layout: default title: Utility Functions and Risk Aversion ——————————————
Utility Functions and Risk Aversion
Sukrit Mittal Franklin Templeton Investments
Outline
- Why utility?
- Preferences and rational choice
- Axioms of expected utility
- Expected utility maximization
- Common utility functions
- Risk aversion
- Utility, mean–variance, and CAPM
- Exercises
1. Why Utility?
Up to now, we used mean–variance preferences.
They were convenient.
They were also restrictive.
Utility theory answers a deeper question:
What does it mean to choose rationally under uncertainty?
What Utility Is Not
Utility is not:
- Money
- Happiness
- Psychological satisfaction
Utility is a numerical representation of preferences.
Nothing more. Nothing less.
2. Preferences and Choice
Consider uncertain outcomes (lotteries).
An investor can rank them:
- $A \succ B$: $A$ preferred to $B$
- $A \sim B$: indifference
Utility assigns numbers consistent with these rankings.
Rationality as Consistency
Rationality does not mean intelligence.
It means:
- Consistent choices
- Stable preferences
Economics asks for coherence, not wisdom.
3. Axioms of Expected Utility
To represent preferences by expected utility, we assume:
- Completeness: choices are comparable
- Transitivity: no preference cycles
- Continuity: no infinite jumps
- Independence: irrelevant alternatives do not matter
These axioms are demanding.
They are also explicit.
Expected Utility Theorem
If preferences satisfy the axioms, then:
There exists a utility function $u(\cdot)$ such that preferences are represented by expected utility.
That is:
\[U = \mathbb{E}[u(W)]\]This is a representation result.
Not a psychological claim.
4. Expected Utility Maximization
Given wealth $W$ as a random variable:
The investor chooses portfolios to:
\[\max ; \mathbb{E}[u(W)]\]Subject to budget constraints.
This is the most general formulation of choice under uncertainty.
Utility Is Ordinal
If $u$ represents preferences, so does:
\[\tilde u = a u + b \quad (a>0)\]Only rankings matter.
Levels do not.
This subtle point is often forgotten.
5. Common Utility Functions
Quadratic Utility
\[u(W) = W - \frac{\gamma}{2}W^2\]- Leads to mean–variance preferences
- Analytically convenient
- Implies increasing absolute risk tolerance (problematic)
Useful. Not realistic.
Exponential Utility (CARA)
\[u(W) = -e^{-\gamma W}\]- Constant absolute risk aversion
- Wealth-independent risk attitudes
- Tractable under normality
Popular in theory.
Power Utility (CRRA)
\[u(W) = \frac{W^{1-\gamma}}{1-\gamma}\]- Constant relative risk aversion
- Scale-invariant behavior
- Widely used in macro and asset pricing
This one survives contact with data better.
6. Risk Aversion
Risk aversion captures dislike for uncertainty.
Formally:
An investor is risk-averse if she prefers the expected value to the lottery.
Mathematically:
\[\mathbb{E}[u(W)] \le u(\mathbb{E}[W])\]Concavity and Risk Aversion
Risk aversion is equivalent to:
\[u''(W) < 0\]Utility is concave.
This single inequality drives everything.
Arrow–Pratt Measures
- Absolute risk aversion:
- Relative risk aversion:
These quantify attitudes toward risk.
7. Utility and Mean–Variance
Under:
- Quadratic utility, or
- Normally distributed returns
Expected utility reduces to:
\[\mathbb{E}[W] - \frac{\gamma}{2}\text{Var}(W)\]Mean–variance is not a heuristic.
It is a special case.
8. Utility and CAPM
CAPM assumes:
- Mean–variance behavior
- Homogeneous beliefs
These are justified by:
- Quadratic utility, or
- Normal returns
CAPM rests on fragile foundations.
But foundations nonetheless.
What Changes with General Utility?
- Market portfolio need not be mean–variance efficient
- Pricing relations become nonlinear
CAPM is elegant.
Reality is not.
9. Exercises
Exercise 1
Show that concavity of $u$ implies risk aversion using Jensen’s inequality.
Exercise 2
Compute absolute and relative risk aversion for:
\[u(W) = -e^{-\gamma W}\]Interpret the result.
Exercise 3
Explain why mean–variance preferences may fail for skewed return distributions.
Final Takeaways
- Utility formalizes rational choice under uncertainty
- Risk aversion is concavity
- Mean–variance is a special case of expected utility
- CAPM relies on strong utility assumptions
Next, we move beyond expected utility.
Because investors do.