Chapter 6
Chapter 6
Chapter 6
Uncertainty
Expected
value
Expected
utility
Attitudes
toward Risk Rational Choice Under Risk and Uncertainty
Discussion
1 / 41
2019-09-20
Rational Choice Under Risk and Uncertainty
Next time when teaching this material, consider adding some stuff about
maxmin and alpha-maxmin expected utility from Just’s book, chapter 9.
Introduction: Risk vs Uncertainty
Introduction
• Chapters 4 and 5 left the theory of decision aside for a moment.
Uncertainty
In this chapter we return to questions of decision, and
Expected
value specifically rational decision.
Expected • In this module we will explore the theory of rational choice
utility
under risk and uncertainty.
Attitudes
toward Risk
• A choice under risk is made when the probabilities of the
Discussion outcomes are meaningful and known.
• A choice under uncertainty is made when the probabilities are
unknown or not meaningful.
• We want a theory that gives us answers to the question of what
choice to make in any given decision problem.
• We begin by discussing uncertainty and then proceed to
expected value, before getting to expected utility.
• Ultimately, expected-utility theory combines the concept of
utility from Chapter 2 with the concept of probability from
Chapter 4 into an elegant and powerful theory of choice under
risk.
2 / 41
Decision Strategies under Uncertainty:
Probabilities are Unknown
Introduction
Uncertainty • The maximin criterion specifies that the best choice is the
Expected
value one with the greatest minimum utility payoff. (The
Expected maximum minimum)
utility
Attitudes
- This strategy is exceedingly cautious.
toward Risk
• The maximax criterion specifies that the best choice is the
Discussion
one with the greatest maximum payoff. (The maximum
maximum)
– This strategy is exceedingly reckless.
• The α-maximin criterion specifies that the best choice is
the one with the greatest weighted average of the
maximum and minimum payoffs.
– This strategy is a good middle ground between the
maximin and maximax criteria.
3 / 41
One More Criterion
Introduction
Uncertainty
Expected
value
Expected
utility
• The minimax-risk (regret) criterion specifies that the best
Attitudes
toward Risk choice is the one that minimizes the maximum risk or
Discussion regret.
• The regret equals the best payoff you could have had if
you had acted differently minus your actual payoff. The
choice with the least maximum regret is the one to choose.
4 / 41
Example
Introduction
Uncertainty
Expected Economy
utility
Attitudes
Growing Stable Declining
toward Risk
Discussion
Bonds 40 45 5
Stocks 70 30 -13
Mutual Funds 53 45 -5
Introduction
Uncertainty The maximin criterion specifies that the best choice is the one
Expected with the greatest minimum utility payoff. (The maximum
value
minimum)
Expected
utility
Attitudes
toward Risk Table: Utility Payoff Matrix: Maximin
Discussion
Economy
Growing Stable Declining Minimum
Bonds 40 45 5 5
Stocks 70 30 -13 -13
Mutual Funds 53 45 -5 -5
6 / 41
Decision Strategies under Uncertainty: Maximax
Introduction
Uncertainty
The maximax criterion specifies that the best choice is the one
Expected
value with the greatest maximum payoff. (The maximum maximum)
Expected
utility
Discussion Economy
Growing Stable Declining Maximum
Bonds 40 45 5 45
Stocks 70 30 -13 70
Mutual Funds 53 45 -5 53
7 / 41
Decision Strategies under Uncertainty: Maximax
Uncertainty
2019-09-20
The maximax criterion specifies that the best choice is the one
with the greatest maximum payoff. (The maximum maximum)
Economy
The watch Example Having just bought a brand new watch, you are
asked if you also want the optional life-time warranty.
(a) Would a maximin reasoner purchase the warranty?
(b) What about a maximax reasoner?
ANSWER(a) A maximin reasoner would purchase the warranty. (b) A
maximax reasoner would not.
Decision Strategies under Uncertainty: α-Maximin
Introduction
• The α-maximin criterion specifies that the best choice is the one with
Uncertainty
the greatest α-weighted average of the maximum and minimum
Expected
value payoffs.
Expected • α ∈ [0, 1] is the (Leonid Hurwicz) optimism-pessimism index, with
utility
α = 0 for an extremely optimistic individual and α = 1 for an
Attitudes extremely pessimistic individual.
toward Risk
• Assume α = 0.8 (relatively pessimistic), then the choice metric is
Discussion
α × min +(1 − α) × max = 0.8 × min +0.2 × max
Economy
Growing Stable Declining α-maxmin
Bonds 40 45 5 12
Stocks 70 30 -13 3.6
Mutual Funds 53 45 -5 6.6
8 / 41
Decision Strategies under Uncertainty:
Minimax-risk (regret)
Introduction
Uncertainty
Expected
value
Expected
• Regret equals the best payoff you could have had if you
utility
had acted differently minus your actual payoff. The choice
Attitudes
toward Risk with the least maximum regret is the one to choose.
Discussion • The term regret aversion is sometimes used when
discussing people’s tendency to behave in such a way as to
minimize anticipated regret. Regret aversion may be
driven by loss aversion, since regret is due to the loss of a
payoff that could have resulted from the state that
obtains, had the agent acted differently.
9 / 41
Decision Strategies under Uncertainty:
Minimax-risk
Introduction
Uncertainty
Table: Utility Payoff Matrix
Expected
value
Economy
Expected
utility Growing Stable Declining
Attitudes
toward Risk Bonds 40 45 5
Discussion Stocks 70 30 -13
Mutual Funds 53 45 -5
Economy
Growing Stable Declining Max Regret
Bonds 30 0 0 30
Stocks 0 15 18 18
Mutual Funds 17 0 10 17
10 / 41
Decision Strategies under Uncertainty:
Minimax-risk
Uncertainty
2019-09-20
Table: Utility Payoff Matrix
Economy
Growing Stable Declining
Bonds 40 45 5
Stocks 70 30 -13
Minimax-risk Bonds
Growing
30 0
Economy
Stable Declining
0
Max Regret
30
Stocks 0 15 18 18
Mutual Funds 17 0 10 17
Exercise: Rational choice under uncertainty: What course of action would be favored by
(a) the maximin criterion, (b) the maximax criterion, (c)α-maximin criterion assuming
that α = 0.3, and (d) the minimax-risk criterion? As part of your answer to (d), make
sure to produce the risk(regret)-payoff matrix.
Uncertainty
2019-09-20
Table: Utility Payoff Matrix
Economy
Growing Stable Declining
Bonds 40 45 5
Stocks 70 30 -13
Minimax-risk Bonds
Growing
30 0
Economy
Stable Declining
0
Max Regret
30
Stocks 0 15 18 18
Mutual Funds 17 0 10 17
The dating game under uncertainty Imagine that you are considering
whether or not to ask somebody out on a date.
(a) Given your utility function, what course of action would be favored
by (i) the max- imin criterion, (ii) the maximax criterion, and (iii) the
minimax-risk criterion?
(b) In the words of Alfred, Lord Tennyson, “’Tis better to have loved and
lost / Than never to have loved at all.” What decision criterion do these
lines advocate?
Decision Strategies under Uncertainty: Discussion
and Criticism
Introduction
Uncertainty
• Of all criteria for choice under uncertainty, the maximin
Expected criterion is the most prominent. It reflects the general
value
notion of ambiguity aversion.
Expected
utility
• One objection about the maximin (or maximax) reasoning
Attitudes is that it fails to consider relevant utility information, since
toward Risk
for each act, it ignores all payoffs except the worst (or the
Discussion
best). Consider the following example of two payoff
matrices.
S1 S2 S1 S2
A 1 1 A 1 1
B 0 10 B* 0 1010
• Maximin reasoning favors A in either scenario. Yet it does
not seem irrational to favor A over B and B* over A, since
B* but not B upholds the prospect of ten billion utiles.
11 / 41
Decision Strategies under Uncertainty: Discussion
and Criticism
Decision Strategies under Uncertainty: best). Consider the following example of two payoff
matrices.
S1 S2 S1 S2
For the first bullet point the following passage in the book is very interesting: The
maximin theory is, among other things, an important part of the philosopher John
Rawls’s theory of justice. In Rawls’s theory, the principles of justice are the terms of
cooperation that rational people would agree to follow in their interactions with each
other if they found themselves behind a “veil of ignorance,” mean- ing that they were
deprived of all morally relevant information about them- selves, the society in which
they live, and their place in that society. Suppose, for example, that you have to choose
whether to live either in a society with masters and slaves or in a more egalitarian society,
without knowing whether (in the former) you would be master or slave. According to
Rawls, the rational procedure is to rank societies in accordance with the worst possible
outcome (for you) in each society – that is, to apply the maximin criterion – and to
choose the more egalitarian option. Rawls took this to constitute a reason to think that
an egalitarian society is more just than a society of masters and slaves.
Decision Strategies under Uncertainty: Discussion
and Criticism
Introduction
Uncertainty
2019-09-20
• Another objection is that maximin reasoning fails to take into
account the chances that the various states of the world will
obtain. A Nobel prize-winning economist John C. Harsanyi
offered the following example:
• Example: Suppose you live in New York City and are offered
two jobs at the same time. One is a tedious and badly paid job
in New York City itself, while the other is a very interesting and
Decision Strategies under Uncertainty: well-paid job in Chicago. But the catch is that, if you wanted
the Chicago job, you would have to take a plane from New York
to Chicago (for example, because this job would have to be
taken up the very next day). Therefore, there would be a very
Discussion and Criticism small but positive probability that you might be killed in a plane
accident.
• Maximin reasoning would favor the tedious NYC job, no matter
how much you prefer the Chicago job and no matter how
unlikely you think a plane accident might be.
Perhaps there are scenarios in which the probabilities of the relevant out-
comes are completely unknown or not even meaningful, and perhaps in
those scenarios maximin reasoning – or one of the other criteria discussed
earlier, in this section – is appropriate. Yet the upshot is that, whenever
possible, it is perfectly reasonable to pay attention to all possible payoffs
as well as to the probabilities that the various states might obtain. When
considering to take an umbrellla on the way out for example, it seems
rational to take into account all payoffs in all states of the world as well
as the probability of rain.
Choice under Risk: Expected Value
Introduction
Uncertainty
Expected
value • If it is both meaningful and possible to assign probabilities
Expected
utility to outcomes, then the expected value approach is a
Attitudes straight-forward way of taking the entire payoff matrix as
toward Risk
well as probabilities into account.
Discussion
• The expected value of a gamble is the amount you can
expect to win on the average, in the long run, when you
play the gamble.
• Example: We flip a fair coin. I give you $10 if it is heads
and nothing if it is tails. What is the expected value of
this gamble?
13 / 41
Choice under Risk: Expected Value
Introduction
The gamble above can be easily represented in both tree and
Uncertainty
Expected
table form:
value
Expected
utility
Attitudes
toward Risk
Discussion
H T
Offer $10 $0
On the average, in the long run, you would get $5 when playing
this gamble; ⇒ the expected value of the gamble is $5.
14 / 41
Choice under Risk: Expected Value
Expected value The gamble above can be easily represented in both tree and
On the average, in the long run, you would get $5 when playing
this gamble; ⇒ the expected value of the gamble is $5.
Example:
(a) Represent the gamble accepted by someone who plays Lotto 6/49 as a tree
and table. Assume that the grand prize is a million dollars.
(b) What is the expected value of a Lotto 6/49 ticket, if the grand prize is a
million dollars?
(c) What would you pay for the gamble?
(a)
(b) 1/13, 983, 816 × $1, 000, 000 = $0.07 (c) No more than 7 cents.
Representing More Complex Gambles in Tree Form
Introduction
Imagine you are given a choice between the gamble in the
Uncertainty
Expected
previous example or $4 for sure. You either have to accept (A)
value the gamble or reject it (R):
Expected
utility
Attitudes
toward Risk
Discussion
OR
H T
Accept $10 $0
Reject $4 $4
15 / 41
Representing More Complex Gambles in Tree Form
Expected value Imagine you are given a choice between the gamble in the
2019-09-20 previous example or $4 for sure. You either have to accept (A)
the gamble or reject it (R):
Form H T
Accept $10 $0
Reject $4 $4
Expected
opportunity to play the gamble above. If it come up heads then
value yes, otherwise no :
Expected
utility
Attitudes
toward Risk
Discussion
OR
HH ¬HH
Gamble $10 $0
16 / 41
Representing More Complex Gambles in Tree Form:
One More Example
The key to analyzing more complex, multi-stage gambles like this is to use
one of the and rules to construct a simpler one.
More Formally
Introduction
Uncertainty
• In general, there can be more than two acts or more than
Expected two states. So, we end up with a matrix like the one
value
below:
Expected
utility
S1 S2 ... Sn
Attitudes
toward Risk
A1 C11 C12 ... C1n
Discussion .. .. .. .. ..
. . . . .
Am Cm1 Cm2 ... Cmn
• Given the decision problem in the matrix above, the
expected value of act i is , EV (Ai )
P r(S1 ) × Ci1 + P r(S2 ) × Ci2 + . . . + P r(Sn ) × Cin
• P r(Sj ) is the probability that state Sj obtains.
• Cij is the consequence of choosing Ai in state Sj .
• Act Ai is optimal if it results in maximum expected value.
17 / 41
More Formally
Expected value
2019-09-20
• In general, there can be more than two acts or more than
two states. So, we end up with a matrix like the one
below:
S1 S2 ... Sn
A1 C11 C12 ... C1n
.. .. .. .. ..
More Formally Am
. .
Cm1
.
Cm2
.
...
.
Cmn
• Given the decision problem in the matrix above, the
expected value of act i is , EV (Ai )
P r(S1 ) × Ci1 + P r(S2 ) × Ci2 + . . . + P r(Sn ) × Cin
• P r(Sj ) is the probability that state Sj obtains.
• Cij is the consequence of choosing Ai in state Sj .
• Act Ai is optimal if it results in maximum expected value.
One more example:You are offered the following gamble: if a (fair) coin
comes up heads, you receive $10; if the coin comes up tails, you pay $10.
What is the expected value of this gamble?
ANSWER:The expected value of this gamble is 1/2 × 10 + 1/2 × (−10) =
0.
More Formally
Expected value
2019-09-20
• In general, there can be more than two acts or more than
two states. So, we end up with a matrix like the one
below:
S1 S2 ... Sn
A1 C11 C12 ... C1n
.. .. .. .. ..
More Formally Am
. .
Cm1
.
Cm2
.
...
.
Cmn
• Given the decision problem in the matrix above, the
expected value of act i is , EV (Ai )
P r(S1 ) × Ci1 + P r(S2 ) × Ci2 + . . . + P r(Sn ) × Cin
• P r(Sj ) is the probability that state Sj obtains.
• Cij is the consequence of choosing Ai in state Sj .
• Act Ai is optimal if it results in maximum expected value.
Example: Suppose somebody intends to roll a fair die and pay you $1 if
she rolls a one, $2 if she rolls a two, and so on. What is the expected value
of this gamble?
ANSWER: 3.5
More Formally
Expected value
2019-09-20
• In general, there can be more than two acts or more than
two states. So, we end up with a matrix like the one
below:
S1 S2 ... Sn
A1 C11 C12 ... C1n
.. .. .. .. ..
More Formally Am
. .
Cm1
.
Cm2
.
...
.
Cmn
• Given the decision problem in the matrix above, the
expected value of act i is , EV (Ai )
P r(S1 ) × Ci1 + P r(S2 ) × Ci2 + . . . + P r(Sn ) × Cin
• P r(Sj ) is the probability that state Sj obtains.
• Cij is the consequence of choosing Ai in state Sj .
• Act Ai is optimal if it results in maximum expected value.
Exercise, Lotto 6/49, cont. Suppose a Lotto 6/49 ticket costs $1 and that
the winner will receive $1,000,000. What does the probability of winning
need to be for this lottery to be actuarially fair, that is, for its price to
equal its expected value?
ANSWER: 1/1,000,000
Exercise Warranties A tablet computer costs $325; the optional one-year
warranty, which will replace the tablet computer at no cost if it breaks,
costs $79. What does the probability p of the tablet computer breaking
need to be for the expected value of purchasing the optional warranty to
equal the expected value of not purchasing it?
ANSWER: The probability of losing 325 is p. The p of losing nothing=0.
Thus, −325p + 0(1 − p) = −79 ⇒ p = −79/ − 325 ≈ 0.24
Some Important Applications of Expected Value
Introduction
Uncertainty
Expected
(a) If a parking ticket costs $30, what does the probability of
utility getting a ticket need to be for the expected value of
Attitudes parking legally to equal the expected value of parking
toward Risk
illegally?
Discussion
(b) If the parking ticket costs $100, what does the probability
need to be for the expected value of parking legally to
equal the expected value of parking illegally?
(c) What if the ticket costs $10?
(d) Given what you pay for parking and given what parking
fines are in your area, what does the probability of getting
a ticket need to be for the expected value of parking
legally to equal the expected value of parking illegally?
18 / 41
Some Important Applications of Expected Value
Expected value
2019-09-20 • Parking Example: Assume that parking legally costs $5.
(a) If a parking ticket costs $30, what does the probability of
getting a ticket need to be for the expected value of
parking legally to equal the expected value of parking
illegally?
Some Important Applications of Expected (b) If the parking ticket costs $100, what does the probability
need to be for the expected value of parking legally to
equal the expected value of parking illegally?
(c) What if the ticket costs $10?
Value (d) Given what you pay for parking and given what parking
fines are in your area, what does the probability of getting
a ticket need to be for the expected value of parking
legally to equal the expected value of parking illegally?
ANSWERS
(a) Assume the probability of getting a ticket when when you park
illegally is p. Then p × (−$30) + (1 − p) × $0 = −$5 ⇒ p = 1/6.
This means that if the probability of getting a ticket is 1/6, the
expected values are identical. If p is greater than 1/6, the expected
value of parking legally is greater than the expected value of parking
illegally; if p is lower than 1/6, the expected value of parking legally
is smaller than the expected value of parking illegally.
(b) p × (−$100) + (1 − p) × $0 = −$5 ⇒ p = 1/20
(c) p × (−$10) + (1 − p) × $0 = −$5 ⇒ p = 1/2
Some Important Applications of Expected Value
Introduction
Uncertainty
19 / 41
Some Important Drawbacks of Using Expected
Value as a Decision Criteria
Introduction
Uncertainty
Expected
value
• Unfortunately, when used as a guide in life, expected-value
Expected
calculations have drawbacks.
utility – Obviously, we can only compute expected values when
Attitudes consequences can be described in terms of dollars, lives
toward Risk
lost, or similar. The definition of expected value makes no
Discussion
sense if the consequences Cij are not expressed in
numbers.
– Under many realistic conditions, expected-value
considerations give apparently perverse advice, and
therefore cannot serve as a general guide to
decision-making in real life. The St. Petersburg Paradox
explained on the next slide is one example of such
conditions.
20 / 41
Some Important Drawbacks of Using Expected
Value as a Decision Criteria
Expected value
2019-09-20 • Unfortunately, when used as a guide in life, expected-value
calculations have drawbacks.
– Obviously, we can only compute expected values when
consequences can be described in terms of dollars, lives
lost, or similar. The definition of expected value makes no
Some Important Drawbacks of Using Expected sense if the consequences Cij are not expressed in
numbers.
– Under many realistic conditions, expected-value
considerations give apparently perverse advice, and
For the second sub-bullet point above use the following example: what to
do if you have 30 minutes in a casino before the mafia comes after you to
reclaim your debts. Assuming that you will be in deep trouble unless you
come up with, say,$10,000 before they show up, gambling can be a very
reasonable thing to do, even if the expected values are low.
Some Important Drawbacks of Using Expected
Value as a Decision Criteria: St Petersburg Paradox
Introduction
Uncertainty
Expected
value
• St. Petersburg Paradox A gamble is resolved by tossing
Expected an unbiased coin as many times as necessary to obtain
utility
heads. If it takes only one toss, the payoff of the gamble is
Attitudes
toward Risk $2; if it takes two tosses, it is $4; if it takes three, it is $8;
Discussion and so forth.
H TH TTH ...
St. Peterburg Gamble $2 $4 $8 ...
21 / 41
Some Important Drawbacks of Using Expected
Value as a Decision Criteria: St Petersburg Paradox
Expected value
2019-09-20 • St. Petersburg Paradox A gamble is resolved by tossing
an unbiased coin as many times as necessary to obtain
heads. If it takes only one toss, the payoff of the gamble is
$2; if it takes two tosses, it is $4; if it takes three, it is $8;
and so forth.
Value as a Decision Criteria: St Petersburg • How much would you be willing to pay for this gamble?
What is the expected value of the gamble?
Paradox
The Answer to the second question is as follows:
the probability of getting heads on the first flip (H) is 1/2; the probability
of getting tails on the first flip and heads on the second (TH) is 1/4; the
probability of getting tails on the first two flips and heads on the third
(TTH) is 1/8; and so on. Thus, the expected value of the gamble is:
1/2 × $2 + 1/4 × $4 + 1/8 × $8 + =$1 ˙ + $1 + · · · = ∞
Expected Utility
Introduction
Uncertainty • How does one resolve the St. Petersburg Paradox. There are
Expected few approaches. Daniel Bernoulli’s (a Swiss mathematician and
value physicist) solution lay in the realization that people’s marginal
Expected
utility
utility of additional unit of money is diminishing. This means
that the more money you acquire, the less extra utility you get
Attitudes
toward Risk from additional money.
Discussion • Following Bernoulli, let’s assume that the marginal utility for
money is diminishing. Mathematically, the utility of a given
amount of money x might equal the logarithm (say to base 10)
of x, so that u(x) = log(x).
• If so, we can transform the St. Petersburg gamble into a table in
which consequences are expressed in utilities instead of dollars:
H TH TTH ...
St. Peterburg Gamble log(2) log(4) log(8) ...
22 / 41
Expected Utility
Introduction
Uncertainty • How does this help? We can compute the expected utility of
Expected the gamble.
value
• The expected utility of a gamble is the amount of utility you
Expected
utility can expect to gain on the average, in the long run, when you
Attitudes play the gamble.
toward Risk
• In the case of the St Petersburg gamble, the expected utility is:
Discussion
1 1 1
2 × log(2) + 4 × log(4) + 8 × log(8) + · · · ≈ 0.602 < ∞
• This way, the expected utility of the St Petersburg gamble is
well-defined and finite.
• Examples like the St Petersburg paradox suggest that
expected-utility maximization is both a better guide to behavior,
and a better description of actual behavior, than expected-value
maximization. That is, the theory of expected utility is a better
normative theory, and a better descriptive theory, than the
theory of expected value.
23 / 41
Expected Utility
Introduction
Uncertainty
Expected
• Generally, given a decision matrix
value
Expected S1 S2 ... Sn
utility
24 / 41
Example
Introduction
Uncertainty
Expected
value
√
Expected
utility
• Suppose your utility function is u(x) = x. Consider the
Attitudes
following two games:
toward Risk
1. You have a 1/4 chance of winning $25 and a 3/4 chance
Discussion of winning $1.
2. You have a 2/3 chance of winning $9 and a 1/3 chance of
winning $4.
• What is the expected value and the expected utility of
these games?
25 / 41
Example
Expected utility
2019-09-20 • Suppose your utility function is u(x) =
√
x. Consider the
following two games:
1. You have a 1/4 chance of winning $25 and a 3/4 chance
of winning $1.
ANSWER:
√ √
1. 1/4 × 25 + 3/4 × 1 = 1/4 × 5 + 3/4 × 1 = 8/4 = 2
√ √
2. 2/3 × 9 + 1/3 × 4 = 2/3 × 3 + 1/3 × 2 = 2 + 2 = 4
One More Example
Introduction
Uncertainty
Expected
√
value • Suppose your utility function is u(x) = x. Consider the
Expected
utility
following proposition considered earlier:
Attitudes 1. Accept the gamble and get a 50% chance of winning $9
toward Risk and a 50% chance of winning $0.
Discussion 2. Reject the gamble and get $4 for sure.
• Should you accept or reject the gamble?
• Now suppose instead that your utility function is
u(x) = x2 . What is the expected utility of accepting the
gamble. What is the expected utility of rejecting the
gamble? What should you do?
26 / 41
One More Example
Expected utility
2019-09-20 √
• Suppose your utility function is u(x) = x. Consider the
following proposition considered earlier:
1. Accept the gamble and get a 50% chance of winning $9
and a 50% chance of winning $0.
2. Reject the gamble and get $4 for sure.
Introduction
Discussion
27 / 41
The Shape of the Utility Function, Risk Aversion
and Proneness
Introduction
Uncertainty
Expected
value
Expected • When the curve bends downwards as you move from left
utility
Attitudes
to right, like the utility function u(x) = x0.5 does, the
toward Risk curve is said to be concave.
Discussion
• Conversely, When the curve bends upwards as you move
from left to right, like the utility function u(x) = x2 does,
the curve is said to be convex.
• As we will discuss below, convexity implies risk proneness
while concavity risk aversion.
28 / 41
Attitudes Toward Risk
Introduction
Uncertainty
Expected
value
Attitudes
with comparable expected value, you are risk averse. As
toward Risk we will demonstrate below, you prefer a sure amount over
Discussion
a comparable gamble when the utility function is concave.
• If you prefer the gamble over the sure amount, you are risk
prone. This happens when your utility function is convex.
• If you are indifferent, you are risk neutral. In this case, the
utility function is a straight line.
29 / 41
Graphical Representation: Expected Value,
Expected Utility, and Risk Aversion
Introduction
Uncertainty
Expected
value
Expected
utility
Attitudes
toward Risk
Discussion
33 / 41
Graphical Representation: Expected Value,
Expected Utility, and Risk Aversion
Next time when teaching this material, redo the utility graphs below to
reflect the examples of the gamble above, G=0,0.5,9,0.5 with square root
and squared utility functions.
Graphical Representation: Expected Value,
Expected Utility, and Risk Aversion
Introduction
Uncertainty
Expected
value
Expected
• As the graph on the previous slide demonstrates, for a risk
utility
averse individual, the utility of a certain amount of money
Attitudes
toward Risk that is equal to the expected value of the gamble must be
Discussion greater than the expected utility of the gamble itself.
• For a risk prone individual, the expected utility of the
gamble is greater than the utility of a certain dollar
amount that is equal to the expected value of the gamble
itself. The graph on the next slide demonstrates this
notion.
34 / 41
Graphical Representation: Expected Value,
Expected Utility, and Risk Aversion
Introduction
Uncertainty
Expected
value
Expected
utility
Attitudes
toward Risk
Discussion
35 / 41
What’s the Most You’d Pay for a Gamble:
Certainty Equivalent
Introduction
Introduction
Uncertainty
Expected
value
Expected
utility
Attitudes
toward Risk
Discussion
37 / 41
An Example
Introduction
Uncertainty
38 / 41
An Example
√
– Utility √
of $4 is 4 = √ 2. Expected utility of the gamble is
1/4 × 9 + 3/4 × 1 = 3/4 + 3/4 = 6/4 = 1.5. You would
definitely
√ pick the certain $4. Certainty equivalent of the gamble is
x = 1.5 ⇒ x = 1.52 = 2.25. Thus, even though the expected
value of the gamble is 1/4 × 9 + 3/4 × 1 = 9/4 + 3/4 = 12/4 = 3,
you woud pay only $2.25 for the gamble.
– Utility of $4 is 42 = 16. Expected utility of the gamble is
1/4 × 92 + 3/4 × 12 = 81/4 + 3/4 = 82/4 = 20.5. You would
definitely pick the gamble because it provides you with greater
expected utility. Certainty
√ equivalent of the gamble is
x2 = 20.5 ⇒ x = 20.5 = 4.53. Thus, even though the expected
value of the gamble is 1/4 × 9 + 3/4 × 1 = 9/4 + 3/4 = 12/4 = 3,
you woud pay more than that, $4.53, for the gamble.
An Example
Introduction
Introduction
Uncertainty
Expected
value
• The distinction between risk and uncertainty is far from
Expected sharp. In real life, it may not be obvious whether to treat
utility
a decision as the one under risk or uncertainty or both.
Attitudes
toward Risk • You cannot judge whether a decision was rational or not
Discussion
by examining the outcome alone. A rational decision, as
you know, is a decision that maximizes expected utility
given your beliefs at the time when you make the decision.
Such a decision might lead to adverse outcomes.
• If something bad happens as a result of your decision, that
does not mean you acted irrationally: you may just have
been unlucky.
40 / 41
Discussion
Discussion
2019-09-20 • The distinction between risk and uncertainty is far from
sharp. In real life, it may not be obvious whether to treat
a decision as the one under risk or uncertainty or both.
• You cannot judge whether a decision was rational or not
by examining the outcome alone. A rational decision, as
Discussion you know, is a decision that maximizes expected utility
given your beliefs at the time when you make the decision.
Such a decision might lead to adverse outcomes.
• If something bad happens as a result of your decision, that
does not mean you acted irrationally: you may just have
been unlucky.
• For the first bullet poinr. Sometimes decision theorists use the term
“right” to denote the decisions that lead to the best possible
outcome. The fact that good decisions can have bad outcomes
means that decisions can be rational but wrong. They can also be
irrational but right, as when you do something completely reckless
but see good results anyway; buying a lottery ticket as a means to
get rich and winning truckloads of money might fall in this category.
It goes without saying that we always want to make the right
decision. The problem, of course, is that we do not know ahead of
time which decision is the right one. That is why we aim for the
rational decision – being the one with the greatest expectation of
future utility.
Discussion
Discussion
2019-09-20 • The distinction between risk and uncertainty is far from
sharp. In real life, it may not be obvious whether to treat
a decision as the one under risk or uncertainty or both.
• You cannot judge whether a decision was rational or not
by examining the outcome alone. A rational decision, as
Discussion you know, is a decision that maximizes expected utility
given your beliefs at the time when you make the decision.
Such a decision might lead to adverse outcomes.
• If something bad happens as a result of your decision, that
does not mean you acted irrationally: you may just have
been unlucky.
Introduction
Uncertainty
Expected
value
Attitudes
probabilities to states of the world, that these probabilities
toward Risk
satisfy the axioms of the probability calculus, that people
Discussion
assign utilities to outcomes, and that they choose that
alternative which has the greatest expected utility given
the probabilities and utilities.
• In the next module, we consider some conditions under
which these assumptions appear to fail.
41 / 41