EC201: Microeconomic Principles I

1.8 Decision-making with risk

Decision-making under uncertainty
• Many decisions are made without full knowledge; we regularly choose
between risky alternatives
• Whether to buy a lottery ticket, or insurance
• Which stocks to buy
• Whether to adopt new technology

• We can extend our models of behaviour to understand how agents choose

between risky alternatives – expected utility theory; prospect theory

• Enables us to think about aversion to risk and what judgements are

needed to make choices
2
Decision-making under uncertainty
Although ‘choice under uncertainty’ is used quite loosely, it is useful to
make a distinction between:

• risk: facing an uncertain but quantifiable future

• uncertainty: facing an uncertain and unquantifiable future

3
Decision-making under uncertainty
Reports that say that something hasn’t happened are always
interesting to me, because as we know, there are known knowns;
there are things we know we know. We also know there are known
unknowns; that is to say we know there are some things we do not
know. But there are also unknown unknowns – the ones we don’t
know we don’t know. And if one looks throughout the history of our
country and other free countries, it is the latter category that tend to
be the difficult ones.

Donald Rumsfeld, US Secretary of Defence, February 12, 2002

4
Quantifying a risky future
• How do agents choose between risky alternatives?

• Agents choose between ‘lotteries’

• A lottery is a random variable with a probability distribution:

• List of different outcomes that might happen (in different ‘states of nature’)
• Probability associated with each outcome (or state)

𝑙 = 𝑥! , 𝑥" … 𝑥# ; π! , π" … π#

• Note: people may be able to take actions which affect the likelihood of
different events occurring (LT)
5
Examples of lotteries
• Example 1: flip a coin; Heads you win £10, Tails you lose £1.
1 1
𝑙 = 10, −1; ,
2 2

• Example 2: you have £100 and are considering paying £1 for a lottery ticket.
There is a one in a thousand chance of winning, in which case you win £1000.
Otherwise, you get nothing!
999 1
𝑙 = 99, 1099; ,
1000 1000

• Key questions: what determines whether someone would take these gambles?
Can we rank different lotteries?
6
Evaluating a risky future
• Consider lottery 𝑙 = 𝑥! , 𝑥" … 𝑥# ; π! , π" … π#

• Suppose that each outcome 𝑘 has a utility associated with it: 𝑢 𝑥$

• Then expected utility is:

#

𝐸𝑈 𝑙 = 1 π% 𝑢 𝑥%
%&!

• Expected utility theory: an individual will optimise by choosing the lottery

that maximizes his expected utility
7
You have £100 and can take one (or none) of these lotteries.
! ! ! ! ! !
𝑙! = 21, −19; " , " , 𝑙" = 69, −36; " , " , 𝑙' = 125, −84; " , "
Which do you prefer?

A. Lottery 1?
B. Lottery 2?
C. Lottery 3?
D. None – keep the £100?

8
Example

• Consider the three lotteries:

! ! ! ! ! !
𝑙! = 21, −19; , , 𝑙" = 69, −36; , , 𝑙' = 125, −84; ,
" " " " " "

• There is no right or wrong answer…which you prefer depends on your

preferences and your initial level of wealth

• But people care about the spread of outcomes and are often concerned more
by the downside risk than the upside (more on that later)

9
Example
! ! ! ! ! !
𝑙! = 21, −19; , , 𝑙" = 69, −36; , , 𝑙' = 125, −84; ,
" " " " " "

• Consider David, who has an initial wealth of £100 and his utility function is
U 𝑊 = 𝑊. How would he rank 𝑙! , 𝑙" , 𝑙' ?

! ! !! (
• 𝐸𝑈 𝑙! = 121 + 81 = + = 10
" " " "
! ! !' )
• 𝐸𝑈 𝑙" = " 169 + " 64 = "
+ " = 10.5
• 𝐸𝑈 𝑙' =
!
225 +
!
16 =
!* +
+ = 9.5 Ranks 𝑙" top
" " " "
• If he takes none: 𝑈 100 = 100 = 10

10
Expected utility is subjective
• Outcomes and probabilities often based on individual beliefs about the
future

• People can have different views about what is going to happen

• Sometimes probability is based on an objective process (but often not –

much scope for psychology)
• e.g. a fair coin has a probability of 0.5 heads and 0.5 tails

• We examine outcomes as income, but 𝑥$ could be something multi-

dimensional, which gives rise to utility 𝑢 𝑥$

11
Expected utility is cardinal
• 𝐸𝑈 is a cardinal concept; we rely on the shape of the utility function
• Preferences over lotteries are preserved only with linear transformations
• If David has utility function 𝑢 ; and prefers lottery A to lottery B, then:
# -

𝐸𝑈 𝑙, = 1 π%, 𝑢 𝑥%, > 1 π.% 𝑢 𝑥%. = 𝐸𝑈 𝑙.

%&! %&!

• If Diana has utility function 𝑣 ; = 𝑎 + 𝑏𝑢 ; , then she also prefers lottery A

to lottery B, since:
# -

𝑎 + 𝑏 1 π%, 𝑢 𝑥%, > 𝑎 + 𝑏 1 π.% 𝑢 𝑥%.

%&! %&!
12
Attitudes to risk
• Classification of risk attitudes:
• Risk averse
• Risk loving
• Risk neutral

• Utility function U 𝑊 :
• Risk averse – concave e.g. U 𝑊 = 𝑊
• Risk loving – convex e.g. U 𝑊 = 𝑊 "
• Risk neutral – linear e.g. U 𝑊 = 5𝑊

13
With lottery 2, your wealth will be £169 with probability 1/2 or £64 with
probability 1/2, giving an expected payoff of £116.5
If I offer to swap this uncertain income for a certain amount, what is the
minimum amount you would be willing to accept?

A. Less than £116.5

B. Exactly £116.5
C. More than £116.5

14
Attitudes to risk
• Your certainty equivalent (CE) to a lottery is the certain amount that makes
you indifferent between CE and the lottery

• For a risk averse individual: 𝐶𝐸 𝑙 < 𝐸 𝑙

• For a risk loving individual: 𝐶𝐸 𝑙 > 𝐸 𝑙
• For a risk neutral individual: 𝐶𝐸 𝑙 = 𝐸 𝑙

• If you chose less than £116.5, you are risk averse

• If you chose more than £116.5, you are risk loving
• If you chose exactly £116.5, you are risk neutral

15
Risk Averse ! !
initial wealth £100; 𝑙" = 69, −36; ,
" "
Utility

13 ● U 𝑊 = 𝑊

𝐸𝑈= 10.5 ●
8 ●

Wealth
0 64 116.5 169
=𝐸 𝑙 16
Risk Averse ! !
initial wealth £100; 𝑙" = 69, −36; ,
" "
Utility

13 ● U 𝑊 = 𝑊
𝑈𝐸 𝑙 ●
𝐸𝑈= 10.5 ●
𝑈 𝐸 𝑙 > 𝐸𝑈 𝑙
8 ●
Jensen’s inequality

Wealth
0 64 116.5 169
=𝐸 𝑙 17
Jensen’s inequality: 𝑈 𝐸 𝑙 > 𝐸𝑈 𝑙

For any concave function 𝑈 𝑊 – i.e. for any risk averse

individual:

The utility of the expected value of a lottery is always higher

than the expected utility from playing the lottery

18
Risk Averse initial wealth £100; 𝑙" = 69, −36; ,
! !
" "
Utility
CE < 𝐸 𝑙

13 ● U 𝑊 = 𝑊

𝐸𝑈= 10.5 ● ● Risk Premium =

8 ● 𝐸 𝑙 − 𝐶𝐸 > 0
RP Amount willing to
sacrifice for certainty

Wealth
0 64 CE 116.5 169
=𝐸 𝑙 19
Suppose we gather a group of people and offer them two different lotteries
A and B. Some people prefer A and others prefer B.
Does this necessarily mean those who selected A have a different utility
function to those who prefer B?

A. Yes
B. No

20
Measuring risk aversion
• We can construct measures of risk aversion:
• Exploit the fact that the curvature of the utility function is related to
risk aversion

• (Arrow-Pratt) coefficient of absolute risk aversion:

𝑈′′(𝑊)
𝐴𝑅𝐴 = −
𝑈′(𝑊)

• (Arrow-Pratt) coefficient of relative risk aversion:

𝑈′′(𝑊)
𝑅𝑅𝐴 = −𝑊
𝑈′(𝑊) 21
Example: U 𝑊 = 𝑊
!
U 𝑊 = 𝑊= 𝑊"
!
! /
⇒ U′ 𝑊 = " 𝑊 "

#
! /
⇒ 𝑈 00 𝑊 = − 𝑊 "
+

1 '
/
𝑈 00 𝑊 −4𝑊 " 1
𝐴𝑅𝐴 = − 0 =− ! =
𝑈 𝑊 1 /" 2𝑊
𝑊
2

𝑈 00 𝑊 1
𝑅𝑅𝐴 = −𝑊 0 =
𝑈 𝑊 2
22
Which of the following lotteries do you prefer?
! ! ! ! " *
𝑙! = 10,20,30; ' , ' , ' , 𝑙" = 10,20,30; ) , ) , )

A. Lottery 1
B. Lottery 2
C. I am indifferent
D. I don’t know

23
Which of the following lotteries do you prefer?
! ! ! !
𝑙! = 50, 150; " , " , 𝑙" = 20, 180; " , "

A. Lottery 1
B. Lottery 2
C. I am indifferent
D. I don’t know

24
How can we compare lotteries more generally?
• Can we more broadly say that one lottery is ‘better’ than another?

! ! ! ! " *
• e.g. 𝑙! = 10,20,30; , , , 𝑙" = 10,20,30; , ,
' ' ' ) ) )
• Every expected utility maximizer would prefer 𝑙!
• 𝑙!‘first order stochastically dominates’ 𝑙" (Appendix for the curious)

! ! ! !
• e.g. 𝑙! = 50, 150; , , 𝑙" = 20, 180; ,
" " " "
• 𝑙! is a ‘mean-preserving spread’ of 𝑙"; same mean, wider variance
• Every risk averse expected utility maximizer would prefer 𝑙"
• 𝑙"‘second order stochastically dominates’ 𝑙! (Appendix for the curious)

25
Mean preserving spread 𝑙! = 50, 150; ,
! !
" "
, 𝑙" = 20, 180; ,
! !
" "

Utility

● ● U 𝑊
𝐸𝑈 𝑙" ●
𝐸𝑈 𝑙! ● Every risk averse
● expected utility
maximizer prefers 𝑙!
●

𝐸 𝑙! =
Wealth
0 20 50 𝐸 𝑙" = 100 150 180 26
Mitigating risk
• If people are risk averse then they will be willing to pay to eliminate or reduce
it - creates demand for insurance

• There might also be incentives for reducing risk by sharing it or pooling it with
others (example in class)
• Risk sharing: decision-makers share a single lottery e.g. joint venture
• Risk pooling: decision-makers each take a lottery and share the proceeds
e.g. agricultural cooperative

• To analyse we first need to illustrate choice of a risk averse individual in state

contingent income space
27
State contingent income space
State of 𝑥" 𝑥! = 𝑥"
nature 2
Certainty line

𝑥!
0 State of nature 1 28
State contingent income space
State of 𝑥" 𝑥! = 𝑥"
nature 2 𝐸𝑈 𝑙 = 𝜋! 𝑢 𝑥! + 𝜋" 𝑢 𝑥" Certainty line

● 𝑙 = 𝑥! , 𝑥" ; 𝜋! , 𝜋"

𝑥!
0 State of nature 1 29
State contingent income space
State of 𝑥" 𝑥! = 𝑥"
nature 2 𝐸𝑈 𝑙 = 𝜋! 𝑢 𝑥! + 𝜋" 𝑢 𝑥"

Indifferent between
𝑙' 𝑙! , 𝑙" , and 𝑙'
●
𝐶𝐸
𝑥67 ● Certainty
𝑙" Equivalent 𝐶𝐸 of 𝑙! ,
● 𝑙!
● 𝑙" , and 𝑙'

𝑥!
0 𝑥67 State of nature 1 30
State contingent income space
State of 𝑥" 𝑥! = 𝑥"
nature 2 𝐸𝑈 𝑙 = 𝜋! 𝑢 𝑥! + 𝜋" 𝑢 𝑥"

𝜕𝑢 𝑥!
𝜋!
𝜕𝑥!
𝑀𝑅𝑆 =
𝜕𝑢 𝑥"
𝜋"
𝐶𝐸 𝜕𝑥"
●
8!
𝑙! 𝑀𝑅𝑆 = 8 when 𝑥! = 𝑥"
● "

𝜋"
𝑀𝑅𝑆 =
𝜋! 𝑥!
0 State of nature 1 31
State contingent income space
State of 𝑥" 𝑥! = 𝑥"
nature 2 𝐸𝑈 𝑙 = 𝜋! 𝑢 𝑥! + 𝜋" 𝑢 𝑥"
Expected payoff of 𝑙"

𝐸 𝑙" = 𝜋"𝑥" + 𝜋!𝑥!

𝐸 𝑙"
Hence:
𝐶𝐸 ● 𝐸 𝑙" 𝜋"
● 𝑥! =
𝜋!
− 𝑥"
𝜋!
𝑙!
●

Risk Premium > 0

𝑥!
0 𝑥67 𝐸 𝑙" State of nature 1 32
Insurance
• Buying insurance involves paying a premium and receiving a payout if
the bad state occurs (and nothing if the good state occurs)

• A risk averse agent values risky income less than its expected value:
𝐶𝐸 < 𝐸 𝑙

• If the insurance company is less risk averse than the consumer then it
values the random income being offered at close to its expected value

• Scope for trade; consumer pays a premium to shift risk to the less risk
averse insurance company
33
Insurance
• Suppose Julie, a risk-averse agent, has wealth 𝑤

• With probability 𝜋, she suffers a loss 𝐿

• 𝑙 = 𝑤, 𝑤 − 𝐿; 1 − 𝜋, 𝜋
• e.g. burglary, suffers an illness, flood, fire

• Julie can buy insurance: she pays a premium and receives a payout in the
event of loss

• Key questions: What determines how much insurance she will buy? Will
she fully insure i.e. achieve certainty?

34
Insurance
• Let 𝑝 be the premium for every unit of insurance purchased 𝑋
• If things go well: 𝑥! = 𝑤 − 𝑝𝑋
• If things go badly: 𝑥" = 𝑤 − 𝐿 − 𝑝𝑋 + 𝑋

• Julie chooses how much insurance to buy at price 𝑝

• If she chooses 𝑋 = 𝐿, then she is fully insured

• If things go well: 𝑥! = 𝑤 − 𝑝𝐿
• If things go badly: 𝑥" = 𝑤 − 𝐿 − 𝑝𝐿 + 𝐿 = 𝑤 − 𝑝𝐿

• If she chooses 𝑋 < 𝐿, then she is partly insured

35
Premium p per unit of insurance
State of 𝑥" 𝑥! = 𝑥"
nature 2
Certainty line

If buys X units of
insurance, moves to B
𝐵
●
𝑋 − 𝑝𝑋
𝑤−𝐿 ●
𝑙 = 𝑤, 𝑤 − 𝐿; 1 − 𝜋, 𝜋
𝑝𝑋
𝑥!
0 𝑤 State of nature 1 36
Premium 𝑝 > 𝜋 per unit of insurance
State of 𝑥" 𝑥! = 𝑥"
nature 2

If can choose X, choice set

is green line through B

!"#
Gradient:−
𝐵 #
●
𝑋 − 𝑝𝑋
𝑤−𝐿 ●
𝑙 = 𝑤, 𝑤 − 𝐿; 1 − 𝜋, 𝜋
𝑝𝑋
𝑥!
0 𝑤 State of nature 1 37
 Optimal choice of insurance (𝑝 > 𝜋)
State of 𝑥" 𝑥! = 𝑥"
nature 2

Here Julie is partially insured

!"#
𝑤 − 𝐿 − 𝑝𝑋 ∗ + 𝑋 ∗ ● Gradient:− #

𝑤−𝐿 ●
𝑙 = 𝑤, 𝑤 − 𝐿; 1 − 𝜋, 𝜋

𝑥!
0 𝑤 − 𝑝𝑋 ∗ 𝑤 State of nature 1 38
Optimal choice of insurance if 𝑝 = 𝜋
State of 𝑥" 𝑥! = 𝑥"
nature 2

Here Julie is fully insured!

𝑤 − 𝑝𝐿 ● Gradient:−
!"#
#

𝑤−𝐿 ●
𝑙 = 𝑤, 𝑤 − 𝐿; 1 − 𝜋, 𝜋

𝑥!
0 𝑤 − 𝑝𝐿 𝑤 State of nature 1 39
Key result

• If 𝑝 = 𝜋, full insurance:
• competitive insurance market; “actuarially fair premium”

• If 𝑝 > 𝜋, partial insurance

• monopolistic insurance market

40
Back to behavioural economics
• Expected utility theory is the conventional model of decision-making
with risk

• It is simple and useful in many settings

• Also fertile ground for behavioural economics because of anomalies in

decision-making with risk which seem to violate the core model

• We want to:
• Explore some of these anomalies
• Study an alternative theoretical approach: prospect theory
41
Johnny is a risk-averse, expected utility maximiser; he will always turn down
! !
𝑙 = −10,11; " , " .
! !
What is the biggest 𝑌 for which he will turn down 𝑙 = −100, 𝑌; , ?
" "

A. £110
B. £1,100,000
C. £2.5billion
D. Johnny will reject the bet no matter how high 𝑌 is
E. We cannot say without knowing his utility function

42
Johnny’s behaviour…
• Johnny would turn down any bet - but surely he is insane to turn down a
gamble where he stands to gain £2.5bn!

• Why does he act this way?

• Risk aversion comes solely from the concavity of 𝑈(𝑊)
• The fact that Johnny rejects the small gamble, means his marginal utility for
wealth decreases very rapidly
• Even the chance for enormous wealth gains provides such little marginal utility
that Johnny would not risk anything significant to get these gains

General concern: Under EU theory, risk aversion over modest stakes gives rise
to an absurd degree of risk aversion over very large stakes
43
Johnny’s behaviour…
Utility

●
●

Wealth
0 𝑤 − 10 𝑤 𝑤 + 11 44
Overconfidence
• Often described as the “mother of all psychological biases"
• 93% of drivers in the US think that they are better than average drivers!

• Examples of things that have been blamed on overconfidence:

• Sinking of the Titanic
• The nuclear accident at Chernobyl
• The loss of Space Shuttles Challenger and Columbia
• The sub-prime mortgage crisis of 2008

• General concern: People have beliefs that put too much weight on states of
the world they believe that they “know”
45
Overconfidence
• Think of as “re-weighting” probabilities in favour of certain outcomes

• Biases 𝐸𝑈 leading to excess risk-taking or generally faulty decisions

• Thought to be important for the financial crisis of 2008
• Investors were over-confident about the returns on the financial products they
were creating/marketing
• Moral hazard also important – studied in LT

• Policy response?
• Post 2008: greater regulation of financial products
• Can people be educated to reduce overconfidence? Can some gambles
that people make be banned or taxed?
46
You face the following concurrent decisions:
Decision 1. Choose Between:
A. A sure gain of £240
B. 25% chance to gain £1000 and 75% chance to gain £0

Decision 2. Choose Between:

C. A sure loss of £750
D. 75% chance to lose £1000 and 25% chance to lose £0

Which would you choose?

• A and C
• A and D
• B and C
• B and D 47
Decision 3. Choose Between:
E. 25% chance to gain £240 and 75% chance to lose £760
F. 25% chance to gain £250 and 75% chance to lose £750

Which would you choose?

• E preferred
• F preferred

48
What is going on?
You face the following concurrent decisions:
Decision 1. Choose Between:
A. A sure gain of £240
B. 25% chance to gain £1000 and 75% chance to gain £0

Decision 2. Choose Between:

C. A sure loss of £750
D. 75% chance to lose £1000 and 25% chance to lose £0

The majority of those tested (N = 150) chose A and D

• A and D: 25% chance to gain £240 and 75% chance to lose £760 (E)
• B and C: 25% chance to gain £250 and 75% chance to lose £750 (F)
• And yet… most people prefer F to E in experiments!
General concern: reversal when faced with conjunction of risky prospects 49
The UK is facing an outbreak of a dangerous virus, which is expected to kill 600
people. Two programmes have been proposed to deal with the crisis.

”Lives saved frame”

Programme 1: 200,000 people will be saved
! "
Programme 2: probability that 600,000 people will be saved; probability
' '
that nobody will be saved

”Lives lost frame”

Programme 1: 400,000 people will die
! "
Programme 2: probability that nobody will die ; probability that 600,000
' '
people will die

50
What is going on?
• Adapted from Tversky and Kahneman (1986), “Rational Choice and the
Framing of Decisions”, Journal of Business

• Their results on the Progamme 1 vs Programme 2 decision:

• “Lives saved frame”: N = 152, 72% chose Programme 1
• “Lives lost frame”: N = 155, 22% chose Programme 1

• Renowned for their contributions to the psychology of decision-making

• Kahneman won the 2002 Nobel prize in Economics; Tversky died in 1996
51
Framing and Loss Aversion:
Amos Tversky and Danny Kahneman highlighted the importance of:

• Framing: how you present a choice to decision-makers matters

• People respond positively to words like “live” and negatively to words
like “die”
• Framing can have real effects when governments make risky decisions
e.g. military intervention

• Loss aversion: response to losses is more extreme than to gains

• But what is the reference point?
• If ‘all dead’ is our reference point then lives saved are gains; if ‘all alive’
is the reference point, then lives lost are losses.
52
Prospect theory
• Tversky and Kahneman developed an alternative way of thinking about choice
under uncertainty

• For n states of the world:

#

1 𝑝(π% )𝑉 𝑥% − 𝑟
%&!
where:
• π! , π" , π' … are probabilities as before
• function 𝑝 is a probability weighting function; no need for ∑#%&! 𝑝(π% ) = 1
• 𝑉 𝑥% − 𝑟 is a value function which diverges from utility
• value of outcome 𝑥$ is evaluated relative to a reference point 𝑟 53
Reference points matter…

• But where do they come from?

• Do people choose them or are
they a fixed part of preferences?
• Are they socially determined?
• Can we change them?
• What are the implications?

54
Example:
• Let 𝑈 𝑥 = 𝑥 (risk neutral agent)

• Suppose the value function is:

𝑥−𝑟 𝑖𝑓 𝑥 ≥ 𝑟
𝑉 𝑥−𝑟 =W
𝜆 𝑥−𝑟 𝑖𝑓 𝑥 < 𝑟

where 𝜆 > 1

• What does this look like?

55
The value function
𝑉 𝑥−𝑟
𝑉 𝑥−𝑟 =𝑥−𝑟

𝑥
𝑟
Region of losses Region of gains

𝑉 𝑥−𝑟 =𝜆 𝑥−𝑟 56
Example:
! !
• Consider lottery 𝑙 = 60,40; " , " , where 𝑟 = 50
• Gamble is around the reference point
• Let 𝑝(π$ ) = π$

• In the good state there are gains: 𝑉 𝑥 − 𝑟 = 60 − 50 = 10 > 0

• In the bad state there are losses: 𝑉 𝑥 − 𝑟 = 𝜆 40 − 50 = −10𝜆 < 0

10 − 10𝜆
( 𝑝(π! )𝑉 𝑥! − 𝑟 = <0
2
!
• So rejects the gamble!
57
The value function
𝑉 𝑥−𝑟
𝑥−𝑟
Risk neutral with gambles
above and below 𝑟 𝑟 = 50
Behaves as if risk averse
with gambles around 𝑟 10 ●
−10
𝑥
𝑟 10
Region of losses ● Region of gains

● −10𝜆
𝜆 𝑥−𝑟
58
What have we achieved?
• We have presented the standard expected utility model for choice under
uncertainty

• Allows us to think about:

• how probabilities of future events matter for decision making
• how the shape of the utility function captures risk aversion
• how to compare lotteries

• We have also considered how to mitigate risk through insurance

59
What have we achieved? (continued)
• We have explored a range of possible biases in decision-making that
arising in the context of decision-making with risk
• Risk aversion over modest stakes implies absurd risk aversion over very
large stakes
• People make inconsistent decisions
• People choose differently depending on how choices are framed
• People tend to be loss-averse; reference point matters

• It is an area where policy is particularly active e.g. regulation of financial

products

• In general people find it hard to make reliable probabilistic judgements

60
Appendix 1: First and second order stochastic dominance
(for the curious)
 ! ! !
First order stochastic dominance 𝑙! = 10,20,30; , ,
' ' '
! " *
𝑐𝑑𝑓 𝑙" = 10,20,30; ) , ) , )
1 ●
𝑙!
Lottery whose cdf lies
2: entirely underneath first
3 ● 𝑙" order stochastically
dominates
3:
8 ● Every expected utility
1: ● maximizer prefers 𝑙"
3

1: ●
8
Wealth
0 10 20 30 62
Second order stochastic dominance ! !
𝑙! = 50, 150; " , "
! !
𝑐𝑑𝑓 𝑙" = 20, 180; ,
" "
1 ● ●
𝑙!
Lottery whose cdf crosses
once from underneath
second order
𝑙" stochastically dominates
1:
2 ● ● Every risk averse
expected utility
maximizer prefers 𝑙!

Wealth
0 20 50 150 180 63
Appendix 2: Risk loving and risk neutral preferences
(for completeness and for those interested)
Risk Loving initial wealth £100; 𝑙" = ! !
69, −36; ,
" "

Utility

U 𝑊

𝐸𝑈 ●

●
Wealth
0 64 116.5 169 65
Risk Loving initial wealth £100; 𝑙" = ! !
69, −36; ,
" "

Utility
𝑈𝐸 𝑙 < 𝐸𝑈 𝑙

U 𝑊

𝐸𝑈 ●
𝑈𝐸 𝑙 ●
●
Wealth
0 64 116.5 169 66
Risk Loving initial wealth £100; 𝑙" = ! !
69, −36; ,
" "

Utility

U 𝑊
Risk Premium =
𝐸 𝑙 − 𝐶𝐸 < 0
● Compensation
required to accept
𝐸𝑈 ● ● certainty
RP
●
Wealth
0 64 116.5 CE 169 67
Risk loving
𝑥" 𝑥! = 𝑥"
𝐸𝑈 = 𝜋"𝑢 𝑥" + 𝜋!𝑢 𝑥!

𝑥67 ● 𝐶𝐸
𝐸 𝑙" ●

𝑙"
●

Risk Premium < 0

𝑥!
0 𝐸 𝑙" 𝑥67 68
Risk neutral initial wealth £100; 𝑙" = ! !
69, −36; ,
" "
Utility
U 𝑊
𝑈𝐸 𝑙 = 𝐸𝑈 𝑙
● 𝐶𝐸 = 𝐸 𝑙
Risk Premium =
𝐸𝑈 ● 𝐸 𝑙 − 𝐶𝐸 = 0
=𝑈 𝐸 𝑙
●

Wealth
0 64 116.5 169
= CE = 𝐸 𝑙 69
Risk neutral
𝑥" 𝑥! = 𝑥"
𝐸𝑈 = 𝐸 𝑙" = 𝜋"𝑥" + 𝜋!𝑥!

𝐸 𝑙" = 𝑥$% ● 𝐶𝐸

𝑙"
●

Risk Premium = 0
𝑥!
0 𝐸 𝑙" = 𝑥$% 70