Extremal behavior of stochastic integrals driven by regularly varying Lévy processes (Q879257): Difference between revisions

Consider a \(d\)-dimensional Lévy process \(X= (X^1_t,\dots, X^d_t)_{0\leq t\leq 1}\), which is regularly varying with index \(\alpha> 0\), and a càdlàg predictable stochastic process \(Y= (Y^1_t,\dots, Y^d_t)_{0\leq t\leq 1}\), which satisfies the momentum condition \(\mathbb{E}[\sup_{0\leq t\leq 1}\| Y_t\|^{\alpha+\varepsilon}]< \infty\), for some \(\varepsilon> 0\). It is known that the extremal behaviour of \(X\) is due to one large jump. The aim of this work is to establish that the same holds for the stochastic integral \[ (Y\cdot X)_{0\leq t\leq 1}:= \Biggl(\int^t_0 Y^1_s \,dX^1_s,\dots, \int^t_0 Y^d_s \,dX^d_s\Biggr)_{0\leq t\leq 1}. \] Precisely, the main result states that \[ \lim_{n\to\infty}\,\mathbb{P}\Biggl[\delta\Biggl({1\over n} Y\cdot X,{1\over n} Y_\tau\Delta X_\tau 1_{[\tau, 1]}\Biggr)> \varepsilon/\| Y\cdot X\|_\infty> n\Biggr]= 0 \] and \[ \lim_{n\to\infty}\, \mathbb{P}\Biggl[\delta\Biggl({1\over n} Y\cdot X,{1\over n} Y_\tau\Delta X_\tau 1_{[\tau, 1]}\Biggr)> \varepsilon/|Y_\tau\Delta X_\tau|> n\Biggr]= 0, \] where \(\delta\) denotes a metric on the space of càdlàg trajectories, and \(\tau\) the time of the largest jump of \(X\). Moreover, it is also shown that the stochastic integral \(Y\cdot X\) is regularly varying with index \(\alpha\), too, and its limit measure is calculated.