It is important that banks have processes in place to ensure that their internal models have been adequately validated by suitably qualified parties independent of the development process to ensure that they are conceptually sound and adequately capture all material risks. This validation should be conducted when the model is initially developed and when any significant changes are made to the model. The validation should also be conducted on a periodic basis but especially where there have been any significant structural changes in the market or changes to the composition of the portfolio which might lead to the model no longer being adequate. More extensive model validation is particularly important where specific risk is also modelled and is required to meet the further specific risk criteria. As techniques and best practices evolve, banks must avail themselves of these advances. Model validation should not be limited to back-testing, but should, at a minimum, also include the following:
(a) Tests to demonstrate that any assumptions made within the internal model are appropriate and do not underestimate risk. This may include the assumption of the normal distribution, the use of the square root of time to scale from a one day holding period to a 10 day holding period or where extrapolation or interpolation techniques are used, or pricing models;
(b) Further to the regulatory back-testing programmes, testing for model validation must use hypothetical changes in portfolio value that would occur were end-of-day positions to remain unchanged. It therefore excludes fees, commissions, bid-ask spreads, net interest income and intra-day trading. Moreover, additional tests are required, which may include, for instance:
(i) Testing carried out using hypothetical changes in portfolio value that would occur were end-of-day positions to remain unchanged. It therefore excludes fees, commissions, bid-ask spreads, net interest income and intra-day trading;
(ii) Testing carried out for longer periods than required for the regular back-testing programme (e.g. 3 years). The longer time period generally improves the power of the back-testing. A longer time period may not be desirable if the VaR model or market conditions have changed to the extent that historical data is no longer relevant;
(iii) Testing carried out using confidence intervals other than the 99 percent interval required under the quantitative standards;
(iv) Testing of portfolios below the overall bank level;
(c) The use of hypothetical portfolios to ensure that the model is able to account for particular structural features that may arise, for example:
(i) Where data histories for a particular instrument do not meet the quantitative standards and where the bank has to map these positions to proxies, then the bank must ensure that the proxies produce conservative results under relevant market scenarios;
(ii) Ensuring that material basis risks are adequately captured. This may include mismatches between long and short positions by maturity or by issuer;
(iii) Ensuring that the model captures concentration risk that may arise in an undiversified portfolio.
Amended: January 2012
Apr 08
