Pandemic Stress Test
The corona pandemic scenario as a key stress event
As we live through the coronavirus crisis, people have started to gauge the economic consequences that the measures taken to combat the crisis may have. These are likely to be severe, with a significant increase in unemployment and company defaults.
Neither financial firms nor their supervisors seem to have considered a pandemic flu scenario as a key stress event before the coronavirus outbreak, despite the implementation of extensive stress testing capabilities in the past decade. This raises a number of questions:
- Why was this not considered as stress event by financial firms or their supervisors?
- How useful is stress testing in light of this?
- How can stress testing frameworks benefit from this experience?
We will review each of these questions in turn.
Why was this pandemic not considered as stress scenario?
There have been a number of ‘warnings’ in the past decades that could have put a worldwide pandemic flu scenario on the table. For example:
- The SARS-coronavirus spread to at least 29 countries between November 2002 and July 2003, leading to over 8,000 confirmed infections and almost 800 confirmed deaths, according to WHO statistics. Most affected were China, where it originated, and Hong Kong. GDP in China temporarily stalled as a result but retook its growth path in the second half of 2003
- The H1N1 swine flu in 2009 had over 1.6 million confirmed infections and over 18,000 confirmed deaths, according to WHO statistics. Later it was estimated that between 700 million and 1.4 billion people may have been infected worldwide, of which between 175,000 and 550,000 died
The worst pandemic flu on record is the Spanish flu. This saw major outbreaks in both 1918 and 1920, of which it is estimated that one-third of the world’s population of around 1.5 billion at the time was infected, and between 20 and 100 million people died. Other flu pandemics during the past century are the Asian flu in 1958-1959 and the Hong Kong flu in 1968-1969.
So why was a severe pandemic flu not part of the regular set of stress scenarios for financial institutions or their supervisors, despite the fairly regular occurrence of a pandemic during the past century? On the one hand, the economic impact of the past pandemics was not nearly as severe as the economic impact from the current coronavirus pandemic is likely to be.
This past experience may have shaped expectations (too much) about the economic impact of future pandemics. On the other hand, there may have been a (as it turns out, misplaced) trust in the medical profession and industry to prevent or limit the impact of any such future pandemic. In addition, a combination of short memory, limited historical awareness, and optimism may have played a role as well.
To be fair, several firms and organizations actually did identify a pandemic flu as potential stress scenario. For example, Swiss Re reported the estimated impact of a lethal pandemic stress scenario on their insurance business in its annual report up to 2016.
Willis Towers Watson described a storyline of a worldwide flu pandemic in October 2019 that does have similarities with the current spread of the coronavirus. EIOPA, the association of European insurance and pension supervisors, mentioned it as a potential stress scenario in a document on “Methodological principles for insurance stress testing” in December 2019. The Basel Committee referred to it in a 2017 document on supervisory stress testing of central counterparties. In none of these instances, however, was the scenario developed and the financial consequences analyzed to consider the full economic impact that we are likely to face now as a result of the corona crisis.
How useful is stress testing in light of this?
As few firms and supervisors seem to have included a severe pandemic flu as one of their key stress scenarios, one could ask whether this undermines the value of stress testing.
Since the 2007-2008 financial crisis, stress testing has become a popular tool for financial institutions to assess their ability to withstand severe but plausible stress scenarios. Compared to the economic capital models that were popular before the crisis, it explicitly allows for the possibility to consider hypothetical, adverse scenarios that did not occur in the past, but could occur in the future. But now, a severe pandemic flu stress scenario seems to have been generally missed. And as highlighted, there was in fact evidence from the past that it could happen.
By its very nature, an event only turns into a crisis if it comes as a surprise. It is, of course, impossible for anyone to foresee and prepare for all possible future events. Even if we would have considered a pandemic flu as a stress scenario, it is unlikely that it would develop in exactly the way we have specified in our stress scenario. Ultimately, it is key that we have tools at hand that allow us to obtain timely insight into the impacts of an event as it unfolds, to react quickly, and to assess potential further adverse developments in the future. In this perspective, stress testing capabilities that financial institutions have implemented offer various benefits.
First of all, a stress testing framework, when implemented adequately, should enable the quick aggregation of exposure data across the firm. This allows institutions to calculate the impact of new stress scenarios, including those that describe the possible further development of an adverse event that is unfolding at present, within a relatively short time frame.
It is vital to specify any new stress scenario in sufficient detail so that stress testing models can be executed. This first requires a narrative on how the scenario may develop over time, and then subsequently a translation of this scenario narrative to appropriate shocks for individual risk factors that form the parameters in the stress testing models (such as unemployment, interest rates, GDP growth, credit spreads, equity and commodity prices).
During the past decade, many institutions have built up extensive experience with this, especially larger ones. Moreover, for the generation of individual shocks based on a high-level narrative, some institutions have automated tools in place that can generate these quickly. This is a second benefit from the stress testing capabilities that firms have implemented.
Finally, although a severe pandemic flu may not have been considered a stress scenario by most institutions and supervisors, important shocks in such a scenario may resemble those in existing stress scenarios. For example, many stress scenarios will encompass significant drops in equity prices and increases in credit spreads. Results from existing stress scenarios may therefore yield a rough indication of the potential impact. This will be useful while time is spent to define the details of the actual stress scenario in sufficient detail to run through the stress testing models, as described in the previous paragraphs.
How can stress testing frameworks benefit from this experience?
Almost by implication, any new crisis exposes us to events that we did not consider possible before the crisis. The financial crisis in 2008 led, for example, to a dry-up of interbank-funding liquidity that had not been experienced before. As a consequence, financial firms have strengthened their liquidity position significantly and regulators introduced strict liquidity and funding requirements. In addition, we have now experienced a long period with negative interest rates in many countries that most people did not consider possible before the crisis.
Similarly, the coronavirus crisis has features that we have not experienced before. For example, the fall in equity markets by more than 30% occurred in a much shorter timeframe than we have experienced in earlier crises. Furthermore, there was a severe lack of liquidity in the US Treasury market, usually the most liquid of markets, as well as other US bond markets. This forced the US Fed to provide unprecedented amounts of liquidity, leading to an increase in the Fed’s balance sheet by over US$2 trillion in a matter of weeks.
Contrary to the pattern seen in other crises, prices of US Treasuries fell more in March than those of corporate bonds. Simultaneously, there was a perfect storm in the oil markets as an extreme drop in demand because of the worldwide economic standstill resulting from the measures taken by governments to contain the spread of the coronavirus, coincided with a supply shock due to the disagreement between Russia and Saudi-Arabia about restricting supply.
This highlights the necessity to think out-of-the-box when defining stress scenarios to test the resilience of an organization. Guidance from historical crises may be of limited value when exploring potential future stress scenarios. It is therefore advisable to consider many more stress scenarios than the handful that most institutions typically consider at the moment. For the generation of such a larger set of scenarios, a more randomized approach could help to generate ones with unseen and economically less intuitive combinations of shocks.
The most immediate challenge for firms is clearly to assess and manage the impacts of the current corona crisis, and to prepare for the possible longer-term economic impact. This will involve an analysis of the impact on clients and exposures in sectors that are directly affected, but also an assessment of the ramifications that the economic downturn may have on other sectors and markets over time. Since this longer-term impact is obviously uncertain, a number of potential adverse scenarios are typically analyzed.
For firms that are in the early stages of implementing a stress testing framework, this may involve a substantial amount of manual work and ad-hoc judgements. Firms that have a sound stress testing framework in place can benefit from this in many respects:
Aggregation of exposure data across the firm in a quick and consistent manner
Validated stress models that can be run efficiently once exposure data are available and stress shocks have been defined
Tools that can generate all required shocks for all stress models in an automated manner, based on a small set of key shocks that is defined by the scenario narrative
Implementation of all models on a central infrastructure that allows additional stress scenarios to be run without much additional effort
Standardized reporting tools that can quickly summarize the relevant information for senior management
Zanders has extensive experience in stress testing, both in supporting firms to carry out specific stress testing analyses and in defining and implementing stress frameworks. This experience also encompasses the development and validation of stress models and scenario generation tools. Do you want to learn more? Please contact Pieter Klaassen.