On the depletion problem for an insurance risk process: new non-ruin quantities in collective risk theory (Q903681)

The authors study the depletion problem of the reserve for Lévy insurance risk processes, i.e., how fast and how frequent drawdowns of a certain size occur, and derive expressions for the distribution of several depletion-related random variables. Let \(X = (X_{t})_{t \geq 0}\) be a spectrally negative Lévy process starting at an initial surplus \(x \geq 0\). Consider the following notions: {\parindent=0.6cm \begin{itemize}\item[{\(\bullet\)}] The running infinum and supremum of \(X\): \[ \underline X_{t}=\inf_{0 \leq s \leq t} X_{s},~\overline X_{t}=\sup_{0 \leq s \leq t} X_{s}. \] \item[{\(\bullet\)}] The drawdown process \(Y\): \(Y_{t} = \overline X_{t} - X_{t}\) \((t \geq 0)\). \item[{\(\bullet\)}] The first-passage time over a level \(a > 0\) of the drawdown process \(Y\): \[ \tau_{a} = \inf\{t \geq 0 : Y_{t} > a\}. \] \item[{\(\bullet\)}] The last time before \(t\) that \(X\) reaches its running supremum: \[ \overline G_{t}= \sup\{s \leq t : X_{s} \text{ or } X_{s-} = \overline X_{s}\}. \] \item[{\(\bullet\)}] The last time the reserve was at its maximum level prior to critical drawdown: \(\overline G_{\tau_{a}}\). \item[{\(\bullet\)}] The speed of depletion: \(\tau_{a} - \overline G_{\tau_{a}}\). \item[{\(\bullet\)}] The maximum reserve level attained before critical drawdown is observed: \(\overline X_{\tau_{a}}\). \item[{\(\bullet\)}] The minimum reserve level prior to critical drawdown: \(\underline X_{\tau_{a}}\). \item[{\(\bullet\)}] The largest drawdown observed before critical drawdown of size \(a\): \(Y_{\tau_{a}-}.\) \item[{\(\bullet\)}] The overshoot of the critical drawdown over level \(a\): \(Y_{\tau_{a}} - a.\) \end{itemize}} The authors give general expressions for the probability density functions of \(Y_{\tau_{a}-}\), \(Y_{\tau_{a}} - a\), and \(\overline X_{\tau_{a}}\); establish that \(\overline G_{\tau_{a}}\) and \(\tau_{a} - \overline G_{\tau_{a}}\) are independent; and compute the bivariate Laplace transform of \(\tau_{a}\) and \(\overline G_{\tau_{a}}\). The distributions of \(Y_{\tau_{a}-}\), \(Y_{\tau_{a}} - a\), and \(\overline X_{\tau_{a}}\) are also computed conditional on \(\{\underline X_{\tau_{a}} \geq 0\}\), i.e.\ that ruin does not occur before the critical drawdown time. These quantities are explicitly computed in three special models: the Cramer-Lundberg model with exponential claims, the gamma risk process, and the spectrally negative stable risk process.