Note on FSD and SSD
Shuoxun Hellen Zhang
WISE & SOE
XIAMEN UNIVERSITY
March, 2015
1 / 17
First-order stochastic dominance
Definition
First-order stochastic dominance(FSD)
Let Fx (W ) and Gy (W ) be two cumulative probability
distributions for random payoffs in [a, b]. We say that
x FSD y if and only if for all W :
Fx (W ) ≤ Gy (W )
and for some Wi
Fx (W ) < Gy (W )
In other words, the pdf defined on wealth for asset y always
lies to the left of the pdf of x, then x FSD y .
2 / 17
First-order stochastic dominance
Figure: An example of FSD assuming that the distribution of
wealth provided by both assets is a normal distribution.
3 / 17
Risk-Aversion
First-order stochastic dominance applies to all nondecreasing
utility functions.
Figure: Three utility functions with positive marginal utility: (a)
risk lover; (b) risk neutral; (c) risk averter
4 / 17
Risk-Aversion
I If U[E (W )] < E [U(W )] or the utility function is strictly
convex, then the individual is a risk lover;
I If U[E (W )] = E [U(W )] or the utility function is linear,
then the individual is risk neutral;
I If U[E (W )] > E [U(W )] or the utility function is strictly
concave, then the individual is risk averse;
5 / 17
First-order stochastic dominance
Theorem
Let Fx (W ) and Gy (W ) be two cumulative probability
distributions for random payoffs in [a, b]. x FSD y if and only
if
E [u(x)] ≥ E [u(y )]
for any nondecreasing Bernoulli utility function u.
6 / 17
First-order stochastic dominance
Figure: First-order stochastic dominance and expected utility.
7 / 17
First-order stochastic dominance
For a given frequency of wealth, fi (W ), the increasing utility
function assigns higher utility to the level of wealth offered by
asset x than by asset y . This is true for every frequency.
The expected utility is defined by,
Z ∞
E [u(W )] = u(W )f (W )dW
−∞
8 / 17
Second-order stochastic dominance
Definition
second-order stochastic dominance(SSD)
Let Fx (W ) and Gy (W ) be two cumulative probability
distributions for random payoffs in [a, b]. We say that
x SSD y if and only if for any W :
RW
[G (W ) − Fx (W )]dW ≥ 0
−∞ y
Gy (Wi ) 6= Fx (Wi ) for some Wi
9 / 17
Second-order stochastic dominance
second-order stochastic dominance not only assumes utility
functions where marginal utility of wealth is positive, but also
assumes that total utility must increase at a decreasing [Link]
other words, utility functions are nondecreasing and strictly
concave. Thus individuals are assumed to be risk averse.
10 / 17
Second-order stochastic dominance
Figure: An example of SSD
11 / 17
Second-order stochastic dominance
Asset x will dominate asset y if an investor is riskaverse
because they both offer the same expected level of wealth
µx = µy because y is riskier.
This means that in order for asset x to dominate asset y for
all risk-averse investors, the accumulated area under cdf of y
must be greater than the accumulated area for x, below any
given level of wealth.
This implies that, unlike FSD, the cdf can cross.
12 / 17
Second-order stochastic dominance
SSD requires that the difference in areas under the cdf be
positive below any level of wealth Wi .
Figure: Graphic representation of the sum of the differences in
cumulative probabilities
13 / 17
Second-order stochastic dominance
Theorem
Let Fx (W ) and Gy (W ) be two cumulative probability
distributions for random payoffs in [a, b]. x SSD y if and only
if,
E [u(x)] ≥ E [u(y )]
for any nondecreasing and concave Bernoulli utility
function u
14 / 17
Second-order stochastic dominance
15 / 17
Second-order stochastic dominance
The concave utility function of a risk averter has the property
of decreasing marginal utility.
If we select a given frequency of wealth it maps out equal
changes in wealth 4W1 and 4W2 . The difference in utility
between x and y below the mean is much greater than the
difference in utility for the same change in wealth above the
mean. Thus, if we take the expected utility by pairing all such
differences with equal probabilities, the expected utility of x is
seen as greater than the expected utility of y .
If the individual were risk neutral, with a linear utility function,
the differences in utility above and below the mean would
always be equal. Hence a risk-neutral investor would be
indifferent between x and y .
16 / 17
Stochastic dominance
Stochastic dominance is founded on the basis of expected
utility maximization and it applies to any probability
distribution. This is because it takes into account every point
in the probability distribution.
We can be sure that if an asset demonstrates SSD, it will be
preferred by all risk-averse investors, regardless of the specific
shape of their utility functions.
We can use stochastic dominance as the basis of the complete
theory of how risk-averse investors choose among various risky
assets. All we need to do is to find the set of portfolios that is
stochastically dominance.
17 / 17