Risk & Modeling

Student's t-Distribution

TL;DR

The Student's t-distribution is a bell curve with heavier tails than the normal distribution, controlled by a degrees-of-freedom (DOF) parameter. It's the standard model for capturing the fat-tailed behavior of real financial markets in Monte Carlo simulations.

The Student's t-distribution resembles the familiar normal (Gaussian) bell curve but with heavier tails — meaning extreme values occur more frequently. The thickness of the tails is controlled by a single parameter: degrees of freedom (DOF). Lower DOF produces fatter tails; as DOF approaches infinity, the distribution converges to the normal.

How It Works

The key relationship between DOF and tail behavior:

DOFExcess KurtosisBehavior
4∞ (undefined)Extremely fat tails
56.0Fat tails — matches equity markets
83.0Moderate tails
300.2Nearly normal
0Exactly normal

Retirement Lab generates t-distributed returns using the Marsaglia-Tsang gamma sampling method — an O(1) algorithm that efficiently produces Student-t random variables from a seeded PRNG.

Interactive chart: students-t-dof

Student-t distributions at DOF = 4, 5, 8, and 30 vs. normal

Coming soon

Why It Matters for Retirement Planning

The Student-t distribution is the workhorse model for fat-tailed retirement simulations. Combined with the Fernandez-Steel transformation for skewness, it captures two critical properties of real markets:

  • Excess kurtosis: extreme events occur more often than a bell curve predicts
  • Negative skewness: large losses are more likely than large gains of equal magnitude

This combination produces more realistic worst-case scenarios, typically reducing success rates by 5-10 percentage points compared to normal distribution assumptions — revealing hidden risk in retirement plans.

Frequently Asked Questions

What degrees of freedom should I use for stock market returns?
Empirical research suggests DOF values of 4-6 for monthly equity returns. Retirement Lab defaults to DOF=5, which produces excess kurtosis of 6 — closely matching observed market behavior. Lower values (DOF=4) model even more extreme tail events; higher values (DOF=10+) approach a normal distribution.
How is the Student-t distribution different from the normal distribution?
Both are symmetric and bell-shaped, but the Student-t has heavier tails. At DOF=5, extreme events (3+ standard deviations) occur roughly 4 times more frequently than under a normal distribution. As DOF increases toward infinity, the t-distribution converges to the normal distribution.