| Investment Knowledge |
 |
 |
 |
 |
|
Greek Test The P&L distributions and VaR values were compared for two cases, with the instrument rates coming from di erent sources in each case. In the first case, the rates were sourced from the USD GOVT curve and the USD AGY credit spread for the bond, which were obtained from the VaR system. The GOVT curve and the AGY credit spread were added together to give the all-in zero curve. Since the term structure of the GOVT curve (which contains 13 tenor points) is different from that of the AGY credit spread curve (which contains 15 tenor points), the GOVT rates were interpolated to the points on the AGY curve. A set of 20,000 scenarios was then generated with respect to the all-in zero curve, and the base rates and scenarios were transformed to rates on the USD STUB time bucket curve (which contains 27 tenor points) and rates on the Swap time bucket curve (which contains 31 tenor points). Within the existing modeling framework, the trade is implemented by resetting generated default times in each Monte Carlo (MC) scenario to be the maturity date. Compared with the standard first loss trade model, two trends of change can be perceived. First, when the hazard rate of the obligor becomes smaller, the B/E spreads for both models would converge to zero. Second, because in the new model the cash flow pattern for default events has changed, the sensitivity to the interest rate should be different. Generally the new model is not only more sensitive to the interest rate but the sensitivity to its term structure changes dramatically as well. The delta and gamma sensitivities were generated with respect to each of the three curves. The base rates and sensitivities with respect to the Zero, STUB, and Swap time bucket spaces are shown. In the second case, the instrument rates were sourced from the Bloomberg curve YCCF0084 and transformed to all-in zero rates. The zero rates were then transformed to rates on the USD STUB and Swap time bucket curves, and the delta and gamma sensitivities were generated with respect to each curve. We note here that the 20,000 scenarios were not generated for the zero curve at this stage, but the zero curve was decomposed into a GOVT curve and an AGY credit spread curve, which was then re-combined back into an all-in zero curve in the same manner described previously for the first case. The 20,000 scenarios were then generated for the reconstituted all-in zero curve, and were transformed to scenarios for the USD STUB and Swap time bucket curves. (N.B. There is no significant difference in the rates on the reconstituted zero curve and those on the zero curve directly transformed from the source YCCF0084 curve.)
|