Three types of stress test
If you run a Google search for “stress testing” you will find over 900m results and one near the top is from the Bank of England:
“Banking stress tests assess how banks can cope with severe economic scenarios. We look at banks’ resilience, making sure they have enough capital to withstand extreme shocks and are able to support the economy.”
I’ve always felt uneasy about stress tests, it’s a bit surprising, after all, they should provide comfort that, if the unexpected happens we have cover. Are we right to think they strengthen what we do?
Tests fall into three categories:
Those where the results are certain. By applying a load, you can test a material to destruction. Repeating the test will confirm your results. “Everything breaks at some point – the important thing is to know when that will happen.” TÜV Rheinland.
Those where subjectivity must be applied to the results. A patient that has an irregular heartbeat would require an expert diagnosis to determine the cause. “A stress test is often used as a screening test. If the result is normal, then typically you won’t need further tests. If it’s not, you may require further tests, depending on the type of problem identified”. British Heart Foundation
Those where subjectivity exists beforehand and afterwards. “Nevertheless, given that our measurement of risk is at best a prediction - hopefully a well-founded prediction, but nevertheless still involving a high degree of judgment - we will not always get it right. In such an environment, when leverage is high and the room for error is low, we need a strong safety net in case risk may be underestimated”. Mr Stefan Ingves, Governor of Sveriges Riksbank, 2014.
Our uncertainties
Victorian bridges (or any other long-lasting structure) remain functional because the engineering was known. Herein lies a problem our business contains a lot of uncertainty and subjectivity:
What triggers customers to withdraw money?
Are wholesale markets riskier than retail?
Are government guarantees like the FSCS helpful, if so, how do they affect withdrawal rates? Is it the full balance or the excess over £85,000?
How does the channel effect the capacity to withdraw?
How price sensitive are customers and how is this affected by competitor’s rates?
Are new customers more unpredictable than old?
We look back at what happened and guess about the future, there is little of known fact. At best we get an approximation of what may happen and whether this is “severe but plausible” is a matter of opinion. All we know for certain is that, when customers perceive there are problems, they will want their money back sooner rather than later. To assume anything else would be rash.
The economic uncertainties
Furthermore, if we don’t know where the economy is going from one day to the next how can we verify the amount of capital we need in a leveraged business?
“Any changes made to ECL (Expected Credit Losses) to estimate the overall impact of Covid-19 will be subject to very high levels of uncertainty as so little reasonable and supportable forward-looking information is currently available on which to base those changes. ………such an implementation ought to reduce the risk of firms recognising inappropriate levels of ECL, which is very important bearing in mind that a significant overstatement of ECL could prompt behaviour that leads to unnecessary tightening in credit conditions.” Sam Woods, Deputy Governor and CEO of the PRA
Is it just speculation?
The way engineers look at the world, indeed have to look at the world, differs from ours. Theirs is a one of “pass” or “fail” but for us this would be a big mistake as it fails to acknowledge the uncertainty we face. For this reason, looking for red, amber or green may be expedient but it can get us in trouble. It also assumes the calculations are correct in the first place……
“…. the question is whether the numbers are now so layered in detail that it’s not straightforward to identify an error in the first place.” Financial Times correspondent, Thomas Hale, on 13th December 2019, commenting or errors in capital calculations at Metro Bank and Coventry Building Society.
I’ll draw a few strands together:
There are three types of stress test and you need to be aware of which you are using.
In our business we have assumptions and guesses about people’s behaviour and the future.
This means our tests are approximations of risk.
We aren’t dealing with construction and we aren’t engineers therefore red, amber and green paints a simple but potentially misleading picture.
In Part Two I’ll consider some of the beneficial things we can do, in particular the importance of responding flexibly.