Created by Turbolearn AI

Chapter 2: Stochastic Dominance

This chapter delves into scenarios where we can unequivocally say one risky asset is
preferred over another, focusing on situations beyond simple risk aversion. We'll
explore the concepts of stochastic dominance, which provide tools for comparing
risky assets.

2.1 Introduction to Stochastic Dominance

Comparing Risky Assets

Goal: To determine when one risky asset is unambiguously preferred to
another.
Focus: Identifying conditions under which all individuals with certain
preferences (e.g., risk aversion) would prefer asset A over asset B.
Tool: Stochastic dominance concepts.
Limitation: These concepts don't allow comparison between all risky assets,
and (in the terminology of Section 1.3) does not define a complete ordering of
risky assets.

Portfolio Allocation
Analyzes how an individual divides their portfolio between a risky and a risk-
less asset.
Examines how changes in the riskiness of the risky asset affect the portfolio
composition.
Considers the impact of an individual's Arrow-Pratt measure of risk aversion
on these decisions.

2.2 First-Degree Stochastic Dominance (FSD)

Definition
Asset A first-degree stochastically dominates asset B (denoted as A ⪰ B) if, for all
1

individuals with utility functions that are increasing in wealth, A is preferred to B.

Page 1
 Created by Turbolearn AI

Intuition
The probability of asset A's return exceeding any given level is always greater than
or equal to the probability of asset B's return exceeding the same level.

Formal Condition
For all possible returns z, the cumulative distribution function (CDF) of A, F A
, is
(z)

less than or equal to the CDF of B, F (z). B

F A (z) ≤ F B (z) ∀z ∈ [0, 1]

This means the probability of the rate of return on asset A being less than or equal to
z is no more than the probability of the rate of return on asset B being less than or
equal to z.

Example
Consider two assets, A and B, with the following cumulative probabilities:

z F A (z) F B (z)

0 0 1

1 1 4

2 4 5

1 1 1

Asset A first-degree stochastically dominates asset B because for every value of z,

F (z) ≤ F (z).
A B

Necessary and Sufficient Condition

For a non-satiable individual, A ⪰ B if and only if their utility function u is strictly
1

increasing. For all continuous utility functions:

E[u(1 + r A )] ≥ E[u(1 + r B )]

where r and r are the rates of return on assets A and B respectively.

A B

Proof

Page 2
 Created by Turbolearn AI

The proof involves integrating by parts and demonstrating that if F (z) ≤ F (z) for A B

all z, and u(z) is increasing, then the expected utility from asset A is greater than or
equal to that from asset B.

Conversely, if A is preferred to B, then it must be true that F A

(z) ≤ F B (z) for all z.

2.3 Characterization of First-Degree Stochastic Dominance

Asset A first-degree stochastically dominates asset B if and only if the rate of return
on asset A can be expressed as the rate of return on asset B plus a non-negative
random variable ϵ.

r A = r B + ϵ, where ϵ ≥ 0

This means that for any increasing utility function u(⋅), E[u(r A )] ≥ E[u(r B )] .

Combining the results from Sections 2.2 and 2.3, the following statements are
equivalent:

1. A ⪰ B 1

2. F (z) ≤ F (z) ∀z ∈ [0, 1]

A B

3. r = r + ϵ, where ϵ ≥ 0
A B

If A ⪰ B, then asset A must have at least as high an expected return as asset B. The
1

converse is not true.

2.5 Second-Degree Stochastic Dominance (SSD)

Definition
Asset A second-degree stochastically dominates asset B (denoted as A ⪰ B) if all 2

risk-averse individuals prefer asset A to asset B.

Condition
z
∫ [F A (y) − F B (y)]dy ≤ 0 ∀z ∈ [0, 1]
0

Define S(z) = ∫ z

0
. If S(z) ≤ 0 for all z, then A ⪰ B.
[F A (y) − F B (y)]dy 2

Page 3
 Created by Turbolearn AI

2.6 Proof of Sufficiency

The proof involves using integration by parts and the properties of risk-averse utility
functions (concavity) to show that if S(z) ≤ 0 for all z, then
E[u(1 + r )] ≥ E[u(1 + r )].
A B

2.7 Proof of Necessity

The proof involves assuming that A ⪰ B and showing that it must be the case that
2

S(z) ≤ 0 for all z. This is done by contradiction; if S(z) > 0 for some z, a concave

utility function can be constructed such that the expected utility from B is greater
than that from A, contradicting the assumption that A ⪰ B. 2

2.8 Rothchild and Stiglitz's Characterization

Rothschild and Stiglitz (1970) provided another characterization of second-degree
stochastic dominance:

Asset A second-degree stochastically dominates asset B if and only if asset B has

the same distribution as asset A plus noise.

Formal Definition
A ⪰ B if and only if:
2

1. E[r ] = E[r
A B
]

2. S(z) = ∫ [F
z

0
A
(y) − F B (y)]dy ≤ 0 ∀z ∈ [0, 1]

An equivalent condition is that r B

= rA + ϵ , where E[ϵ] = 0.

Implications
If A ⪰ B, then asset A must have at least the same expected return as asset B, and
2

V ar(r ) ≤ V ar(r ). However, the converse is not necessarily true.

A B

When one of the above conditions is satisfied, it can be said that asset B is more
risky than asset A.

Page 4
 Created by Turbolearn AI

2.9 Monotonic Dominance

A related concept is that of monotonic dominance, where A monotonically
dominates B if all risk-averse and non-satiable individuals prefer A to B.

2.10 Investment in a Risky Asset

Consider an individual investing in a risky asset A with return r and a riskless asset
A

with return r.

The optimal amount to invest in the risky asset is determined by the individual's
utility function u(⋅) and is found where:
′
E[u (1 + r + α(r A − r))(r A − r)] = 0

where α is the amount invested in the risky asset.

Scenario
Imagine there is another risky asset B that is more risky than asset A (A ⪰ B). 2

Question
If an individual invests in risky asset B and the riskless asset, will the optimal
investment in the risky asset increase or decrease?

If it is given that:
′′
V (z) < 0

where V (z) = u (1 + r)/[1 + α(z − r)], the utility is concave.

′

Taking the derivative V (z) twice:

′′ α ′ ′ ′′′
V (z) = [1 − R R (z)]u (z) + [R R (z) − R A (z)]u (z) − u (z)r
1+α(z−r)

Where:

R A (z) is the absolute risk aversion

RR (z) is the relative risk aversion

Page 5
 Created by Turbolearn AI

2.11 More Risk Averse Individuals

A more risk-averse individual requires a higher risk premium to be equally willing to
invest in the same risky asset as a less risk-averse individual.

2.12 Ross' Stronger Measure of Risk Aversion

Ross (1981) proposed a stronger measure of risk aversion. An individual is more risk-
averse in Ross's sense if:
′ ′
u (z 1 ) u (z 1 )
k j

′ ≥ ′ f or all z1 < z2
u (z 2 ) u (z 2 )
k j

where u and u are utility functions for individuals k and j.

k j

2.13 Characterization of Stronger Risk Aversion

An individual k is more risk-averse than individual j if there exists a decreasing and
concave function G such that:
′ ′
u (z) = G(u j (z)) + c
k

where c is a strictly positive constant.

Page 6