I will pick the stress test comment to begin my thoughts. I believe that defining the impact of potential stress events for both market and credit risk will become increasingly important as we move forward. If we divide risk measures into marginal risk measures, stress/scenario tests and probablilistic measures, several things seems to stick out. We will always need the marginal risk measures simply because we use them in our daily thinking about risk. Probabilistic risk measures, while useful for many things, suffer from the simple fact that we are not 100% sure of the shape and location of the underlying distribution and cannot (at this time) tell the difference between a Type 1 error and a Type 2 error with complete confidence. “Did we really get a four standard deviation move last week or did the distribution move in the short or long term?” is a question we cannot fully answer at this moment. I know of several institutions that perform VaR simply for regulatory purposes (my point is NOT to dismiss probabilistic measures here at all).

Which brings me to stress and scenario tests. The institutions to which I refer above also noted that the results of stress and scenario tests are actionable, i.e., must be noted and discussed whenever some pre-set loss is reached under a specific scenario. By correctly (or maybe just sufficiently) defining the scenario and stress tests, an activity which requires significant human capital and experience more so than the other two measures (in my opinion) we can bring a combination of quantitative ability AND business experience and judgement to the risk management problem.