The New Basel II Accord (the Accord) refers to “stress testing” numerous times and in relation to each of its three pillars. Nowhere, however, does it attempt to define what constitutes an effective stress test. To echo one of the writers on the subject, the Accord is saying “we can’t define it, but we know it when we see it”. This introduces a problem for both the regulators and the regulated in attempting to agree on what a stress test is. This paper examines the current state of stress testing in relation to the Accord framework and arrives at a suggested approach to stress testing based on what is felt to be common use.
Support for this approach to stress testing is given in the paper discussed below on business continuity.
The Stress Testing Requirements in the Accord
The Accord mentions “stress test” a total of 27 times, 25 of those in the body of the Accord. As mentioned above, however, nowhere does it define the term. In relation to credit risk (at para. 434 and subsequent) it gets closest by describing what sort of scenarios could be included in a stress test. The Accord also makes clear that there needs to be stress testing done in relation to:
- credit risk (again, para. 434);
- liquidity risk in relation to collateral (para. 158); and
- market risk (para. 738).
Regulators are also required to ensure that institutions conduct “[r]igorous and forward-looking stress testing…” to identify factors “…could adversely affect the bank….”(para. 726). In fact, the entire text of Principle 1 of Pillar II (paras. 726 to 745) is centred on the question of stress testing.
What is Stress Testing?
A generally agreed definition of stress testing is “stress-testing means choosing scenarios that are costly and rare, and then putting them to a valuation model.” In this case, the valuation models will be those used by a bank to calculate its economicimportant point to note here is that both the Accord and this definition are calling for multiple scenarios. In fact, the literature makes it abundantly clear that multiple scenarios, of varying likeliness, are required to constitute an effective stress test and that these tests and scenarios need to be re-run and re-evaluated on a regular basis to accommodate both changes in the Bank’s asset mix and in future expectations.
Interestingly, some of the literature suggests that, not only should expected or likely scenarios be tested, but some unusual or unexpected scenarios should also be tested to gauge their likely effects. This would involve four possibilities:
- Simulating shocks which we suspect are more likely to occur than historical observation suggests;
- Simulating shocks that have never occurred;
- Simulating shocks that reflect the possibility that statistical patterns could break down in some circumstances; and
- Simulating shocks that reflect some kind of structural break that could occur in the future.
A Possible Stress Testing Regime
An effective, rigorous and forward looking stress-testing regime would therefore include multiple likely scenarios – including examining movements in all the major macro-economic variables over the past 10 to 20 years and an examination of their likelihood. Capital should be held to allow for these types of events and, in the event that they occur, the minimum capital to be held, as mandated by the regulators, should be allowed to fall to the predicted level. The bank should also have plans to re-establish appropriate levels of capital during the recovery phase from the stress event, with the understanding that other banks in the market will also be attempting similar strategies.
Stress-testing should also include unlikely, but possible, events, along the lines of the four points above. As an example of each one, stress tests could involve such things as the effects of:
- Asian Crisis (1997 – 1998)
- Russian Debt Default
- US Terrorist Attacks
- Hostilities in Iraq in 2003-04
None of these scenarios may be statistically likely, but scenarios similar to these would form a part of a rigorous and forward-looking stress-testing regime. In the event that these scenarios show a possible capital or liquidity deficiency, a bank should have a plan to address the scenario and should be prepared to present and discuss these plans with the regulators.
It is also important, for model verification purposes, that back testing is also carried out on stress scenarios as and when they occur. If the stress testing has been sufficiently broad there should be the capacity to back test stress scenarios on a regularone of the unlikely events may not occur, the regular stress test results on likely scenarios should occur with sufficient regularity to allow for reasonable model verification. Should one of the unlikely events occur, or an event with similar macro economic effects to the stress scenario then the scope for model verification is greatly increased and extensive back testing should be carried out; both on the model and its predictions and on the basis for the assumed macro economic effects.
A rigorous and forward-looking stress-testing regime would consist of a combination of both likely and unlikely events. Sufficient capital should be held to ensure that a capital deficiency does not occur under any of the likely scenarios. Capital need not be held against events judged to be possible but unlikely. Where these show a capital deficiency a plan should be developed to counter the effects of the event.
Back testing should also be regularly carried out for model verification purposes and all of the stress- and back-test results and resulting plans will need to be documented, agreed with management and the Board and discussed with the regulators.