A comprehensive review of a comprehensive review (Louie’s Version)

This is a free-to-view edition of Unpacking Climate Risk. Become a paying subscriber today to get articles sent direct to your inbox three times a week.

The review is in!  Courtesy of the United Nations Environment Programme Finance Initiative (UNEP FI), we now have a definitive rundown of global supervisory climate stress tests, covering their design, implementation, and outputs.

Tony offered his “review of the review” on Wednesday, arguing that the UNEP FI doc dodged the key question: have climate stress tests achieved what they set out to do?  My take is a little different.  I’d argue that the review actually has resolved this question, while posing another – far thornier – one. Namely, are current approaches to climate stress testing fit for purpose, or do we need to try new tools?

Allow me to unpack the two parts of this take one by one.  First, on climate stress tests achieving their goals.  What I found interesting in the review was the assertion that stress tests have “primarily focused on raising awareness of climate risks and fostering climate action within financial institutions.”  If this was indeed the overarching objective, I’d say “job done.” 

Prior to what was (arguably) the first major climate stress test – the Dutch central bank’s 2018 energy transition risk stress test – I’d wager few lenders had dug deep into climate risk.  Property and casualty insurers may well have, but it was certainly not a mainstream concern across the financial sector writ large.  Fast forward to today, and virtually all the world’s major banks and insurers cite climate risk in their public reporting (some exhaustively), and many have initiated net-zero aligned financing plans.   

Of course, it’s impossible to disentangle causality here.  Around the same time supervisors started ordering stress tests en masse, climate finance groups – notably the Glasgow Financial Alliance for Net Zero (GFANZ) – were gaining momentum.  Was it the stress tests that brought awareness to climate risks, or these groupings?  Or were stress tests a lagging indicator of an awareness of climate risks conjured up by the Paris Agreement, COP26, the Task Force on Climate-related Financial Disclosures, etc cetera?  Perhaps this is all academic. For sure, the rollout of climate stress tests certainly helped achieve this awareness objective. 

The review cites other “key objectives” of climate stress tests, too: assessing potential financial exposure to climate risk under various scenarios, enhancing climate risk assessment capabilities across financial institutions, and understanding how to adjust business models in response to different scenarios making up the big three.  I’d argue the latter two are covered under the “raising awareness” theme.  

As for “assessing potential financial exposure”, while I agree with Tony that more thought should go into researching actual rather than potential impacts, I’d also argue that many stress tests have achieved this particular objective.  The problem is that their findings are not the kind to get the pulse racing.

Just read the case studies cited in the review itself.  The De Nederlandsche Bank’s 2021 flood risk stress test? The review says it found the financial system “is likely to remain resilient to physical risks.”  The Australian Prudential Regulatory Authority’s 2022 Climate Vulnerability Assessment?  Again, the finding was that financial systems are likely to remain resilient to physical risks.  The same was said of the Bank of England’s 2022 exercise.  When it came to the European Central Bank’s 2022 test, the script was different, but not excessively so.  The finding was that “[V]ulnerabilities to physical risks … were determined to be non-negligible in some cases.”  None of these findings suggest potential physical risks pose an almighty challenge to the institutions assessed.   

As for climate transition risks, the ECB exercise found the short-term costs of a rush to net-zero would “be overshadowed by costs of climate change without stringent action”, and that long-term transition risks would depend on the transition plans of financial institutions’ counterparties.  Again, there was no indication from any of the case studies in the review that climate risks would unleash the next global financial crisis.

It’s fair to say, then, that potential financial exposures to climate risks have been successfully assessed via these tests – and found to be, well, little to shout about.

Let me be clear – that’s ok!  It’s not inherently “bad” that climate risks aren’t an existential threat to the financial system.  Indeed, as Tony has said before, it’s actually a positive thing, as it suggests banks can actually afford to take on more risks to advance the climate transition – for example, by investing in lots of risky cleantech startups.

However, here’s where we get to the second part of my take – the question of whether current climate stress testing tools are fit for purpose. Are they actually capable of assessing the potential impacts of climate change on financial institutions in the first place?

To be clear, this isn’t a question directly posed by the UNEP FI authors.  But the answer is there in the subtext – a resounding “no.” The review states that “the climate scenarios that are currently used do not reflect extreme, tail-risk events” and that the go-to modeling approach “involves combining different models that were not designed to be used together.” In addition, financial supervisors have used 30-year time horizons for their exercises, a choice that  introduces “significant uncertainties” in the outputs. Moreover, a “static” balance sheet approach to stress testing has been “more widely adopted” by supervisors – one which Tony has rightly lambasted as being completely unrealistic and unhelpful. After all, it assumes that financial institutions don’t change their portfolios at all in response to a climate-transformed economy.  

All these findings imply that mainstream stress testing approaches simply aren’t up to scratch.  

The review indirectly suggests some ways to improve matters.  For example, it explains that the two most widely-used climate scenario models, those that underpin popular scenarios generated by the Network for Greening the Financial System, are both “general equilibrium models.” These are based on the idea that all markets in an economy simultaneously reach a state of balance where supply equals demand.  Useful for sketching out the overall functioning of the economy, for sure, but idealistic and incapable of capturing real-world complexities.

Perhaps what we need are more non-equilibrium models in climate stress testing.  Unlike their staid counterparts, these models are designed to assess how the economy may react to the sort of shocks that climate change could bring about – from a sudden carbon price increase to a blitz of extreme weather events. They may be better placed to capture potential impacts to financial institutions that would be missed by their general equilibrium cousins. 

Of course, these models have their downsides, too.  For one, they are computationally intensive.  For another, their outputs may be all over the map, making it difficult for users to determine signal from noise.  Still, their absence from the mainstream of climate stress testing has led to noisy complaints from some sections of the climate risk community, notably the Real World Climate Scenarios folks and the University of Exeter.  The latter produced the “No Time To Lose” report last year, featuring “decision-useful scenarios” that reflected the volatility, uncertainty, complexity, and ambiguity of the real world. (Disclaimer: yours truly contributed to the scenarios and consulted with the University of Exeter team earlier this year).

Perhaps the UNEP FI review will prompt greater engagement by the climate risk community with these alternative approaches.

One other facet of the review suggests a different way forward. The authors explain that some supervisors have bolted on qualitative modules to their stress testing exercises, typically in the form of questionnaires directed at the tests’ participants.  These have been used to better understand the challenges institutions face changing their business models in light of climate risks, and to gain information on their operational and reputational exposures to physical and transition shocks.

It’s a low-tech means of gathering information from banks and insurers, no doubt, but as a way to raise climate risk awareness across the financial sector and to force firms to focus attention on how their businesses are set up to deal with a hotter world, I don’t see why it’s any better or worse than subjecting them to flawed stress tests.

Perhaps, then, financial supervisors’ time would be more fruitfully spent crafting better questionnaires, capable of producing more informationally dense outputs.  Maybe these could be combined with narrative climate scenarios – hypotheticals that senior bankers could puzzle over without having to go ten rounds with a general equilibrium model.  Given that current stress testing approaches are arguably failing to produce useful outputs, why not save time, money, and hassle by going the low-tech route?

Ultimately, climate stress tests are simply stories that supervisors tell financial institutions; stories they ask institutions to help write the endings.  If the current manner of finishing these stories is inadequate, we should all be willing to show a little more imagination.