FRM 2006-Quantitative analysis
發布時間：2010-01-19 共2頁
Not all FRM readings are created equally. In the Quantitative section, only one reading is new to the 2006 FRM (Quantifying Volatility in VAR Models by Linda Allen et al) but it is the most important reading in the first section. This reading previews themes that re-emerge in later readings.
Allen declares “normality cannot be salvaged.” In other words, we cannot after all assume that asset returns conform to a normal distribution. Asset returns are:
1. fat-tailed
2. skewed, and
3. unstable
Fat or heavy tails (i.e., fatter than implied by a normal distribution) are often observed in actual returns.
Allen also shows that volatility may be “regime-switching:” low one day but it may switch into high gear the next day–a new and different regime. The idea of regime-switching is thematic. A classic problem with correlations is their refusal to remain static. They often don’t hold up under stress events; e.g., in a crisis, if everybody runs for the exit, correlations may spike as asset classes plummet in lock-step. This is not trivial because the whole idea of risk measures is to help us get a grip on bad scenarios; if the models break down precisely when situations go sour, they lose their purpose.
What does Allen recommend for “this striking result [that normality fails] with critical implications”? No panacea but rather a need to supplement (go beyond) value-at-risk with (i) stress testing and (ii) scenario analysis. Also thematic. VAR is not a solution to end all risk measurement solutions. (this even before Dowd reminds us that VAR is “not a proper risk measure at all” because it is not sub-additive).
Allen declares “normality cannot be salvaged.” In other words, we cannot after all assume that asset returns conform to a normal distribution. Asset returns are:
1. fat-tailed
2. skewed, and
3. unstable
Fat or heavy tails (i.e., fatter than implied by a normal distribution) are often observed in actual returns.
Allen also shows that volatility may be “regime-switching:” low one day but it may switch into high gear the next day–a new and different regime. The idea of regime-switching is thematic. A classic problem with correlations is their refusal to remain static. They often don’t hold up under stress events; e.g., in a crisis, if everybody runs for the exit, correlations may spike as asset classes plummet in lock-step. This is not trivial because the whole idea of risk measures is to help us get a grip on bad scenarios; if the models break down precisely when situations go sour, they lose their purpose.
What does Allen recommend for “this striking result [that normality fails] with critical implications”? No panacea but rather a need to supplement (go beyond) value-at-risk with (i) stress testing and (ii) scenario analysis. Also thematic. VAR is not a solution to end all risk measurement solutions. (this even before Dowd reminds us that VAR is “not a proper risk measure at all” because it is not sub-additive).
