Market risk has been the proving ground for the VaR methodology. In all the major markets for equities, debt and foreign exchange, as well as for many commodities markets, there is a huge amount of publicly available daily price data^.. Pre-processing techniques have been developed which allow transactions to be aggregated into fewer positions to facilitate a modelling approach.

VaR models have been increasingly used and accepted as a risk management tool in the area of market risk. Many of the largest investment banks have now accumulated significant experience in the design, operation, and management of models for the purpose of controlling risk and pricing products. A number of supervisors have for some years been developing expertise in reviewing the integrity of individual models (e.g. option pre-processing models and foreign exchange VaR models) and the effectiveness of the associated control systems. Other supervisors have worked closely with major firms in improving their understanding of the strengths and weaknesses of market risk modelling.

The VaR approach to modelling market risk has a number of attractions as a basis for regulatory capital charges. The supervisory framework based on VaR should not require frequent updating to take account of market evolutions. It bases risk measures on an extensive, and continuously updated, dataset of empirical observations. The dataset allows for backtesting of models and hence assessment of their accuracy.

It is recognised that the output of VaR models does vary (as shown by the studies cited in the appendix); it will be a task for supervisors to design a framework to ensure that models are used appropriately and that the level and bias of variations is not of economic significance. ⁴ The aim of using VaR models (subject to supervisory oversight) would be to achieve a better relationship between regulatory capital and the relative risks of portfolios.

To the extent that models rely on historical data (and in practice most do), they will be faced with the problem that the future will not always resemble the past. However, this is a problem for all methods of quantifying risk, and therefore for any supervisory approach that attempts to measure risk. Again, it is a matter for supervisory judgement how, and to what extent, capital charges have to be increased to compensate for this.

Market Liquidity Risk

In the normal course of events, market risks can be captured by VaR models. The danger lies in extreme market movements when correlations and other assumptions break down. For example, in a serious crisis buyers may desert the market and stay away for an extended period. This market liquidity risk is difficult to capture with current VaR methodologies (and to that extent it shares similarities with credit risk). The prior provision of secure funding arrangements is critical, but attempts to quantify the risk and provide capital against it can also be made. Stress testing of portfolios would be one approach to assessing the possible impact of extreme events and calculating an additional capital requirement.

Footnote:

4. Some firms currently use models that assume a normal distribution, whereas market returns are not normally distributed. However, the experience of the Basle Models Taskforce, for instance, was that there was no systematic difference between the results of banks using the historical simulation approach and variance / co-variance.