QN 4 Jensens Ineq04lecture

Fin 501: Asset Pricing
Lecture 04: Risk Preferences and

Expected Utility Theory
• Prof. Markus K. Brunnermeier
Slide 04
04--1
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance
2. Stochastic dominance [DD4]
3. vNM expected utility theory

a) Intuition [L4]
b) Axiomatic
A i ti foundations
f d ti [DD3]
4. Risk aversion coefficients and portfolio choice [DD5,L4]

5 Prudence coefficient and precautionary savings [DD5]
5.
6. Mean
Mean--variance preferences [L4.6]
Slide 04
04--2
St t -by
State-
State b -state
by- t t Dominance
D i
- State-by-state
y dominance Ä incomplete
p rankingg
- « riskier »
Table 2.1 Asset Payoffs ($)
t=0 t=1
Cost at t=0 Value at t=1
π1 = π2 = ½
s=1 s=2
investment 1 - 1000 1050 1200
i
investment
t t2 - 1000 500 1600
investment 3 - 1000 1050 1600
- investment 3 state by state dominates 1.

Slide 04
04--3
St t -by
State-
State b -state
by- t t Dominance
D i ((ctd.)
td )
Table 2.2 State Contingent ROR (r )
State
St t CContingent
ti t ROR (r
( )
s=1 s=2 Er σ
Investment 1 5% 20% 12.5% 7.5%
Investment 2 -50% 60% 5% 55%
Investment 3 5% 60% 32.5% 27.5%
- Investment 1 mean-variance dominates 2

- BUT investment 3 does not m-v dominate 1!
Slide 04
04--4
St t -by
State-
State b -state
by- t t Dominance
D i ((ctd.)
td )
Table 2.3 State Contingent Rates of Return
State Contingent Rates of Return

s=1 s=2
investment 4 3% 5%
investment 5 3% 8%
π1 = π2 = ½
E[r4] = 4%; σ4 = 1%
5 5%; σ5 = 2.5%
E[r5] = 5.5%; 2 5%
- What is the trade-off between risk and expected return?

- Investment
I 4 hhas a hi
higher
h Sharpe
Sh ratio (E[ ] f)/σ
i (E[r]-r )/ than
h iinvestment 5
for rf = 0.
Slide 04
04--5
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance

a) Intuition [L4]
b) Axiomatic
A i ti foundations
f d ti [DD3]
c) Risk aversion coefficients [DD4,L4]
4 Risk aversion coefficients and portfolio choice [DD5,L4]

4. [DD5 L4]
5. Prudence coefficient and precautionary savings [DD5]

6. Mean
Slide 04
04--6
St h ti Dominance
Stochastic D i
Still incomplete
p ordering
g
“More complete” than state-by-state ordering
State-by-state dominance ⇒ stochastic dominance
Risk preference not needed for ranking!
independently of the specific trade-offs (between return, risk and other
characteristics of probability distributions) represented by an agent
agent'ss
utility function. (“risk-preference-free”)
Next Section:
Complete preference ordering and utility
representations
H
Homework:
k Provide
P id an example l which
hi h can be
b ranked
k d
according to FSD , but not according to state dominance.
Slide 04
04--7
Table 3-1 Sample Investment Alternatives Fin 501: Asset Pricing
States of nature 1 2 3
Payoffs 10 100 2000
Proba Z 1 .4 .6 0
Proba Z 2 .4 .4 .2
EZ 1 = 64, σ z 1 = 44
EZ 2 = 444, σ z 2 = 779
Pr obability
F1
10
1.0
0.9
F2
0.8
07
0.7
0.6
0.5
F 1 and F 2
0.4
0.3
0.2
01
0.1
Payoff
0 10 100 2000
Slide 04
04--8
Fi t Order
First O d Stochastic
St h ti Dominance
D i
Definition 33.1
1 : Let FA(x) and FB(x) , respectively,
respectively
represent the cumulative distribution functions of two
random variables ((cash payoffs)
p y ff ) that,, without loss off
generality assume values in the interval [a,b]. We say
that FA(x) first order stochastically dominates (FSD)
FB(x) if and only if for all x ∈ [a,b]
FA(x) ≤ FB(x)
H
Homework:
k Provide
P id an example l which
hi h can be
b ranked
k d
according to FSD , but not according to state dominance.
Slide 04
04--9
Fi t Order
First O d Stochastic
St h ti Dominance
D i
1
0.9
0.8 FB
0.7 FA
0.6
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
X
Slide 04
04--10
Table 3-2 Two Independent Investments
Investment 3 Investment 4
Payoff Prob. Payoff Prob.
4 0 25
0.25 1 0 33
0.33
5 0.50 6 0.33
12 0.25 8 0.33
1
0.9
0.8
0.7
0.6 investment 4
0.5
0.4
0.3
0.2
investment 3
0.1
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Figure 3-6 Second Order Stochastic Dominance Illustrated
Slide 04
04--11
S
Second
dOOrder
d StStochastic
h ti Dominance
D i
Definition 3.2:
3 2: Let FA ( ~x ) , FB ( ~x ) , be two
cumulative probability distribution for
random payoffs in [a, b]. We say that FA ( ~x )
second order stochastically dominates
(SSD) FB ( ~x ) if and only if for any x :
[ FB (t) - FA (t) ] dt ≥ 0
x
∫
-∞
(with strict inequality for some meaningful
interval of values of t).
Slide 04
04--12
Mean Preserving Spread

xB = xA + z ((3.8))
where z is independent of xA and has zero mean
for normal distributions fA (x)
f B (x)
~
x , Payoff
μ = ∫ x fA (x)dx
d = ∫ x f B (x)dx
d
Figure 3-7 Mean Preserving Spread

Slide 04
04--13
Mean Preserving Spread & SSD

Theorem 3.4 : Let FA(•) and FB(•) be two distribution
functions defined on the same state space with identical
means. Then the follow statements are equivalent :
FA ( ~
x ) SSD FB ( ~ x)
FB ( ~
x ) is a mean p
preserving sp ead of FA ( ~x )
ese ving spread
in the sense of Equation (3.8) above.
Slide 04
04--14
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance

a) Intuition [L4]
b) Axiomatic
A i ti foundations
f d ti [DD3]
4. Risk aversion coefficients and portfolio choice [DD4,5,L4]

5.
6. Mean
Slide 04
04--15
AH
Hypothetical
th ti l Gamble
G bl
Suppose someone offers you this gamble:
"I have a fair coin here. I'll flip it, and if it's tail I pay
you $1 and the gamble is over. If it's head, I'll flip
again. If it's tail then, I pay you $2, if not I'll flip again.
With every round, I double the amount I will pay to
you if it's
it s tail
tail."
Sounds like a good deal. After all, you can't loose.
So here
here'ss the question:
How much are you willing to pay to take this
gamble?
g
Slide 04
04--16
P
Proposal
l 1:
1 Expected
E t d Value
V l
With probability 1/2 you get $1. (12 )1 times 20
With probability 1/4 you get $2.
$2 (12 )2 times 21
With probability 1/8 you get $4. (12 )3 times 22
etc.
Ê The expected payoff is the sum of these
payoffs weighted with their probabilities,
payoffs, probabilities
so 1
=∞
t ∞
∞
⎛1⎞
∑ ⎜ ⎟ ⋅ 2
t −1 = ∑2
t =1 2⎠
⎝{ 123 t =1
1
probability payoff
Slide 04
04--17
An Infinitely Valuable Gamble?

probability
You should pay
0.5
everything you own and
0.4
more to purchase the right
to take this gamble! 0.3
0.2
Yet, in practice, no one is
0.1
prepared to pay such a
high price. Why? 0
0 20 40 60 $
E
Even though
th h the
th expected
t d With 93% probability
payoff is infinite, the we get $8 or less, with
distribution of payoffs is 99% probability
b bili we get
not attractive… $64 or less. Slide 04
04--18
What Sho
Should
ld We Do?
How can we decide in a rational fashion about such
gambles (or investments)?
Proposal 2: Bernoulli suggests that large gains should be
weighted less. He suggests to use the natural logarithm.
[Cremer - another great mathematician of the time - suggests the
square root
root.]]
t
∞
⎛1⎞ expected utility
∑ ⎜ ⎟ ⋅ ln(2 ) = ln(2) =
2⎠
⎝{
t −1
of gamble
<∞
t =1 14243
probability utility of payoff
Ê Bernoulli would have paid at most eln(2) = $2 to

participate in this gamble.
Slide 04
04--19
Ri
Risk
Risk-
k-Aversion
A i andd C
Concavity
it
u(x)
( )
The shape of the von
Neumann Morgenstern (NM)
utility function contains a lot
of information. u(x1)
Consider a fifty-fifty lottery

with final wealth of x0 or x1
E{u(x)}
u(x0)
x0 E[x] x1 x
Slide 04
04--20
Ri
Risk
Risk-
k-aversion
i andd concavity
it
u(x)
( )
Risk-aversion means that the
certainty equivalent is smaller
than the expected prize.
prize
Ê We conclude that a u(x1)
risk-averse vNM u(E[x])
utility function
must be concave. E[u(x)]
u(x0)
x0 E[x] x1 x
u-1(E[u(x)]) Slide 04
04--21
J
Jensen’s
’ IInequality
lit
Theorem 3.13 1 (Jensen
(Jensen’ss Inequality):
Let g( ) be a concave function on the interval
[[a,b],
] and be a random variable such that
Prob (x ∈[a,b]) =1
Suppose the expectations E(x) and E[g(x)] exist;
then
E [g ( ~
x )] ≤ g [E ( ~
x )]
Furthermore, if g(•) is strictly concave, then the

inequality is strict.
Slide 04
04--22
R
Representation
t ti off Preferences
P f
A preference ordering is (i) complete, (ii) transitive, (iii)
continuous and [(iv) relatively stable] can be represented
by a utility function, i.e.
(c0,cc1,…,ccS) Â (c
(c’0,cc’1,…,cc’S)
⇔ U(c0,c1,…,cS) > U(c’0,c’1,…,c’S)
(preference ordering over lotteries –

(S+1)-dimensional
(S ) space)
p )
Slide 04
04--23
P f
Preferences over P
Prob.
b Di
Distributions
t ib ti
Consider c0 fixed, c1 is a random variable
Preference ordering over probability distributions
Let
P bbe a sett off probability
b bilit distributions
di t ib ti with
ith a finite
fi it
support over a set X,
Â a (strict) preference ordering over P P, and
Define % by p % q if q ¨ p
Slide 04
04--24
S states of the world

Set of all possible lotteries
Space with S dimensions
CCan we simplify
i lif the
th utility
tilit representation
t ti off
preferences over lotteries?
Space
S with
ith one dimension
di i – income
i
We need to assume further axioms
Slide 04
04--25
E
Expected
t d Utility
Utilit Th
Theory
A binary relation that satisfies the following
three axioms if and onlyy if there exists a function
u(•) such that
p Â q ⇔ ∑ u(c) p(c) > ∑ u(c) q(c)
i.e. preferences correspond to expected utility.
Slide 04
04--26
vNM
NM Expected
E t d Utility
Utilit Theory
Th
Axiom 1 ((Completeness
p and Transitivity):
y)
Agents have preference relation over P (repeated)
Axiom 2 (Substitution/Independence)
For all lotteries p,q,r ∈ P and α ∈ (0,1],
p < q iff α p + ((1-α)) r < α q + ((1-α)) r ((see next slide))
Axiom 3 (Archimedian/Continuity)
For all lotteries p,q,r
p q r ∈ P,
P if p Â q Â r,
r then there
exists a α , β ∈ (0,1) such that
α p + (1-
( α) r Â q Â β p + ((1 - β) r..
Problem: p you get $100 for sure, q you get $ 10 for sure, r you are killed
Slide 04
04--27
I d
Independence
d Axiom
A i
Independence of irrelevant alternatives:
p q
π π
p<q ⇔ <
r r
Slide 04
04--28
Allais Paradox –
Violation of Independence Axiom
10% 10’ 9% 15’

≺
0 0
Slide 04
04--29
Allais Paradox –
10% 10’ 9% 15’

≺
0 0
100% 10’ 90% 15’

Â
0 0
Slide 04
04--30
Allais Paradox –
10% 10’ 9% 15’

≺
0 0
100% 10’ 90% 15’
10%
Â 10%
0 0
0 0
Slide 04
04--31
vNM
NM EU Th
Theorem
A binary relation that satisfies the axioms 1-3 if
and only
y if there exists a function u(•)
( ) such that
p Â q ⇔ ∑ u(c) p(c)) > ∑ u(c)

( ) p( ( ) q(
q(c))
i.e. p
preferences correspond
p to expected
p utility.
y
Slide 04
04--32
E
Expected
t d Utility
Utilit Th
Theory
U(Y)
U( Y0 + Z 2 )
~
U( Y0 + E( Z))
~
EU( Y0 + Z)
U( Y0 + Z1 )
~
CE( Z) Π
Y0 Y0 + Z1 ~ ~ Y0 + Z 2
CE ( Y0 + Z) Y0 + E(Z)
Y
Slide 04
04--33
Expected Utility & Stochastic Dominance
Theorem 3. 2 : Let FA ( ~x ) , FB ( ~x ) , be two cumulative

probability
b bilit distribution
di t ib ti for f random d payoffs ff ~ x ∈ [a , b].
Then FA ( ~x ) FSD FB ( ~x ) if and only if
for all non decreasing utility functions U( U(•)).
E A U(~
x) ≥ E B U(~
x)
Slide 04
04--34
Expected Utility & Stochastic Dominance

Theorem 3. 3 : Let FA ( ~x ) , FB ( ~x ) , be two cumulative
probabilityy distribution
p
for random payoffs ~ x defined on [a , b] .
Then, FA ( ~x ) SSD FB ( ~x ) if and only if E A U(~x) ≥ E B U(~x)
f all
for ll non decreasing
d andd concave U.
Slide 04
04--35
Digression: S bj ti EU Theory
Subjective Th
Derive perceived probability from preferences!
Set S of prizes/consequences
Set Z of states
Set of functions f(s) ∈ Z
Z, called acts (consumption plans)
Seven SAVAGE Axioms
Goes
G bbeyond
d scope off thi
this course.
Slide 04
04--36
Digression: Ell b
Ellsberg Paradox
P d
10 balls in an urn
Lottery 1: win $100 if you draw a red ball
L tt
Lottery 2:
2 wini $100 if you ddraw a bl blue bball
ll
Uncertainty: Probability distribution is not known
Risk: Probability distribution is known
(5 balls are red, 5 balls are blue)
Individuals are “uncertainty/ambiguity averse”

(non-additive probability approach)
Slide 04
04--37
Digression: Prospect
P t Th
Theory
Value function ((over ggains and losses))
Overweight low probability events

Experimental evidence
Slide 04
04--38
I diff
Indifference curves
Any point in
x2
this plane is
a particular
lottery.
Where is the
set of risk-
ffree
lotteries?
If x1=x x2,
then the
45°
x1
lottery
contains no
risk.Slide 04
04--39
I diff
Indifference curves
Where is the
x2
set of lotteries
with expected
π prize E[L]=z?
It's a
straight line,
z and the
slope is
given by the
relative
45° probabilities
x1 of the two
z
states.
Slide 04
04--40
Suppose the
agent is risk
I diff
Indifference curvesaverse. Where
is the set of
x2
??? lotteries which
are indifferent
π to (z,z)?
That's not
right! Note
z th t there
that th
are risky
lotteries
with smaller
45°
expected
x1 prize and
z
which are
preferred.
Slide 04
04--41
I diff
Indifference curves
x2
So the
indifference
π curve must be
tangent to the
iso-expected-
prize line.
z
This is a direct
p
implication of
risk-aversion
45°
alone.
x1
z
Slide 04
04--42
I diff
Indifference curves
x2
But risk-
risk
aversion does
π not imply
convexity.
This
z indifference
d ff
curve is also
compatible
p
with risk-
45° aversion.
x1
z
Slide 04
04--43
I diff
Indifference curves
x2
∇ V(z,z)
The tangency
implies that
the gradient
of V at the
π
point (z,z) is
collinear to π.
z
Formally,
∇ V(z,z) = λπ,
for some λ>0.
45°
x1
z
Slide 04
04--44
C i
Certainty Equivalent
E i l andd Risk
Ri k Premium
P i
~ ~
(3.6) EU(Y + Z ) = U(Y + CE(Y, Z ))
~ ~
(3.7) = U(Y +E Z - Π(Y, Z ))
Slide 04
04--45
C i
Certainty Equivalent
E i l andd Risk
Ri k Premium
P i
U(Y)
U( Y0 + Z 2 )
~
U( Y0 + E( Z))
~
EU( Y0 + Z)
U( Y0 + Z1 )
~
CE( Z) Π
Y0 Y0 + Z1 ~ ~ Y0 + Z 2
CE ( Y0 + Z) Y0 + E(Z)
Y
Figure
g 3-3 Certainty
y Equivalent
q and Risk Premium: An Illustration
Slide 04
04--46
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance

a) Intuition [L4]
b) Axiomatic
A i ti foundations
f d ti [DD3]

5.
6. Mean
Slide 04
04--47
M
Measuring
i RiRiskk aversion
i
U(W)
tangent lines
li
U(Y+h)
U[0.5(Y+h)+0.5(Y-h)]
0.5U(Y+h)+0.5U(Y-h)
U(Y-h)
Y-h Y Y+h
W
Figure 3-1 A Strictly Concave Utility Function

Slide 04
04--48
A
Arrow-
Arrow-Pratt
P tt measures off risk
i k aversion
i
and their interpretations
absolute risk aversion = - U" ((Y)) ≡ R A (Y)

U' (Y)
Y U" (Y)
relative risk aversion =- ≡ R R (Y)
U' ((Y))
risk tolerance =
Slide 04
04--49
Ab l t risk
Absolute i k aversion
i coefficient
ffi i t
π Y+h
Y
1−π Y-h
Slide 04
04--50
R l ti risk
Relative i k aversion
i coefficient
ffi i t
π Y(1+θ)
Y
1−π Y(1-θ)
Homework: Derive this result.
Slide 04
04--51
CARA and
d CRRA
CRRA--utility
tilit functions
f ti
Constant Absolute RA utility function
Constant
C R
Relative
l i RA utility
ili function
f i
Slide 04
04--52
Investor ’s
’ Levell off Relative
l i Risk
i k Aversion
A i
1− γ
( Y + CE) ( Y + 50,000 )1 − γ 12 ( Y + 100,000 )1 − γ
1
2
= +
1- γ 1- γ 1- γ
γ=0 CE = 75,000 (risk neutrality)

γ=1 CE = 70,711
Y=0 γ=2 CE = 66,246
66 246
γ=5 CE = 58,566
γ = 10 CE = 53,991
γ = 20 CE = 51,858
γ = 30 CE = 51,209
Y=100,000 γ=5 CE = 66,530

Slide 04
04--53
Ri k aversion
Risk i andd Portfolio
P tf li Allocation
All ti
No
N savings
i ddecision
i i (consumption
( i occurs only
l at t=1)
1)
Asset structure
One risk free bond with net return rf
One risky asset with random net return r (a =quantity of risky assets)
Slide 04
04--54
• Theorem 4.1: Assume U'( ) > 0,, and U"( ) < 0 and let â
denote the solution to above problem. Then
â > 0 if and only if E~r > rf
â = 0 if and only if E~r = r f
â < 0 if and only if E~r < rf .
• Define W(a ) = E{U (Y0 (1 + rf ) + a (~r − rf ))}. The FOC can

then be written W′(a ) = E[U′(Y0 (1 + rf ) + a (~r − rf ))(~r − rf )] = 0 .
(U''<0), W′′(a ) = E[U′′(Y0 (1 + rf ) + a (~r − rf ))(~r − rf )2 ]
By risk aversion (U''<0)
< 0, that is, W'(a) is everywhere decreasing. It follows that
â will be positive if and only if W′(0) = U′(Y0 (1 + rf ))E(~r − rf ) > 0
(since then a will have to be increased from its value of 0 to
achieve equality in the FOC). Since U' is always strictly
positive this implies aâ > 0 if and only if E(~r − rf ) > 0 .
positive, W’(a)
W (a)
The other assertion follows similarly. a
0 Slide 04
04--55
P tf li as wealth
Portfolio lth changes
h
• Theorem 4.4 (Arrow, 1971): Let be the

solution to max-problem above; then:
(i)
(ii)
(iii) .
Slide 04
04--56
P tf li as wealth
Portfolio lth changes
h
• Theorem 4.5 (Arrow 1971): If, for all wealth levels Y,
(i)
(ii)
( )
(iii)
da/a (elasticity)
where η= dY /Y
Slide 04
04--57
L utility
Log tilit & Portfolio
P tf li All
Allocation
ti
U(Y) = ln Y.
states, where r2 > rf > r1

2 states
Constant fraction of wealth is invested in risky asset!
Slide 04
04--58
Portfolio of risky assets as wealth changes

Now -- many risky assets
Theorem 4.6 (Cass and Stiglitz,1970). Let the vector
⎡ â1 ( Y0 ) ⎤
⎢ . ⎥ denote the amount optimally
p y invested in the J risky
y assets if
⎢ ⎥
⎢ . ⎥
⎢ ⎥ ⎡ â1 ( Y0 ) ⎤ ⎡ a1 ⎤
⎣ J 0 ⎦
â ( Y ) ⎢ . ⎥ ⎢.⎥
the wealth level is Y0.. Then ⎢ ⎥ = ⎢ ⎥f ( Y0 )
⎢ . ⎥ ⎢.⎥
⎢ ⎥ ⎢ ⎥
⎣ â J ( Y0 ) ⎦ ⎣a J ⎦
if andd only
l if either
ith
Δ
(i) U ' ( Y0 ) = ( θY0 + κ ) or
(ii) U ' ( Y ) = ξe − νY0 .
0
In words, it is sufficient to offer a mutual fund.

Slide 04
04--59
LRT/HARA--utility
LRT/HARA tilit functions
f ti
Linear Risk Tolerance/hyperbolic absolute risk aversion
Special Cases
B=0, A>0 CARA
B ≠ 0, ≠1 Generalized Power
B=1 Log utility u(c) =ln (A+Bc)
B=-1 Quadratic Utility u(c)=-(A-c)2
B ≠ 1 A=0
A 0 CRRA Utilit
Utility function
f ti
Slide 04
04--60
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance

a) Intuition [L4]
b) Axiomatic
A i ti foundations
f d ti [DD3]

5.
6. Mean
Slide 04
04--61
I t d i Savings
Introducing S i
• Introduce savings decision: Consumption at tt=00 and tt=11
• Asset structure 1:
– risk free bond Rf
– NO risky asset with random return
– Increase Rf:
– Substitution effect: shift consumption from t=0 to t=1
⇒ save more
– Income
I effect:
ff t agentt is
i “effectively
“ ff ti l richer”
i h ” andd wants
t to
t
consume some of the additional richness at t=0
⇒ save less
– For log-utility (γ=1) both effects cancel each other
Slide 04
04--62
P d
Prudence and
dPPre-cautionary
Pre- ti Savings
S i
• IIntroduce
t d savings
i decision
d ii
Consumption at t=0 and t=1
• Asset structure 2:
– NO risk free bond
– One risky asset with random gross return R
Slide 04
04--63
P d
Prudence and
dSSavings
i B
Behavior
h i
Risk aversion is about the willingness to insure …
… but not about its comparative statics.
How
H ddoes th the behavior
b h i off an agentt change
h when
h
we marginally increase his exposure to risk?
An old hypothesis (going back at least to
J.M.Keynes) is that people should save more now
when they face greater uncertainty in the future.
The idea is called precautionary saving and has
intuitive appeal.
Slide 04
04--64
P d
Prudence and
dPPre-cautionary
Pre- i Savings
S i
Does not directlyy follow from risk aversion alone.
Involves the third derivative of the utility function.
Kimball
ba ((1990)
990) de
defines
es abso
absolute
ute p prudence
ude ce as
P(w) := –u'''(w)/u''(w).
Precautionaryy savingg if anyy onlyy if theyy are prudent.
p
This finding is important when one does comparative
statics of interest rates.
Prudence seems uncontroversial, because it is weaker
than DARA.
Slide 04
04--65
Pre--cautionary Saving (extra material)

Pre
(+) ((-)) in s
Is saving s increasing/decreasing in risk of R?

Is RHS increasing/decreasing is riskiness of R?
Is U’() convex/concave?
Depends on third derivative of U()!
N.B: For U(c)=ln c, U’(sR)R=1/s does not depend on R.

Slide 04
04--66
P -cautionary
Pre-
Pre ti Saving
S i (extra material)
2 effects: Tomorrow consumption is more volatile
• consume more today, since it’s not risky
• save more for precautionary reasons
~ ~
R R
Theorem 4.7 (Rothschild and Stiglitz,1971) : Let A , B
be
~ two~ return distributions with identical means such that
RB = RA + e, (where e is white noise) and let s and s be,
A B
respectively, the savings out of Y0 corresponding to the
~ ~
return distributions R A
and R .B
If R ' R ( Y ) ≤ 0 and RR(Y) > 1, then sA < sB ;
If R ' R ( Y ) ≥ 0 and RR(Y) < 1, then sA > sB
Slide 04
04--67
Prudence & Pre-

Pre-cautionary Saving
− U ' ' ' (c)
P(c) =
U ' ' (c)
− cU ' ' ' (c)

P(c)c =
U ' ' (c)
~ ~
Th
Theorem 4.8
4 8 : Let
L tR , be
A RB
b two
t return
t distributions
di t ib ti suchh
~ SSD ~ , and let s and s be, respectively, the
that R A RB A B
savings out of Y0 corresponding to the return distributions
~ ~
R Aand R B. Then,
s A ≥ s B iff cP(c) ≤ 2, and conversely,
sA < sB iff cP(c) > 2
Slide 04
04--68
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance

a) Intuition [L4]
b) Axiomatic
A i ti foundations
f d ti [DD3]

5.
6. Mean
Slide 04
04--69
M -variance
Mean-
Mean i Preferences
P f
Early researchers in finance, such as Markowitz
and Sharpe, used just the mean and the variance
of the return rate of an asset to describe it.
Mean-variance characterization is often easier
than using an vNM utility function
But is it compatible with vNM theory?
The answer is yes … approximately … under
some conditions.
Slide 04
04--70
M -Variance:
Mean-
Mean V i quadratic
d ti utility
tilit
Suppose utility is quadratic, u(y) = ay–by2.

Expected utility is then
E [u ( y )] = aE [ y ] − bE [ y ]
2
= aE
E [ y ] − b ( E [ y ] 2 + var[[ y ]).
])
Thus, expected utility is a function of the

Thus
mean, E[y], and the variance, var[y], only.
Slide 04
04--71
M -Variance:
Mean-
Mean V i joint
j i t normals
l
Suppose all lotteries in the domain have normally
distributed prized. (independence is not needed).
This
Thi requires
i an infinite
i fi i state space.
Any linear combination of normals is also normal.
The normal distribution is completely described by its
first two moments.
Hence, expected utility can be expressed as a function of
just these two numbers as well.
Slide 04
04--72
Mean-
Mean
e -V Variance:
ce:
linear distribution classes
Generalization of joint nomarls.
Consider a class of distributions F1, …, Fn with the
f ll i property:
following
for all i there exists (m,s) such that Fi(x) = F1(a+bx) for all x.
Thi
This iis called
ll d a li
linear di
distribution
t ib ti class.
l
It means that any Fi can be transformed into an Fj by an
appropriate shift (a) and stretch (b).
(b)
Let yi be a random variable drawn from Fi. Let μi = E{yi}
and σi2 = E{(yi–μi)2} denote the mean and the var of yi.
Slide 04
04--73
Mean-
Mean
e -V Variance:
ce:
Define then the random variable x = (yi–μi)/σi. We
denote the distribution of x with F.
Note that the mean of x is 0 and the variance is 1, and F
is p
part of the same linear distribution class.
Moreover, the distribution of x is independent of which
i we start with.
Ê We want to evaluate the expected utility of yi ,
+∞
∫− ∞ v(z)dFi (z).
Slide 04
04--74
Mean-
Mean
e -V Variance:
ce:
But yi = μi + σi x, thus
+∞ +∞
∫− ∞ v(z)dFi (z) = ∫− ∞ v(μi + σi z)dF (z)
=: u(μ i , σi ).
The expected utility of all random variables

d
drawn ffrom the
h same linear
l distribution
d b
class can be expressed as functions of the
mean and the standard deviation only.
only
Slide 04
04--75
M -Variance:
Mean-
Mean V i small
ll risks
ik
Justification for mean-variance for the case of small risks.

use a second order (local) Taylor approximation of vNM
U(c).
If U(c)
( ) is concave,, second order Taylor
y approximation
pp is a
quadratic function with a negative coefficient on the
quadratic term.
Expectation
E i off a quadratic
d i utility
ili function
f i can beb
evaluated with the mean and variance.
Slide 04
04--76
Mean--Variance: small risks

Mean
Let f : R Æ R be a smooth function.
function The Taylor
approximation is
( x − x0 )1 ( x − x0 )2
f ( x ) ≈ f ( x0 ) + f ' ( x0 ) + f ' ' ( x0 ) +
1! 2!
( x − x0 )3
f ' ' ' ( x0 ) +L
3!
Use Taylor approximation for E[u(x)].
Slide 04
04--77

Mean
Since E[u(w+x)] = u(cCE), this simplifies to
var( x)
w − cCE ≈ RA( w) .
2
Ê w – cCE is the risk premium.
Ê We see here
W h that
th t the
th risk
i k premium
i is
i
approximately a linear function of the
variance of the additive risk
risk, with the
slope of the effect equal to half the
coefficient of absolute risk.
Slide 04
04--78

Mean
The same exercise can be done with a multiplicative risk.
Let y = gw, where g is a positive random variable with
unit mean.
Doing
D i theth same steps
t as before
b f leads
l d tot
var( g )
1 − κ ≈ RR ( w) ,
2
where κ is the certainty equivalent growth rate,
u( w) = E[u(gw)].
u(κw) E[u(gw)]
Ê The coefficient of relative risk aversion is
relevant for multiplicative risk
risk, absolute risk
aversion for additive risk.
Slide 04
04--79
12:30 Lecture Risk Aversion
E t material
Extra t i l follows!
f ll !
Slide 04
04--80
Joint saving-
saving-portfolio problem
• Consumption at t=0 and t=1. (savings decision)
• Asset structure
– One risk free bond with net return rf
– One riskyy asset ((a = quantity
q y of riskyy assets))
max U ( Y0 − s ) + δEU ( s(1 + rf ) + a ( ~r − rf )) ((4.7))

{ a ,s }
FOC:
s: U’(ct) = δ E[U’(ct+1)(1+rf)]
a: E[U’(c
E[U (ct+1)(r-r
)(r rf)] = 0
Slide 04
04--81
for CRRA utility functions
s: ( Y0 − s) − γ ( −1) + δE ([s(1 + rf ) + a ( ~r − rf )]− γ (1 + rf ) ) = 0

a: [ ]
E (s(1 + rf ) + a ( ~r − rf )) − γ ( ~r − rf ) = 0
Where s is total saving and a is amount invested in risky asset.
asset
Slide 04
04--82
Multi-
M lti-period
Multi i d Setting
S tti
Canonical framework (exponential discounting)
U(c) = E[ ∑ βt u(ct)]
prefers earlier uncertainty resolution if it affect action
indifferent, if it does not affect action
Time-inconsistent (hyperbolic discounting)
Special case: β−δ formulation
U(c) = E[u(c0) + β ∑ δt u(ct)]
Preference for the timing of uncertainty resolution
recursive utility formulation (Kreps-Porteus
(Kreps Porteus 1978)
Slide 04
04--83
Multi-
M lti-period
Multi i d Portfolio
P tf li Choice
Ch i
Theorem 4.10 ((Merton, 1971):) Consider the above canonical

multi-period consumption-saving-portfolio allocation problem.
Suppose U() displays CRRA, rf is constant and {r} is i.i.d.
Then a/st is time invariant.
invariant
Slide 04
04--84
Preference
Digression:
g ess o :e e e ce for
o thee
timing of uncertainty resolution
$100 $150
π
$100 $ 25
0
π $150
$100
Kreps-Porteus $ 25
Early (late) resolution if W(P1,…) is convex (concave)
Marginal rate of temporal substitution risk aversion

Slide 04
04--85

QN 4 Jensens Ineq04lecture

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

QN 4 Jensens Ineq04lecture

Uploaded by

Copyright:

Available Formats

Fin 501: Asset Pricing

Lecture 04: Risk Preferences and

3. vNM expected utility theory

4. Risk aversion coefficients and portfolio choice [DD5,L4]

- investment 3 state by state dominates 1.

Table 2.2 State Contingent ROR (r )

- Investment 1 mean-variance dominates 2

State Contingent Rates of Return

- What is the trade-off between risk and expected return?

3. vNM expected utility theory

4 Risk aversion coefficients and portfolio choice [DD5,L4]

5. Prudence coefficient and precautionary savings [DD5]

Table 3-2 Two Independent Investments

Figure 3-6 Second Order Stochastic Dominance Illustrated

Mean Preserving Spread

for normal distributions fA (x)

Figure 3-7 Mean Preserving Spread

Mean Preserving Spread & SSD

3. vNM expected utility theory

4. Risk aversion coefficients and portfolio choice [DD4,5,L4]

An Infinitely Valuable Gamble?

Ê Bernoulli would have paid at most eln(2) = $2 to

 Consider a fifty-fifty lottery

Furthermore, if g(•) is strictly concave, then the

(preference ordering over lotteries –

 S states of the world

 Space with S dimensions

p Â q ⇔ ∑ u(c) p(c) > ∑ u(c) q(c)

i.e. preferences correspond to expected utility.

10% 10’ 9% 15’

10% 10’ 9% 15’

100% 10’ 90% 15’

10% 10’ 9% 15’

100% 10’ 90% 15’

p Â q ⇔ ∑ u(c) p(c)) > ∑ u(c)

Expected Utility & Stochastic Dominance

 Theorem 3. 2 : Let FA ( ~x ) , FB ( ~x ) , be two cumulative

Expected Utility & Stochastic Dominance

 Individuals are “uncertainty/ambiguity averse”

 Overweight low probability events

3. vNM expected utility theory

4. Risk aversion coefficients and portfolio choice [DD4,5,L4]

Figure 3-1 A Strictly Concave Utility Function

 absolute risk aversion = - U" ((Y)) ≡ R A (Y)

Homework: Derive this result.

γ=0 CE = 75,000 (risk neutrality)

Y=100,000 γ=5 CE = 66,530

â < 0 if and only if E~r < rf .

• Define W(a ) = E{U (Y0 (1 + rf ) + a (~r − rf ))}. The FOC can

• Theorem 4.4 (Arrow, 1971): Let be the

states, where r2 > rf > r1

Constant fraction of wealth is invested in risky asset!

Portfolio of risky assets as wealth changes

 In words, it is sufficient to offer a mutual fund.

3. vNM expected utility theory

4. Risk aversion coefficients and portfolio choice [DD4,5,L4]

Pre--cautionary Saving (extra material)

Is saving s increasing/decreasing in risk of R?

N.B: For U(c)=ln c, U’(sR)R=1/s does not depend on R.

Prudence & Pre-

− cU ' ' ' (c)

3. vNM expected utility theory

Consider a fifty-fifty lottery

S states of the world

Space with S dimensions

Theorem 3. 2 : Let FA ( ~x ) , FB ( ~x ) , be two cumulative

Individuals are “uncertainty/ambiguity averse”

Overweight low probability events

absolute risk aversion = - U" ((Y)) ≡ R A (Y)

In words, it is sufficient to offer a mutual fund.

Justification for mean-variance for the case of small risks.

Marginal rate of temporal substitution risk aversion