INSTRUCTOR: Next, we are going to develop a concept called

risk aversion, or to be more precise,

relative risk aversion.
This is the key property of the utility
function, which reflects the property of investor behavior.
How averse is this investor to risky outcomes?
We are going to start with the basic fundamental properties
of the utility function itself.
Some desirable properties.
For example, if we want the investor to prefer more
to less, which is very common, mathematically it's
going to imply that, when confronted
with a choice between gamble x or gamble
x plus positive epsilon, the investor
is going to prefer x plus epsilon.
What it means is that to capture preferences like these,
preference for more to less, an investor
would have to have an increasing utility function.
So far, so good.
Utility function has to be increasing.
What more can we say about it's shape?
As it happens, if an investor is averse to risk,
the utility function also has to be concave.
Now why is that?
Suppose an investor is presented with a risky gamble, x.
The choice is, take the risk, play out, gamble x,
collect the outcomes.
Or receive a sure thing, a pay off without risk
equal to the expected value of x.
Between x random, expected value of x, sure thing.
What would the investor choose?
Now most investors are averse to risk.
They would rather collect a sure thing,
a fixed payoff equal to the expected value of the gamble.
Mathematically, it means that the utility function
has to be concave.
If the utility function is concave,
then the mean of the gamble is always
preferred to the gamble itself.
To illustrate the concept of risk aversion graphically,
and to see how this is related to concavity of the utility
function, look at the graph.
The blue line is the utility function
plotted against the outcome of the gamble.
This function is concave.
Now imagine that we are playing a binary gamble, as before.
Two outcomes, 1.2, 2.8.
Both outcomes are equally likely.
If the investor evaluates this gamble using
the expected utility model, the expected utility
is going to be seen on the graph as the dot in the middle
of the red segment.
This is the line connecting two points, utility of 1.2,
utility of 2.8.
Now this is the expected utility of the random payoff.
What if instead, the investor was
presented with the sure thing.
The expected value of the gamble.
In this case, the expected value is 2.
The utility evaluated at the expected value of the gamble
is shown as the blue dot on the utility curve.
The blue dot, of course, is above the red dot.
It has to be, because, the utility function is concave.
This is the essence of what a concave function is.
So what we can see is that, based
on this graphical representation,
it's quite clear that concavity of the utility function
is equivalent to risk aversion in behavior.
Now we're going to develop a concept of risk premium.
Suppose that an investor is evaluating an investment
strategy with return x, which is random.
And an investor starts with wealth w,
invests, and collects a payoff equal to w times 1 plus x.
Let's assume that the expected return is 0.
And we are going to call the variance of chance sigma
x squared.
The risk averse investor, of course,
would prefer to receive a 0 return for sure,
a risk-free return of 0, rather than playing this gamble.
Rather than making this investment.
Now we are going to define the concept of risk premium
as a number, pi.
It's a fixed number, it's not a random variable.
Such that an investor would be indifferent between collecting
a return of minus pi or investing
into a strategy with the risk return x.

To determine what pi is, how it reflects

the properties of the utility function
and the risk of the gamble, we need
to think about investors' preferences.
Being indifferent between investing
into an asset with a random return
x versus getting a sure risk-free return of minus pi
means mathematically that expected utility
of w times 1 plus x is equal to the utility evaluated
at w times 1 minus pi, which is a riskless return of minus pi.
To analyze this equation further,
we need to make some approximations.
We are going to use a Taylor expansion to simplify
the problem.
So to start with, let's assume that
the range of outcomes of this gamble--
the support of the return x--
is relatively narrow.
So that the range of possible outcomes of x
is concentrated somewhere around 0.
Let's say between minus and plus epsilon where
epsilon is a small number.
If that is the case, then there is premium,
pi, is going to be small in magnitude as well,
which allows us to use a Taylor expansion
to approximate the problem.
With this idea in mind, we're going
to expand both sides of the equation around 0.
Let's start with the left-hand side.
Expected utility of w times 1 plus x.
We are going to expand it around x equal to 0.
First, let's expand the utility function under the expectation.
We are going to keep three leading terms in the expansion.
First, utility of w.
Then the margin utility, u prime,
evaluated to w times w times x.
That's the linear term in the Taylor expansion.
Then comes the second order, a quadratic term.
1/2 of the second derivative of u times w
squared times x squared.
When we take the expected value of this expression,
we are left with utility of w u prime times w times
the expected value of x, which by assumption is 0, plus--
and here is the important term--
1/2 of the second derivative of u,
times w squared times the variance of the return,
which is sigma x squared.
This is the approximation of the left-hand side.
The expected utility of the risk to return.
On the right-hand side, we have the utility
of w times 1 minus pi, which we also expand.
Here, we only need to keep the two leading terms, 0 order
and the first order.
We end up with u of w minus u prime times w times pi.
We end up with inequality.
On the left-hand side, we have an approximate expression
for the expected utility of the risky investment.
On the right-hand side, it's the utility of w times 1 minus pi,
the certainty equivalent, which defines the risk premium.
When we solve this equation, we see that the risk premium
is given by minus 1/2 times the ratio of the second derivative
of the utility function times w divided by u prime of w
times sigma x squared.
We see that the risk premium is a product of two objects.
The first is what we call relative risk aversion
coefficient.
This is minus w times the second derivative
of the utility function divided by the first derivative.
This object captures what we know
is the curvature of the utility function.
If the utility function were to be a straight line,
the second derivative would be 0.
Relative risk aversion would be 0.
For a properly concave utility function,
this number is positive.
The second piece, the second term, in the risk premium
is the measure of return risk.
In this case, we see that its return variance.
The conclusion is that the risk premium
that investor demands, based on the riskiness of the gamble
of the investment strategy, and based
on his or her own preferences, is comprised of two terms.
Investor's relative risk aversion,
which is the property of the utility function,
and return variance, which is the measure of return risk.
Let's consider a couple of examples of utility functions.
First, the simplest case is a linear utility.
In this case, u of w is an affine function of w.
Linear utility is special because it implies
0 relative risk aversion.
An investor with linear utility does not care about risk.
This kind of investor is going to evaluate each investment
opportunity based on the expected value of the return.
The second example of the utility function
that we want to introduce is the so-called power utility
function.
It's also called constant relative risk aversion.
We will see in a second why this name.
A power utility function is given by a power function
of investor's wealth.
And its coefficient is a positive number.
The case of the coefficient equal to 1 is special.
In this case, the utility function becomes logarithmic.
The key property of the power utility function,
and the reason why it's so commonly used,
is that the relative risk aversion
coefficient of this utility function is constant.
It's equal to the exponent in the power,
and it's independent of the level of wealth.
The implication of this is that an investor following a power
utility function is going to have the same risk
aversion, regardless of the level of wealth.
Which means that making an investment decision
with a million dollars or 5 or 10,
this investor is going to choose the same portfolio
of investments.
The level of initial wealth doesn't effect the choice.