Multivariate CLT for Lévy processes: convergence rates without moment assumptions ↗
Authors: J. González Cázares, D. Kramer-Bang, and A. Mijatović (Bernoulli, 2026+)
Mathematical Snippet & Key Result
For a $d$-dimensional Lévy process $X_t$ with zero mean and unit covariance matrix $I_d$, the Central Limit Theorem holds as $t \to \infty$ ($t^{-1/2}X_t \xrightarrow{d} N_d(0,I_d)$). Without assuming $2+\delta$ moments, we prove that finite second moment is equivalent to the Dini-type integrability of the convex distance $d_c(t)$ under a modified jump-truncated covariance scaling $\sqrt{t}\Delta_t$:
$$\mathbb{E}\big[\|X_1\|^2\big] < \infty \iff \int_1^\infty d_c^{\mathrm{mod}}(t)\,\frac{dt}{t} < \infty.$$
In contrast, under the classical $\sqrt{t}$ scaling, $\int_1^\infty d_c(t)\,\frac{dt}{t} < \infty$ requires an extra $x^2 \log(1+x)$ moment.