As a Market Risk analyst, one of the main responsibilities of the department I work in is to carry out periodic stress tests on the bank’s portfolio. This is to ensure that the models used for ensuring adequate capital calculations are sound, and that the likelihood of a severe market move damaging the bank beyond expectations is minimized as much as possible within typical constraints (such as capital reserves and provisions).
The Value at Risk calculation of many financial institutions (already covered extensively in previous blog posts) failed in the crash of 2007, and portfolio loss expectations were being breached far more regularly than the confidence interval would suggest (with far greater losses than were predicted).
Stress testing at the time was clearly inadequate as it had previously failed to recognize that the VaR was liable to fail more regularly due to the effects that were observed in the crash; the scenarios defined were not severe enough.
Instead of discarding stress testing as a defunct and pointless exercise after its critical failure, it was retained and made even better to ensure that banks are better equipped when the market moves significantly against them.
What is Stress Testing?
Stress testing is an attempt to mimic certain market movements which follow a path to their worst outcome (rather than just changing a number of factors, and seeing the immediate theoretical change).
One of the main changes in philosophy since the financial crisis has been to change the thought from “what if”, to “when it happens, how bad?”
By accepting that adverse market moves are a real phenomenon, financial institutions can act with greater pragmatism, rather than just riding their luck.
What Type of Stress Tests are there?
One of the main stress tests (albeit not very comprehensive) is sensitivity analysis. Sensitivities of financial instruments and assets are measured by changes in a risk metric.
An example of a sensitivity test would be to check the change of value of an option for a 1% increase in the volatility (also known as the Vega – a ‘Greek’ used to measure sensitivities).
The benefit of sensitivity analysis is to see by how far the value of instruments and assets on the balance sheet are likely to change for a given move in a certain factor (such as a 1 basis point change in the underlying asset price, or as mentioned, a 1% increase in the underlying asset price volatility).
While sensitivity analysis can measure individual changes, in reality we expect to see changes to more than one factor simultaneously – this is where scenario analysis comes in.
Examples of typical scenario stress tests could be an increase in the volatility of the share price of a stock, with a large reduction in dividend payout.
Scenarios don’t necessarily have to only focus on “bottom up” situations; they could also define a hypothetical rise in interest rates or fall in GDP (also known as top down). Scenario stress tests are designed to be simple enough to execute, and with an output which is easily quantifiable.
All that is required from a planning point of view is to create the scenarios that can provide coverage for all situations & magnitudes, and to understand the value of current positions (which should be simple if adequate risk management systems are already in place).
Once the planning has been finalized, running the scenarios should provide value outputs which can be used for performance analysis. There are two main types of Stress Test:
Internal Stress Testing
This refers to the testing which a firm conducts on its own models, placed under the stresses that it feels could potentially happen.
The reason for conducting internal stress tests is to assess the suitability of its own risk models, and to make the necessary adjustments where economic capital is not sufficient to withhold the strains of these tests.
By conducting regular stress tests, the firm can increase its confidence in its own ability to pass external mandated tests.
External Stress Testing
Regulators in different markets may decide that institutions must undergo a prescription of its own defined stress tests to assess the capital adequacy of that institution.
These tests will typically assess that an institution is able to meet the minimum requirements of the regulators’ main objectives, otherwise minimum capital reserve ratios which the institution benefits from by using effective models may be increased to an arbitrary amount.
In addition to the individual stress testing, regulators may also find it necessary to test the ability of the industry as a whole to withstand potential stresses – the test requiring all relevant institutions performing the same tests concurrently.
The FDSF Programme (Firm Data Submission Framework) is the perfect example of a mandated coordination of bank-wide stress testing carried out at least monthly (and sometimes for ad-hoc requests from regulators).
The data it gathers monthly from standardized stress testing carried out by all Systemically Important Financial Institutions (SIFIs) is key in order for the analysts working at the Bank of England to assess the health of the banking system in the UK.
The defined framework which each bank must follow is standardized, allowing risk assessment and management to be carried out far easier than previous (especially for smaller institutions which may not have the size of budget to cope with frequent changes to the submission).
With the introduction of the standardized framework, the regulatory authorities in the UK will have a better understanding of the health of the banking industry, and theoretically should be able to mitigate any risk of systemic damage caused by undesirable shocks.
With smaller SIFIs who may have less to spend on implementing an effective risk management regime, the FDSF should allow a cost effective way for those institutions to measure their risk too, hopefully moving us to a more controlled financial industry.
It is certainly prudent to test against potential firm specific and industry wide shocks, but when recessions are touted as being unpredictable and inevitable, one would certainly fear the impact that the next great recession could have on the international economy; even despite the best efforts of stress testing.