Practical Applications from the Experts - February 2009

categories: Product Documentation, general

Predictive Stress Testing Diagnostics

Since market variables tend to move together, one should take into account the correlation between risk factors in order to generate realistic stress scenarios. One can express expert views by defining changes for a subset of risk factors (core factors), and then making predictions for the rest of the factors (peripheral factors) based on user-defined variables. The magnitude of the change in the peripheral factor corresponds to the correlation between the core and the peripheral factor scaled by the ratio of their volatilities (or Betas). RiskMetrics recently added Stress Test Diagnostics, which can be used to analyze a predictive stress test. Factor Levels provide the portfolio risk factor levels generated by a stress test while the Beta is the regression coefficient for the portfolio risk factors (peripheral) against the stressed risk factor (core). This option will help users better understand the relationships between the factors. Let's examine a predictive stress test that shocks the USD Index up 5% and its effect on the PV of a convertible bond - "SLB 2.125% 06/01/202":

Under this scenario, the price of the convertible bond will drop to 139.11. It is important to understand what caused this change given the number of underlying risk factors. Below are the initial levels of the main underlying risk factors used in pricing ignoring the risk free curve (Equity, Volatility, and CDS Spreads):

After running the stress testing diagnostics, one obtains the following results:

The negative beta between the USD Index and SLB of -1.78 causes the equity to drop to 41.53 in the stressed scenario and alone would lead to a drop of -7.19 in bond price. The positive beta between the USD Index and the volatility of SLB of 2.90 leads to a jump in volatility to 65.77% and alone would lead to an increase of +4.53 in bond price. The positive beta between the USD Index and CDS spreads leads to small jumps in spreads (about 4 BP on the 5 year node) and alone would lead to a decrease of about -0.40 in bond price. In aggregate the bond has lost -3.10 stemming mainly from the opposing effects of the equity and volatility.

Download file