Simone De Bonis
Simone De Bonis
← Back to Blog
StatisticsJuly 28, 2024· 4 min read

Fat and Heavy Tails: Differences and Implications

Understanding the distinction between heavy-tailed, fat-tailed, and light-tailed probability distributions and their implications for the Central Limit Theorem.

#Statistics#Probability#Quantitative Finance#Theory

Heavy-tailed distributions form a broad class of probability distributions that contrast sharply with light-tailed distributions—the family of distributions whose tails decay at least as fast as an exponential or Gaussian distribution. In heavy-tailed distributions, tail probabilities decay much more slowly, giving extreme events a non-negligible probability of occurring.

Defining Heavy Tails

A random variable XX with cumulative distribution function F(x)F(x) and survival function Fˉ(x)=P(X>x)=1−F(x)\bar{F}(x) = P(X > x) = 1 - F(x) is said to have a heavy right tail if its moment generating function is infinite for all t>0t > 0:

lim sup⁡x→∞etxFˉ(x)=∞∀t>0\limsup_{x \to \infty} e^{tx} \bar{F}(x) = \infty \quad \forall t > 0

Equivalently, the tail decays slower than any exponential distribution e−txe^{-tx}.

Heavy Tails vs. Fat Tails

While the terms are sometimes used interchangeably, in quantitative finance and extreme value theory a useful distinction is made:

  1. Heavy-Tailed Distributions (Sub-exponential decay): Any distribution whose tail is not exponentially bounded (e.g., Log-normal, Weibull with shape k<1k < 1, Pareto).
  2. Fat-Tailed Distributions (Power-law decay): A stricter subset of heavy-tailed distributions whose survival function decays asymptotically as a power law (regularly varying tail):
Fˉ(x)∼L(x)x−αas x→∞,α>0\bar{F}(x) \sim L(x) x^{-\alpha} \quad \text{as } x \to \infty, \quad \alpha > 0

where L(x)L(x) is a slowly varying function and α\alpha is the tail index. For a fat-tailed distribution, all moments of order m≥αm \ge \alpha are infinite:

E[∣X∣m]=∞for m≥α\mathbb{E}[|X|^m] = \infty \quad \text{for } m \ge \alpha

For example, the Log-normal distribution is heavy-tailed (slower than exponential) but not fat-tailed because all of its moments are finite. By contrast, the Student-tt (with ν\nu degrees of freedom), Cauchy (α=1\alpha = 1), and Pareto distributions are genuinely fat-tailed.

Central Limit Theorem Implications

When the tail index satisfies α≤2\alpha \le 2, the variance Var(X)\text{Var}(X) is infinite (α<2\alpha < 2) or borderline, and the classical Central Limit Theorem no longer guarantees rapid convergence of sample means to a Gaussian distribution:

  • Slow Convergence of Sample Moments: Even when 2<α<42 < \alpha < 4 (finite variance, infinite kurtosis), sample variance and empirical risk metrics converge extremely slowly and remain dominated by single extreme observations.
  • Generalized Central Limit Theorem: For α∈(0,2)\alpha \in (0, 2), properly normalized sums of i.i.d. random variables converge not to a Normal distribution, but to an α\alpha-stable (Lévy stable) distribution.
  • Risk Modeling Takeaway: Assuming Gaussianity in financial returns or energy price spikes leads to severe underestimation of tail events (>5σ> 5\sigma moves).
← Back to Blog
Share article: