Perpetual integrals for Lévy processes (Q325924)

The aim of this short but dense paper is to extend a few previous results concerning the finiteness of \textit{perpetual integrals} \[ \int_0^\infty f(\xi_s)ds \] to more general Lévy processes. Such integrals arise in financial mathematics, where -- with \(f\) is a positive function and \(\xi_s\) a process -- they are interpreted ``as the present value of a continuous stream of perpetuities''. It is indeed acknowledged today that it would be of great importance to have clear and easy-to-apply criteria (in terms of \(f\)) under which a perpetual integral is finite, and several results are known (and quoted by the authors) when \(\xi\) is either a Brownian motion or, under special conditions, a Lévy process. The present paper, as clearly stated by the authors from the beginning, extends such results in the following way: if {\parindent=0.6cm\begin{itemize}\item[--] \(\xi\) is a Lévy process with strictly positive mean \(\mu<\infty\) and local times, and is not a compound Poisson process \item[--] \(f\) is a measurable, locally integrable, positive function, \end{itemize}} then the following \(0\)-\(1\) law holds: \[ \mathbb{P}\left(\int_0^\infty f(\xi_s)ds<\infty\right)=1 \;\Longleftrightarrow\; \int_0^\infty f(x)dx<\infty,\qquad (\mathrm{T}) \] while the same probability vanishes when the integral of \(f\) diverges. The rest of the paper is finally devoted to a complete discussion of this theorem -- in particular of the \textit{necessity} of the integral test \((\mathrm{T})\) -- whereas the authors also claim in a preliminary \textit{Remark} that ``It is not clear whether or not the assumption that \(\xi\) has local time is necessary for \((\mathrm{T})\) to hold. For (ultimately) decreasing \(f\), the existence of local time is clearly not needed, whereas we have no conjecture for general \(f\).'' The proof is based on an adapted (weaker) version of the Jeulin lemma which is duly discussed in this second part of the article.