TPPE24 Ekonomisk Analys:

Besluts- och Finansiell Metodik

Lecture 4

Applications and criticism of risk theory

1
 Contents

 Insurance
 Risk sharing
 Diversification
 Portfolio

 Framing effect
 Allias paradox
 Ellsberg paradox
 Others

2
 Insurance

Individual person, risk avert and concave utility function (why?)

u (Insurance) ≥ u( no insurance)
1-q W

q – probability of accident
No q 0
W –wealth
r – insurance

Insurance W-r

u(W-r) ≥(1-q)u(W) +q u(0)

Let u(W)=1 and u(0)=0
u(W-r) ≥(1-q)×1+q×0=1-q
W-r ≥u-1(1-q)
r≤W- u-1(1-q)=rmax 3
 Insurance

W-r ≥u-1(1-q)
r≤W- u-1(1-q)=rmax

u(x)

1-q

rmax
x
0 W-rmax W
4
 Insurance

Insurance company, risk neutral and EMV (why?)

EMV (Insurance) ≥ EMV( no insurance)
1-q r

Insurance
q r-W

No 0

(1-q)r+q(r-W) ≥0
r-rq+rq-qW ≥ 0
r-qw ≥ 0
r ≥ qW=rmin
5
 Insurance

r≤W- u-1(1-q)=rmax r ≥ qW=rmin

u(x)

1-q
rmin

profit

rmax
x
0 W-rmax (1-q)W W
6
 Price of insurance - an example (1)

An individual owns a house worth of 1,000 K SEK. If a fire accident

occurs, the rest value of the house will reduced to 0 and she believes
this has a probability 0.002. What is the highest premium she is willing
to pay? Her utility function looks like u(x) = x0.5.

u(No insurance)=(1-0.002)u(1000) + 0.002u(0)=31.56

u(Insurance)=u(1000-r) ≥ u(No insurance)= 31.56

r ≤ 4 K SEK

7
 Price of insurance - an example (2)

If the same person has won another property worth of 1,000 K SEK,
would this change her highest premium for house insurance?

u(No insurance)=(1-0.002)u(1000+1000) + 0.002u(0+1000)=44.70

u(Insurance)=u(2000-r) ≥ u(No insurance)=44.70

r ≤ 2.34 K SEK
Risk premium

Wealth

8
 Risk sharing (1)

If you have an opportunity to play the following lottery, will you accept it?

0.5 x Accepting condition

lottery ~ no lottery
Eu(lottery)=Eu(no lottery)
0.5 -y
y g-curve for an 0.5u(x)+0.5u(-y) = u(0)
EMV’er
Reject Let u(0)=0
u(x) = -u(-y)

Hence we can define a g-curve

Accept

x
9
 Risk sharing (2)

A A: lottery
500
0.5 1000
400

300 0.5 -500

y B
200 B: 50% of the lottery

100
0.5 500

0 250 500 750 1000 0.5 -250

x

You will reject lottery A, but you will accept if you have an opportunity to
take 50% of this lottery.
10
 Risk sharing (3)

Should a γ sharing of the following project acceptable?

(assuming the initial wealth is zero)

q W

1-q L

Accepting condition: qu (γW ) + (1 − q )u (γL) ≥ u (0)

First condition: EMV (W , L, q ) = qW + (1 − q ) L ≥ 0

Maximising: U (γ ) = qu (γW ) + (1 − q )u (γL)

Optimal sharing: U ' (γ *) = qWu ' (γ *W ) + (1 − q ) Lu ' (γ * L) = 0

11
 Effective risk sharing

Sharing a project is possible only if

γ i ∈ [0,1]

∑γ
i
i =1

U i (γ i ) = qu (γ iW ) + (1 − q )u (γ i L) ≥ 0

We may share a project in different ways

U i (α i , β i ) = qu (α iW ) + (1 − q )u ( β i L) ≥ 0

12
 Effective risk sharing - example

Kalle and Olle are sharing a project which has a positive EMV. Let Olle’s
share be γ
U (Olle) = qu (γW ) + (1 − q )u (γL)
U ( Kalle) = qu ((1 − γ )W ) + (1 − q )u ((1 − γ ) L)

Utility

Olle’s utility

Kalle’s utility

Olle’s share γ
0 1

13
 Pareto optimal

For any two states S and S’, S is said to be Pareto superior to S’ if and only
if under S at least one individual is better off, and none is worse off, than
under S’.

A state S* is said to be Pareto optimal (efficient or Pareto efficient) if and

only if it follows two conditions
a) S* is a feasible state
b) there is no feasible state which is Pareto superior to S*

14
Effective risk sharing - example cont.

Utility

Olle’s utility

Kalle’s utility

Olle’s share γ
0 1
Pareto optimal

Pareto superior

15
 Diversification (1)

You have the following choices for investment

Investment A1: Invest all your money in one big project. The payoff is 2x
with probability q and 2y with probability 1-q.

Investment A2: Split your investment in two independent small projects. For
each project, the payoff is x with probability q and y with probability 1-q.

You also have a utility function u(z).

Which investment is better?

16
 Diversification (2)

q 2x

1-q 2y
A1
q x

A2 q x 1-q y

y q x
1-q

1-q y

u(A1)=qu(2x)+(1-q)u(2y)
u(A2)=q2u(2x)+2q(1-q)u(x+y)+(1-q)2u(2y)
17
 Diversification (3)

u(A2)-u(A1)= q2u(2x)+2q(1-q)u(x+y)+(1-q)2u(2y) - qu(2x)-(1-q)u(2y)

=2q(1-q)[u(x+y)-(u(2x)+u(2y))/2]
u(z)

u(x+y)

[u(2x)+u(2y)]/2

z
0 2y x+y 2x
18
 Diversification (4)

Risk avert : investment 2

Risk neutral: indifference with two investments
Risk seeking: investment 1

Lessons to learn for risk avert person:

Do not put all eggs in one basket !!

19
 Portfolio

State
S1 ... Sj ... Sn
q1 ... qj ... qn
Security I1 γ1 U11 ... U1j … U1n
... ... ...
Ii γ i Ui1 ... Uij … Uin
... ... ...
Im γ m Um1 ... Umj … Umn

There are m investment opportunities I1 , I2 …and n states of nature S1, S2.

We are selecting γi so that to maximise investor’s expected utility

E (U (γ i )) = ∑∑ q jγ iU ij
j i
This is so called portfolio selection problem - Markowitz 1959
20
 Portfolio

When the number of security increases, the risk is reduced, due to

the fact that the variance of the expected return has reduced
whereas the expected return remains the same.
Risk

No. of security

More in the course Financial markets and instruments

21
 Investment under risk (1)

You are an EMVer and you are evaluating a project with G as the initial
investment in year 0 and possibly (1+r)NG as the payoff in year N. However,
in each year, this project may entirely stop with a zero payoff. The probability
for this to happen is q.

Determine the size of risk rate of return r so that the project is profitable.

Invest -G
1-q 1-q 1-q
S
No q q q
invest
0 0 0 0

1 2 year
0 N

22
 Investment under risk (2)

S = (1 + r ) N (1 − q) N G

( )
S − G ≥ 0 ⇒ (1 + r ) N (1 − q) N − 1 G ≥ 0

(1 + r ) N (1 − q) N ≥ 1
1
(1 + r ) N ≥
(1 − q) N
1
(1 + r ) ≥
(1 − q)
q
r≥
1− q
This is the rate of return compensating the risk. The interest rate
is not included here.
23
 Investment under risk (2)

q
r≥
1− q
r

q = 0.2 r= 0.25
q = 0.5 r= 0.1
q = 0.8 r= 4
q = 0.9 r= 9 q

q ∞
r≥ = q + q + q + ... = ∑ q i
2 3

1− q i =1

24
 Selling price of a lottery

Consider the same lottery that pays W with probability q or 0.

What is the selling price?

q W

u(s)≥ qu(W)+(1-q)u(0)
hold 1-q 0
Let u(0)=0
u(s) ≥ qu(W)

sell
s

25
 Buying price of a lottery

Consider the same lottery. What is the buying price?

q W-b

qu(W-b)+(1-q)u(-b)≥ u(0)
buy 1-q -b
Let u(0)=0
qu(W-b)+(1-q)u(-b) ≥ 0

not
buy 0

26
 Selling and buying price

Selling and buying price will NOT in general be the same.

u(s) ≥ qu(W)
qu(W-b)+(1-q)u(-b) ≥ 0

Selling and buying price will be the same if we have

Lineal utility function u(x)=a+bx

Exponential utility function u(x) = a+be -λx

Minimum selling price is CME

27
 Double inflection utility function

Friedman and Savage: it is not necessarily true that an individual’s utility

function has the same kind of curvature everywhere. It should depends on
the level of wealth.
u(x)

A x

Roller Coaster curve ?

28
 Double inflection utility function

Phenomena 1: people may take low probability, high payoff risks (lottery
tickets) while at the same time insuring against mild risks with mild payoff
(bicycle insurance)

Phenomena 2. It is not uncommon for an individual at the lower end of his

wealth to take risk. This is in order to extricate himself from this region of
relative poverty, but once reach a reasonable wealth, the individual become
risk averse in order to maintain this position.

29
 Framing effects (1)

People do not necessarily make choices in according to the utility

theory.

Tversky and Kahneman (1981) show how an individual’s risk attitude

can change depending on the way the decision problem is posed - frame
effect.

A small country is preparing for an outbreak of an usual influenza.

Experts expect 600 people to die from the disease. Several programs are
available to combat the disease, but due to the limits of resources only
one can be implemented.

30
 Framing effects (2)

Program A: 400 people will be saved.

Program B: There is an 80% chance that 600 people will be saved and a
20% chance no one will be saved.

Program C: 200 people will die.

Program D: There is a 20% chance that 600 people will die and an 80%
chance that no one will die.

Many people prefer A on one hand, but D on the other.

31
 Framing effects (3)

Results from empirical studies: people tend to be risk-averse in dealing

with gains but risk-seeking in deciding about losses.

Utility

Losses Gains

32
 Allais paradox (1)

Decision 1 A: win $1 billion with probability 1

B: win $5 billion with probability 0.1
win $1 billion with probability 0.89
win $0 billion with probability 0.01

Decision 2 C: win $1 billion with probability 0.11

win $0 billion with probability 0.89
D: win $5 billion with probability 0.10
win $0 billion with probability 0.90

Experimentally, as many as 82% subjects prefer A over B and 83%

prefer D over C. (Allais, 1979, Allais and Hagen, 1979)

33
 Allais paradox (2)

Let u(0)=0 and u(5)=1

0.11
1
1

A C
0.10 0.89 0
5
0.89 5
B 1 0.10
D
0.01 0 0
0.90

u(1)>0.1u(5)+0.89u(1)+0.01u(0) 0.1u(5)+0.9u(0)>0.11u(1)+0.89u(0)
u(1)>0.1+0.89u(1) 0.1>0.11u(1)

u(1)>0.91 u(1)<0.91
34
 Allais paradox (3)

Kahneman and Tversky (1981) attribute this common inconsistency to

the certainty effect, where individuals tend to place too much weight on
a certain outcome relative to the uncertain outcomes.

In the Allais paradox, the certainty effect would tend to make individual
overvalue A in the Decision 1.

Sure-thing principle.

35
 Ellsberg Paradox (1)

A barrel contains a mixture of 90 red, blue and yellow balls

30 of the balls are red,
the remaining is a mixture of blue and yellow.

Decision 1 A: win 1000 if you get a red ball.

B: win 1000 if you get a blue ball.

Decision 2 C: win 1000 if either a red or yellow ball is chosen.

D: win 1000 if either a blue or yellow ball is chosen.

36
 Ellsberg Paradox (2)

Red Blue Yellow

1/3 ? ?
*A 1000 0 0
B 0 1000 0

C 1000 0 1000
*D 0 1000 1000

Experimentally, many people prefer A over B and prefer D over C.

(Ellsberg, 1961).

37
 Ellsberg Paradox (2)

Red Blue Yellow

1/3 ? ?
*A 1000 0 0
B 0 1000 0

C 1000 0 1000
*D 0 1000 1000

Experimentally, many people prefer A over B and prefer D over C.

(Ellsberg, 1961).

38
 Ellsberg Paradox (3)

The choice of A over B and D over C violates “Event Monotonicity”

- adding a common event to all alternatives in a decision should not
affect the decision.

People often shy away from the risk with vague probability and
defend their choices using a worst-case scenario.

A person’s choice behavior may not always be rational, the

assumption that economists often have.

39
 Discrete and continuous payoff

Discrete payoff Continuous payoff

p(X) u(X) f(X) u(X)

E [U ( X )] = ∑ p ( x)u ( x) E [U ( X )] = ∫ p( x)u ( x)dx

x∈ X x∈ X

Same principle for the expected utility.

40
 Decision analysis - summary

Method
Normative, economic decisions under risk

Model
decision under strict uncertainty
EMV model
Normal form
Extensive form - decision tree
Extensive form to normal form
Reference lottery
Utility function - substitution, reduction
Risk attitude

41
 Decision analysis - summary

Applications
Financial market
Risk management
Insurance
Diversification, portfolio
Risk sharing

Other important terms

EMV, EPC, EOL, EVPI, EVSI
CME, risk premium, ARA, RRA, g-curve etc

42