Perpetual Integrals for Lévy Processes

Leif Döring, Andreas E. KyprianouJournal of Theoretical Probability 2015 · DOI

TL;DR

This paper tackles the question of when the "perpetual integral" of a Lévy process, representing the total accumulated value over an infinite horizon, remains finite. The authors establish a precise condition for this finiteness, linking it to the integral of a specific function over the process's state space, and prove a 0-1 law for this property. This work is crucial for understanding the long-term behavior of stochastic systems in finance, insurance, and physics.

What Problem Does This Paper Address?

Imagine a financial model where you're continuously earning income from an investment that follows a Lévy process (a type of random walk with jumps). The "perpetual integral" represents the total accumulated wealth over an infinitely long time. A critical question is: under what conditions will this total wealth not grow infinitely large? This is akin to asking if a perpetuity in finance has a finite present value. While this problem has been studied for simpler processes like Brownian motion, extending it to the more general and complex Lévy processes, which can exhibit jumps, presents significant challenges. The authors aim to provide a precise, universally applicable condition for the finiteness of this perpetual integral for a broad class of Lévy processes.

Methods

The paper employs a sophisticated blend of probabilistic tools. A cornerstone of their approach is the occupation time formula, which relates the integral of a function over the process's path to the integral of that function against the process's local times. They leverage Jeulin's lemma, a powerful result that connects the finiteness of an integral of a random variable against a function to the integral of that function itself. Crucially, they also utilize the Hewitt-Savage 0-1 law, which states that any "exchangeable" event in a sequence of independent and identically distributed random variables must have a probability of either 0 or 1. This allows them to prove a definitive 0-1 law for the finiteness of the perpetual integral. The proof also involves concepts like spatial stationarity and randomized initial conditions to handle the complexities of Lévy processes.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • The Integral Test is Key: The most striking insight is the direct link established between the finiteness of the perpetual integral and the finiteness of the integral of the function f over the process's state space. This provides a clear and actionable criterion.
  • A Definitive 0-1 Law: The paper proves that for a broad class of Lévy processes, the perpetual integral is either almost surely finite or almost surely infinite. There's no in-between. This simplifies the analysis significantly.
  • Leveraging Lévy Process Structure: The authors cleverly exploit the specific properties of Lévy processes, such as their independent increments and the existence of local times (under certain conditions), to adapt existing tools like Jeulin's lemma.
  • The Role of Local Times: While the paper proves the result for processes with local times, they acknowledge that it's unclear if this assumption is strictly necessary for all cases, hinting at potential avenues for generalization.

Key Results

The main result is Theorem 1, which states that for a Lévy process with a strictly positive mean, local times, and that is not a compound Poisson process, the perpetual integral $\int_0^\infty f(\xi_s) ds$ is finite if and only if $\int_{-\infty}^\infty f(x) dx < \infty$. This is accompanied by a 0-1 law: $P(\int_0^\infty f(\xi_s) ds < \infty) \in \{0, 1\}$. The paper also provides a detailed comparison of their results with existing literature for Brownian motion and spectrally negative Lévy processes, highlighting the generality of their findings.

Limitations and Future Work

The authors note that the assumption of the Lévy process having local times might not be strictly necessary for the result to hold, particularly for decreasing functions f. They express uncertainty about whether this assumption is essential for general f, suggesting this as an open question. Future work could explore relaxing this assumption and investigating the behavior of perpetual integrals for Lévy processes with different properties, such as those with infinite mean or those that are compound Poisson processes.

Summary updated

BibTeX
@article{doring2015perpetual,
  author = {Leif Döring and Andreas E. Kyprianou},
  title = {{Perpetual Integrals for Lévy Processes}},
  journal = {Journal of Theoretical Probability},
  year = {2015},
  doi = {10.1007/s10959-015-0607-y},
}

Loading related content...

Figure