In the world of mechanical engineering, stress testing involves subjecting a mechanism to extreme conditions, considerably beyond the intended operating environment, in order to determine the robustness of the device and the circumstances under which it might fail. Financial stress testing is much the same.
What is a Financial Stress Test?
Generally speaking, a stress test is an assessment of the financial impact of changing a specific variable, without regard to the likelihood of this change.
Often, all other factors remain constant (even if this is not especially realistic). Sometimes the point of the test is to determine failure modes: A reverse stress test determines the magnitude of change necessary to induce financial ruin.
The term scenario test is often used to describe an assessment of the financial impact of a specific event (again, without regard to that event’s likelihood), in which the testers seek to reflect realistically the impact of this event on all aspects of the firm.
So a scenario test involves a more holistic look at possible circumstances rather than altering a specific variable in isolation.
Unlike probabilistic simulation modeling, stress testing:
- Is concrete and intuitive
- Does not require selection of probability levels
- Does not require understanding of overall dependencies among interlinked risks
- Avoids “black-box syndrome”
Stress tests can be used as a primary risk measure: assessing the level of a specific risk, measuring aggregate risk level, setting risk tolerances, or evaluating the benefit of risk mitigation. They can also be used to verify the calibration of more complex risk models.
Examples of Stress Tests
Stress tests and scenario tests have a long history and have been broadly applied. Deterministic financial projections readily lend themselves to stress testing.
For example, Willis Re’s eNVISION financial forecasting model allows users to easily change the value of a single parameter and see how that change affects key metrics.
An example of scenario testing is Standard & Poor’s use of past market stress events, pegging them to a rating level. In other words, a company with a BB rating should be able to get through a “BB event” without defaulting.
A blend of stress and scenario testing can be seen in the A.M. Best approach. Since 2011, the rating agency has asked insurers to estimate the impact of the largest potential threats to the firm arising from six different types of risk: market risk, credit risk, underwriting risk, operational risk, strategic risk, and liquidity risk – each using a specific “Risk / Event / Scenario” combination designed by the company.
For example, in terms of market risk one could consider a stock market scenario based on the events of 2008, and/or a 3% rise in interest rates such as that experienced in 1994.
The lessons of recent events have also led regulators to look to stress tests to assess the how well the market could stand up to adverse events.
The Solvency II process has seen the European Insurance and Occupational Pensions Authority (EIOPA) run such a stress test in 2011, which examined resilience under 3 scenarios of varying severity.
Each included deterioration in market, credit and insurance risk variables. Regulators will increasingly expect insurers to evidence such stress testing as part of their overall solvency management.
While it is easy to develop scenarios which reflect prior experience it is a much more difficult proposition to consider scenarios which factor emerging or as yet unknown risks.
The Lloyd’s emerging risk reports provide interesting examples of the extensive work that is being carried out to try and increase understanding and awareness of risk.
Natural Catastrophe Analysis
Another example of stress testing can be seen in the realm of natural catastrophe analysis. While sophisticated simulation models are quite well accepted for certain perils and regions (such as U.S. hurricane and earthquake), other catastrophe models are not so far advanced.
For example, the modeling of severe convective storm—tornado and hail—still faces significant shortcomings and is subject to significant model risk; for other perils, such as brushfire and sinkhole subsidence, there may be no model at all.
That’s why many companies prefer to use stress tests and scenario tests to assess their catastrophe exposure, supplementing stochastic models in some cases.
Willis Re’s SpatialKey geospatial platform, including stress testing apps such as eXTREME Tornado, is one example of a tool that facilitates this approach.
We understand that the International Association of Insurance Supervisors (IAIS) is considering a scenario test approach for its developing insurance capital standards for Globally Systemically Important Insures (G-SIIs) and Internationally Active Insurance Groups (IAIGs).
Calibration and Interpretation
When creating a stress test, analysts typically calibrate by ensuring that it ranks among real events of appropriate magnitude—and, while likelihood is not necessarily considered in stress testing, the frequency of real events of comparable magnitude may guide the design of the stress test.
An understanding of this calibration provides context for the numerical results of the stress test.
When reviewing the results of a stress test or scenario test, the first question to ask is: What does this say about the firm’s resiliency? As in the Standard & Poor’s example, the results may indicate a level of security that is either higher or lower than desired.
Given the concrete, intuitive nature of stress tests and scenario tests, these results facilitate communication with senior managers, the board of directors, and other stakeholders.
When only a single variable is stressed, the explanatory power of the test is clear. And when using a scenario test, the “story” of the scenario enables company leaders to think concretely about its financial effects, how the firm could respond, and what might be done to prevent a loss that large in the first place.
Overall, stress tests and scenario tests deserve a prominent place in a strong enterprise risk management program: they do much to foster a healthy risk culture.