Uncertainty and Information

Dr. Bryony Reich

Topic 3: Attitudes Towards Risk

To avoid distracting others, please turn off or silence your

phone and put it in your bag. Laptops/tablets should be in
airplane mode and not distract others in the class.
Optional Textbook Reading

• The lecture notes draw from Steve Tadelis “Game Theory: An Introduction” Chapter 2

• They also draw from [Link]

• And prior lecture notes.

Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
 Which would you choose?

ENTER A GAMBLE
7 5
WITH PROBABILITY 12 ≈ 0.58 YOU WIN £400 WITH PROBABILITY 12 ≈ 0.42 YOU WIN £1600

RECEIVE £900
Which would you choose?
• Write the example in terms of our uncertainty framework.

• A lottery with three monetary outcomes, 𝑥1 = £400, 𝑥2 = £900, 𝑥3 = £1600, where the probability of an
outcome is denoted 𝑝1 , 𝑝2 , 𝑝3 respectively.

• You have a choice of two different lotteries.

7 5
• Lottery p′ = 𝑝1′ , 𝑝2′ , 𝑝3′ = (12 , 0, 12) or lottery p′′ = 𝑝1′′ , 𝑝2′′ , 𝑝3′′ = 0,1,0 .

• Which one would you choose?

Which would you choose?
• Work out the average monetary payoff of each lottery.

7 5
• Lottery 𝑝′ average monetary payoff is × £400 + × £1600 = £900.
12 12

• Lottery 𝑝′′ average monetary payoff is £900.

• One lottery gives you £900 on average and the other gives you £900 for sure.

• Which do you prefer?

Which would you choose?
• Which lottery you prefer tells us about your attitudes towards risk.

• If you prefer getting £ 900 for sure to a risky gamble where you receive £ 900 on average, then we say you
are risk averse. You dislike risk.

• If you prefer the risky gamble with average payout of £ 900 (where you are more likely to receive £ 400 but
have a chance of £ 1600) to getting £ 900 for sure, then we say you are risk loving. You enjoy the risk!

• If you are indifferent between the two lotteries then we say you are risk neutral.
The St. Petersburg Paradox (Bernoulli)
Question: how much are you willing to pay to play this game?

The Game
• You pay a fixed fee £ X to play the game.
• A fair coin is tossed repeated until it lands on tails.
• The “pot” starts at £ 1 and is doubled each time the coin lands on heads.
• The game ends when the coin first lands on tails.
• You win the money in the pot when the game ends.
The St. Petersburg Paradox
Question: how much are you willing to pay to play this game?

The Game
• You pay a fixed fee £ X to play the game.
• A fair coin is tossed repeated until it lands on tails.
• The “pot” starts at £ 1 and is doubled each time the coin lands on heads.
• The game ends when the coin first lands on tails.
• You win the money in the pot when the game ends.

£1 Toss coin
Game ends.
Pot starts Win £1
at £1
The St. Petersburg Paradox
Question: how much are you willing to pay to play this game?

The Game
• You pay a fixed fee £ X to play the game.
• A fair coin is tossed repeated until it lands on tails.
• The “pot” starts at £ 1 and is doubled each time the coin lands on heads.
• The game ends when the coin first lands on tails.
• You win the money in the pot when the game ends.

£1 Toss coin £2 Toss coin Game ends.

Win £2
Pot starts
at £1
 The St. Petersburg Paradox
Question: how much are you willing to pay to play this game?

The Game
• You pay a fixed fee £ X to play the game.
• A fair coin is tossed repeated until it lands on tails.
• The “pot” starts at £ 1 and is doubled each time the coin lands on heads.
• The game ends when the coin first lands on tails.
• You win the money in the pot when the game ends.

£1 Toss coin £2 Toss coin Toss coin £8 Toss coin

£4 Game ends.
Win £8
Pot starts
at £1
The St. Petersburg Paradox
Question: how much are you willing to pay to play this game?

The Game
• You pay a fixed fee £ X to play the game.
• A fair coin is tossed repeated until it lands on tails.
• The “pot” starts at £ 1 and is doubled each time the coin lands on heads.
• The game ends when the coin first lands on tails.
• You win the money in the pot when the game ends.
• E.g. you win £ 1 if it lands on tails the first toss, £ 2 if it lands on tails the 2nd, £ 4 if it lands on tails the
3rd, and £2𝑘−1 if it lands on tails on the 𝑘th toss.

• How much are you willing to pay to play?

The St. Petersburg Paradox
• Calculate the expected monetary value of this lottery.

• If it lands on tails on the 𝑘th toss you win £2𝑘−1 .

• The probability that the coin lands on tails on the 𝑘th toss is the same as the probability that it lands on
heads 𝑘 − 1 times in a row and then lands on tails the 𝑘th time.
1 𝑘
• The probability of this happening is .
2
1
• Why? Each toss is independent and the probability of any particular side coming up is 2.

∞ 1 𝑘 1 1 1 1 1 1
• The expected monetary value of the lottery is σ𝑘=1 2𝑘−1 = σ∞
𝑘=1 = + + + + … = ∞.
2 2 2 2 2 2 2

• There are very tiny probabilities of winning big amounts of money.

• Are you willing to pay big sums to play? Were you willing to pay even £ 10 to play this game?

• (Bernoulli early on made the point that instead of looking at the expected monetary value of the lottery, we
should be looking at a player’s expected payoffs.)
Concept here: How do people evaluate and choose risk?
• Last class we leant how economics embeds uncertainty into a decision-maker’s problem.

• This class we want to think about how a decision maker evaluates that uncertainty, how much he dislikes it.

• The decision maker’s attitude towards risk is described by his preferences over different lotteries.

• That is, his attitude towards risk is given by how much a decision maker prefers one lottery versus another.
(Recall cardinality matters in expected utility theory.)

• Beware, the outcome needs to be money when we talk about risk.

Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Risk Aversion
• Take a simple lottery with two outcomes X = {𝑥1 , 𝑥2 }, where 𝑝 𝑥1 = 𝑝 and 𝑝 𝑥2 = 1 − 𝑝. Let 𝑢(𝑥) be the
player’s payoff over outcomes 𝑥 ∈ 𝑋.

• The expected monetary value of the lottery is E 𝑥 = 𝑝𝑥1 + 1 − 𝑝 𝑥2 .

• The player’s expected payoff from the lottery is 𝐸[𝑢(𝑥)] = 𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 ).

• Question: what is the player’s payoff from the expected monetary value of the lottery? How would we write
it?

• The player’s payoff from the expected monetary value of the lottery is 𝑢 𝐸 𝑥 = 𝑢 𝑝𝑥1 + 1 − 𝑝 𝑥2 .
Risk Aversion
• Typically, for most people, the following inequality does not hold:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 = 𝑢 𝑝𝑥1 + 1 − 𝑝 𝑥2 .

• Question: What does this equality mean?

Risk Aversion
• Typically, for most people, the following inequality does not hold:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 = 𝑢 𝑝𝑥1 + 1 − 𝑝 𝑥2 .

• Question: What does this equality mean?

• 𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 is the player’s expected payoff from the lottery.

• 𝑝𝑥1 + 1 − 𝑝 𝑥2 is the expected monetary value of the lottery.
• 𝑢 𝑝𝑥1 + 1 − 𝑝 𝑥2 is the player’s payoff from the expected monetary value of the lottery.

• 𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 = 𝑢 𝑝𝑥1 + 1 − 𝑝 𝑥2 means the player is indifferent between playing the lottery

and getting the expected monetary value of the lottery for sure.

• Typically, people are not indifferent. People are risk averse in most situations.

• If a player is risk averse, he prefers the expected value of the lottery for sure to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).
 Risk Aversion: Which would you choose?

ENTER A GAMBLE
7 5
WITH PROBABILITY 12 ≈ 0.58 YOU WIN £400 WITH PROBABILITY 12 ≈ 0.42 YOU WIN £1600

RECEIVE £900
Example: Risk Aversion
• A lottery with three monetary outcomes, 𝑥1 = £400, 𝑥2 = £900, 𝑥3 = £1600, where the probability of an
outcome is denoted 𝑝1 , 𝑝2 , 𝑝3 respectively.

• You have a choice of two different lotteries:

7 5
➢ Lottery p′ = 𝑝1′ , 𝑝2′ , 𝑝3′ = (12 , 0, 12)
➢ Or lottery p′′ = 𝑝1′′ , 𝑝2′′ , 𝑝3′′ = 0,1,0 .

7 5
• Player’s expected payoff from lottery 𝑝′: 12 𝑢 400 + 12 𝑢 1600

• Player’s expected payoff from lottery 𝑝′′: 𝑢 900

Example: Risk Aversion
• Which lottery does the player choose?

7 5
• If 𝑢 400 + 𝑢 1600 = 𝑢 900 the player is risk neutral. He is indifferent between a sure payout of $𝑀 and a
12 12
lottery that gives an expected monetary payout of $𝑀.

7 5
• If 12 𝑢 400 + 12 𝑢 1600 < 𝑢 900 the player is risk averse. He strictly prefers a sure payout of $𝑀 to a lottery
that gives an expected monetary payout of $𝑀.

7 5
• If 12 𝑢 400 + 12 𝑢 1600 > 𝑢 900 the player is risk loving. He strictly prefers a lottery that gives an expected
monetary payout of $𝑀 to a sure payout of $𝑀.
Risk Aversion
• If a player is risk averse, he prefers the expected value of the lottery for sure to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• But this inequality is the definition of a concave function!!

• This means that: A player is risk averse if and only if 𝑢() is a concave function.

• So we are saying a player is risk averse if he has concave utility over money.

• To understand this well, let’s recap concave functions…

Risk Aversion
• If a player is risk averse, he prefers the sure value of the lottery to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• 𝑢(𝑥) is a concave function if no line segment joining any two

points of the graph lies above the graph.

𝑢(𝑥)

𝑥
Risk Aversion
• If a player is risk averse, he prefers the sure value of the lottery to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• 𝑢(𝑥) is a concave function if no line segment joining any two

points of the graph lies above the graph.

𝑢(𝑥)
• Using our graph, take 2 points, say 𝑥1 and 𝑥2 . 𝑢(𝑥2 )

• Draw the line segment between 𝑥1 and 𝑥2 . ҧ 1 + (1 − 𝜆)𝑥

ҧ 2)
𝑢(𝜆𝑥

• This line draws 𝜆𝑢 𝑥1 + 1 − 𝜆 𝑢 𝑥2 , ҧ 𝑥1 + 1 − 𝜆ҧ 𝑢(𝑥2 )

𝜆𝑢
for all 𝜆 ∈ 0,1 .
𝑢(𝑥1 )
• We can see that for all 𝜆 ∈ 0,1 ,
𝑢 𝜆𝑥1 + 1 − 𝜆 𝑥2 ≥ 𝜆𝑢 𝑥1 + 1 − 𝜆 𝑢 𝑥2 .

𝑥1 ҧ 1 + (1 − 𝜆)𝑥
𝜆𝑥 ҧ 2 𝑥2 𝑥
Risk Aversion
• If a player is risk averse, he prefers the sure value of the lottery to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• 𝑢(𝑥) is a concave function if no line segment joining any two

points of the graph lies above the graph.

𝑢(𝑥)
• This is true for any two points 𝑥1 , 𝑥2 on 𝑢(𝑥2 )
our graph.
ҧ 1 + (1 − 𝜆)𝑥
𝑢(𝜆𝑥 ҧ 2)
• Therefore, for all 𝑥1 , 𝑥2 ∈ 𝑋 and for all
𝜆 ∈ 0,1 , ҧ 𝑥1 + 1 − 𝜆ҧ 𝑢(𝑥2 )
𝜆𝑢
𝑢 𝜆𝑥1 + 1 − 𝜆 𝑥2
≥ 𝜆𝑢 𝑥1 + 1 − 𝜆 𝑢 𝑥2 .
𝑢(𝑥1 )

• Thus a player with a concave payoff

function always prefers the sure value
of the lottery to the lottery itself. 𝑥1 ҧ 1 + (1 − 𝜆)𝑥
𝜆𝑥 ҧ 2 𝑥2 𝑥
Risk Aversion
• If a player is risk averse, he prefers the sure value of the lottery to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• 𝑢(𝑥) is a concave function if no line segment joining any two

points of the graph lies above the graph.

• Plug in the lottery above: outcomes 𝑢(𝑥)

X = {𝑥1 , 𝑥2 }, with 𝑝 𝑥1 = 𝑝 and 𝑢(𝑥2 )
𝑝 𝑥2 = 1 − 𝑝.
𝑢(𝑝𝑥1 + (1 − 𝑝)𝑥2 )
• You can see that
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 )
< 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 )
𝑢(𝑥1 )

𝑥1 𝑝𝑥1 + (1 − 𝑝)𝑥2 𝑥2 𝑥
Risk Aversion
• If a player is risk averse, he prefers the sure value of the lottery to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• A player is risk averse if and only if 𝑢() is a concave function.

• What is the intuition for concave payoffs and 𝑢(𝑥)

risk aversion? 𝑢(𝑥2 )

• A concave payoff function implies a 𝑢(𝜆𝑥1 + (1 − 𝜆)𝑥2 )

diminishing marginal utility of wealth – the
benefit from each additional unit of wealth 𝜆𝑢 𝑥1 + 1 − 𝜆 𝑢(𝑥2 )
decreases as you have more.
𝑢(𝑥1 )
• Thus, I prefer a small amount of wealth for
sure rather than a gamble with the same
value. The reason is that the benefit from a
big gain is not large enough to outweigh the 𝑥1 𝜆𝑥1 + (1 − 𝜆)𝑥2 𝑥2 𝑥
cost of the potential loss from the gamble.
Risk Aversion
• If a player is risk averse, he prefers the sure value of the lottery to the lottery itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 < 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

Definition.
A player is risk averse if for any (non degenerate) lottery on 𝑋, the player prefers the expected monetary value of the
lottery to the lottery itself. That is for any lottery 𝑢 𝐸 𝑥 ≥ 𝐸[𝑢(𝑥)].
Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Certain Equivalent
• How much is a lottery worth to a player? The certain equivalent measure this.

• The certain equivalent (for a given lottery) is the amount of money that makes the player indifferent between
getting the certain equivalent for sure or participating in the lottery.

• In notation. Take a risky lottery denoted 𝓁. (Notice we write |𝓁 below. This makes clear we are referring to
lottery 𝓁.)

• The expected monetary value of the lottery 𝓁 is denoted 𝐸[𝑥|𝓁].

• The player’s expected payoff from the lottery 𝓁 is 𝐸 𝑢 𝑥 |𝓁

• The certain equivalent is the amount of money 𝐶𝐸(𝓁) such that the player’s payoff from getting this amount
for sure is equal to the player’s expected payoff from participating the lottery 𝓁. That is, 𝐶𝐸(𝓁) satisfies
𝑢 𝐶𝐸 𝓁 = 𝐸 𝑢 𝑥 |𝓁 .

• A player is risk averse if the certain equivalent is less than the expected monetary value of the lottery,
𝐶𝐸(𝓁) ≤ 𝐸[𝑥|𝓁], for all lotteries 𝓁.
Certain Equivalent
• Consider again the simple lottery with two outcomes X = {𝑥1 , 𝑥2 }, where 𝑝 𝑥1 = 𝑝 and 𝑝 𝑥2 = 1 − 𝑝. Let
𝑢(𝑥) be the player’s payoff over outcomes 𝑥 ∈ 𝑋.

• This is the same as the previous graph, but I have plugged in the probability 𝑝.

• Illustrate the certain equivalent on this graph.

𝑢(𝑥)
𝑢(𝑥2 )

𝑢(𝑝𝑥1 + (1 − 𝑝)𝑥2 )

𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 )

𝑢(𝑥1 )

𝑥1 𝑝𝑥1 + (1 − 𝑝)𝑥2 𝑥2 𝑥
Certain Equivalent
• Consider again the simple lottery with two outcomes X = {𝑥1 , 𝑥2 }, where 𝑝 𝑥1 = 𝑝 and 𝑝 𝑥2 = 1 − 𝑝. Let
𝑢(𝑥) be the player’s payoff over outcomes x ∈ 𝑋.

• This is the same as the previous graph, but I have plugged in the probability 𝑝.

• Illustrate the certain equivalent on this graph.

• The certain equivalent satisfies 𝑢 𝐶𝐸 = 𝐸 𝑢 𝑥 . 𝑢(𝑥)

𝑢(𝑥2 )

𝑢(𝑝𝑥1 + (1 − 𝑝)𝑥2 )

𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 )

𝑢(𝑥1 )

𝑥1 𝐶𝐸 𝑝𝑥1 + (1 − 𝑝)𝑥2 𝑥2 𝑥
Certain Equivalent
• Consider again the simple lottery with two outcomes X = {𝑥1 , 𝑥2 }, where 𝑝 𝑥1 = 𝑝 and 𝑝 𝑥2 = 1 − 𝑝. Let
𝑢(𝑥) be the player’s payoff over outcomes x ∈ 𝑋.

• This is the same as the previous graph, but I have plugged in the probability 𝑝.

• Illustrate the certain equivalent on this graph.

• The certain equivalent satisfies 𝑢 𝐶𝐸 = 𝐸 𝑢 𝑥 . 𝑢(𝑥)

𝑢(𝑥2 )

𝑢 𝐸(𝑥) = 𝑢(𝑝𝑥1 + (1 − 𝑝)𝑥2 )

𝑢 𝐶𝐸 = 𝐸[𝑢(𝑥)] = 𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 )

𝑢(𝑥1 )

𝑥1 𝐶𝐸 𝑝𝑥1 + (1 − 𝑝)𝑥2 𝑥2 𝑥
= E(x)
 Certain Equivalent
• Consider again the simple lottery with two outcomes X = {𝑥1 , 𝑥2 }, where 𝑝 𝑥1 = 𝑝 and 𝑝 𝑥2 = 1 − 𝑝. Let
𝑢(𝑥) be the player’s payoff over outcomes x ∈ 𝑋.

• This is the same as the previous graph, but I have plugged in the probability 𝑝.

• Illustrate the certain equivalent on this graph.

• The certain equivalent satisfies 𝑢 𝐶𝐸 = 𝐸 𝑢 𝑥 . 𝑢(𝑥)

𝑢(𝑥2 )

player’s payoff from the expected 𝑢 𝐸(𝑥) = 𝑢(𝑝𝑥1 + (1 − 𝑝)𝑥2 )

monetary value of the lottery
𝑢 𝐶𝐸 = 𝐸[𝑢(𝑥)] = 𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 )

𝑢(𝑥1 )
player’s expected payoff from the lottery

𝑥1 𝐶𝐸 𝑝𝑥1 + (1 − 𝑝)𝑥2 𝑥2 𝑥
= E(x)
expected monetary value of the lottery
Example: What is the Certain Equivalent in the St. Petersburg Paradox?
Question: how much are you willing to pay to play this game?

The Game
• You pay a fixed fee £ X to play the game.
• A fair coin is tossed repeated until it lands on tails.
• The “pot” starts at £ 1 and is doubled each time the coin lands on heads.
• The game ends when the coin first lands on tails.
• You win the money in the pot when the game ends.
Example: What is the Certain Equivalent in the St. Petersburg Paradox?
• Calculate the expected monetary value of this lottery.

• The probability that the coin lands on tails on the 𝑘th toss is the same as the probability that it lands on
heads 𝑘 − 1 times in a row and then lands on tails the 𝑘th time.
1 𝑘
• The probability of this happening is .
2
1
• Why? Each toss is independent and the probability of any particular side coming up is .
2

• If it lands on tails on the 𝑘th toss you win $2𝑘−1 .

∞ 1 𝑘 1
• The expected monetary value of the lottery is σ𝑘=1 2𝑘−1 = σ∞
𝑘=1 = ∞.
2 2

• There are very tiny probabilities of winning big amounts of money.

• Are you willing to pay big sums to play? Were you even willing to pay £ 10 to play this game?

• (Bernoulli early on made the point that instead of looking at the expected monetary value of the lottery, we
should be looking at expected payoffs.)
Example: What is the Certain Equivalent in the St. Petersburg Paradox?
• Calculate the certain equivalent for this lottery.

• You already provided your own certain equivalent. Recall, the amount £ X you are willing to pay to play the
game.

• The certain equivalent 𝐶𝐸 is the amount of money that makes the player indifferent between getting the
certain equivalent for sure or participating in the lottery, that is 𝑢 𝐶𝐸 = 𝐸 𝑢 𝑥 .

• The maximum you would be willing to pay to play the game is your certainty equivalent (and you would be
even happier to pay a lower amount).
 Example: What is the Certain Equivalent in the St. Petersburg Paradox?
• Calculate the certain equivalent for this lottery.

• Let’s work out the certain equivalent for a given player. To do this we need to know his payoff function over
money 𝑢(𝑥).

• Suppose 𝑢 𝑥 = log 2 x . (This is the binary logarithm where y = log 2 x ⟺ 2𝑦 = 𝑥)

• Calculate the player’s expected payoff from the lottery.

1 1 1 1
𝐸𝑢 𝑥 = log 2 1 + log 2 (2) + log 2 (4) + log 2 (8) + ⋯
2 4 8 16
∞ 𝑘 ∞
1 log 2 2𝑘−1
=෍ log 2 2𝑘−1 =෍
2 2𝑘
𝑘=1 𝑛=1

∞ ∞ ∞ 𝑘 𝑘
(k − 1)log 2 (2) 𝑘−1 1 1
=෍ = ෍ = ෍ 𝑘 −
o Using log 2 ( 2) = 1
2𝑘 2𝑘 2 2
𝑘=1 𝑘=1 𝑘=1
o Log power rule
log 𝑏 (𝑚𝑛 ) = 𝑛 × 𝑙𝑜𝑔𝑏 (𝑚)
Example: What is the Certain Equivalent in the St. Petersburg Paradox?
• Calculate the certain equivalent for this lottery.

∞ 1 𝑘 1 𝑘
• Next work out the geometric series in the expected payoff: 𝐸 𝑢 𝑥 = σ𝑘=1 𝑘 2 − 2
.

• Using properties of geometric series if 𝑟 < 1

∞ ∞
𝑘
1 𝑘
𝑟
෍𝑟 = ෍𝑟 −1= −1=
1−𝑟 1−𝑟
𝑘=1 𝑘=0
∞ ∞
𝑘−1
1 𝑘
𝑟
෍ 𝑘𝑟 = 2
⟺ ෍ 𝑘𝑟 = 2
1−𝑟 1−𝑟
𝑘=1 𝑘=1
• Plugging this into the expression we derived for expected utility we have
∞ 𝑘 𝑘1 1
1 1
𝐸 𝑢 𝑥 = ෍𝑘 − = 2−2 =2−1=1
2 2 1 1
𝑘=1 4 2
• His expected utility from the game is 1 and so the certain equivalent satisfies 𝑢 𝐶𝐸 = log 2 CE = 1.

• Since y = log 2 x ⟺ 2𝑦 = 𝑥 then 21 = 2 and so 𝐶𝐸 = 2. He is willing to pay £2 to play.

Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Risk Premium
• The risk premium measures how much the player needs to be paid to take on the risk.

• Take a risky lottery 𝓁.

• The expected monetary value of the lottery is 𝐸[𝑥|𝓁].

• The player’s expected payoff from the lottery is 𝐸 𝑢 𝑥 |𝓁

• The certain equivalent is the amount of money 𝐶𝐸(𝓁) that satisfies 𝑢 𝐶𝐸 𝓁 = 𝐸 𝑢 𝑥 |𝓁 .

• The risk premium is 𝑟 = 𝐸 𝑥 𝓁 − 𝐶𝐸 𝓁 .

Risk Premium
• Consider again the simple lottery with two outcomes X = {𝑥1 , 𝑥2 }, where 𝑝 𝑥1 = 𝑝 and 𝑝 𝑥2 = 1 − 𝑝. Let
𝑢(𝑥) be the player’s payoff over outcomes x ∈ 𝑋.

• This is the previous graph, but I have plugged in probability 𝑝.

• Illustrate the risk premium on this graph.

• The risk premium is 𝑟 = 𝐸[𝑥] − 𝐶𝐸. 𝑢(𝑥)

𝑢(𝑥2 )

𝑢(𝑝𝑥1 + (1 − 𝑝)𝑥2 )

𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 )

𝑢(𝑥1 )

𝑥1 𝐶𝐸 𝑝𝑥1 + (1 − 𝑝)𝑥2 𝑥2 𝑥
Risk Premium
• Consider again the simple lottery with two outcomes X = {𝑥1 , 𝑥2 }, where 𝑝 𝑥1 = 𝑝 and 𝑝 𝑥2 = 1 − 𝑝. Let
𝑢(𝑥) be the player’s payoff over outcomes x ∈ 𝑋.

• This is the previous graph, but I have plugged in probability 𝑝.

• Illustrate the risk premium on this graph.

• The risk premium is 𝑟 = 𝐸[𝑥] − 𝐶𝐸. 𝑢(𝑥)

𝑢(𝑥2 )

𝑢(𝑝𝑥1 + (1 − 𝑝)𝑥2 )

𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢(𝑥2 )

Intuition. 𝑢(𝑥1 )
• Risk is costly.
• The risk premium is how much extra you have to pay me for
me to be willing to take on the risk.
𝑥1 𝐶𝐸 𝑝𝑥1 + (1 − 𝑝)𝑥2 𝑥2 𝑥
𝑟 = 𝐸[𝑥] − 𝐶𝐸
Example: Risk Premium
• What is the amount $x for which the car salesman is indifferent between getting $100,000 on average per
year and getting $x per year for sure?

• Job 1: $100,000 on average (some years you make a lot some years very little)
• Job 2: $x for sure

• This is asking what is his expected utility from the risky job that gives him $100,000 on average and what
guaranteed salary $x gives him the same utility.

• The value $x is the certain equivalent.

• Suppose I am indifferent between getting $100,000 on average per year in the risky job and getting $80,000
per year for sure. Then $80,000 is the certain equivalent.

• If x < 100,000, the salesman is risk averse.

• In the real world, why don’t people like this kind of risk on their salary?
Example: Risk Premium

• What is the amount $x for which the car salesman is indifferent between getting $100,000 on average per
year and getting $x per year for sure?

• Suppose I am indifferent between getting $100,000 on average per year in the risky job and getting $80,000
per year for sure. Thus $80,000 is the certain equivalent.

• The amount $100,000 - $x is the “risk premium” of the $100,000 pay package, r = 𝐸[𝑥] − 𝐶𝐸

• What does this mean?

• The firm has to compensate the worker for the risk costs with this risk premium. The firm has to pay the
worker $20,000 more on average to get him to take the risky job over a certain one.

• This is why, other things equal, jobs that involve more financial risk pay more overall. Those jobs need to pay
a risk premium (since typically most people are risk averse when it comes to their salary).
Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Measuring Risk Aversion – the Certain Equivalent
• The certain equivalent (and risk premium) is one way to measure the degree of a player’s risk aversion.

• Take two players. Player 1 has payoff function 𝑢(⋅) over outcomes and Player 2 has payoff function 𝑣(⋅).

• If the certain equivalent for Player 1 is lower than the certain equivalent for Player 2, 𝑪𝑬𝟏 < 𝑪𝑬𝟐 who is
more risk averse?
Measuring Risk Aversion – the Certain Equivalent
• The certain equivalent (and risk premium) is one way to measure the degree of a player’s risk aversion.

• Take two players. Player 1 has payoff function 𝑢(⋅) over outcomes and Player 2 has payoff function 𝑣(⋅).

• If the certain equivalent for Player 1 is lower than the certain equivalent for Player 2, 𝑪𝑬𝟏 < 𝑪𝑬𝟐 who is
more risk averse?

• The certain equivalent (for a given lottery) is the amount of money that makes the player indifferent between
getting the certain equivalent for sure or participating in the lottery, 𝑢 𝐶𝐸 𝓁 = 𝐸 𝑢 𝑥 |𝓁 . It tells us how
much a lottery is worth to the player.

• For Player 1 the lottery is worth less. He dislikes risk more. He is more risk averse.

• The risk premium, 𝑟 = 𝐸[𝑥] − 𝐶𝐸, gives us the same information. Player 1’s risk premium is 𝐸 𝑥 − 𝐶𝐸1 >
𝐸 𝑥 − 𝐶𝐸2 .

• What does this mean?

Measuring Risk Aversion – the Certain Equivalent
• The certain equivalent (and risk premium) is one way to measure the degree of a player’s risk aversion.

• Take two players. Player 1 has payoff function 𝑢(⋅) over outcomes and Player 2 has payoff function 𝑣(⋅).

• If the certain equivalent for Player 1 is lower than the certain equivalent for Player 2, 𝑪𝑬𝟏 < 𝑪𝑬𝟐 who is
more risk averse?

• The certain equivalent (for a given lottery) is the amount of money that makes the player indifferent between
getting the certain equivalent for sure or participating in the lottery, 𝑢 𝐶𝐸 𝓁 = 𝐸 𝑢 𝑥 |𝓁 . It tells us how
much a lottery is worth to the player.

• For Player 1 the lottery is worth less. He dislikes risk more. He is more risk averse.

• The risk premium, 𝑟 = 𝐸[𝑥] − 𝐶𝐸, gives us the same information. Player 1’s risk premium is 𝐸 𝑥 − 𝐶𝐸1 >
𝐸 𝑥 − 𝐶𝐸2 .

• What does this mean? Player 1 has to be paid more to be willing to take on the risk than Player 2.
Measuring Risk Aversion – Absolute Risk Aversion
• There are other ways to measure risk aversion.

• Suppose 𝑢(⋅) is a twice differentiable utility function.

Definition.
𝑢′′ (𝑥)
The (Arrow-Pratt) coefficient of absolute risk aversion is 𝐴 𝑥 = − 𝑢′ (𝑥) .

Example.
• Take the concave utility function 𝑢 𝑥 = 1 − 𝑒 −𝑎𝑥 .
• First derivative: 𝑢′ 𝑥 = 𝑎𝑒 −𝑎𝑥 .
• Second derivative: 𝑢′′ 𝑥 = −𝑎2 𝑒 −𝑎𝑥
𝑎2 𝑒 −𝑎𝑥
• The coefficient of risk aversion in this case is a constant: 𝐴 𝑥 = 𝑎𝑒 −𝑎𝑥
= 𝑎.
• This player has constant absolute risk aversion 𝑎.
Measuring Risk Aversion – Absolute Risk Aversion
• An important consideration that arises when modeling risk aversion is how this aversion to risk changes with
wealth.

• Suppose I prefer some risky gamble to getting £1000 for sure. Then we might think that someone wealthier
than me would also take the gamble over £1000 for sure.

• Suppose 𝑢(⋅) is a twice differentiable utility function.

Definition.
𝑢′′ (𝑥)
The payoff function 𝑢(𝑥) has decreasing (constant, increasing) absolute risk aversion if 𝐴 𝑥 = − 𝑢′ (𝑥) is a
decreasing (constant, increasing) function of 𝑥.

• Intuitively, if your coefficient of absolute risk aversion goes down as your wealth increases, we say you have
decreasing absolute risk aversion, and it means you find the same level of risk less costly as you get wealthier.
Measuring Risk Aversion – Absolute Risk Aversion Example
• Use the utility function 𝑢 𝑥 = 1 − 𝑒 −𝑎𝑥 to illustrate why we say a player with this utility function has
constant absolute risk aversion.

Example. The concave utility function 𝑢 𝑥 = 1 − 𝑒 −𝑎𝑥 with 𝑎 > 0. (Graph plots 1 − 𝑒 −𝑥 )
Measuring Risk Aversion – Absolute Risk Aversion Example
• Use the utility function 𝑢 𝑥 = 1 − 𝑒 −𝑎𝑥 to illustrate why we say a player with this utility function has
constant absolute risk aversion.

Example. The concave utility function 𝑢 𝑥 = 1 − 𝑒 −𝑎𝑥 with 𝑎 > 0.

• Suppose a player with this utility function and wealth 𝑤 is offered a bet with a 50% chance of winning or
losing $1.

• What does the lottery he is facing look like? Outcomes are 𝑋 = {𝑤 − 1, 𝑤 + 1} with probability 0.5 of each
outcome.

• His utility from each outcome is 𝑢 𝑤 − 1 = 1 − 𝑒 −𝑎 𝑤−1

= 1 − 𝑒 𝑎 𝑒 −𝑎𝑤 and 𝑢 𝑤 + 1 = 1 − 𝑒 −𝑎 𝑤+1
=
1 − 𝑒 −𝑎 𝑒 −𝑎𝑤 .

• His expected payoff is

E 𝑢 𝑥 = 0.5 1 − 𝑒 𝑎 𝑒 −𝑎𝑤 + 0.5 1 − 𝑒 −𝑎 𝑒 −𝑎𝑤 = 1 − 0.5 e𝑎 + 𝑒 −𝑎 𝑒 −𝑎𝑤 = 1 − 𝑒 −𝑎(𝑤−𝑐)

log(0.5 e𝑎 +𝑒 −𝑎 )
where 𝑐 = .
𝑎
Measuring Risk Aversion – Absolute Risk Aversion Example
• Use the utility function 𝑢 𝑥 = 1 − 𝑒 −𝑎𝑥 to illustrate why we say a player with this utility function has
constant absolute risk aversion.

Example. The concave utility function 𝑢 𝑥 = 1 − 𝑒 −𝑎𝑥 .

−𝑎(𝑤−𝑐) log(0.5 e𝑎 +𝑒 −𝑎 )
• His expected payoff is E 𝑢 𝑥 = 1−𝑒 , where 𝑐 = .
𝑎

• What is his certain equivalent? That is the value 𝐶𝐸 such that u CE = 1 − 𝑒 −𝑎𝐶𝐸 = E 𝑢 𝑥 = 1−
𝑒 −𝑎(𝑤−𝑐) .

• Thus 𝐶𝐸 = 𝑤 − 𝑐. The risk premium is 𝑟 = 𝐸 𝑥 − 𝐶𝐸 = 𝑤 − 𝑤 − 𝑐 = 𝑐.

• He is indifferent between a bet than earns him $0 on average or being paid $𝑐. In other words, you have to
log(0.5 e𝑎 +𝑒 −𝑎 )
pay him 𝑐 = dollars to take this risky bet.
𝑎

• Notice the amount you have to pay him to take this risky bet does not depend on his wealth 𝑤. The person’s
risk aversion is the same whatever his wealth. Hence the term “constant absolute risk aversion”.
Measuring Risk Aversion – Relative Risk Aversion
• There are other ways to measure risk aversion. Relative risk aversion weights by wealth.

• Suppose 𝑢(⋅) is a twice differentiable utility function.

Definition.
𝑢′′ (𝑥)
The coefficient of relative risk aversion is R 𝑥 = −𝑥 𝑢′ (𝑥) .

Example.
𝑥 1−𝑟
• Take the utility function u 𝑥 = , 𝑟 > 1.
1−𝑟
• First derivative: 𝑢′ 𝑥 = 𝑥 −𝑟 .
• Second derivative: 𝑢′′ 𝑥 = −𝑟𝑥 −𝑟−1
−𝑟𝑥 −𝑟−1
• The coefficient of relative risk aversion in this case is a constant: R 𝑥 = −𝑥 𝑥 −𝑟 = 𝑟.
• This player has constant relative risk aversion 𝑟.
Measuring Risk Aversion – Relative Risk Aversion Example
• Go back to the utility function and illustrate why we say a player with this utility function has constant
relative risk aversion.

𝑥 1−𝑟
Example. The concave utility function u 𝑥 = .
1−𝑟

• Suppose a player with this utility function and wealth 𝑤 is offered a bet with a 50% chance of increasing his
wealth by 10% to 1.1𝑤 and a 50% chance of decreasing his wealth by 10% to 0.9𝑤.

• What does the lottery he is facing look like? Outcomes are 𝑋 = {0.9𝑤, 1.1𝑤} with probability 0.5 of each
outcome.

1 1
• His utility from each outcome is u 1.1𝑤 = 1−𝑟 𝑤 1−𝑟 1.11−𝑟 and u 0.9𝑤 = 1−𝑟 𝑤 1−𝑟 0.91−𝑟 .

• His expected payoff is

1 1 1
E𝑢 𝑥 = 0.5 𝑤 1−𝑟 1.11−𝑟 + 0.5 𝑤 1−𝑟 0.91−𝑟 = 0.5 𝑤 1−𝑟 (1.11−𝑟 +0.91−𝑟 ).
1−𝑟 1−𝑟 1−𝑟
Measuring Risk Aversion – Relative Risk Aversion Example
• Go back to the utility function and illustrate why we say a player with this utility function has constant
relative risk aversion.

1
Example. The concave utility function u 𝑥 = 1−𝑟 𝑥 1−𝑟 .

1
• His expected payoff is E 𝑢 𝑥 = 0.5(1.11−𝑟 +0.91−𝑟 ) 𝑤 1−𝑟 .
1−𝑟

1 1
• His certain equivalent satisfies 𝑢 𝐶𝐸 = 𝐶𝐸1−𝑟 =E𝑢 𝑥 = 0.5(1.11−𝑟 +0.91−𝑟 ) 𝑤 1−𝑟 .
1−𝑟 1−𝑟

1
1−𝑟 1−𝑟
• 𝐶𝐸 = [0.5(1.1 +0.9 )]
1−𝑟 ×𝑤

• His certain equivalent is a fraction of his wealth 𝑐𝑤 where the fraction 1 > 𝑐 > 0 does not depend on his
wealth.

• He values the lottery at 𝐶𝐸 and thus as he gets wealthier he is putting more value on a risky lottery but he
always values the lottery at a constant fraction of wealth (hence “relative” risk aversion).
Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Attitudes Towards Risk
• If a player is risk neutral, he is indifferent between the sure value of the lottery and participating in the lottery
itself:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 = 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• This equality is the definition of a linear function!

• That is, a player that is risk neutral has a linear payoff function, 𝑢().

Definition.
A player is risk neutral if for any (non degenerate) lottery, the player is indifferent between the expected monetary value
of the lottery and the lottery itself. That is for any lottery 𝑢 𝐸 𝑥 = 𝐸[𝑢(𝑥)].
Attitudes Towards Risk
• If a player is risk loving, he prefers participating in the lottery itself to the sure value of the lottery:
𝑝𝑢 𝑥1 + 1 − 𝑝 𝑢 𝑥2 ≥ 𝑢(𝑝𝑥1 + 1 − 𝑝 𝑥2 ).

• This equality is the definition of a convex function!

• That is, a player that is risk loving has a convex payoff function, 𝑢().

Definition.
A player is risk loving if for any (non degenerate) lottery, the player prefers the lottery itself to the expected monetary
value of the lottery. That is for any lottery 𝑢 𝐸 𝑥 ≤ 𝐸[𝑢(𝑥)].
Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Continuous case
• Take a simple lottery over an interval X, with cdf 𝐹(). Let 𝑢(𝑥) be the player’s payoff over outcomes x ∈ 𝑋.

• Suppose 𝐹() is differentiable with derivative 𝑓, the pdf.

• The expected monetary value of the lottery is E 𝑥 = ‫𝑥𝑑 𝑥 𝑓𝑥 𝑋∈𝑥׬‬.

• The expected payoff from the lottery is 𝐸[𝑢(𝑥)] = ‫𝑥𝑑 𝑥 𝑓 𝑥 𝑢 𝑋∈𝑥׬‬.

• The player’s utility from the expected monetary value of the lottery is 𝑢 𝐸 𝑥 = 𝑢 ‫ 𝑥𝑑 𝑥 𝑓𝑥 𝑋∈𝑥׬‬.
Agenda

• Introductory Examples

• Risk Aversion

• Certain Equivalent

• Risk Premium

• Measuring Risk Aversion

• Attitudes Toward Risk

• Continuous Case

• Summary
Summary
• In this topics we addressed attitudes towards risk within the expected utility framework.

• How we think about dislike of risk – risk aversion.

• How we measure dislike of risk – certain equivalent.

• How we measure the cost of risk – risk premium.

• How we compare degrees of risk aversion between different players.

• Risk neutrality, risk aversion, risk loving.

End Notes
• An important thing is that when talking about risk preferences the outcome 𝑥 ∈ 𝑋 needs to be money.

• A non-degenerate lottery involves at least two different payouts 𝑥1 ≠ 𝑥2 with positive probabilities.

• Useful notes from Martin Osborne on concavity

[Link]