Thinking the unthinkable – Staying ahead of the crisis curve

Industry leaders discuss the increased value of stress-testing in a world rocked by its second financial crisis in 12 years, the likely emergence of non-financial risks, and how financial institutions can establish efficient and effective stress-testing frameworks in the future

The panel

• Thomas Verbraken, Executive Director, MSCI Research
• Laurent Birade, Senior Director, Risk and Accounting Solutions, Moody’s Analytics
• Vivian Chan, Executive Director, Head of Finance Nomura Asset Management  
• Model risk and stress-testing specialist at an international banking group

How has the crisis brought about by the Covid‑19 pandemic highlighted the value of stress-testing for financial services firms?

Thomas Verbraken, MSCI Research

Thomas Verbraken, MSCI: It is precisely during periods such as the recent Covid-19 crisis – when historical analysis has its limits – that stress-testing proves to be most useful because of its forward-looking nature and the ability to inject expert knowledge into the analysis. That makes stress-testing a great tool for modelling unprecedented scenarios.  

Having a stress-testing framework in place may leave investors more prepared for market turmoil. Of course, it is unlikely that an exact pandemic-like stress test would have featured on the list of scenarios run by investors prior to Covid-19. But running other crisis scenarios – even if not capturing the exact nature of the pandemic – may help get a sense of tail risk and the vulnerabilities and exposures of portfolios. 

Furthermore, meaningful stress-test scenarios are often characterised by economically sound narratives, expressed as a set of shocks on major market and macroeconomic variables. A solid understanding of how macroeconomic variables are linked to financial markets is crucial, especially in cases such as the Covid-19 crisis, from which an unprecedented economic shock originated. Understanding those relationships can also be used in a reverse stress-testing exercise and help answer such questions as: which levels of economic growth, real rates and equity risk premia could explain an observed market downturn? In particular, during periods of high uncertainty, such analysis can provide an anchor for a better understanding of what is happening.

Finally, it is not only useful to focus on drawdown-type scenarios. If Covid-19 taught us one lesson, it is that it may be equally important to think about possible recovery scenarios during market turmoil so the impact of potential tactical shifts can be understood. For example, the swift ‘V-shaped’ recovery seen in the past months might have been an unpleasant surprise if de-risking near the bottom meant partly missing out on the rebound, ending up with underperforming the market and certain investors able to ride out the crisis.

Laurent Birade, SAS

Laurent Birade, Moody’s Analytics: The stress-testing framework established after the 2007–08 financial crisis prepared firms to understand the impact of new stress events and to have a well-rehearsed process for determining potential impacts on capital and firm viability. Because of those improvements to stress-testing, many banks were already maintaining substantially higher capital and thus have been better positioned for the economic shocks caused by Covid‑19.

The ongoing Covid‑19 crisis reminds us that having a decision-ready, stress-testing framework – and being more prepared for the next crisis – is of value well beyond the cost of implementing that framework.  

Vivian Chan, Nomura Asset Management: Stress‑testing is very useful for financial firms, but everyone has their own approach. Regulatory tests are necessary but highly prescriptive by nature, and are therefore less useful for commercial purposes. Internal stress tests allow you to adapt the scenarios to better reflect your business – Nomura, for example, looks at whether the bond or equities markets will move against us.

The Covid‑19 pandemic is a fascinating scenario. When we were reforecasting post-Covid‑19, and using the pandemic as the baseline, we expected to see a fall in assets under management, as well as bonds and equites markets, but we didn’t. Instead, they dipped briefly and then rose. It’s important we have the stress tests, particularly for capital adequacy, but Covid‑19 proved the scenarios and economy are very difficult to predict, with little consistency in its impacts across industries and regions. Covid‑19 isn’t the only reason the value of stress‑testing has increased. Most companies stress-test a global economic downturn, and there are also reverse stress tests as part of firms’ risk and control self-assessment processes, where it is considered what could ‘break’ the organisation.

Model risk and stress-testing specialist at an international banking group: There’s value in thinking about what might happen in stress tests, and Covid‑19 doesn’t change this. However, the elephant in the room is whether current stress-test figures are remotely credible.

Models have proven far more accurate than even the best human intuition, but the results are now being challenged by senior managers and others. 

When stress tests are dominated by models, you need confidence in your model results. With both a Covid‑19 pandemic-induced recession materialising extremely quickly – much faster than previous recessions – and with results way outside of the typical calibration range, you have to question whether the models are delivering credible results. 

The biggest challenge is knowing the limitations of the models and the level of human overrides required, but it’s uncharted territory. It is not like previous recessions where industries behaved in a reasonably similar way. There are so many things outside the usual range that we do not have the economic or historic data to model with confidence what might happen. 

Even if it was a typical recession, it’s atypical in terms of the impact. Some industries are doing very well while others are experiencing extreme difficulties. Government intervention is also far more pronounced in this case, so some of the data is distorted as you cannot see some of the defaults. For example, how do you assess the future stresses in the personal sector when there are few defaults with people being furloughed or benefiting from temporary payment holidays? 

Models can only predict what they are calibrated for. When you go outside that range, you need management adjustments to reflect the perceived limitations and weaknesses in the models. The question is how to perform the overrides in a structured, rational way. For example, an International Financial Reporting Standard 9 submission might incorporate relationship managers’ views on individual companies’ performance alongside the model results. This helps reflect the idiosyncratic effect of this recession where some industries are booming and others have been shut down.

 

A recent regulatory review of banks’ internal stress-testing programmes found deficiencies such as siloed operations, overly lenient scenarios and a paucity of ad hoc stress-testing capability. Are these criticisms fair? What are the key tenets of a robust stress-testing framework?

Vivian Chan: I think this is mostly fair. Some larger banks have their own scenario expansion teams and undertake a diligent governance process to understand the outlook and macroeconomic behaviour for each market. For the Internal Capital Adequacy Assessment Process, for example, the UK Financial Conduct Authority has guidelines on how banks should stress the budget or business plan. It’s not just that it’s siloed, but that each bank will have a very prescriptive view. There are no publicly published scenarios each bank should be using, so it is difficult for anybody trying to compare or assess whether banks could withstand a certain economic impact. One firm could decide a particular market will drop 5% while another thinks it will be nearer to 10%. 

Laurent Birade: Although financial institutions are relatively better prepared than they were before the previous financial crisis, their work is not complete. The evolving Covid‑19 crisis has revealed some weaknesses in the regulatory framework that are typically run once a year and supported by many governance activities. The deficiencies we’re seeing today have only come to light due to the rapid evolution of this year’s unforeseen events. Ad hoc but timely stress-testing must supplement the regulatory framework to give banks the ability to run any scenario, at any time, for any set of assumptions. A robust stress-testing framework must incorporate:

• A wide range of risk evaluation (accounts for emerging risks)
• Consistent loss estimation 
• Scenario analyses (quantitative and qualitative, plus outer-bounds estimates)
• Recent data for more informed decision-making (data from the past few days, not months).

Moody’s Analytics provides rapid-fire scenario analysis capabilities to clients through its award-winning Capital Risk Analyzer platform, which can be used as a complement to a robust Comprehensive Capital Analysis and Review/Dodd-Frank Act Stress Tests framework.

 

The recent crisis has brought non-financial risks sharply into focus. To what extent will climate, geopolitical, technological and pandemic risks climb the stress-testing agenda?

Laurent Birade: Before the previous financial crisis, home price declines were not considered a serious risk. Similarly, Covid‑19 now moves pandemic risks into the spotlight. Climate, geopolitical and technological risks follow closely behind, now that these elements are surfacing more frequently – particularly climate risks.

 The speed and depth of the disruption caused by Covid‑19 is unprecedented. How can risk managers become more adept at ‘thinking the unthinkable’ and creating effective future scenarios? Scenario design is a crucial part of risk managers’ jobs – thinking outside the box and looking at risk in the following paradigms:

• Known
• Known unknowns
• Unknown unknowns.

Thomas Verbraken: Even before the Covid-19 crisis we had noticed non-financial scenarios becoming increasingly important to investors, for example the Brexit referendum, the US-China trade war and the US elections. Their concern is not only about the turbulence that may occur around these events themselves, but also the potential longer-lasting effects they may have through their macroeconomic impact. At MSCI we are spending a lot of time and effort on understanding the links between macroeconomic shocks and their consequences for financial markets. It is a complicated problem because of the different timescale in which macroeconomic changes and financial market shocks play out, and because financial markets may react to perceived changes in long-term macroeconomic expectations that ultimately may not materialise. Having a credible model to connect macroeconomic shocks to financial markets has proven very helpful for creating meaningful scenarios about geopolitical events. 

The second major trend we see is around climate change. Scenario analysis lends itself well to this topic for various reasons. First, there is no real precedent for climate change, so the analysis must be forward-looking in nature and stress-testing allows for that. Second, there is a large degree of uncertainty around future outcomes, not only in the evolution of the climate itself but also because of the offsetting nature of two types of climate risk for financial portfolios: transition and physical risk. The former is caused by the efforts undertaken to slow down climate change and can potentially impact carbon-intense companies negatively while creating opportunities for renewable energy and other climate-friendly technological innovations. Physical risk, on the other hand, is the cost of inaction, such as increased wildfires, floods, rising sea levels, and so on. Each climate scenario represents a trade-off between these two risks, whereby the balance between transition and physical risk varies. Being able to run a range of plausible climate scenarios (for example, from 1.5º to 3º Celsius global warming scenarios) and assessing their possible impact on financial portfolios is an increasingly important exercise for many investors.

Model risk and stress-testing specialist: Stress-testing is dominated by the needs of regulators to ensure banks are well capitalised to survive extreme stress. However, the level of detail and associated cost of stress-testing is increasing for firms. As a consequence, there is less appetite to work beyond the regulatory requirements. 

Stress-testing processes need to become much more efficient before a wider range of scenarios can be considered. Climate change and global warming will certainly be among these scenarios, but there are many more that could be valuable to banks – for example, wars or conflicts that disrupt the global supply chain.

This risk classification will enable risk managers to think more actively about what may trigger the next crisis.

Vivian Chan: Some of these elements are already considered, such as the impact of elections or trade wars within certain scenarios, or how a pandemic would trigger financial risk. But I think climate and technological risk will need to feature more. More firms are committing to reporting on climate risk. Environmental, social and governance is also a factor that could be added into stress‑testing scenarios as it could have a major impact in terms of credit risk. For example, in current circumstances, we’ve seen some retail firms exposed as having workforces operating outside Covid‑19 guidelines to maintain production.  

Technology has changed the way the world operates but again it’s difficult to predict the behaviour. I’ve seen reports that more people are dabbling in share dealing as a way to supplement reduced income under Covid‑19, and chasing futuristic tech stocks such as Tesla, so there’s potential for a bubble to emerge if prices are artificially inflated. 

 

The speed and depth of disruption caused by Covid-19 was unprecedented. How can risk managers become more adept at ‘thinking the unthinkable’ and creating effective future scenarios?

Vivian Chan: It’s very hard. Most banks are using severe adverse scenarios but, with Covid‑19, the impact on such areas as unemployment is extremely hard to predict from country to country. What could be worse than a third world war? A data war perhaps? Banks are using reverse stress tests of the most severe scenarios such as a US, China or European Union crash. But every bank could do something completely different. 

Thomas Verbraken: We have learned through conversations with our clients that risk management becomes more useful when you can actively involve more stakeholders in the exercise. Portfolio managers, senior executives and board members can all provide meaningful input into stress-test scenarios. Drawing on more diversified expertise may not only improve the scenarios themselves but could also make the stress-testing exercise more impactful, because stakeholders other than risk managers will have bought into the analysis early on. 

We have observed it is useful to have a framework that allows investors to express stress-test narratives quantitatively at a high level, with a relatively limited set of shocks on major market and macroeconomic variables (for example, MSCI USA, the 10-year Treasury yield and US GDP growth). Once the narrative is specified, the risk system should be able to take care of the rest and propagate such views to granular portfolios in a robust way. However, most effort – in particular from the other stakeholders – should go into the definition of economically sound scenario narratives at a high level.

Investors with the risk management processes and systems in place to do such stress-testing exercises efficiently might benefit on two fronts. Not only can they draw on much more expertise to ‘imagine the unthinkable’ during calmer times but also – and perhaps more importantly, once a crisis starts to unfold – they will be able to react quickly by defining stress-test scenarios early on in the crisis and stay on top of things by continuously refining the analysis as more information flows in.

Model risk and stress-testing specialist: I remember a quote about it being impossible to convince someone of a fact if their livelihood depends on not believing it. Similarly, in stress-testing, if a scenario suggests a profitable business strategy could cause large losses in the future, it will be difficult to obtain engagement from senior management if their rewards are tied to the strategy.

Thinking about the different responses to the Covid‑19 pandemic, you realise people don’t always act logically. Nor are all the facts readily available to make well-informed decisions when a scenario such as Covid‑19 is unfolding. History is not necessarily a good guide. Even as the pandemic was unfolding, I would not have believed the response would be a lockdown of the economy. The response to Covid‑19 appears far more draconian than the response to the Spanish flu, which was much more lethal, with most deaths occuring in the healthiest adults in the prime of their lives. Remember, when using historic scenarios, peoples’ values and behaviour were different in the past. Hence, you need to consider how attitudes have changed to assess the impact of future potential scenarios.

 

Stress-testing places huge demands on data, systems and resources, but too often the resulting insights are left on the table. How can firms derive more meaningful action from their endeavours?  

Laurent Birade: Having appropriate stress-testing based on needs is the right starting point. The regulatory stress test emphasises compliance with a strictly governed process that must meet qualifications of regulatory reviews. However, a more nimble top-down stress-testing process should accompany that infrastructure to ensure results can be produced in a timely manner, and that those results are congruent with the regulatory stress test. Incorporating the process as part of a regular strategic plan is only possible if your process is nimble enough to accommodate expert judgement analysis that enables managers to conduct rapid expert scenario analysis, to complement any quantitative approach.

Model risk and stress-testing specialist: The real question is, what is the purpose of stress-testing? If you’re saying it’s just a numeric exercise, and a tick-box exercise based on the maths, then I would question the value of the stress-testing.

I always remind myself that ‘failing to plan is planning to fail’. Similarly, the objective of stress-testing should be to stimulate debate so that management can make quicker and better decisions when a scenario is crystallising. 

Vivian Chan: Banks in particular spend time spent trying to pull a coherent dataset together. You need people who understand the finance and requirements of the stress test, as well as the technology and the data language. For example, I had to hire a specialist team that could enter the data into a system that uses Python, and then process it so we could run the necessary analytics. You often find gaps in the data, and measurement criteria change frequently. You put so much effort into data cleaning, creating the perfect system and looking at the requirements of a particular scenario that it is easy to lose focus on the commercial value. 

 

Considering the operational challenges caused by the Covid‑19 pandemic, how can firms drive efficiencies in their stress-testing procedures in the future?

Laurent Birade: By streamlining the stress-testing process and ensuring you can run the right scenario on the right process framework. This means adopting a bottom-up regulatory stress test with all the governance bells and whistles, complemented by a top-down approach that offers directionally correct answers in a timely manner for rapid decision-making. The regulatory stress test provides a robust governance process that, while required, may not accommodate daily scenario requests in times of economic upheaval.

Vivian Chan: One of the most time-consuming aspects is working out what data you need and applying the correct filters to enhance your core datasets. The challenge lies in identifying the dataset you need to apply the correct drivers to perform the stress test. In the scenarios, depending on the product, I might be looking at a bear market in bonds or equities, for example, and whether we have enough data to look at underlying assets in bond markets in particular countries. 

It’s a process of evolution, deciding what data you might need on a consistent basis and ensuring it is readily available and fit for purpose. The systems are evolving, but currently too much is Excel spreadsheet-driven and manually built. I’d like to be able to pull the data into something like Tableau to give a better visualisation of the impacts, and make it more commercialised to help inform business strategy. Technology has meant people have been able to transition seamlessly to working from home so, from that perspective, the challenge hasn’t changed.

The panellists’ responses to our questionnaire are in a personal capacity, and the views expressed herein do not necessarily reflect or represent the views of their employing institutions

Stress-testing – Special report 2020
Read more 

Sponsored content