Extended creditRisk+ with guarantees / Lukas Fabrykowski
Weitere Titel
Extended creditRisk plus with guarantees
Verfasser / Verfasserin Fabrykowski, Lukas
Begutachter / BegutachterinSchmock, Uwe
ErschienenWien, 2017
Umfangv, 95 Seiten : Diagramme
HochschulschriftTechnische Universität Wien, Diplomarbeit, 2017
Schlagwörter (DE)Erweitertes CreditRisk+ / Risikoaggregation / kollektives Risikomodell / Kreditgarantien und Approximationen / gewichtete Verteilungen von Risikofaktoren / Poisson-Mischverteilung / Abhängigkeitsmodellierung / Panjer-Rekursion / Gammaverteilung / zusammengesetzte Poisson-Mischverteilung
Schlagwörter (EN)extended CreditRisk+ / risk aggregation / collective risk model / credit guaranties and approximations / biased distributions of risk factors / Poisson mixture distribution / dependence modelling / Panjer recursion / gamma distribution / compound Poisson mixture distribution
URNurn:nbn:at:at-ubtuw:1-98629 Persistent Identifier (URN)
 Das Werk ist frei verfügbar
Extended creditRisk+ with guarantees [1.07 mb]
Zusammenfassung (Englisch)

The current Extended CreditRisk+-model has a rather rudimentary methodology to consider guarantees, which, for example, does not take any dependence between guarantors into account. This drastically limits its applicability on the market. This thesis proposes several possible approaches how to incorporate guarantees - be it credit guarantees, reinsurance contracts or government subsidies - into the Extended CreditRisk+ framework. We first adapt the current notation of the model to allow for the securitisation of the exposure. Subsequently we propose three different methods to include the additional information in the computation of the potential portfolio loss. Finally we apply all these approaches to several exemplary portfolios and benchmark them agains known reference distributions. Additionally we give a short presentation of a software library developed to model various distributions and in particular used to implement the proposed methods.

Das PDF-Dokument wurde 71 mal heruntergeladen.