Stress testing in a crisis - how effective is it?
Over the past 20 years we have become accustomed to constant crisis phenomena in our lives and in general we have learned to live in this reality, keeping track of oil and currency exchange rates, political events and hedging various risks. The COVID-19 pandemic was a new experience for the global economy. Hardly anyone anticipated in 2018-2019 that 2020 would pass under the banner of a worldwide pandemic that would lead to a significant drop in business activity in Russia and the world's largest economies, an increase in costs caused by it, while simultaneously causing a drop in income at all levels. The current situation has shown the importance of a scenario approach in planning and budgeting, including those whose probability we expect little. Correct construction of stress forecasts in this case becomes an important task of business.
For the first half of 2020 to update internal forecasts of the impact of the pandemic on the implementation of the business plan, banking regulations, internal capital adequacy ratios, in general the risk appetite of the bank in all possible areas: corporate and retail customers, subsidiaries, industry affiliation, etc. However funny it may sound, playing with figures within the framework of these stress tests allowed us to clarify significantly the notion and objectives of stress testing: what matters is not so much the process or even the result, as understanding of what's going on.
We came up with a definition of stress testing from the contrary, that is, we established exactly what it is not.
1. Stress testing is not a computational task: trying to maximize the accuracy and completeness of the report has no independent value. It is not a "Let's calculate, then draw some report - that will be the most important result" task. In fact, technologically, this is a very individual task, and often the accuracy of testing is much less important than its meaning.
2. Stress testing is not an industrial task: inputs change faster than an industrial system is refined. Like many market players, we often hear about ready-made "boxed" solutions, capable of satisfying all stress testing needs at the push of a button, about professional consultants ready to do stress testing for a bank. But in fact, this is usually more of an exploratory question: it is important to understand what to count, and identifying the essence is often more difficult than making calculations.
Of course, technological support for stress testing is important. At the same time, the quality of calculations and model risks come to the fore. If there are no tools, no proper technical support, if analytical services do not have the appropriate technology, it is too difficult and often impossible to account for and calculate many scenarios. That said, the idea that there is some ready-made tool, a product that will solve all your stress testing tasks, is too good to be true. Rather, it will require internal research and the development of a sufficiently flexible, open system to bring together the knowledge of different experts and analytical departments.
3. Stress testing is not a stationary task. Approaches to stress loss assessment depend on the scenario used and are weakly reused. In fact, there are additional factors to consider each time, and often for each scenario the data and dashboard (actual portfolio, forecast, KPIs we look at, etc.) are assembled from scratch. Trying to do it once and apply it always is a waste of work, such a desire will only increase the labor and cost of finding a solution to the problem, but will not provide comparable process efficiency.
4. Stress testing is not a stand-alone task. The quality of a stress test depends directly on the skills and predictive abilities of the analyst. However, stress testing is not a matter of selecting the right expert when one out of ten can handle the task.
5. Stress-testing is not a task of finding the only right solution. Such a solution does not exist, and having an n-dimensional domain of choices does not help. Stress-testing usually doesn't answer the question, it doesn't help to find some particular indicator, on which everything depends. Its task is to show the interaction of factors, the logic of balance sheet mutation and changes in financial indicators as a result of risk exposure.
We have come to the conclusion that the objective of stress testing is to ensure readiness not for any particular scenario, but for changes in general, and this is the approach we are trying to turn into part of the bank's risk culture. It's essentially building a simulator where you can feel the mechanics of how risk factors affect financial performance, feel the logic, rather than calculate it mathematically, prepare yourself for the potential impact of external factors, get specific solutions and response tools that you can use in different situations.