Problem Set 1 - Solutions

Econ 119 - Summer 2017

Due: 4pm July 6

Problem 1. This question is motivated by a story from the New York Times, which is
worth reading as background. Click here to read the story.

Consider a drug o↵ender who is on probation and regularly monitored for drug use. He
is deciding whether to take a hit and receive an immediate instantaneous utility of h = 1,
knowing that he will be punished with certainty for violating the terms of his probation.

The criminal justice system would like to deter drug use, and is contemplating two options
for punishing probation violators. Option one: give violators a short jail sentence of js days
immediately, yielding a utility of js immediately. Option two: give violators a long jail
sentence of jl days in the future, yielding a utility of jl at the beginning of the sentence
(i.e. all of the utility impact of the sentence is felt in a lump sum at the beginning of the
sentence).

The o↵ender’s utility is zero if he neither takes a hit nor serves any jail time.

1. Assume drug o↵enders are exponential discounters (i.e. = 1) with  1. Let

js = 1.01 and jl = 2, to be served a month in the future.
(a) What is the largest monthly discount factor m such that the short jail sentence
successfully deters drug use while the long jail sentence does not?
(b) What is the largest yearly discount factor y where this condition holds? [Hint:
1
since a month is 12 of a year, the yearly discount factor is a simple function of
the monthly discount factor.]
When faced with the short but immediate punishment, the o↵ender decides to use
drugs if:

U (takingdrugs) = h js 0 = U (abstaining)

When h = 1 and js = 1.01 this condition never holds, so the o↵ender will never take
drugs when faced with the short but immediate punishment (no matter what his value
of m might be).

When faced with jl = 2 a month from now, an o↵ender who is an exponential dis-
counter decides to use drugs if:

1
U (takingdrugs) = h m jl 0 = U (abstaining)
h m jl
h 1
m  =
jl 2
So the largest monthly discount factor, m , such that the o↵ender might do drugs
when faced with jl = 2 a month from now is m = 0.5 (he would be exactly indi↵erent
between the two options).

The yearly discount factor, y, is simply the monthly discount factor multiplied by
12
itself twelve times: y = ( m) . In this case, the yearly discount factor is y =
(0.5)12 ⇡ 0 : 000244.
2. Now assume drug o↵enders are quasi-hyperbolic discounters with beta  1 and y =
0.99. If js = 1.01 and jl = 2 is to be served a month in the future, what is the largest
such that the immediate short jail sentence successfully deters drug use while the
delayed long jail sentence does not?
As seen in part (a), the short immediate sentence will deter drug use no matter how
the o↵ender discounts the future. An o↵ender who is a hyprbolic discounter with
y = 0.99 will take drugs when faced with jl = 2 a month from now if:

U (takingdrugs) = h m jl 0 = U (abstaining)
h m jl
h

m jl
h
p
 12
y · jl
h
p
 12 ⇡ 0.5
0.99 · 2
The largest such that an o↵ender with y = 0.99 might do drugs when faced with
jl = 2 a month from now is approximately = 0.5.
3. Consider an exponential discounter with yearly discount factor, y, found in part (1).
(a) Would a massive sentence of jl = 20, 000 starting two years from now be e↵ective
at deterring drug use?
(b) Compare the e↵ectiveness of the massive sentence to the e↵ectiveness of a 1.01-
day sentence served immediately, which you determined in part (a). Explain the
psychological intuition for the similarities or di↵erences between the e↵ectiveness
of these two punishments.
As seen in part (a), the short immediate punishment will perfectly deter drug use. An
o↵ender who is an exponential discounter with y = (0.5)12 = 0.000244 will do drugs
when faced with a lifetime sentence of jl = 20, 000 starting two years from now if:

2
U (takingdrugs) = h y · jl 0 = U (abstaining)
2
1 (0.000244) · 20, 000 ⇡ 0.9988 > 0

2
Thus the o↵ender will choose to do drugs. For the yearly discount factor found in
part (a) a lifetime sentence starting two years from now does not deter drug use more
e↵ectively than a sentence of 1.01 days served immediately.

This does not seem particularly reasonable. An addict that’s will- ing to go to jail
for the rest of his life to take a hit now is con- ceivable, but this same addict would
presumably also be willing to endure just over a day of jail now for a hit now.
4. Consider the quasi-hyperbolic discounter, with = 0.99 and the found in part (2).
(a) Will the massive delayed sentence deter drug use?
(b) Compare your answer to your answer from part (c). Explain the psychological
intuition for the similarities or di↵erences between these answers.
The hyperbolic discounter with = 0.5 and y = 0.99 would do drugs when faced
with a lifetime sentence of jl = 20, 000 starting two years from now if:

2
U (takingdrugs) = h y jl 0 = U (abstaining)
2 2
h y jl = 10.5 · 0.99 · 20, 000 = 9, 800 ⇤ 0

The hyperbolic discounter in part (b) responds as we might expect to the prospect of
a lifetime sentence starting years from now: he decides not to use drugs. Since he has
revealed himself unwill- ing to endure just over a day of jail now to take drugs now,
it seems unlikely that he would be willing to endure a lifetime of jail starting in two
years.

Problem 2. Nigel and Sophie both like to play a single-player casual computer game. It’s
the kind of game in which there are multiple rounds, and once you decide to start playing,
the decision of when to stop is determined by a series of stop/don’t-stop decisions made
at the end of each round. The motivation for this problem—and the research agenda it is
based on—is the so-called “one more game” phenomenon that so many gamers experience,
in which players find themselves repeatedly saying “I’ll just play one more game.” This
problem is intended to explore how we might model that phenomenon, and how naı̈veté vs.
sophistication are implicated.

Let’s start by constructing a (very) simplified model of the decisions involved in playing
a casual game. It’s a four-period model, t = 1, 2, 3, 4. In each of the first three periods,
the player decides whether or not to play the game. Thus, in period one they are deciding
whether to start playing, and in periods two and three they decide whether to play another
round. Not playing is an absorbing state, which means that once you decide to not play, you
can’t start again. In other words, deciding to not play in period two means that you played
for one round and then stopped. Period 4 comes after they stop playing, and is intended
to capture any long-term or delayed costs or benefits of playing in periods 1 through 3.
One way to think of this is that you play the game in the evening, and the time you spend
playing is time you could have been sleeping, so period 4 is the next day, when you are too
tired to perform well on an exam. (Note that in this simplified model we are going to set
= 1 so it doesn’t matter that periods 1 through 3 are much shorter than period 4, because
from the perspective of any prior period, all period-4 costs and benefits are just going to be
discounted by .)

For each period that a player plays the game they receive an enjoyment benefit of bt which
they experience immediately, and they pay a delayed cost of c, which they experience in

3
period four. (As above, c represents something like decreased performance on an exam the
next day. The total delayed cost depends on how long they played the game.

Nigel and Sophie have identical preferences over the game—meaning identical costs, benefits,
and discounting—as follows:

c = 4, b1 = 5, b2 = 3, b3 = 1, = 1/2, = 1

Thus, if a player plays in period one they receive a benefit of b1 = 5 in period one, and
a cost of c = 4 in period four. If a player plays in both period one and period two, they
receive a benefit of b1 = 5 in period one, a benefit of b2 = 3 in period two, and a total cost
of 2 ⇥ c = 8 in period four. Etc.

However, Nigel is a naif, while Sophie is a sophisticate.

1. How many rounds will Nigel play? [Hint: you can probably do this without a game
tree, by simply computing Nigel’s utility maximization problem for playing versus not
playing, given what he believes his future selves will do. But you should be able to
draw the appropriate game tree, and you will need to for part b anyway.]
The game tree is at the end of the solutions, below.
Nigel, being naive, always thinks he will stick with his preferred plan. He plays in
period one, planning to stop in period two. When period two comes he plays again,
thinking he will stop in period three. In period three he actually does stop.
2. How many rounds will Sophie play? [Hint: This time it will really help to start by
drawing a game tree. It should have three players, each of which represents Sophie in
a di↵erent time period. Think carefully about how to represent the utility outcomes
of each choice Sophie makes for each of her three “selves.”]
Refer again to the game tree at the end of this document.
Sophie, being sophisticated, looks forward to period three and foresees that she will
not play in period three. Thus, looking forward to period two, she foresees that her
period-two self will foresee not playing in period three, and that her period-two self
will therefore play in period two, correctly foreseeing that there’s no risk in doing so,
because she knows she’ll stop in period three. Now, from the perspective of period
one, Sophie would really like to play only in period one, and then stop in period two,
but she knows that not a realistic option. So she weighs the total discounted utility
from playing in periods one and two, which is 5 5/2 = 21/2 to the total discounted
utility of not playing in period one, which is 0 and decides to play, even though she
knows it means she play longer than she would really like.

Now assume that there is a period before the beginning of the game, t = 0. In period
zero players cannot play the game, but they can pay a flat-rate price, p, for a binding
commitment device that allows them to decide in advance how many rounds they want to
limit themselves to, and prevents them from playing more than that.
3. In period zero, how many rounds of play will Sophie choose to limit herself to?

From the perspective of period zero, all costs and benefits are discounted by ,
so from the perspective of period zero, the payo↵s from playing 1, 2, or 3 games
are 1/2, 0 and 3/2 respectively.

4
If Sophie uses the commitment device she will choose to commit herself to play
in only period one

4. [Harder] Assuming that all costs and benefits are measured in dollars, what is the
maximum that Sophie will pay to use the commitment device?

From the perspective of period zero Sophie foresees that the commitment device
will allow her to play one period instead of two, and from the perspective of
period zero this gives her discounted utility of 1/2 instead of 0, so she will be
willing to pay p = $0.50 for the commitment device.

Problem 3. Consider a di↵erent pair, Nigel and Sophie, who live for three periods, t =
0, 1, 2, and have to make some decisions about saving, borrowing, and consumption. (They
make all their decisions separately.) All consumption happens in periods 1 and 2. Period 0
is just a decision-making period. Both Nigel and Sophie have instantaneous utility function
u(ct ) = ct where ct is the amount they consume in period t. In addition, they both have
present-biased preferences with = 12 and = 1. However, Nigel is a naif ( ˆ = 1), whereas
Sophie is a sophisticate ( ˆ = ).

Nigel and Sophie each start o↵ with a wealth of $20 at t = 0, and earn additional income of
$20 per period at t = 1 and t = 2. At t = 0 each of them can keep his or her initial wealth
in a checking account, in which case it earns no interest, but is available for consumption in
either period 1 or period 2. Alternatively, they can choose to put part or all of their initial
wealth in a savings account.

There are two types of savings account available. Type-A accounts work like this: if you
put money in at t = 0 you can take it out again at t = 1 but you won’t earn any interest.
If you leave it in until t = 2 you will earn a total of 20% interest. (That is, if you put $10
into the account at t = 0 and leave it until t = 2 you get back $12.) Type-B accounts are
similar but di↵erent: if you put money in at t = 0 you cannot take it out again until t = 2,
and furthermore it will only earn a total of 10% interest.

In addition to these savings options, Nigel and Sophie can each choose to borrow $10 on
their credit card in period 1 at a per-period interest rate of 50%, which means that if they
decide to borrow $10, they have to pay back $15 in period 2.

[Important note: from the perspective of t = 0, the individual doesn’t care which period
they spend their future wealth. The only thing they care about is the sum of consumption
over the two future periods. In other words, from the perspective of t = 0, Nigel and Sophie
are just trying to maximize their total future consumption. To convince you that this is
true, notice that their utility function at t = 0 is U (c1 , c2 ) = c1 + c2 . Maximizing that
function is exactly the same as maximizing c1 + c2 . They both just want to maximize their
total future consumption, without caring which period the consumption comes in.]

1. Let’s figure out what Nigel will do.

(a) Will Nigel choose to save money at t = 0? If so, how much, and what type of
savings account will he choose?
(b) At t = 0, what will Nigel plan to do at t = 1? Will he plan to borrow the $10 on
his credit card? If he saved in a type-A savings account, how much will he plan
to withdraw?

5
(c) When t = 1 arrives, will Nigel follow through on his plan? If not, what will he
do instead?
(d) Explain the psychology behind Nigel’s behavior.
Looking forward from t = 0, Nigel believes his future self will follow through on
whatever plan he currently prefers. (That is because he believes his future self will be
time consistent, and a time consistent person is defined as a person who will always
follow through on their plans.) Since his goal at this stage is simply to maximize his
future income, his preferred plan is to save all of his initial wealth in a type-A savings
account and leave it there until t = 2, and to not borrow on his credit card. And since
he believes he will follow through on that plan, he goes ahead and saves $20 in the
type-A account.

At t = 0 he believes he will not withdraw his savings, or borrow on his credit card,
because either of those actions would reduce his total future income. Thus, is plan
will be to neither withdraw or borrow.

When period 1 arrives, Nigel stops discounting period-1 consumption, though he still
discounts period-2 consumption. Thus, for each dollar he withdraws he will get im-
mediate utility of 1, and lose discounted future utility of 12 · $1.2 = $0.6 due to losing
the interest he would have earned at t = 2. Thus, he will withdraw all of his savings.
Similarly, for each dollar he borrows he will get immediate utility of 1, and lose dis-
counted future utility of 12 · $1.5 = $0.75, so he will borrow the entire $10 available to
him.

The psychology is simple. When looking forward, Nigel imagines that he will be patient
in the future, so even though he discounts all future utility, he behaves like a perfectly
rational utility maximizer at first, saving all his money at the highest possible rate of
return. An imporant part of this situation is the fact that all of Nigel’s consumption
utility lies in the future. In other words, at t = 0 there’s no way for Nigel to get
any immediate gratification, by, for example, borrowing all his future income and
spending it right away. Thus, by taking away all temptation in the decision-making
period, we’ve made it possible for Nigel to make a “good” choice at first.

However, once period 1 arrives, Nigel is exposed to temptation, and because he did not
correctly predict that he would be tempted, he did not take steps to protect himself
from temptation by putting his savings in the type-B account. Instead, he just spends
everything he possibly can.
2. Now let’s consider Sophie. [Hint: remember to use backwards induction.]

(a) Will Sophie choose to save money at t = 0? If so, how much, and what type of
savings account will she choose?
(b) At t = 0, what will Sophie plan to do at t = 1? Will she plan to borrow the $10
on her credit card? If she saved in a type-A savings account, how much will she
plan to withdraw?
(c) When t = 1 arrives, will Sophie follow through on her plan? If not, what will she
do instead?
(d) Explain the psychology behind Sophie’s behavior.
Just like Nigel, Sophie would like to maximize her total future income. However, when
considering her available options, she looks forward to predict what her future self will

6
do in each of the possible scenarios. (This is what we mean by backwards induction.)
When she considers putting all her initial wealth into a type-A account, she foresees
that her period-1 self will withdraw the money. Thus, in terms of total future income,
she sees that there will be no di↵erence between saving the money in a type-A account
and simply keeping it in checking. Putting her initial wealth into a type-B account,
however, will earn her $2 in period 2, which is better than nothing, so she puts her
entire initial wealth into a type-B account.

At t = 0 Sophie would prefer not to borrow on her credit card, but she foresees that
her period-1 self will borrow, and since there is nothing she can do at t = 0 to remove
thae temptation to borrow from her period-1 self, or to “bind” herself to her preferred
plan of not borrowing, she plans to borrow the money.

Once period 1 arrives, Sophie will stick to her plan, which is to borrow on her credit
card. We know this without having to solve any new math, because a sophisticate will
only ever choose a plan that she knows she will follow through on, which means that
whatever she plans to do, by definition she will follow through on the plan.

The psychology is straightforward. Sophie is tempted in exactly the same way as

Nigel, but unlike Nigel she foresees the negative future consequences of temptation
and thus rules out plans that are unfeasible, selects her favorite plan from among the
ones that are feasible, and then follows through on it consistently. She binds herself
where she can, which is in the savings domain. She would like to be able to bind
herself in the borrowing domain, but she can’t.
3. [Harder] Now let’s consider the policy preferences of Nigel and Sophie, as expressed by
their willingness to pay for di↵erent policies. Let’s begin by assuming that at t = 0, in
addition to the $20 they have to save, Nigel and Sophie also have immediate income
of $20 which (for some reason) they cannot save.
(a) At t = 0, how much would Sophie be willing to pay to have the government make
credit cards illegal?
(b) At t = 0, how much would Nigel pay to prevent the government from making
credit cards illegal?
(c) If the government were to decide at t = 0 whether to make credit cards illegal
purely on the basis of willingness to pay (in other words, on the basis of benefit
cost analysis, would the resulting policy make Sophie and Nigel better o↵ or
worse o↵? Given your answers, how do you explain the fact that credit cards are
not illegal?
At t = 0 Sophie realizes that if credit cards were illegal it would save her $5 because
she would be protected from compulsive borrowing in period 1. Thus, she will be
willing to pay up to the value of that $5 savings to have the policy put into place.
Now, Sophie will discount the $5 savings by because it is in the future, so that
savings will only be worth · 5 = 2.5, and that is how much she will be willing to pay
to have credit cards made illegal.

At t = 0 Nigel thinks that he will follow through on his preferred plan to not borrow, so
he does not think he will benefit from making credit cards illegal. Thus, his willingness
to pay is zero.

Clearly the government will get more money out of making credit cards illegal, and
will impose that policy. This will increase total income for both Sophie and Nigel by

7
$5, so in some sense it makes them both better o↵. However, it also deprives them
both of period-1 temptation utility, and it is unclear whether we should prefer their
long-term utility or their short-term utility when deciding what makes them better
o↵. One approach would be to treat each of their “selves” as if they were di↵erent
members of society. In this approach the policy is making them worse of in period one
and better o↵ in period two. It isn’t clear which self we should care about most. If
we treat each period as equally important, and just sum up the gains and losses over
all of their future selves, then making credit cards illegal increases the sum of per-self
utility over all selves. In this example, that is identical to just maximizing their total
income.

In addition to credit cards, payday lending is a form of extremely costly short-term

credit, and if people have self-control problems, our model suggests that it might make
society better o↵ if the government banned them. The fact that government has not
banned credit cards (or, indeed, payday lending) suggests that there is something
wrong with our model. One problem with our model is the existence of the unrealistic
t = 0, in which people get to make decisions about public policy without any immediate
temptation, and put o↵ the costs of implementing the public policy into the future.
In reality, people are faced with the temptation to borrow money against their future
income all the time, and they have to pay the cost of political activism to change
public policy in the immediate present, when that cost is undiscounted.

8
Self-1 Game tree for playing a casual game.
[Inst: 𝑏𝑏1 = 5, 𝑏𝑏1 = 3, 𝑏𝑏1 = 1, c = 4]

Self-2

Self-1: 0

Self-3

1
Self-1: 5 − � 4 = 3
2

Self-2: 0

1 1 1
Self-1: 5 + (3 − 8) = 2 …………………...5 + (4 − 12) = 1
2 2 2
1 1 1
Self-2: 3 − � 4 = 1 ……………………3 + 1−8 =−
2 2 2

Self-3: 0 1
…………………………..1 − � 4 = −1
2