01 Preference, Risk Aversion
Finance Theory I 01 Preference, Risk Aversion 1 / 24
Preference, Risk Aversion, and Expected Utility
Recall: uncertainty problem in microeconomics
Consider two choices:
r1 : Receive $2m for certain.
r2 : A lottery with a 1/9 probability of winning $10m and a 8/9
probability of winning $1m.
Which one will you choose?
They have the same expected value: E[r1 ] = E[r2 ] = $2m.
A risk-averse person might choose r1 . We assume that individuals
have a utility function of wealth, u(w), which is concave.
u′ (w) > 0 (More wealth is better)
u′′ (w) < 0 (Diminishing marginal utility of wealth)
Finance Theory I 01 Preference, Risk Aversion 2 / 24
Preference, Risk Aversion, and Expected Utility
Because the utility function is concave, the utility of the expected
value is greater than the expected utility of the gamble.
1 8
u(E[r2 ]) = u(2) > E[u(r2 )] = u(10) + u(1)
9 9
A function f is concave if and only if for all λ ∈ [0, 1]:
λf (a) + (1 − λ)f (b) ≤ f (λa + (1 − λ)b)
Finance Theory I 01 Preference, Risk Aversion 3 / 24
Preference: Basic Assumption
Basic Assumptions:
Two dates: time 0 and 1.
A single consumption good: utility is derived from this good.
States of nature: which can be discrete or continuous.
Ω = {ω}
Outcomes: xω , represent the number of units of the single
consumption good.
A consumption plan specifies the outcome for each state of nature.
Probabilities are assigned to each state, P(ωi ) for discrete states or
dP(ω) for continuous states.
A utility function U : X → R maps consumption outcomes to a real
number.
Finance Theory I 01 Preference, Risk Aversion 4 / 24
Expected Utility
An individual prefers consumption plan xω to xω′ if the expected
utility of xω is greater than or equal to the expected utility of xω′ .
Z Z
E[u(xω )] ≥ E[u(xω′ )] ⇐⇒ u(xω )dP(ω) ≥ u(xω′ )dP(ω)
Ω Ω
For example, if the end-of-period wealth is x with distribution F(x)
and density f (x):
Z ∞ Z ∞
E[u(x)] = u(x)dF(x) = u(x)f (x)dx
−∞ −∞
This is the Von Neumann-Morgenstern expected utility
formulation.
Finance Theory I 01 Preference, Risk Aversion 5 / 24
Axioms of Preference
Let x, y, z be three consumption plans. The preference relation ⪰
(“is preferred to”) must satisfy:
1. Completeness: For any two plans x and y, either x ⪰ y or y ⪰ x.
They can be compared.
2. Transitivity: If x ⪰ y and y ⪰ z, then x ⪰ z.
3. Reflexivity: x ⪰ x.
4. Independence of Irrelevant Alternatives: If x ≻ y, then for any z
and any a ∈ (0, 1], ax + (1 − a)z ≻ ay + (1 − a)z.
5. Continuity (Archimedean Axiom): If x ≻ y ≻ z, then there exist
a, b ∈ (0, 1) such that ax + (1 − a)z ≻ y ≻ bx + (1 − b)z.
Finance Theory I 01 Preference, Risk Aversion 6 / 24
More on utility functions
Given these axioms, there exist a utility function u(·) and a utility
function v(·) such that for any two consumption plans x and y:
x ⪰ y ⇐⇒ E[u(x)] ≥ E[u(y)]
⇐⇒ E[v(x)] ≥ E[v(y)]
This utility function is unique up to a positive affine transformation.
That is, if u(x) is a valid utility function, then v(x) = a · u(x) + b with
a > 0 is also a valid utility function representing the same
preferences.
Finance Theory I 01 Preference, Risk Aversion 7 / 24
Allais Paradox: common violation of the independence
axiom.
Choice 1: Choose between P1 and P2.
P1: $1m with 100% probability.
P2: $5m with 10% probability, $1m with 89% probability, $0 with 1%
probability.
Most people choose P1 ≻ P2 . Assuming u(0) = 0:
u(1) > 0.1u(5) + 0.89u(1) =⇒ 0.11u(1) > 0.1u(5)
Choice 2: Choose between P3 and P4.
P3: $5m with 10% probability, $0 with 90% probability.
P4: $1m with 11% probability, $0 with 89% probability.
Most people choose P3 ≻ P4 .
0.1u(5) > 0.11u(1)
This contradicts the preference derived from the first choice.
Finance Theory I 01 Preference, Risk Aversion 8 / 24
Time-Additive Utility and Boundedness
A common assumption is that utility is time-additive or
time-separable:
T
X
u(Z0 , Z1 , ..., ZT ) = ut (Zt )
t=0
A very popular form is ut (·) = β t u(·), where β is a discount factor.
Bounded Utility: The von Neumann-Morgenstern utility theory
requires utility to be bounded.
Mas-Colell et al. P. 176 Prop 6.B.3
X X
L ⪰ L′ iff un pn ≥ un p′n
∞ cannot be greater than ∞.
Finance Theory I 01 Preference, Risk Aversion 9 / 24
Time-Additive Utility and Boundedness
However, many popular utility functions are not bounded above.
Power utility: u(z) = z1−b is not bounded above if b < 1.
Log utility: u(z) = ln(z) is not bounded above.
Getting Around Boundedness:
1. If the state space Ω is finite, utility will be bounded.
2. For a concave utility function, u(x) ≤ u(b) + u′ (b)(x − b) for any b.
Then:
E[u(x)] ≤ u(b) + u′ (b)(E[x] − b) < ∞
As long as the consumption plan x has a finite expectation, E[u(x)]
is bounded above.
Finance Theory I 01 Preference, Risk Aversion 10 / 24
Concaveness and Risk Aversion
Jensen’s Inequality: For a random variable z̃ and a concave
function f , E[f (z̃)] ≤ f (E[z̃]).
Fair gamble:
ph1 + (1 − p)h2 = 0
Initial wealth w0
u(p(w0 +h1 )+(1−p)(w0 +h2 )) = u(w0 ) ≥ pu(w0 +h1 )+(1−p)u(w0 +h2 )
A risk-averse individual will reject a fair gamble. A gamble ε̃ is fair
if E[ε̃] = 0. For a risk-averse individual with utility function u:
E[u(w0 + ε̃)] = E[u(w0 + Eε̃ + (ε̃ − Eε̃)]
= E[u(w0 + Eε̃ + Ṽ]
≤ u(w0 + E[ε̃]) = u(w0 )
Finance Theory I 01 Preference, Risk Aversion 11 / 24
Absolute and Relative Risk Aversion
Arrow-Pratt absolute risk aversion coefficient:
u′′ (w)
A(w) = −
u′ (w)
Arrow-Pratt relative risk aversion coefficient:
u′′ (w)
R(w) = −w = w · A(w)
u′ (w)
w ↑ ⇒ u′ (w) ⇓
+
⇒ u′′ (w)?
−
Finance Theory I 01 Preference, Risk Aversion 12 / 24
Examples of Utility Functions and Risk Aversion
Coefficients
Exponential utility (CARA - Constant Absolute Risk
Aversion):
u(w) = −e−Aw
u′ (w) = Ae−Aw
u′′ (w) = −A2 e−Aw
A(w) = A (constant)
R(w) = Aw (increasing with w)
Finance Theory I 01 Preference, Risk Aversion 13 / 24
Examples of Utility Functions and Risk Aversion
Coefficients
Power utility (CRRA - Constant Relative Risk Aversion): For
R > 0, R ̸= 1:
w1−R
u(w) =
1−R
u (w) = w−R
′
u′′ (w) = −Rw−R−1
A(w) = R/w (decreasing with w)
R(w) = R (constant)
Finance Theory I 01 Preference, Risk Aversion 14 / 24
Examples of Utility Functions and Risk Aversion
Coefficients
Logarithmic utility (CRRA, with R=1):
u(w) = ln(w)
This is the limit of the power utility function as R → 1.
u′ (w) = 1/w
u′′ (w) = −1/w2
A(w) = 1/w (decreasing with w)
R(w) = 1 (constant)
Finance Theory I 01 Preference, Risk Aversion 15 / 24
Examples of Utility Functions and Risk Aversion
Coefficients
Quadratic utility:
b
u(w) = w − w2
2
′
u (w) = 1 − bw
u′′ (w) = −b
Requires 1 − bw > 0.
b
A(w) = 1−bw (increasing with w)
bw
R(w) = 1−bw (increasing with w)
Finance Theory I 01 Preference, Risk Aversion 16 / 24
Portfolio Choice with Expected Utility
An investor has initial wealth w0 and can invest in N risky assets and
one risk-free asset.
aj : dollar amount invested in risky asset j.
r̃j : return on risky asset j.
rf : risk-free rate.
The end-of-period wealth w̃ is:
N
X N
X N
X
w̃ = (w0 − aj )(1 + rf ) + aj (1 + r̃j ) = w0 (1 + rf ) + aj (r̃j − rf )
j=1 j=1 j=1
The investor’s problem is to choose a1 , ..., aN to maximize expected
utility:
max E[u(w̃)]
a1 ,...,aN
Finance Theory I 01 Preference, Risk Aversion 17 / 24
Portfolio Choice with Expected Utility
The first-order condition (FOC) for each asset j is:
E[u′ (w̃)(r̃j − rf )] = 0
u′ > 0 ⇒Prob[(r̃j − rf ) > 0] > 0 (1)
Prob[(r̃j − rf ) < 0] > 0 (2)
Suppose (1) does not hold, then
Prob[(r̃j − rf ) < 0] = 1.
⇒ We can short sell r̃j by an infinite amount. It’s an arbitrage
opportunity.
Finance Theory I 01 Preference, Risk Aversion 18 / 24
Further Application of the Portfolio Choice Problem
P
If all aj ≥ 0 ⇒ rule out short sales; if aj ≤ W0 ⇒ rule out borrowing.
The reason you won’t hold long position on risky assets ( aj < 0 ∀j). We
start with the FOC:
P E[u′ (w̃)(r̃j − rf )] = 0.
w̃ = w0 (1 + rf ) + aj (r̃j − rf )
If the optimal choice is aj ≤ 0 ∀j, but we set aj = 0 ∀j, then we are not at
the optimal point.
E[u′ (w0 (1 + rf ))(r̃j − rf )] ≤ 0 ∀j ← Max E(u(w))
| {z }
non-stochastic
⇒ u′ (w0 (1 + rf )) E[r̃j − rf ] ≤ 0, ∀j
| {z }
positive
⇒ E[r̃j − rf ] ≤ 0 ∀j
Finance Theory I 01 Preference, Risk Aversion 19 / 24
Further Application of the Portfolio Choice Problem
What if a1 = W0 : you might benefit from borrowing but can’t.
E[u′ (W0 (1 + r̃1 ))(r̃1 − rf )] ≥ 0
Taylor expansion: r̃1 around rf .
E[u′ (W0 (1 + r̃1 ))(r̃1 − rf )]
| {z }
h i
=E u′ (W0 (1 + rf ))(r̃1 − rf ) + u′′ (W0 (1 + rf ))W0 (r̃1 − rf )2 + o(r̃1 − rf ) ≥ 0
Assuming u′ (W0 (1 + rf )) is a positive constant, we can divide by it:
u′′ (W0 (1 + rf )) h i
E(r̃1 − rf ) + ′
W0 E (r̃1 − rf )2 ≥ 0
u (W0 (1 + rf ))
h i
⇒ E(r̃1 − rf ) ≥ RA (W0 (1 + rf )) · W0 E (r̃1 − rf )2
The higher an individual’s absolute risk aversion coefficient is, ⇒ The higher the
risk premium she requires.
Finance Theory I 01 Preference, Risk Aversion 20 / 24
Another side of Risk Premium: certainty equivalent
Certainty equivalent:
E[u(w + ε̃ − Eε̃)] = u(w − π)
A first-order Taylor expansion of the right hand side gives:
u(w − π) ≈ u(w) − πu′ (w) + · · ·
A second-order Taylor expansion of the left hand side gives:
(ε̃ − Eε̃)2 ′′
E[u(w + ε̃)] = E u(w) + (ε̃ − Eε̃)u′ (w) + u (w) + · · ·
2
var(ε̃) ′′
= u(w) + u (w)
2
Finance Theory I 01 Preference, Risk Aversion 21 / 24
Another side of Risk Premium: certainty equivalent
Equating the two approximations:
var(ε̃) ′′
u(w) − πu′ (w) ≈ u(w) + u (w)
2
var(ε̃) ′′
−πu′ (w) ≈ u (w)
2
u′′ (w)
var(ε̃)
π≈ − ′
2 u (w)
u′′ (w)
var(ε̃)
π≈ −
2 }
| {z u′ (w)
| {z }
↑ ↑
volatility Absolute risk aversion
Finance Theory I 01 Preference, Risk Aversion 22 / 24
Decreasing Absolute Risk Aversion (DARA)
Decreasing Absolute Risk Aversion, A′ (w) < 0, is a common and
plausible assumption. It implies that as an individual’s wealth increases,
they are willing to invest more dollars in risky assets.
That is, risky assets are normal goods: if A′ (w) < 0, then da
dw0 > 0.
The FOC is E[u′ (w̃)(r̃ − rf )] = 0. Differentiating with respect to w0 (using
implicit differentiation):
da −E[u′′ (w̃)(r̃ − rf )(1 + rf )]
=
dw0 E[u′′ (w̃)(r̃ − rf )2 ]
The denominator is positive because u′′ < 0 and (r̃ − rf )2 > 0. Therefore,
the sign is determined by the numerator’s sign.
da
sgn = sgn (E[u′′ (w̃)(r̃ − rf )])
dw0
Finance Theory I 01 Preference, Risk Aversion 23 / 24
Decreasing Absolute Risk Aversion (DARA)
Case 1: r̃ ≥ rf ⇒ w̃ ≥ W0 (1 + rf ).
Decreasing absolute risk aversion implies RA (w̃) ≤ RA (W0 (1 + rf ))
Multiply −u′ (w̃)(r̃ − rf ) on both sides:
u′′ (w̃)(r̃ − rf ) ≥ −RA (w0 (1 + rf ))u′ (w̃)(r̃ − rf )
Case 2: r̃ ≤ rf ⇒ w̃ ≤ W0 (1 + rf )
Decreasing absolute risk aversion implies RA (w̃) ≥ RA (W0 (1 + rf ))
Multiplying by −u′ (w̃)(r̃ − rf ) (a positive quantity) gives:
u′′ (w̃)(r̃ − rf ) ≥ −RA (W0 (1 + rf ))u′ (w̃)(r̃ − rf )
Put these two cases together:
⇒ E[u′′ (w̃)(r̃ − rf )] ≥ −RA (W0 (1 + rf )) E[u′ (w̃)(r̃ − rf )] = 0
| {z }
∥
0 (from F.O.C)
da
⇒ ≥0
dw0
Finance Theory I 01 Preference, Risk Aversion 24 / 24