Path decomposition of ruinous behavior for a general Lévy insurance risk process (Q453239): Difference between revisions
From MaRDI portal
Created a new Item |
Changed an Item |
||
Property / review text | |||
Consider a Lévy process \(\{X_t\}\), such that \(\lim_{t \to \infty} X_t = -\infty\). Let \(\tau(u) = \inf\{t: X_t > u\}\), \(\bar X_t = \sup \{X_s: 0 \leq s \leq t\}\), \(G_t = \sup\{0 \leq s \leq t: X_s = \bar X_t\}\). \(\tau(x)\) is then the time of ruin, and \(G_t\) is the last time before \(t\) at which the running maximum is attained. One is also interested in the shortfall at ruin, \(X_{\tau(u)}-u\), the surplus prior to ruin \(u - X_{\tau(u)-}\), the minimal surplus before ruin \(u - \bar X_{\tau(u)-}\), the time of the minimal surplus prior to ruin \(G_{\tau(u)-}\), and the time remaining to ruin from the time of the minimal surplus \(\tau(u) - G_{\tau(u)-}\). Of interest is also the event of ruin for large initial capital \(u\). One therefore considers the measure \(\operatorname{P}^{(u)}[ \cdot ] = \operatorname{P}[\cdot \mid \tau(u) < \infty]\). These quantities have been investigated for the small claims case and for the subexponential case. In the small claims case, the process behaves like the associated process, that is, there are slightly more and slightly larger claims until ruin occurs. In the subexponential case, suddenly an enormous claim occurs, and the the shortfall becomes enormous, too. In this paper, it is assumed that there is \(\alpha > 0\) such that \(\bar\Pi^+_X \in S^{(\alpha)}\) and \(\operatorname{E}[e^{\alpha X_1}] < 1\). Here, \(\bar \Pi^+_X\) is the tail of the Lévy measure, and \(S^{(\alpha)}\) is the set of measures \(G\) such that, for \(x_0 > 0\), \(F(x) = G(x)/\bar G(x_0) I_{[x_0,\infty)}\) has the property \[ \lim_{u \to \infty} {\bar F(u+x) \over \bar F(u)} = e^{-\alpha x},\qquad \lim_{u \to \infty} {\bar F^{*2}(u) \over \bar F(u)} = 2 \int e^{\alpha x} \,d F(x). \] The main result states that, conditioned on ruin, asymptotically the process behaves first like the Esscher transformed process with parameter \(\alpha\) until an exponentially distributed time. At this time point, a large claim occurs moving the process close to the capital \(u\), so that ruin may occur; and, if ruin has not occurred, the process behaves like the process conditioned on ruin with a moderate initial capital. Using this path decomposition, the asymptotic distribution of the quantities of interest can be found. In particular, a (more or less) explicit formula for the distribution of the time to ruin is found. This means, the time of ruin converges weakly to a finite variable as the initial capital goes to infinity. Not all of the results are new, but many existing results are extended and the proofs are simplified. The method used is very interesting, and gives insight how ruin occurs for the class \(S^{(\alpha)}\). | |||
Property / review text: Consider a Lévy process \(\{X_t\}\), such that \(\lim_{t \to \infty} X_t = -\infty\). Let \(\tau(u) = \inf\{t: X_t > u\}\), \(\bar X_t = \sup \{X_s: 0 \leq s \leq t\}\), \(G_t = \sup\{0 \leq s \leq t: X_s = \bar X_t\}\). \(\tau(x)\) is then the time of ruin, and \(G_t\) is the last time before \(t\) at which the running maximum is attained. One is also interested in the shortfall at ruin, \(X_{\tau(u)}-u\), the surplus prior to ruin \(u - X_{\tau(u)-}\), the minimal surplus before ruin \(u - \bar X_{\tau(u)-}\), the time of the minimal surplus prior to ruin \(G_{\tau(u)-}\), and the time remaining to ruin from the time of the minimal surplus \(\tau(u) - G_{\tau(u)-}\). Of interest is also the event of ruin for large initial capital \(u\). One therefore considers the measure \(\operatorname{P}^{(u)}[ \cdot ] = \operatorname{P}[\cdot \mid \tau(u) < \infty]\). These quantities have been investigated for the small claims case and for the subexponential case. In the small claims case, the process behaves like the associated process, that is, there are slightly more and slightly larger claims until ruin occurs. In the subexponential case, suddenly an enormous claim occurs, and the the shortfall becomes enormous, too. In this paper, it is assumed that there is \(\alpha > 0\) such that \(\bar\Pi^+_X \in S^{(\alpha)}\) and \(\operatorname{E}[e^{\alpha X_1}] < 1\). Here, \(\bar \Pi^+_X\) is the tail of the Lévy measure, and \(S^{(\alpha)}\) is the set of measures \(G\) such that, for \(x_0 > 0\), \(F(x) = G(x)/\bar G(x_0) I_{[x_0,\infty)}\) has the property \[ \lim_{u \to \infty} {\bar F(u+x) \over \bar F(u)} = e^{-\alpha x},\qquad \lim_{u \to \infty} {\bar F^{*2}(u) \over \bar F(u)} = 2 \int e^{\alpha x} \,d F(x). \] The main result states that, conditioned on ruin, asymptotically the process behaves first like the Esscher transformed process with parameter \(\alpha\) until an exponentially distributed time. At this time point, a large claim occurs moving the process close to the capital \(u\), so that ruin may occur; and, if ruin has not occurred, the process behaves like the process conditioned on ruin with a moderate initial capital. Using this path decomposition, the asymptotic distribution of the quantities of interest can be found. In particular, a (more or less) explicit formula for the distribution of the time to ruin is found. This means, the time of ruin converges weakly to a finite variable as the initial capital goes to infinity. Not all of the results are new, but many existing results are extended and the proofs are simplified. The method used is very interesting, and gives insight how ruin occurs for the class \(S^{(\alpha)}\). / rank | |||
Normal rank | |||
Property / reviewed by | |||
Property / reviewed by: Hanspeter Schmidli / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 60G51 / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 60F17 / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 91B30 / rank | |||
Normal rank | |||
Property / Mathematics Subject Classification ID | |||
Property / Mathematics Subject Classification ID: 62P05 / rank | |||
Normal rank | |||
Property / zbMATH DE Number | |||
Property / zbMATH DE Number: 6083939 / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
Lévy insurance risk process | |||
Property / zbMATH Keywords: Lévy insurance risk process / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
convolution equivalence | |||
Property / zbMATH Keywords: convolution equivalence / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
time to ruin | |||
Property / zbMATH Keywords: time to ruin / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
overshoot | |||
Property / zbMATH Keywords: overshoot / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
capital prior to ruin | |||
Property / zbMATH Keywords: capital prior to ruin / rank | |||
Normal rank | |||
Property / zbMATH Keywords | |||
expected discounted penalty function | |||
Property / zbMATH Keywords: expected discounted penalty function / rank | |||
Normal rank |
Revision as of 11:56, 30 June 2023
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Path decomposition of ruinous behavior for a general Lévy insurance risk process |
scientific article |
Statements
Path decomposition of ruinous behavior for a general Lévy insurance risk process (English)
0 references
19 September 2012
0 references
Consider a Lévy process \(\{X_t\}\), such that \(\lim_{t \to \infty} X_t = -\infty\). Let \(\tau(u) = \inf\{t: X_t > u\}\), \(\bar X_t = \sup \{X_s: 0 \leq s \leq t\}\), \(G_t = \sup\{0 \leq s \leq t: X_s = \bar X_t\}\). \(\tau(x)\) is then the time of ruin, and \(G_t\) is the last time before \(t\) at which the running maximum is attained. One is also interested in the shortfall at ruin, \(X_{\tau(u)}-u\), the surplus prior to ruin \(u - X_{\tau(u)-}\), the minimal surplus before ruin \(u - \bar X_{\tau(u)-}\), the time of the minimal surplus prior to ruin \(G_{\tau(u)-}\), and the time remaining to ruin from the time of the minimal surplus \(\tau(u) - G_{\tau(u)-}\). Of interest is also the event of ruin for large initial capital \(u\). One therefore considers the measure \(\operatorname{P}^{(u)}[ \cdot ] = \operatorname{P}[\cdot \mid \tau(u) < \infty]\). These quantities have been investigated for the small claims case and for the subexponential case. In the small claims case, the process behaves like the associated process, that is, there are slightly more and slightly larger claims until ruin occurs. In the subexponential case, suddenly an enormous claim occurs, and the the shortfall becomes enormous, too. In this paper, it is assumed that there is \(\alpha > 0\) such that \(\bar\Pi^+_X \in S^{(\alpha)}\) and \(\operatorname{E}[e^{\alpha X_1}] < 1\). Here, \(\bar \Pi^+_X\) is the tail of the Lévy measure, and \(S^{(\alpha)}\) is the set of measures \(G\) such that, for \(x_0 > 0\), \(F(x) = G(x)/\bar G(x_0) I_{[x_0,\infty)}\) has the property \[ \lim_{u \to \infty} {\bar F(u+x) \over \bar F(u)} = e^{-\alpha x},\qquad \lim_{u \to \infty} {\bar F^{*2}(u) \over \bar F(u)} = 2 \int e^{\alpha x} \,d F(x). \] The main result states that, conditioned on ruin, asymptotically the process behaves first like the Esscher transformed process with parameter \(\alpha\) until an exponentially distributed time. At this time point, a large claim occurs moving the process close to the capital \(u\), so that ruin may occur; and, if ruin has not occurred, the process behaves like the process conditioned on ruin with a moderate initial capital. Using this path decomposition, the asymptotic distribution of the quantities of interest can be found. In particular, a (more or less) explicit formula for the distribution of the time to ruin is found. This means, the time of ruin converges weakly to a finite variable as the initial capital goes to infinity. Not all of the results are new, but many existing results are extended and the proofs are simplified. The method used is very interesting, and gives insight how ruin occurs for the class \(S^{(\alpha)}\).
0 references
Lévy insurance risk process
0 references
convolution equivalence
0 references
time to ruin
0 references
overshoot
0 references
capital prior to ruin
0 references
expected discounted penalty function
0 references