Corrections to the Central Limit Theorem for Heavy-tailed Probability Densities

Henry Lam, José Blanchet, Damian Burch, Martin Z. Bazant

TL;DR

This paper extends the classical Edgeworth expansion—a method for refining the Central Limit Theorem (CLT)—to "heavy-tailed" distributions where higher-order moments diverge. By utilizing special functions like Dawson’s integral and parabolic cylinder functions, the authors provide a new set of correction terms that bridge the gap between the central limit region and large-deviation asymptotics.

What Problem Does This Paper Address?

The Central Limit Theorem is the bedrock of statistics, telling us that the sum of many independent variables tends toward a normal (Gaussian) distribution. However, the CLT is just a first-order approximation. To get more accuracy, mathematicians use Edgeworth expansions, which add "correction terms" to account for skewness and kurtosis.

The catch? Standard Edgeworth expansions require finite higher-order moments (like the 3rd or 4th moment). In the world of "heavy-tailed" or "fat-tailed" distributions—common in finance, physics, and insurance—these moments often diverge (become infinite). When moments diverge, the standard expansion "hits a wall" and cannot provide further refinements. This paper solves that problem by deriving new correction terms specifically for distributions with power-law tails (regularly varying distributions) where higher moments do not exist.

Methods

The authors employ a standard but technically rigorous approach:

  1. Fourier Analysis: They analyze the characteristic function (the Fourier transform of the probability density).
  2. Decomposition: They split the characteristic function into an analytic component (which leads to the standard Edgeworth terms) and a non-analytic component (which arises from the heavy tail).
  3. Inversion: The core innovation lies in how they perform the inverse Fourier transform on this non-analytic part.
  4. Special Functions: Unlike standard expansions that rely on Hermite polynomials, these new corrections are expressed through Dawson’s integral and parabolic cylinder functions.

The derivation covers three main scenarios: symmetric tails, asymmetric tails where one side is "heavier" than the other, and cases where the power-law exponent is an integer versus a non-integer.

Datasets

As a theoretical probability paper, no empirical datasets were used. The authors validate their results through mathematical proofs and by comparing their expansions to known Pareto distributions (a classic example of a heavy-tailed power law).

Key Insights and Ideas

  • The Parity Effect: The nature of the correction changes qualitatively depending on the tail exponent $\alpha$. There are distinct mathematical forms for when $\alpha$ is even, odd, or a non-integer.
  • Logarithmic Corrections: One of the most subtle findings is that if the distribution has an even number of finite moments, the expansion requires additional logarithmic terms ($\log n$) to remain accurate.
  • The "Smooth Transition": Traditionally, the CLT describes the "center" of a distribution, while Large Deviation Theory describes the "far tails." This paper shows that their new correction terms naturally "blend" into large-deviation asymptotics. This provides a unified mathematical bridge between the two regimes.
  • Non-Analytic Dominance: While standard Edgeworth terms eventually vanish as you move further from the mean, the authors' new heavy-tail corrections actually become the dominant source of error in the "moderate deviation" region.

Key Results

  • Theorem 1 & 2: The authors provide explicit, closed-form formulas for the probability density of a normalized sum ($S_n/n^{1/2}$). These formulas include the standard Gaussian term, the existing Edgeworth terms (up to the last finite moment), and the new heavy-tail correction term $F(x, n)$.
  • Uniformity: They prove that for certain classes of distributions (like the symmetric Pareto), their refined expansion is valid uniformly over a wide range of spatial scales, specifically for $x \leq \sqrt{(\alpha - 2) \log n}$.
  • Asymptotic Matching: They demonstrate that their non-analytic correction term matches the large-deviation results of Rozovskii (1989) exactly when applied to the spatial scales of the CLT.

Limitations and Future Work

  • Regular Variation Requirement: The results are specifically tailored for distributions with "regularly varying" tails (power laws). They may not apply to distributions with different tail behaviors (e.g., log-normal or "super-heavy" tails).
  • Smoothness Assumptions: The proofs assume the underlying densities are continuous and satisfy certain differentiability conditions on their slowly varying components.
  • Future Directions: The authors suggest that this framework could be extended to stable limits (where the sum converges to a non-Gaussian Levy distribution) and to more complex stochastic systems like queueing models in heavy traffic.

Summary updated

Abstract

Classical Edgeworth expansions provide asymptotic correction terms to the Central Limit Theorem (CLT) up to an order that depends on the number of moments available. In this paper, we provide subsequent correction terms beyond those given by a standard Edgeworth expansion in the general case of regularly varying distributions with diverging moments (beyond the second). The subsequent terms can be expressed in a simple closed form in terms of certain special functions (Dawson's integral and parabolic cylinder functions), and there are qualitative differences depending on whether the number of moments available is even, odd or not an integer, and whether the distributions are symmetric or not. If the increments have an even number of moments, then additional logarithmic corrections must also be incorporated in the expansion parameter. An interesting feature of our correction terms for the CLT is that they become dominant outside the central region and blend naturally with known large-deviation asymptotics when these are applied formally to the spatial scales of the CLT.
BibTeX
@article{lam2011corrections,
  author = {Henry Lam and José Blanchet and Damian Burch and Martin Z. Bazant},
  title = {{Corrections to the Central Limit Theorem for Heavy-tailed Probability Densities}},
  journal = {Journal of Theoretical Probability},
  year = {2011},
  doi = {10.1007/s10959-011-0379-y},
}

Loading related content...

Figure