An overview of recent climate stress exercises in Europe

Introduction

After several exercises carried out by European regulators, and a displayed intention by the Federal Reserve to tackle the climate stress test of the financial system, it is now established that this subject will be part of the risk studies of banking institutions worldwide, and this for several years.

In broad terms, a climate stress test is always expressed as follows.

1. The anticipation of climate change is carried out by researchers, the results of this research being compiled and published by the IPCC. This work determines the physical risk that the global warming may pose to the world.
2. The anticipation of policy responses to this change, to reduce or contain its effects, is also the subject of research, defining response scenarios. This work determines the risk that the transition to a low carbon economy poses to the world.
3. The combination of a climate change scenario and a response scenario impacts the economy in different ways by region and by sector. This impact is assessed in a more or less precise way by macroeconomic models by identifying the "transmission channels" from the climate to the economy.
4. The financial sector can then use all these elements to assess its resistance to these different scenarios. Even if the research is generally already done outside the financial sector, it is important that regulators and banks understand it well, in particular to assess the uncertainty of anticipations.

This article summarizes the recent exercises carried out in Europe.

The Dutch Central Bank 2018 exercise

In 2018, the Dutch Central Bank DNB launched one of the first exercises “An Energy Transition Risk Stress Test for the Financial System of the Netherlands”.

Interestingly, although the Netherlands are particularly exposed to the anticipated sea level rise, the stress test was focusing only on transition: “The economic losses are brought about by policy measures, technological breakthroughs, or a drop in consumer and investor confidence.”
The Dutch exercise considers 4 scenarios, characterized by 2 variables: is there a technological breakthrough allowing the use of renewable energy? and do the policy makers impose an increase on the carbon price of $100 per ton?

Fig 1 – The transition scenarios according to DNB stress-test

Transition vulnerability factors per industry are calculated by DNB based on an assessment of the embodied emissions of the final goods and services in each industry. The study was conducted by DNB itself, considering its knowledge of the portfolio allocation of Dutch institutions.
The impacts were low as the loss of assets were estimated between 1% and 3% for banks.

Fig 2 – Impact on assets as a percentage of total stressed assets per sector, disaggregated by risk driver

The French Central Bank 2020 exercise

In July 2020, the French Central Bank (Autorité de Contrôle Prudentiel et de Régulation) launched the first stress tests that needed to be carried out by banks and insurers themselves, based on assumptions provided by the regulator: ACPR Pilot Climate Exercise.

As for DNB, the focus was on transition: “One of the objectives of the pilot exercise is to measure, over a long-term time horizon, the consequences of scenarios materialising the transition risk on banking balance sheets.”

At the horizon considered (2050), the physical risks are considered to be incorporated in all scenarios, and are not differentiated. In reality, the impact of physical risks is limited to insurers: “we consider the impact of the physical risk on the frequency of and costs associated with extreme climate events on insurance companies, in particular floods, droughts, sea submersions and cyclone storms (for the overseas departments).”

Three scenarios considered in addition to the Business As Usual (BAU) scenario: orderly transition, delayed transition, sudden transition.

It is interesting to note that since the reference scenario is BAU, in which no transition action is taken, and since the physical risks are assumed to be the same in all scenarios, the only measurable cost for the institutions is therefore the transition cost, without any trade-off with the physical risk. The design of these scenarios therefore measures only the risk of action, not the risk of inaction, which delivers an almost climate-sceptic message.

In this scenario, the French Central Bank provided the estimated impacts of transition on economic sectors, as did the DNB in its first exercise.

Most of the large French banks (BNPP, BPCE, Crédit Agricole, SocGen), and insurers (ACM, AXA, CCR, CNP) took part in the exercise.

The results are shared in a separate document released in May 2021.

On the credit risk side, the results show that the cost of credit risk will increase from 12.4 basis points in 2020, to up to 17.2 basis points in 2050 in the sudden scenario. The difference between the orderly scenario and the sudden scenario is only 1.4 basis points, i.e., less than 10%.

Fig 3 – Evolution of the cost of credit risk for the main French banks according to ACPR scenarios

The Bank of England 2021 exercise

In June 2021, the Bank of England published its 2021 Biennial Exploratory Scenario: Financial risks from climate change (designated as CBES).

This exercise was again considering 3 scenarios, focused on transition: no additional action, early action, late action. This is similar to the French Central Bank scenario. However, it is interesting to note that the CBES exercise differentiates the physical risk associated to each of these scenarios.

Fig 4 – Example of the differentiation of the physical risks in the CBES exercise: annual average precipitation

The European Central Bank 2021 exercise

In September 2021, the European Central Bank published the results of a stress test exercise carried out internally, i.e. without collecting data from the banks: ECB economy-wide climate stress test.

The scenarios are again quite similar to the ones used in the regulators’ exercises: hot house scenario (no additional action), orderly and disorderly transition.

The very interesting feature of this exercise is the explicit integration of physical risk: “The calculation of expected losses from physical risk combines direct impact on firms’ exposure to extreme weather events with indirect impact, such as the expected damage at the regional level as a share of GDP”.

As far as credit risk is concerned, the impact shown is consistent with the results provided by the French Central Bank:

• Increase of PD by 7% (comparison of Hot House vs Orderly)
• Increase of LGD by 1% (comparison of Hot House vs Orderly)
• Increase of Expected Losses by 8.5% (comparison of Hot House vs Orderly)

Fig 5 – Distribution of Expected Losses by 2050 – according to ECB stress test

Comments on scenarios and results

With the exception of the DNB initial exercise, most of the scenarios included in the regulators’ exercises are derived from the NGFS (Network for Greening the Financial System) scenarios. The NGFS is a group of more than 100 regulators including the US Federal Reserve.

Fig 6 – The NGFS scenarios

The scenarios generally include 2 types of impacts on the financial system:

• An impact of climate stress on the general economy,
• A differentiated impact on specific sectors

This allows financial institutions to assess the stress on their portfolio, based on their exposure to sensitive sectors.

In the 2020 version of the NGFS scenarios, the impact on the general economy was based on a simple assumption of a damage to the economy growing as the square of the change of temperature. This assumption is based on Nordhaus initial findings, which are not considered as very robust. The 2021 version of the scenarios has been improved.=

Fig 7 – Damage functions according to the 2020 NGFS scenarios (page 23 of the technical documentation)

Generally speaking, the observed impact on the banks is low. If we consider that the impact on credit risk at the horizon 2050 is an increase of around 10%, as for instance the French Central Bank and the European Central Bank assess it, it is interesting to compare it to the observed evolution of the probability of default for corporate borrowers as reported in the S&P 2020 Annual Global Corporate Default And Rating Transition Study.

Fig 8 – Evolution of corporate PD (probability of default) between 1981 and 2020

There is an apparent trend in the change of default rate, which would represent an increase of 18% in the last 30 years. This trend is not very significant, but of course this shows that the order of magnitude of the climate stress, as reported by the present models does not seem realistic. One point we have to keep in mind is that the climate stress test exercises are conducted with a “closed book assumption”. The assumption is that the climate stress is assumed to be observed instantaneously rather than over time. And yet, no shock is observable. Compare this for instance to the change of PD during the 2008 crisis, where the PD has been multiplied by 10 in 2 years (from 0.37% in 2007 to 4.19% in 2009), or to the Covid crisis, where the PD has been multiplied by 2 between 2019 and 2020.

Conclusion

There is an apparent mismatch between the anticipated severity of climate change, and the assessed impact on the economy and on the financial sector. A recent paper from the New York Federal Reserve even seems to indicate that natural “disasters over the last quarter century had insignificant or small effects on U.S. banks’ performance.”

To put it simply:

• From 1970 to 1999, the mean global temperature indeed increased at 0.17°C per decade, rising to 0.25°C per decade after 2000, which represents an increase of approximately 1°C in the last fifty years. And this had no measurable impact – and even a positive one, according to the NY Fed paper.
• The anticipated 4°C increase, including increased physical risk and transition, will impact the banks by an increase of the cost of risk around 10%, according to the stress test exercises recently published, which is barely noticeable compared to recent financial crises.
• But a 10°C increase would probably make life on earth very challenging, if only possible.

From this, it appears clearly that the impact of global warming on the economy is highly nonlinear. In addition, these impacts will generate shocks.  The current generation of models is certainly missing these points, which are not easy to capture.