Explain the computation of implied correlation for liquidly traded
In this dissertation explain the computation of implied correlation for liquidly traded (standardized) STCDOs, using single-factor Gaussian copula models for the modeling of the statistical dependence of default events. For trading desks an important issue is the risk management of STCDOs. Hedging of STCDOs may require the computation of many Greeks. A simple hedging strategy tries to hedge the systematic spread risk in the portfolio tranches using the related iTraxx index swaps.
This hedging ignores idiosyncratic spread risk and spread dispersion. Refinements of the hedging by index CDS can be thought of. Different hedging schemes seem to exist in the market and their historical performance could be investigated by hedging simulations based on historical data.
Also, the influence of the pricing model chosen could be investigated. More complicated pricing models and different factor-copulae could be used as well and the reverse-engineering of the proprietary pricing models, e.g., of HVB could be carried further.
For the computation of Greeks, a method has been published by Andersen et al. (2003); Andersen and Sidenius (2005), which uses the recursive computation method, which is already implemented in C++ for the SFGC model (cf. Appendix B). The intruding extension of the existing C++ code and the availability of historical spread data suggest to continue the work presented in this thesis with hedging simulations.
Although this is a matter of interest to trading desks and the corresponding risk management units in investment banks, it seems that such simulations have not been made publicly available yet. However, such calculations depend on a good data quality. Therefore, data integrity will be of concern in these investigations, as it was conjectured above that the Bloomberg data may not always be reliable.
