In my text (Investments by BKM), the investor's mean-variance utility (given as $U = E[R] - \frac12A\sigma^2$) is stated to be the objective function we wish to maximize. Upon further digging, it seems that this stems from the assumption of quadratic utility functions ($U = aW - bW^2$). This kind of bothers me since I see two unrealistic properties for quadratic utility functions. (1) They exhibit increasing absolute risk aversion, and (2) they achieve a satiation point, beyond which money/return begins to have negative value.

So why do we assume quadratic utility? Are there no other simple, more realistic functional forms for utility that would still lead to a reasonably clean portfolio optimization theory? Or are the issues I cited about the quadratic just negligible in practice?

  • 5
    $\begingroup$ If utility is quadratic then we can do optimization on it. So it is convenient to have quadratic function. $\endgroup$
    – Naz
    Apr 30, 2016 at 13:10
  • $\begingroup$ It is not $U = E[R] - \frac12A\sigma^2$ but $E(U) = E[R] - \frac12A\sigma^2$. $\endgroup$ Feb 7, 2020 at 9:00

4 Answers 4


The assumption of quadratic utility function is convenient in portfolio theory because it is possible to demonstrate that if the portfolio returns are not normally distributed, the mean-variance approach is still best (best in the sense that any other distributional properties is amenable into mean and variance.) Conversely, if the return are normally distributed, the choice of utility function is irrelevant. More generally, if the portfolio return distributions are not known and we use a general utility function, the mean-variance approach is valid yet, but only as approximation.

  • 2
    $\begingroup$ More precisely, under normality the choice of utility function is irrelevant up to the constants specifying the relative importance of mean vs. variance reflecting the degree of risk aversion. $\endgroup$ Feb 7, 2020 at 9:03

What you learn in school are models, meant to illustrate the concepts and methods of the field. Later you will learn about other forms of utility functions (power utility most prominently). With such families of utility functions the computations aren't as clean as with quadratic utility, but by then you will have understood the concepts and methods, and you will understand the approximate methods that you will need to use at that stage.

With that said, Markowitz tried, in a few papers, to explain why approximating one's utility function by a quadratic utility function makes sense in some cases. Not very convincing in my opinion, but it's out there.

  • $\begingroup$ so you mean this is due to pedagogical purpose rather than pratical one? $\endgroup$
    – SiXUlm
    May 31, 2016 at 12:12
  • $\begingroup$ At this time, pretty much, yes. In real life there are almost no practitioners (people with PhD's in math and physics working in finance) who use quadratic utility functions in their work. Power utility (including logarithmic utility) are the most commonly used family. $\endgroup$
    – user20429
    May 31, 2016 at 12:17
  • $\begingroup$ I do not know myself what practitioners actually use, but in this answer there is a counterargument to yours. See paragraphs starting with First, most quants and Why do the Nobels matter?. $\endgroup$ Feb 7, 2020 at 9:16
  • $\begingroup$ @RichardHardy - I do (or did) know what practitioners actually use. I am a mathematician by training (PhD, taught up to graduate level courses at a top university - including in statistics and in financial math). Math is the avenue by which I was able to enter the actual practice of finance, retiring in 2015 as managing partner of a firm that gives investment advice based on modern portfolio theory and such. It is true that most people use mean-variance models; the Answer you linked to, in some places, seems to confuse between that and "quadratic utility". $\endgroup$
    – user20429
    Feb 7, 2020 at 20:04
  • $\begingroup$ @RichardHardy - Mean-variance models work equally validly with other classes of utility functions - as long as utility is a two-parameter family (so, it is completely determined by its mean and variance). "Power utility", for example, is in this class. But the math calculations can no longer be done in closed form. A large number of "quants" do use power utility; honestly, I can't recall ever meeting one who assumed quadratic utility. And the bit about the Nobels providing a legal safe harbor is pure BS - that is exactly the business I was in, and there is no such safe harbor. $\endgroup$
    – user20429
    Feb 7, 2020 at 20:07

In most settings, utility functions are defined up to an affine transformation: if $u(x)$ defines the preference of an investor, then so does $a*u(x)+b.$ This implies, you can normalize the Taylor expantion of any smooth utility function to $u(x)=x+a*x^2+\ldots$ around 0. So the next step is just to drop off higher order terms. The investor is also usually assumed to be risk-averse, which implies, that $a < 0.$ You can check the details, e.g. here: https://www.empiwifo.uni-freiburg.de/lehre-teaching-1/winter-term-10-11/materialien-portfolio-analysis/utility.pdf


The capital asset pricing model (CAPM) is based on mean-variance utility; investors choose their portfolio based only on its mean and variance.

This is an entirely different approach than expected utility maximisation.

You can express the expected utility using its Taylor approximation around the expected wealth, with $\mu = E[W], \Delta = W - \mu$ (note that $E[\Delta]=0$):

\begin{align} E[U(W)] &= E[U(\mu + \Delta)] \\ &= E\left[U(\mu) + \Delta \cdot U' + \frac 1 2 \Delta^2 U'' + \frac 1 6 \Delta^3 U''' + ...\right] \\ &= U(\mu) + \frac 1 2 \sigma^2 U''(\mu) + \frac 1 6 E[\Delta^3] U''' + ... \end{align}

Now, to turn this into a mean-variance utility (ie only a function of $\mu, \sigma$), there are 3 ways:

  1. Impose conditions on the utility function: Take an expected utility function $U$ whose higher derivatives (above second derivative) vanish. That is the idea with quadratic utility (which, as you point out, has many problems).

  2. Impose conditions on the returns, ie the distribution of $W$. It can be shown that one can express the above as a function of only $\mu, \sigma$ for elliptical distributions, no matter what $U$ is chosen. Those can have a restricted domain (ie limited liability, though then also limited upside, if I'm not mistaken).

  3. Impose joint conditions on both the expected utility function and the distribution of $W$. That's complicated.

So why do we assume quadratic utility?

Easy to handle analytically.

Are there no other simple, more realistic functional forms for utility that would still lead to a reasonably clean portfolio optimization theory?

Mean-variance utility has many problems. For example, a mean-variance investor could refuse a gift of a limited liability asset, if it's volatile enough. It doesn't make much sense.

Not aware of a clean portfolio optimisation theory under more realistic assumptions.

Or are the issues I cited about the quadratic just negligible in practice?

Oh no, they are real. But mean-variance analysis or CAPM generally are hardly used in practice, but rather as conceptual tools, I'd say.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.