The framework of the required testing for banks in this $10-50B asset range is similar to that of banks with more than $50B in assets - there are three scenarios required: baseline, adverse, and severely adverse. These tests are required to be completed annually, but the banks in this asset range are allowed more time to complete them - the due date is March 31 vs January 5th of each year. Additionally, banks in this asset range are not required to use all of the variables in their tests that are required at the larger size - they CAN use additional variables and cover additional quarterly periods to improve their tests, but it is not REQUIRED.
Scenario descriptions are required to be provided to the banks by the agencies and these scenarios are required to be used by the DFA rules. The time periods for delivery of the descriptions are also spelled out. Scenario descriptions will be provided to each bank by its primary regulator no later than November 15 of each year. Stress tests will be completed using exposures as of September 30 of each year, and will extend with quarterly estimates over a nine quarter period beginning on December 30 of that year, and through December 30 of the second year beyond that date. Each of those estimates will have to address the expected impact of the stress scenario on consolidated earnings, balance sheet and capital.
The model itself is addressed. The agencies expect that companies will have established effective model risk management practices and will validate those models. Use of a vendor provided model does not limit the bank's liability related to the validity of the model. Bottom-up stress testing of individual portfolios is not precluded, and the guidance indicates that the loss estimation practices should be 'commensurate with the materiality of the risks measured'. Data segmentation is considered - banks may address risk with more granularity, looking at business lines and risk segments to more specifically limit distortions that might occur at the higher levels within a more broadly defined portfolio.
The new guidance also covered the idea of using regional variables. Use of regional variables is allowed, but it should not be a variable that is exhibiting limited stress compared to the variables provided by the agencies, and the company should be prepared to confirm that the variables they chose are relevant and relate to company-specific characteristics. The use of company historical data for example, should not be considered an adequate justification on its own, as over time the historical period may not include data that mirrors a stress period or reflects a downturn in the credit cycle.
The rules also spell out the level of responsibility of the board of each bank, and of course of senior management. Having a system of controls over the stress testing process is required, and the board is expected to review and approve that no less than annually. They are expected to receive a summary of the results of the testing, and to consider those results in the capital planning of the company. Results are to be provided to the primary regulator by March 31 each year, along with the qualitative information supporting the projections, descriptions of the methodologies, etc.
Even more controversial is the public disclosure aspect of the tests. It is written in the DFA rules that summary results must be publicly disclosed between June 15 and 30 of each year. These disclosures are required at the company and holding company level, and must be available on the company web site or in some other equally accessible public forum. Essentially the same information given to the regulator is expected to be available to the public.
The first experience with this process is quickly approaching, and there is much work to be done to prepare for successfully navigating this new gauntlet. Creating the models to be effective and accurate will likely require many iterations and the man hours potentially required represent a daunting impact. The issuance of this guidance is timely, but unfortunately many questions remain to be answered, and they will likely arise as bankers begin to create the models and run the tests. Hopefully the public will be sympathetic when the publications of those first tests hit the streets next year!