Volume 3 — Valuation and Risk Metrics © Copyright 2002, CCRO. All rights reserved. 29 III SENSITIVITY ANALYSIS, SCENARIO ANALYSIS, AND STRESS TESTING 1.0 Discussion of Analyses Sensitivity analysis, scenario analysis, and stress testing are conducted by evaluating the effect(s) of user-defined changes to a base set of the value and risk drivers of a portfolio, asset, or position of interest. This is in sharp contrast to the value and risk metrics that are traditionally used, such as VaR or DEaR. The distinction lies in the user’s specification of the value and risk drivers, which may change according to circumstance or perceived need. Also, the normal metrics are typically based on statistical analysis, where different driver values are weighted to correspond to their frequency of occurrence. In contrast, in stress testing, scenario analysis, and sensitivity analysis, the user decides the value chosen for any driver of interest, and its selection may or may not be a function of its relative frequency of occurrence. There is a natural hierarchy in these three types of analyses. The base of the hierarchy is sensitivity analysis, in which the response of an object or variable of interest such as MTM value or P&L is measured against selected changes made to one or more principal drivers, such as the underlying price or volatility. Sensitivity analysis can be performed by standard tests, such as the Greek measures that are generated for any portfolio with option components, or nonstandard, such as measuring the response of the MTM value to a 10% change in the volatility. The nonstandard sensitivity analyses are more user-defined measures. At the next level of the hierarchy is scenario analysis, in which either a predefined or a context- specific set of potential evolutions of the underlying variable or variables of interest is used to evaluate changes in the object or variable of interest. One common characteristic of a scenario is that it is tied to some descriptive set of facts that occur to cause the risk drivers to behave in a particular way. For instance, one simple scenario analysis could be to examine the impact on a generation asset’s operating revenues from a shift in the forward price curve for natural gas or a shift in the interest rate curve across the board. In scenario analysis, non-market drivers of value and risk can also be suitably perturbed to examine the effect on value and/or risk. An example would be to evaluate the change in the economic revenue stream from a generation plant due to a change in the economic growth rate or the demand growth rate for electricity. Another scenario analysis could focus on the impact of generation economics due to specified changes to the emission standards. Thus, scenario analysis is a form of sensitivity analysis wherein the impact of selected drivers on a variable or object of interest is evaluated. Stress testing is at the top of the hierarchy because it enables management to examine the results of selected scenario analyses. The rationale for stress tests lies in the choice of events or values for the drivers, which are not usually captured by the more traditional value and risk analyses. Thus, under stress testing we would typically evaluate the variable or object of interest under more extreme values of the underlying variables. For instance, a simple stress test may involve examining the changes in the operating revenues from a generation asset in the