Abstract
Dynamic Portfolio Credit Risk Measurement
by
Sandeep Juneja
We consider the problem of measuring credit risk, particularly portfolio credit risk, as it evolves over time. Our analysis relies on viewing the problem in discrete time in an asymptotic regime where the defaults become rarer asymptotically. We observe that this view leads to considerable simplification in typical calibration techniques that are known to be computationally demanding. In particular, we arrive at approximate closed form solutions of the underlying parameters in a popular regime and observe that they perform quite well in practice. Further, in our asymptotic regime, we conduct large deviations analysis of large losses where the model explicitly allows contagion effect, and other dependencies. This large deviations analysis provides insights into occurrence of large losses in a portfolio and facilitates development of provably fast simulation techniques for measuring portfolio risk.
Committee
Workshop
Key Dates
Communication
First Conference Link