Although we believe that our proprietary market risk models are of a high standard, we are committed to their ongoing development and allocate substantial resources to reviewing and improving them.
Our stress testing results and economic capital estimations are limited by the obvious fact that our stress tests are necessarily limited in number and not all downside scenarios can be predicted and simulated. While the risk managers have used their best judgment to define worst case losses, with the knowledge of past extreme market moves, it is possible for our market risk positions to lose more value than even our economic capital estimates.
Our value-at-risk analyses should also be viewed in the context of the limitations of the methodology we use and are therefore not maximum amounts that we can lose on our market risk positions. The limitations of the value-at-risk methodology include the following:- The use of historical data as a proxy for estimating future events may not capture all potential events, particularly those that are extreme in nature.
- The assumption that changes in risk factors follow a normal or logarithmic normal distribution. This may not be the case in reality and may lead to an underestimation of the probability of extreme market movements.
- The use of a holding period of one day (or ten days for regulatory value-at-risk calculations) assumes that all positions can be liquidated or hedged in that period of time. This assumption does not fully capture the market risk arising during periods of illiquidity, when liquidation or hedging in that period of time may not be possible. This is particularly the case for the use of a one-day holding period.
- The use of a 99% confidence level does not take account of, nor makes any statement about, any losses that might occur beyond this level of confidence.
- We calculate value-at-risk at the close of business on each trading day. We do not subject intra-day exposures to intra-day value-at-risk calculations.
- Value-at-risk does not capture all of the complex effects of the risk factors on the value of positions and portfolios and could, therefore, underestimate potential losses. For example, the way sensitivities are represented in the value-at-risk model may only be exact for small changes in market parameters.
We believe that the aggregate value-at-risk estimates for our consolidated Group as a whole stand up well against our back-testing procedures (as measured by the number of hypothetical buy-and-hold portfolio losses against the predicted value-at-risk). However, we acknowledge the limitations in the value-at-risk methodology by supplementing the value-at-risk limits with other position and sensitivity limit structures, as well as with stress testing, both on individual portfolios and on a consolidated basis.