Hedge fund needs to capiture credit risk. Credit risk management has two functions: default sensitivity (or default Greek) test and credit spread sensitivity test. It is implemented by the Scenario Manager, an autorun program, which manages the calculation of mark-to-market (MTM) value of the deal and the sensitivity Greeks.
Through testing we have found that this approximation is good when the correlation is weak. In general the Scenario Manager model can be viewed as an intermediate one between the models with 30,000 bps credit spread shock to the perturbed obligor and the model with no correlation between the perturbed obligor and the other obligors in the portfolio.
Default sensitivity test is a kind of stress tests matching the situation that a credit default event of an obligor has occurred or is perceived to be imminent. It is implemented by finding the change of Present Value (PV) for the tranches when the default time of an obligor is perturbed. The Scenario Manager Model is implemented by setting the default time of each obligor to be the valuation date when calculating the present value of each tranche, no matter when that obligor defaults in the generated Monte Carlo (MC) scenarios.
In this measure both the default times of the perturbed obligor seen by other obligors in the portfolio and the default correlations between the perturbed obligor and other obligors remain unchanged. This is not realistic because the MC paths of all the other obligors are generated without the updated information that the perturbed obligor has defaulted. We only perturbed the primitive default time of the perturbed obligor to be the valuation date while the joint default events incurred by this perturbed obligor is not re-simulated accordingly.
The test of the risk engine was conducted in two phases. In the first phase, a test model which serves the same function as Scenario Manager was implemented. The pricing of the Scenario Manager model for three deals was then tested against the test implementation. In the second phase of the test, the default sensitivity was simulated by the test implementation and compared with the results of Scenario Manager. The methodology of default sensitivity test was also discussed and compared with three other possible options of the default sensitivity test to check the effect of correlation and accuracy of the model.
References: