Climate stress tests: How regulators are turning the screw on banks
Annual stress tests of bank balance sheets were one of the last decade’s most obvious supervisory responses to the global financial crisis. With a wave of new bottom-up assessments now getting under way, regulators hope to do something similar with climate risks. Can they do it or will this simply result in a toothless box-ticking exercise?
Can the response to one crisis inform the response to another? Financial regulators across the world are hoping that the answer to that question is a resounding ‘yes’ as they grapple to repurpose the balance sheet stress testing regime that was adopted in the wake of the 2008 global financial crisis to fit the even more awesome challenge posed by a warming planet.
This year marks the moment when regulators have fired their machinery into action. An early effort to quantify climate risks to the banking sector by Dutch authorities in 2018 and a landmark pilot bottom-up test of the French financial sector in 2021 by the Autorité de Contrôle Prudentiel et de Résolution (ACPR) have set the scene for more complex scrutiny being undertaken right now by the Bank of England (BoE) and next year by the European Central Bank together with the European Banking Authority.
The UK and European tests will report their results in 2022.
Similar projects are at various stages in countries including Brazil, Canada, Australia, Hong Kong and Singapore, most of which are also expected to report results in 2022. The US Federal Reserve is watching closely. Chairman Jay Powell has said he supports the concept, making it a virtual certainty that the country will adopt its own process before long.
And as Euromoney went to press, researchers at the Fed published a paper detailing one approach to measuring risk at banks, unveiling a proposed systemic climate risk metric that they dubbed CRISK, being the expected shortfall in capital from a climate stress scenario.
However, the thorny question of what the regulatory consequences for banks will be of such tests remains unanswered. At the time of writing, there are no formal suggestions of explicit capital add-ons to be linked to bottom-up stress test results. Those involved prefer to describe the processes as exploratory, data-gathering exercises, intended to demonstrate to regulators and banks just what they do and do not know about climate risks and to build systems to fix the knowledge gaps.
“The stress tests are coming from regulators and are for regulators, but they are also useful for banks because to have regulators asking for things is a good way to drive internal change,” says Richard Barfield, a prudential risk and regulation specialist at law firm Hogan Lovells, who has spent years advising banks on developing responses to regulatory stress tests. “Banks have a self-interest in understanding the risks, so they can adjust their appetite.”
But many observers hint at an eventual and inevitable feed-through to capital requirements from further iterations of such tests in future years. What is not clear is whether this will be an organic process, with banks adjusting their internal models – and therefore their capital requirements – as a result of being compelled to report to regulators, or whether regulators will be more proactive and impose explicit regulatory add-ons themselves.
In all this, what is often not fully grasped is the true objective of stress testing for climate risks to banks. Such processes are not designed to simply punish banks for lending to dirty industries. Climate stress tests are about identifying, provisioning for or preventing systemic risk in the banking sector.
That means they must evaluate the credit risks that result from the physical consequences of a failure of the world to respond to climate change – borrowers whose businesses or assets are threatened by fires, for instance. But just as importantly, they must also take into account the business risks of actually responding to climate change – borrowers who are threatened by transition policies or business lines at banks that will suffer from reduced lending to those dirty companies.
Set against that must be assessments of the new opportunities emerging from the financing of alternatives like renewable energy – an area that Autonomous Research has estimated could contribute $6 billion of annual revenues for European listed banks.
The ball started rolling with a report in 2017 by the Task Force on Climate-related Financial Disclosures, a body under the aegis of the Financial Stability Board. The report brought to the fore the issue of how to disclose risks related to climate change and the efforts now under way to shoehorn this into a new type of stress test are the result.
It is hard to overstate the challenges. Stress tests imposed on financial institutions around the world in the wake of the 2008 financial crisis have been effective in bolstering banks’ capital levels to the point at which they were able to enter the coronavirus pandemic in early 2020 in good shape – albeit with their balance sheets also protected by the extraordinary government support of economies since then.
What existing stress tests have done is familiarize big financial institutions with the concept of running such tests, scrutinizing their exposures and justifying their internal modelling to regulators. What they have not done is prepare the industry for an assessment of risks that are of a different category and over a very different timescale. While traditional stress testing seeks to model risks that might emerge over a cycle of a few years, to be meaningful, climate stress tests must address potential shifts that will take place over decades.
Some banks will be asking themselves if they should be using an internal risk-weighted assets surcharge for climate
In doing that, they must also cope with a dynamic that Ernest Hemingway captured in his observation of how bankruptcies occur: “Gradually, then suddenly.” Climate issues are full of such tipping points; very few are linear progressions. Physical risks associated with rising sea levels, for example, might increase only slowly until a critical level is breached, like a sea defence, at which point the risk steps up to become catastrophic.
Those tasked with modelling climate risks to the banking sector must factor in many such tipping points, alongside a separate path of government and institutional policy measures that follow their own timetables and which will differ radically between jurisdictions. Global banks already thought themselves hard done by when faced with the multiplicity of regimes for traditional capital regulation. But meeting the requirements of regulators around the world on the issues surrounding climate stress could make that look like a sideshow.
Whatever the difficulties they face, banks have been put on notice. Sarah Breeden, the Bank of England’s executive director for financial stability strategy and risk, said in May 2021 that banks were underestimating the impact of climate change. Christine Lagarde, the ECB president, said in July that most banks the central bank supervises did not meet its expectations on progress towards climate targets. In September, the ECB said that its analysis of four million companies and 1,600 banks showed that European corporate loan defaults could rise by 30% by 2050 if there were no efforts made to slow the physical risks of climate change, making it a much worse outcome than any risks associated with transition.
In May 2020 the ECB published guidelines of the approach it expected banks to take when assessing climate risks internally. While they are non-binding and are intended to form part of the supervisory dialogue, the guidelines require banks to explain why they are not complying if they are not. As is typically the case, if those explanations are unsatisfactory, additional Pillar 2 requirements could result.
The European Commission followed up in April 2021 with the long-awaited publication of its sustainable finance taxonomy – its system for classifying green investments – ahead of final decisions in 2022 on what can be labelled as sustainable investment.
Now the supervisory push across the world is stepping up further through stress tests, in something of a geographic reverse of the post-2008 order that saw the concept of stress testing launched in the US before it migrated to Europe and then the UK.
The ACPR’s test in France, which covered nine banks and 15 insurers, published its conclusions in May 2021. The test involved a 30-year horizon to 2050 and looked at 55 sectors. It used a static balance sheet that ran to 2025, much as traditional stress tests do, but added another analysis based on a dynamic balance sheet through to 2050 that aimed to assess how banks might implement their various climate risk policies over that time.
The ACPR test found that French banks and insurers had a “moderate” exposure to climate risks, with banks in particular suffering a fairly low impact from those sectors that are expected to be most affected by transition to a net-zero target – sectors that accounted for less than 10% of the credit portfolio of the banks concerned.
The risk to insurers, not least from physical impacts, was higher.
For all the attention the ACPR test garnered in prudential circles, its ambitions were necessarily limited. It was pitched as a voluntary pilot exercise, the first bottom-up test of its kind that sought to consider risks of a different order and time horizon to any conducted before. But it is seen as a key benchmark for the tests that will follow it. While French banks might have a relatively low exposure to the most affected sectors, cost of risk was shown to triple in those sectors over the period covered by the test.
The next French assessment will begin in 2023. But for now, all eyes are on the UK and the Climate Biennial Exploratory Scenario (CBES) stress test that the Bank of England launched in June.
The CBES test, which covers the 19 biggest banks and insurance companies in the UK, covers three scenarios that span 30 years. They assume either early action by governments to cut emissions or late action to do so or no action. They cover two types of risks: the physical, like flooding, and those associated with transition to lower-carbon economies, such as shifts in asset values or the effect on business models.
The UK test of banks is focused on credit risks, with banks required to show the progress of their provisioning against lending through the various scenarios. After pushback from the industry, the CBES is explicitly not addressing traded risk and non-traded market risk, after the BoE agreed that dynamic exposures of that nature would be less appropriate for the test – although it has said it could change that assessment in the future. For insurers, the focus is on invested assets and insurance liabilities.
As was the case for the French test, the CBES scenarios are modelled on scenarios drawn up by the clunkily-named Network of Central Banks and Supervisors for Greening the Financial System (NGFS), an initiative created in 2017 that aims to create best practice resources for central banks conducting assessments and whose secretariat sits within the Banque de France. The BoE’s Breeden chairs the macrofinancial workstream within the NGFS.
The NGFS published the second edition of its scenarios on June 7, one day before the BoE published final details of the CBES. The underlying NGFS scenarios the BoE is using are: Net Zero 2050, Delayed Transition and Current Policies.
The early action scenario in the CBES assumes that transition to net zero begins this year, with taxes and policies becoming steadily more onerous over time, until emissions reach net zero by 2050. Late action implies no transition until 2031, but with more dramatic interventions at that point to reach net zero by 2050. The no action scenario, by its nature, is mostly concerned with the physical impacts of unchecked climate change.
Tail risks are a particularly tricky area. Banks are trying to model progressive changes that are material, but what is even more difficult to model is the volatility of climate, which could be high. To get around that the BoE is guiding banks to model 2050 as if it were 2080. Modelling for 2080 necessarily captures more change, meaning that it acts as a proxy for a 2050 result if the reality ends up worse than the prediction.
Banks must make their submissions to the BoE by October 13, but another round is seen as inevitable before final aggregate results are published in May 2022.
“Given the new aspects and the emerging topics relating to climate risk, it is highly likely that the Bank of England will find inconsistencies between submissions and will require changes,” says Fernando De La Mora, head of Spain and Portugal at Alvarez & Marsal, a consultancy advising institutions on climate stress tests.
This is all new for banks, but at the same time it is a long topic that is here to stay
The ECB kicked off its own process confidentially in the summer with a webcast to banks explaining its methodology and templates. The test itself will begin in January 2022 and run through the first half of the year, with aggregate disclosures expected in July.
The European exercise is being seen as particularly demanding, not least because of the multiplicity of time horizons being used. Banks will have to run a one-year simulation of low probability but extreme stress related to physical risks such as floods and heat waves. Then they have a three-year mid-term scenario, where the assumption is for a quicker than anticipated transition to high carbon prices, with an accompanying credit and market shock. The final set uses the NGFS long-term scenarios.
“As with many topics, it seems that regulators are competing with each other to see who can come up with the most comprehensive and sophisticated approach to the topic,” says De La Mora. “Overall, the ECB seems to be trying to be more advanced, with a greater ambition than the UK tests, because it is including more types of risk and different time horizons.”
Bankers also say the ECB is notably interested in the question of reputational risk. “The supervisors are less concerned by loss of revenues caused by exiting certain areas, because there are likely to be other areas where there will be growth,” says one banker. “But we think they are concerned with negative coverage in the media or the potential for funding difficulties that could come with a shaming campaign.”
Many banks don’t want to talk at the moment about their work for the stress tests. But one that did – perhaps unsurprisingly, given its intense focus on the area of sustainability – was BNP Paribas.
Marc Irubetagoyena, head of group stress testing and financial synthesis at BNP Paribas, sounds energized by the process. “We are very keen to work with supervisors to build these exercises,” he says. “Clearly for us it is a way to design our own framework and for supervisors it is a way to assess the right approach to capture climate risks.”
The past is providing limited information on what will really be the key drivers in the next 10 to 15 years
He says the most meaningful risk for a bank like his, given the actions that it has already taken to limit its exposures to those sectors most impacted, is credit risk from transition. With so much science already available on climate issues, modelling tools are not necessarily the problem. It is more about the data.
“The past is providing limited information on what will really be the key drivers in the next 10 to 15 years and the materiality of these drivers on the cost of risk and on capital requirements,” he adds.
It’s clear that all the big firms have invested heavily in their preparations for the climate tests. “This is all new for banks, but at the same time it is a long topic that is here to stay,” says De La Mora. “Banks are obviously not ready, but they are getting ready. They are putting a fair amount of resources into upgrading their data and modelling in this space.”
And because it is bottom-up, the task is big, particularly if banks are creating their models internally from scratch, rather than licensing from third parties. They have to model hundreds or thousands of corporates. They must factor in what might happen to a company’s income and costs, how it might respond to a reduction in demand, what might happen with a carbon tax, how much might be passed on to clients and what the impact on capital expenditure might be.
Such analysis can lead to interesting conclusions. For the oil and gas sector, for example, it does not follow that just because in 30 years’ time there will be less demand for oil, there will be credit problems in that industry.
The biggest international integrated players are highly cash generative with low gearing. They could spend years running parts of their business down, progressively repaying their debts and returning capital to shareholders. Some will shift to renewables or hydrogen, but not adapting in this way does not automatically lead to a failure to repay borrowing, something that bankers argue is not always well understood.
Someone familiar with these challenges is Jean Boissinot, deputy director for financial stability at the Banque de France, who also serves as the head of the NGFS’s secretariat. Speaking to Euromoney in his NGFS capacity, he says he sees the French, UK and European tests as part of a new generation of approaches that go further than mere information gathering.
“The goals are to make sure that banks look very seriously at how to assess these risks and start to gather relevant information,” he says. “To really look at these risks you need to have a much more granular approach than a traditional stress test and you need to see how to manage the longer-term horizon. These are new questions.
“To be honest, this is work in progress and we are not yet completely out of the lab at the moment.”
Boissinot defines the challenges as threefold: the time horizons, the data and the models. With the time horizon, the task is to balance the fact that tipping points mean that climate change is simultaneously happening in different ways over the long and short term.
There are two broad types of data gap. The first is what Boissinot thinks of as a dissemination gap: useful data might exist, but the people looking for it do not know it exists – what one might term an ‘unknown known’. Climate scientists have reams of data that would be useful to banks, as does the insurance community when it comes to physical risks.
Equally, one institution that is particularly focused on one question might find a solution or the relevant data, while others that are focused elsewhere might not. Identifying processes whereby knowledge can be shared is a challenge for a competitive industry.
The other gap is still more fundamental, says Boissinot. “This is a new way of looking at counterparties, so the relevant data doesn’t always exist, and on top of that it is not clear that the impact of these risk factors is clearly understood. If you give me a carbon price for the transport industry, what should be the right metrics to capture its credit risk sensitivity to that carbon price? That kind of question is not settled yet.”
The way Boissinot tells it, it’s not just a case of not having the data – the risk framework is also not ready.
Bridging such gaps is the focus of a specific workstream in the NGFS, but Boissinot argues that institutions cannot wait for perfect answers.
Everybody has been saying that data is an issue for the last 10 years; and sometimes data gaps have been a fig leaf for a slow pace of progress
“Everybody has been saying that data is an issue for the last 10 years; and sometimes data gaps have been a fig leaf for a slow pace of progress,” he says. “But within the NGFS, we strongly believe that a lack of data is not an excuse for not acting now.”
The benefit of the kind of bottom-up approach that the climate tests are generally forcing on banks, where the process starts from a defined set of scenarios and variables that are then applied to different categories of credit to reach an overall result, is that banks and regulators are better able to understand risks at the level of a counterparty because it forces a more disaggregated approach.
“If you are asked to imagine that tomorrow we stop burning coal, then the result is much more precise because you will discuss with your sector experts what would happen to specific counterparties or specific loans in that scenario,” adds Boissinot.
Banks are coming to this conclusion too. When the French regulators began their test, banks resisted the notion that it would be essential to work at a much more granular level than they were used to doing. By the end, many had changed their minds, say those who observed the process.
Disclosures of test results are, for the moment, expected to be limited to the most aggregate level deemed meaningful. There is clearly nervousness about disclosure while so much remains imprecise.
“The scenarios being used are plausible outcomes rather than forecasts,” says one official. “In this pilot phase I am not sure that there is very much value in putting out numbers for specific banks as there is so much that we do not know. But I would not rule it out in a few years.”
Can that approach really stand scrutiny given the severity of the risks?
“People will make the case that even if the data is crap we should disclose it,” says the official. “But what you have to recognize about crappy data is that you can write nice reports based on scenarios, but if something in the methodology was completely off, then disclosing on the basis of that would not be helpful.”
For the moment, there seems little regulatory appetite to disclose information beyond aggregate sector findings. This is in keeping with the objectives of the tests, which are to enable banks and regulators to get a handle on the systemic risks and to educate themselves on the challenges of proper translation of them into credit decisions. Banks have lobbied hard for regulators to steer clear of individual disclosures, say observers of the process.
“There are always two dimensions to stress tests – qualitative and quantitative,” says Barfield at Hogan Lovells. “In the case of CCAR [the Fed’s Comprehensive Capital Analysis and Review] in the US, the numbers were important, but equally important was the ramping up of a bank board’s understanding of risk and the institution’s capabilities of assessing it. It’s exactly the same with climate change.”
The nature of this initial round of climate stress tests, based on static balance sheets at a point in time, means that banks may well steer clear of disclosing much new information to markets, even if it is positive, for fear of confusing investors. But over time, those with particularly strong results may find it tempting to do so. Typically banks are not prevented from disclosing details of traditional stress tests should they choose, provided that they do not get ahead of the regulator’s official disclosures. And if that happens, competitive pressure will build.
“Like anything else, once you start to disclose something, banks will begin a race,” says De La Mora at Alvarez & Marsal. “That could create an acceleration of alignment of banks to a net-zero position, so it may actually take a shorter time. They could sell portfolios or do other transactions to move more quickly.”
Regulators might like stress tests for the information they give them on systemic risks. Banks might end up grudgingly liking them for the way they push them into more sophisticated credit analysis, which will doubtless be able to be used for competitive advantage. Investors like anything that helps differentiate one business from another. But not everyone is convinced by disclosure for disclosure’s sake.
If you are a bank that excels in comprehensive reporting, you will get high marks from the market. But what is missing is a focus on the real narrative
“My main disappointment at this stage is that so much of the debate revolves around reporting,” says Sam Theodore, senior consultant at Scope Group. “If you are a bank that excels in comprehensive reporting, you will get high marks from the market. But what is missing is a focus on the real narrative. Analysts look too much at the number and not enough at what a bank is doing practically right now, how it is addressing the transformation of its portfolio.”
He says it reminds him of credit ratings 40 years ago. What he wants is more specifics and fewer broad declaratory statements. “Banks say they cannot be too specific, but at some point you need more transparency in the narrative.”
Can regulators hope to capture that? “No and they don’t want to,” says Theodore. “For them it is about specific metrics for prudential risk. They are set up for metrics.”
He likens it to how the focus since the global financial crisis has settled on whether banks are sufficiently well capitalized, rather than what they are doing with that capital. “People say that this bank is better because its core equity tier-1 ratio is higher, so there is less of likelihood that a supervisor will intervene,” he adds. “But you hear less and less about what a bank is doing with its capital. I am afraid that the same will apply with climate risk.”
Perhaps the murkiest area of the tests is the issue of whether they will lead to specific regulatory actions at some point. In a report in March 2021, Fitch Ratings said it expected climate stress tests to feed into prudential capital requirements in the longer term and many bankers think this is likely. But it depends on how quickly regulators change their thinking around the objectives.
“The policy discussion is starting, for sure,” says Boissinot, “but one important aspect is to be clear on the policy objective. If your goal is to have the banking sector very well aware of the risks and able to manage them well, when should you rely on capital add-ons?”
Regulators certainly have wide scope to exert pressure even without specific add-ons, such as broad additions to regulatory capital requirements of firms if they are not satisfied with their approach to risk management.
“The Bank of England has been forthright in saying that at this time they don’t expect the results to feed into capital requirements,” says Pauline Lambert, executive director for financial institutions at Scope. “But once the tests are run and if the results show that there are meaningful risks, it will be difficult to ignore them. There may not be a blanket approach as the risks will likely vary by bank, but this could be addressed through the use of Pillar 2 guidance.”
That is an area that BNP Paribas thinks could be impacted. “The position of the bank is to promote a Pillar 2 approach on these matters rather than in the Pillar 1 layer, which would simply add to risks that are already somewhat covered, because we are already running assessments of very extreme situations there, so there would be an element of double-counting,” says Irubetagoyena. “We would favour a Pillar 2 approach that promotes a framework where we link the consequences of climate change to an enriched credit, market and operational risk assessment.”
In June, BoE deputy governor Sam Woods told a conference that the BoE did not yet have the evidence to force banks to hold more capital in respect of climate risks. And he went further, adding that the regulator penalizing banks for lending to carbon-heavy industries might be “over-reach”. He noted, however, that he thought banks might in any case gradually charge more for credit that exposed them to greater risks.
Hogan Lovells’ Barfield thinks that banks will inevitably end up imposing higher capital requirements on themselves as a result of the process. “Inside the banks, their own risk management policies will say that if you have a risk, you need to make sure you have enough capital under stress or an action plan to deal with it,” he says. “And that should come through because of the tests. Some banks will be asking themselves if they should be using an internal risk-weighted assets surcharge for climate.”
What is clear is that Internal models of default probability and loss given default will need re-engineering to accommodate the additional complexity of climate change, particularly as there may be restrictions on the mitigation that banks can obtain through traditional insurance products when it comes to climate issues.
And while a central bank might not have the legal force to impose specific environmental, social and governance (ESG) factors through its capital regulations as things stand, the fact that ultimately all the stress test outputs feed into credit risk still gives regulators scope to act.
Once the tests are run and if the results show that there are meaningful risks, it will be difficult to ignore them
“If you have a bank that is exposed to clients that are at high transition risk because their business model is going to be forbidden by the European taxonomy, then we are not talking just about ESG – we are talking about credit risk,” says Paola Sabbione, co-head of bank research at Barclays. “So in this sense the ECB might say banks’ models need to also capture that discontinuity because this will affect their clients’ future creditworthiness.”
But Sabbione adds that regulators do not only have the threat of sanction at their disposal. Carrots could be just as important as sticks.
“There could be positive things that the ECB could do to support the behaviour that it wants to see,” she says. “You could have a green targeted long-term refinancing operation with benchmarks for lending based on green loans, as we have seen in Japan, or you could have lower haircuts for green collateral placed with the ECB.”
One of the takeaways of the French test was that it was the risk governance in banks that mattered most at this stage. That is not to say that there will be no policy outcome over time, but it might be different to that which results from traditional stress tests.
Others echo that view. “We know that the financial system in the UK is very robust in terms of capital and liquidity, so the immediate issue is not going to be capital,” says one adviser. “It’s much more whether banks understand the risks. If they have a portfolio of clients in an impacted sector, how good is their data? How effectively have they updated their credit assessment process?”
In any case, capital add-ons might not even be an effective answer to whatever the tests throw up in their results.
“Eventually the risks are so big that the idea that you can take them onto balance sheets and then put capital in front of that afterwards might not be the right approach,” says one official. “What’s different about these risks is that they unfold over time and so there is time not to take them on board at all. That’s what transition is about.”
Looked at in this way, the ultimate purpose of climate stress tests might be to make climate stress tests unnecessary. Former BoE governor Mark Carney, in a speech in 2015, spoke of the goal to ensure that climate issues never became financial stability issues.
“The intention is to make sure that the risks are managed upfront,” says Boissinot.
What often gets lost in the debate around supervision of the financial system is that it can only go so far. In the case of climate, there are other ways of changing behaviour within individual economies, but these must involve conversations between households and governments. Regulators can contribute to a smooth transition by making sure that risks are understood and that the financial sector is not pushing in the wrong direction, but that might be the limit of their power.
Many are fond of the crisis-response mantra that monetary policy is not the only game in town. When it comes to climate, central banks will surely not be the only green game in town either.
Those involved in these processes are evidently passionate about their mission, confident that only good can come of such scrutiny. But will climate stress tests simply add another metric? Is it all just more box ticking; for supervisors to say they did their jobs, for banks to say they passed and for investors to say they checked?
Theodore sounds a gloomy note.
“With capital stress tests, all markets care about is who passes and fails,” he says. “For climate risks it will be the same. They will focus on the outliers, without seeing the mainstream.”