risk is increasing as the agent becomes
less risk averse. This amount of money that the agent risks can increase at a faster,
slower, or the same rate as the increase in money. It is the relationship between these
growth rates that will determine how the coefficient of relative risk aversion changes as
wealth changes. Mathematically, the relationship between the coefficient of relative risk
aversion, RRA and absolute risk aversion, ARA is given as follows: RRA = x ARA,
where x is the level of wealth. A counterexample to the claim could be the following:
a coefficient of ARA equal to 1/x is decreasing. However, the coefficient of RRA for
someone with this ARA coefficient is 1.
MIDTERM I
Fall 2013

Question 7: (5 minutes) Suppose that on this exam, true and false answers are equally
likely. If you have answered false to all of the previous questions, then its more likely
that the answer to this question is true.

False. Its just like a coin flip. Knowing the first n observations doesnt give you any
information about observation n+1. Such a belief is what we have referred to as mean
reversion or the gamblers fallacy.
MIDTERM I
Fall 2013

Part II: Long Questions

Question 8: Tax Evasion as a gamble (27 minutes)


Suppose that an individual earns Y of income in a given year. When it comes time
for the individual to do his taxes, the individual chooses how much income to report
to the IRS. Denote this reported amount X. The taxpayer pays a constant tax rate of
on his reported income. (ie. the tax paid is X).
The taxpayers problem is to choose how much income to report to the IRS. The
taxpayer may be truthful, in which case X = Y , or may evade taxes by reporting less
than his true income, i.e., X < Y . If he chooses to evade, there is some probability
that he will be caught. If caught, the taxpayer will have to pay taxes on the reported
amount plus taxes on the undeclared amount. The latter is taxed at a penalty rate
> .

a. (3 minutes) What is the individuals after-tax income if he truthfully reports his


income to the IRS?

If the individual reports X = Y he gets after-tax income of Y (1 ) with probability


1.

b. (3 minutes) What is the individuals after-tax income if he reports X to the IRS


and is not audited?

If the Individual reports X < Y he gets after-tax income of Y X with probability


(1 )
MIDTERM I
Fall 2013

c. (3 minutes) What is the individuals after-tax income if he reports X to the IRS


and is audited?

If the Individual reports X < Y he pays tax of X on the reported amount and (Y X)
on the unreported amount. Therefore he gets after-tax income of Y X (Y X)
with probability .

d. (4 minutes) Assume the individual has a von Neumann-Morgenstern utility function


over after-tax income denoted by U (W ). Write down the condition where an individual
would choose to evade taxes rather than truthfully report his income.

He will choose to evade if he gets a higher expected utility from evading. In equations,
he evades if there exists an X 6= Y such that:

U (Y X (Y X) ) + (1 )U (Y X ) >U (Y (1 ))
MIDTERM I
Fall 2013

e. (5 minutes) Suppose the individual is risk-neutral. Find the range of values for
under which the individual would choose to evade taxes.

If the agent is risk neutral, we know his utility function is linear. We also know that we
can take any positive affine transformation of a utility function and get an equivalent
utility function. Therefore we can consider the function U (W ) = W . In this case we
have:

(Y X (Y X) ) + (1 )(Y X ) >Y (1 )
( (X Y )) >(X Y )
< /

so he will chose to evade if 0 < < / .

f. (5 minutes) If the individual were instead risk-averse, what would happen to the
range of values for under which the individual would choose to evade taxes? What
is the intuition behind this?

A risk-averse individual would have a smaller range of values for under which he
would choose to evade taxes.
MIDTERM I
Fall 2013

g. (5 minutes) For utility functions that are increasing in money, is it always true that
evading is preferred? Why or why not? Include a graph in your argument. [Hint:
What do the CDFs of the choices look like?]

It is not always true that evading is preferred. Depends on parameters. In particular,


neither choice dominates the other in the sense of first order stochastic dominance. To
see this, consider the CDFs of the respective choices:

1
Y (1
X

Y
X
)
(Y
X)

Since the CDFs cross, neither is better in the sense of first order stochastic domi-
nance.
MIDTERM I
Fall 2013

Question 9: Tornado Insurance (18 minutes) Gotta Have It Insurance is consider-


ing selling tornado insurance in the small town of Shawnee, Oklahoma. Their talented
risk assessors have estimated that there is a 1% probability that any given house in
Shawnee will be hit with a tornado in a given year.

a. (2 minutes) Consider an insurance policy with zero load (i.e., is actuarially fair)
that pays $10,000 if the house is hit by a tornado and zero otherwise. What is the
premium for this insurance policy?

P remium = (0.01) $10, 000 = $100

b. (3 minutes) Define a random variable X as the net payout of the insurance company
(i.e., payout minus the premium) of providing the insurance policy. What are the
expected value and variance of X?

E[X] = (0.99) $100 + (0.01) $9, 900 = 0


.

V ar(X) = E[(X E[X])2 ] = (0.99) ( 100)2 + (0.01) (9, 900)2 = 990, 000

.
MIDTERM I
Fall 2013

c. (3 minutes) Assume that the probability that any one house in Shawnee being hit
by a tornado is independent of that of any other house. As the number of houses that
Gotta Have It insures increases, to what value will the sample average net payout for
the insurance company converge? What concept discussed in lecture did you use to
come to this conclusion?

We know from the Law of Large Numbers that the sample average will converge
to the population average as the number of observations grows large. In this case the
population average of the net payout is zero so that is the value to which the sample
average net payout will converge.

d. (5 minutes) Continue to assume that the probability of getting hit by a tornado


is uncorrelated across houses in Shawnee. According to the central limit theorem, to
what distribution does the sample average net payout converge as the number of houses
that Gotta Have It insures increases?

The Central Limit states that the mean of a sequence of i.i.d. random variables {Xn }
with E[Xi ] = and V ar[Xi ] = 2 < 1 then as n approaches infinity, the random
p
variables n(X ) converge in distribution to a normal N (0, 2 ):
Xn
p 1 d 2
n Xi ! N (0, )
n
i=1

In this case we have E[Xi ] = 0 and V ar[Xi ] = 990, 000 so


Xn
p 1 d
n Xi ! N (0, 990, 0000)
n
i=1
MIDTERM I
Fall 2013

It turns out that houses being hit by tornados in Shawnee are not really independent
events. Instead, if one house is hit by a tornado, the probability that other houses in
Shawnee are hit by a tornado increases.

e. (5 minutes) What happens to the variance of Gotta Have Its net payout relative
to the case where getting hit by a tornado is independent? Is it problematic for Gotta
Have It to ignore that tornado damages are related?

Because the Central Limit Theorem assumes that observations are independent, it fails
to hold for the case of correlated outcomes. This is a huge problem for Gotta Have It,
as the variance of the total payout will be much larger than they would in the case of
independence.

Vous aimerez peut-être aussi