Thanks for the invitation to speak today. It is something of an understatement to say that central banks have a lot on their plates these days. Globally this has led many prominent observers to question whether central banks are over-reaching in analyzing climate risk? I am going to explain why a proper analysis of climate risk is absolutely essential from a financial stability perspective. To make this tangible I will illustrate my argument by describing some of the findings from the climate change focused Exploratory Exercise, known as the “CBES”, that the Bank of England has recently completed with the major UK banks and insurers.footnote  The results were published in May, and I am going to try explain why I see them as one of the Bank’s most important publications of this year – perhaps I can even convince some of you to take a look.
I will divide my remarks up into three parts. I will begin with a brief overview of what we did in this exercise. Then I will describe some of the findings that I see as being most interesting and, in a few cases troubling. I will conclude by posing some open questions that I think senior management at firms ought to be asking.
My bottom line is that financial services firms and their customers have a long way to go if we are going to overcome some of the problems surfaced by the CBES. Unless we make significant progress in the near term, we will encounter serious problems for the financial system in the long-term.
The CBES structure
There have already been a number of speeches by colleagues at the Bank who have gone through the structure of the exercise and many of its quantitative findings.footnote  So I am going to focus more on qualitative lessons. To get there though I need to tell you a bit about how this project was set up. The basic idea was to present major banks and insurers with a fully worked out description of how the economy might evolve with respect to various climate scenarios.
A first response to the over-reach critique is to recognize that the industry could not do this on its own, involvement of the regulators was essential. Finalising plausible scenarios required a dedicated team in the Bank, building on the extensive work from the Network for Greening the Financial System (NGFS)footnote , working for over a year and it involved 1000s of hours of work. This included consulting climate scientists, talking to the financial institutions themselves, and drawing on expertise from economists and other Bank staff. Importantly they had to make sure that the various assumptions were internally consistent. Having a regulator do this rather than an industry committee was undoubtedly a more rigorous and efficient way to proceed. Furthermore, because much of the interest is in seeing the aggregate results, making sure that all the major players participated was essential. Of course, we can walk and chew gum, so this work relied on different staff than the ones who support the monetary policy committee and therefore this exercise did not hinder the MPC.
Ultimately, we probed three scenarios. Two had to do with different assumed paths for how the price of carbon changed in order to hit the UK government’s stated objective of net zero greenhouse gas emissions by 2050 (the price for carbon is a good stand in for overall circumstances in these scenarios).footnote  One transition path assumed that carbon prices began rising right away and moved smoothly to hit the required level − we called this the early action (EA) path. The other assumed that policy action was delayed and then had to be abruptly changed to quickly raise prices to meet the objective (the late action, LA path). The delayed scenario causes more disruption to the economy because the rapid shift in policy makes it more difficult for people and businesses to adjust, but under each of these scenarios the objective is obtained. So the economy reaches a new equilibrium with a sustainable climate by the end of the scenarios. Figure 1 shows the assumed price of carbon in the two scenarios.
Figure 1: CBES scenario carbon prices, early and late action
- Sources: Network for Greening the Financial System and Bank calculations.
The third scenario was very different. It supposed no new climate policies are introduced beyond those already implemented – the no additional action, NAA, path. This leads to devastating consequences for the environment and importantly means that the planet is not in equilibrium by the end of the scenario, with physical risks continuing to increase.footnote  So the longer you look out, the more grim the consequences will be (and there are other issues that arise in this scenario such as people losing their houses that are not present in the other scenarios). So if you do decide to look at any of the numbers in the report, keep that in mind, it is not really an apples to apples comparison between the first two scenarios and the third one.
To give some sense of the stakes, the average temperature change in the scenarios that involve a policy response and no additional action scenario are shown in Figure 2.footnote  I trust you all realize that climate scientists are agreed that a 3 degree increase in temperatures would be devastating, and that in the NAA scenario temperatures would still climb further even after the scenario ends.
Figure 2: CBES scenario change in global warming, late and no additional actionfootnote 
The participants all delivered their estimates of what they believed would happen to their loan portfolio and insurance businesses under these scenarios. The Bank staff engaged in dialogue with participants through the design process and after launch to promote consistency of interpretation around the rules of the exercise, and to refine methodology. These conversations were valuable for both sides.
The final and quite novel step in the exercise was to ask participants to update their assumptions, based on what we learned when we analysed participants’ initial proposed responses and projections. The participants were also invited to adjust their submissions to account for aggregate patterns that had been uncovered.
Here an example might be the best way to explain why this is important. A firm might suppose that there would be many exciting investment opportunities that would arise during the transition period and presume it could rotate its portfolio to undertake these investments, while pulling back from business models that might be becoming obsolete. If everyone reasons this way, the supply of funding available for the new opportunities could easily exceed the demand for funds. So this kind of rotation may be more difficult to accomplish than would be foreseen. A second response to the over-reach critique is how could a responsible financial stability regulator not want to know this information?
Some Remarkable Results
With the scene now set, let me turn to some of the main things that we learned. To be clear, given time constraints, I am obviously cherry-picking and you really ought to review the speeches I cited by my colleagues to see the conclusions that others have drawn. Here, I will focus on four areas.
First, it is apparent that modelling climate risk will require financial services firms to look at their customers differently than they do now. For many customers, business as usual will become impossible to sustain. This could occur for several reasons. It is possible that the product being sold is itself inconsistent with a world that features much higher carbon prices. In other words, a direct effect of the transition to the greener economy could render the product unviable.
However, there could also be an indirect reason why the product could become obsolete. A firm could be an input supplier to a business which is directly affected and so even though the input itself can be produced without any required changes, the demand for the input could disappear.
Lenders do not normally need to know every detail about the entire production chain of their borrowers. They also do not necessarily collect information on the carbon content of the products being sold by the borrowers. Insurers also do not currently need to know this information, but will in the future to assess whether the securities that they will invest in will retain their value. So financial services firms will need a lot of non-standard information and reporting requirements will need to change to make this information available. Again the customers will have to cooperate on this because in some cases the data are not collected by anyone.
This leads me to what is probably the most striking finding for me that comes out of this exercise is the degree to which missing data hampered responses. Every single respondent had some examples in their submissions for which they could not fully respond because they did not have the information that they needed available. Here are some of the examples of things the participants said they wished they knew but did not.
- The market value of properties will increasingly depend on their energy efficiency if carbon prices rise significantly. Yet, lenders rarely know anything about the energy efficiency of the properties they lend against given it has historically not been a data item they routinely collect. So judging the collateral value of a house is going be a challenge.
- The physical risks of climate change will impact some areas more than others. Lenders rarely know the physical location of suppliers to a business. So judging whether a borrower maybe at risk for disruption would not be possible.
- Insurers tend to rely on Environment, Social and Governance scores for judging whether their portfolios are on a path that is consistent with their stated aims. Several noted, however, that ratings for some firms are unavailable and in many cases (as is well-known) ratings from different sources are inconsistent. Furthermore, ESG scores pertain to a broader set of factors than just climate issues. These discrepancies caused problems in complying with the test.
Another response to the over-reach critique is how we would verify these kinds of data gaps are present and decide which are most acute if we had not done this exercise?
A second, closely related observation is that the modelling was typically uneven inside each firm. Firms differed greatly in the degree of granularity that they used in their modelling. At a high-level, some firms think about top down approaches that suppose all firms in a sector are identical so that a single loss assumption for the sector is the essential driver of the results. Others use a very detailed bottom up approach that took borrower specific conditions into account. Particularly for the energy sector there was varying degrees of sophistication shown in thinking through the consequences of the various scenarios.
Given that we are in the early stages of modelling climate risk, it is not surprising that firms differ in the degrees of sophistication that they exhibit in modelling. What was surprising to me was that even within a given firm there tended to be a lot of variation in the how different parts of the organization modelled things. Here are some examples.
- One bank was quite careful to account for the financial impacts from the ending of the government’s flood reinsurance program in 2039 and used property flood scores to judge whether a location would be uninsurable from that point onwards. Yet the same bank did not model anything about risks from heat, coastal erosion as a result of sea level changes, windstorm or freezing and thawing.
- Another firm carefully considered the effects that four physical hazards would have on their commercial real estate portfolios but did not consider the impacts on any of parts of its wholesale business. Generally speaking, the ability of most firms to assess the financial impacts from physical impacts on retail and credit books was materially poorer than compared to transition risks.
- For general insurers, modelling today’s risks is already fairly sophisticated – with risk pricing often carried out at a postcode or even property level. However, this exercise identified that many of these models are not sufficiently adaptable to cater for climate-related risks that are expected to increase – for example, investigating the implications of increased surface water flooding, or assessing potential losses from a UK west coast storm surge. In addition, this exercise also highlighted that only a few insurers were able to readily estimate the financial implications of different levels of flood defences.
The uneven responses within the same organization yields an important implication for management: there are limits on how much having the right tone at the top buys in terms of managing climate risk. My takeaway from the CBES is all the major firms have more work that needs to get done.
Third, there was also another important modelling consideration that proved important. All firms relied, to differing degrees, on third party consultants to assist in some aspects of the modelling. Again given that this is a new topic the fact that there is not complete expertise in house is to be expected. It will take time for the financial services firms to hire the new staff that is needed and build up competence.
In a few cases, even the consultant’s models were not flexible enough to fully follow the instructions about what was to be assumed in a given scenario. For instance, one model that was used was overly reliant on a simple treatment of emissions to capture the impact of transition policies. Users of this model would feed in estimates of emissions and then a carbon price would be applied to the emissions to arrive at cost numbers. This meant that for some sectors the impact on a corporate’s customers was poorly assessed. This was a particular issue for modelling the performance of financial firms for whom the success of their business model is intimately linked with the performance of their clients.
More complex sector specific models also made simplifying assumptions that were not always consistent with the scenarios provided. For example, we tried to ask participants to factor in how changing demand for their products/services would impact their customers. We also specified some guidelines about how much adaptation could be plausibly assumed. Some participants instead just assumed that market shares of customers would remain constant -- this implicitly assumes the customers are able to respond to the scenario in a way that was not modelled and could not be assessed.
I expected that the modelling challenges would be large. I also expect that improvements will be incremental. The value in the CBES is showing many tangible places where progress is needed so that both supervisors and firms can benchmark their capabilities.footnote  A fourth response to the over-reach critique, is how else would this progress occur if the supervisors were not engaged in an exercise like the CBES?
A final issue surfaced when we provided the participants with feedback on the responses that we collected and asked them to reconsider their initial answers. One risk that we noted is that many lenders and insurers plan to dial back business with so-called brown industries and actively are looking to establish ties to green industries. This will need to happen, but the overall health of the economy and financial system requires an orderly rotation. In particular, there are immutable constraints on the speed at which green forms of energy production can be ramped up. There needs to be adequate credit made available to brown firms so they can invest in greening during the transition period and so that the overall energy supply for the economy is adequate. If everyone cuts off the brown producers indiscriminately and too quickly, that could be calamitous for the economy. The distress in the economy would also likely have adverse effects for the banks. Some of the participants realized that their plans might not be feasible or at least much more challenging to stick with than they had initially realized. For instance, a bank might need to rotate its customer mix to achieve its lending goals.
Life Insurance firms tended to assume that a simultaneous pullback in asset exposure would be manageable. To the extent divestment was needed the standard assumption is that this could take place via letting bonds mature. Of course, if the debt issuers cannot roll their expiring bonds that will create additional risks, for instance causing impairments on their bonds that have not expired.
I hope these examples explain why I think this exercise has been so valuable, both for the participants and for the Bank of England. This is going to be a long road and everyone involved has already learned a lot. In the spirit of trying to advance the agenda, here are some questions that I think I would be asking if I was on the management team of a financial services firm.
First, how much extrapolation are you having to do because of missing data? Perhaps more importantly, how will we fill in these gaps? It does not seem from afar that progress on this front is likely absent an active plan for improvement that needs to come from the top. Given the massive uncertainty associated with how this will play out, getting much better data is essential managing risks well.
Second, do you know which third parties you’re reliant on and what are their key assumptions/features of their models? It is probably efficient for at least the next few years to outsource some of this work. Imagine trying to decide which consultants to pick: it is quite likely that the most highly rated consultants provide models that require better data to be input and perhaps more expert judgment to take maximum advantage of the models. Do you know if your staff has the needed data and expertise to effectively utilize and eventually challenge the consultants?
Third, what is my firm best and worst at modelling and how will we level up our capability? The Bank of England is meeting with the participants in the CBES to try to share best practices. It will also share these findings with broader financial firms, who will also find it valuable. Firms though must own the responsibility of implementing the changes and this is going to take time, so starting sooner is better.
Fourth, going forward what are the key assumptions your institution is making about the trajectory of climate policy and physical risk and its impact? Are the assumptions associated with your risk management of different credit books internally consistent within your organisation? In some sense, being able to answer these questions is essential for having a fully credible modelling process in place. If the management cannot ask the staff to do scenario analysis that is internally consistent, that is a sign that the modelling capabilities are incomplete.
Finally, if other banks and insurers are expecting to carry out the same strategic changes to their businesses that you are, would your planned response to climate risk still be a workable/sensible one? To me this is the most important question for financial stability. We can’t expect firms to automatically internalize the effects of their choices on the overall system. That is exactly why doing this kind of exercise to uncover the system-wide response is essential.
When Bill Clinton was running for President of the United States in 1992 his campaign advisors famously had a sign in their headquarters to try to make sure they kept the big picture in mind. That sign said “It’s the economy stupid”. I fear that the CBES was so rich with numbers that it is easy to get lost in the details and lose the forest from the trees about what we learned already and what we need to do to move forward. That is why I titled the speech “It’s the risk management stupid”.
My big takeaway is that private sector firms have a big risk management problem regarding climate risk in front of them. It is the job of supervisors to assess the size of the problem and work with the industry to address it. To say this is over-reach is folly. Risks to financial stability can arise from many places, whether its climate change, cyber-attacks or the cryptoassets. If the supervisors were ignoring this problem, a decade on critics would be asking how could we have not seen problems coming and failed to act?
Thanks for your attention and I will now hand back to Simon for the panel.
The views expressed here are my own not those of the Financial Policy Committee. I am grateful to Sarah Breeden, Jon Hall, Carolyn Wilkins, Chris Faint, Neal Kilbane, Ben Westwood, David Wilkinson, Genevieve Monclin Evans, Bridget Clifford, William Metha, Nick Parry, Zane Jamal, Giorgis Hadzilacos, Shreya Biyani, Kara St. John, Cassandra Archer, Andrew Bailey, Ali Moussavi, Jethro Green and Stefan Claus for helpful comments and conversations.
See Results of the 2021 Climate Biennial Exploratory Scenario (CBES), Bank of England, 24 May 2022
See Driving different decisions today: putting climate scenarios into action (Sarah Breeden, 20 October 2021); Balancing on the net-zero tightrope (Sarah Breeden, 7 April 2022); Why macroprudential policy needs to tackle financial stability risks from climate change (Elisabeth Stheeman, 3 May 2022); Climate capital (Sam Woods, 24 May 2022); Climate Biennial Exploratory Scenario: Insurance Insights (Stefan Claus, 8 June 2022)
The NGFS has published a series of freely available climate scenarios. The Bank built on these scenarios for the CBES exercise, and other central banks are also drawing on them for their own exercises.
Carbon price depicts a shadow price of greenhouse gas emissions, i.e. the marginal abatement cost of an incremental tonne of emissions. This is a simplification, intended to capture a range of different policies to reduce greenhouse gas emissions, which may include, for example, carbon taxes, cap-and-trade schemes, green subsidies and environmental regulations.
Climate scientists’ projections conditioned on no further policy action suggest that temperature increases as significant as the exercise would only be likely to occur later in the century. The shifting forward in time of these more severe temperature rises – and associated physical risks – was deliberate, as it allowed the Bank to explore the impact of these more extreme risks within the 30-year time horizon of the exercise.
Our publication also shows outcomes for GDP and many other variables.
Taking policy action sooner in the EA scenario would mean that there was more chance of a lower peak temperature than in the LA scenario. As a prudent and simplifying assumption, however, the warming level incorporated is the same in the EA and LA scenarios.
The CBES also complements other supervisory expectations regarding adjustment to climate change, e.g. the Prudential Regulation Authority Supervisory Statement SS/3-19.