Professional Documents
Culture Documents
Slide 04
04--1
Fin 501: Asset Pricing
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance
2. Stochastic dominance [DD4]
Slide 04
04--2
Fin 501: Asset Pricing
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
St t -by
State-
State b -state
by- t t Dominance
D i ((ctd.)
td )
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%
Slide 04
04--4
Fin 501: Asset Pricing
St t -by
State-
State b -state
by- t t Dominance
D i ((ctd.)
td )
Table 2.3 State Contingent Rates of Return
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance
2. Stochastic dominance [DD4]
Slide 04
04--6
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Slide 04
04--11
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
f B (x)
~
x , Payoff
μ = ∫ x fA (x)dx
d = ∫ x f B (x)dx
d
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
Fin 501: Asset Pricing
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance
2. Stochastic dominance [DD4]
Slide 04
04--15
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
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)
u(x0)
x0 E[x] x1 x
Slide 04
04--20
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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 )]
Slide 04
04--22
Fin 501: Asset Pricing
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)
Slide 04
04--23
Fin 501: Asset Pricing
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,
 a (strict) preference ordering over P P, and
Define % by p % q if q ¨ p
Slide 04
04--24
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Slide 04
04--26
Fin 501: Asset Pricing
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  q  r,
r then there
exists a α , β ∈ (0,1) such that
α p + (1-
( α) r  q  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
Fin 501: Asset Pricing
I d
Independence
d Axiom
A i
Independence of irrelevant alternatives:
p q
π π
p<q ⇔ <
r r
Slide 04
04--28
Fin 501: Asset Pricing
Allais Paradox –
Violation of Independence Axiom
Slide 04
04--29
Fin 501: Asset Pricing
Allais Paradox –
Violation of Independence Axiom
Slide 04
04--30
Fin 501: Asset Pricing
Allais Paradox –
Violation of Independence Axiom
10%
 10%
0 0
0 0
Slide 04
04--31
Fin 501: Asset Pricing
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
i.e. p
preferences correspond
p to expected
p utility.
y
Slide 04
04--32
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
Slide 04
04--34
Fin 501: Asset Pricing
Slide 04
04--35
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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)
Digression: Prospect
P t Th
Theory
Value function ((over ggains and losses))
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance
2. Stochastic dominance [DD4]
Slide 04
04--47
Fin 501: Asset Pricing
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
A
Arrow-
Arrow-Pratt
P tt measures off risk
i k aversion
i
and their interpretations
Y U" (Y)
relative risk aversion =- ≡ R R (Y)
U' ((Y))
risk tolerance =
Slide 04
04--49
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
R l ti risk
Relative i k aversion
i coefficient
ffi i t
π Y(1+θ)
Y
1−π Y(1-θ)
Slide 04
04--51
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
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
Fin 501: Asset Pricing
• 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
P tf li as wealth
Portfolio lth changes
h
(i)
(ii)
(iii) .
Slide 04
04--56
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
L utility
Log tilit & Portfolio
P tf li All
Allocation
ti
U(Y) = ln Y.
Slide 04
04--58
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance
2. Stochastic dominance [DD4]
Slide 04
04--61
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
(+) ((-)) in s
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
Fin 501: Asset Pricing
O
Overview:
i Ri k Preferences
Risk P f
1. State
State--by-
by-state dominance
2. Stochastic dominance [DD4]
Slide 04
04--69
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
M -Variance:
Mean-
Mean V i quadratic
d ti utility
tilit
= aE
E [ y ] − b ( E [ y ] 2 + var[[ y ]).
])
Slide 04
04--71
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
Mean-
Mean
e -V Variance:
ce:
linear distribution classes
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
Fin 501: Asset Pricing
Mean-
Mean
e -V Variance:
ce:
linear distribution classes
But yi = μi + σi x, thus
+∞ +∞
∫− ∞ v(z)dFi (z) = ∫− ∞ v(μi + σi z)dF (z)
=: u(μ i , σi ).
Slide 04
04--75
Fin 501: Asset Pricing
M -Variance:
Mean-
Mean V i small
ll risks
ik
Slide 04
04--76
Fin 501: Asset Pricing
Slide 04
04--77
Fin 501: Asset Pricing
Slide 04
04--78
Fin 501: Asset Pricing
E t material
Extra t i l follows!
f ll !
Slide 04
04--80
Fin 501: Asset Pricing
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))
Slide 04
04--81
Fin 501: Asset Pricing
Slide 04
04--82
Fin 501: Asset Pricing
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
Fin 501: Asset Pricing
Multi-
M lti-period
Multi i d Portfolio
P tf li Choice
Ch i
Slide 04
04--84
Fin 501: Asset Pricing
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