Financial Institutions May Be Lulled Into Complacency By Climate Stress Test Results

The Institute & Faculty of Actuaries (IFoA), a UK professional body involved in training actuaries, together with the University of Exeter has released a highly critical report on the use and misuse of climate scenario analysis. This analysis has grown in popularity with the launch and expansion of the Network for Greening the Financial Sector (NGFS). This is a body composed of central banks, regulators and supervisors from around the world working to share knowledge relating to managing climate-related financial risk.

One of the main successes of the NGFS has been to encourage its members’ development of climate stress testing. This involves a regulator selecting climate scenarios and asking banks and insurers to analyze their own portfolios and how they would be impacted by each scenario. Some central banks include a second-round review that allows for some dynamic response from financial institutions to the aggregate financial sector response to the implications of the chosen scenario.

Most climate stress tests done at the regulatory level do not impact the capital requirements of banks, but are learning exercises for banks to understand their climate risks, and for supervisors to evaluate the readiness for dealing with climate-related stress. From the regulators’ perspective, the value of the climate stress tests is both microprudential (dealing with individual institutions’ vulnerabilities) or macroprudential (dealing with systemic risk vulnerability for the financial sector as a whole).

The report does not criticize the value of doing the stress tests per se. It is much more focused on the “risk of group think, with scenario analysis outcomes being taken too literally and out of context”. The report is clear that what it is concerned about relates to all forms of climate risk, although it delves primarily into problems with the economic implications of a ‘hot house’ world of 3°C global warming, which would occur if transition is not successful (and could lead to much higher temperature change if tipping points are triggered).

The primary issue raised with the economic models, or the ‘hot house’ world, is that the impacts in some cases – it uses several anonymized real-world TCFD reports – assume a limited difference in economic impact between an orderly transition and a failure to do so that leads to a ‘hot house’ scenario. The issue that arises with these types of assumptions being incorporated into models is that the economic implications are at odds with the stark physical impacts that a ‘hot house’ world would imply.

Underpinning the economic models in use today, where financial institutions often rely on outsourced modeling firms, are three elements. First, there is an assumption about the remaining carbon budget that is often presented with limited recognition of the significant uncertainty around that figure.

At the current pace of global emissions, the remaining carbon budget consistent with a 1.5°C increase in global temperatures would last just eight years. The IFoA notes that climate impacts in recent years and faster-than-anticipated global warming may mean a shorter or zero remaining emissions budget. It says the uncertainty is often not highlighted as a limitation of the models being used.

Second, there is an assumption made about how rapidly the GHG emissions in the atmosphere will translate to increases in global average temperatures. Again, the experience of recent years, including the record land and sea temperatures of recent days, could indicate that the models assume a longer lag and give more time for climate mitigation than we actually have, leading to a degree of complacency that is reinforced by not considering the limitations of any model used.

Finally, the models employed relate the path of future emissions and warming to the economic impacts of acute and chronic physical impacts of the climate change that would result, and the impact of changes necessary for the energy transition of each model. This modeling often uses the same types of macroeconomic models used in policymaking over a much shorter time-frame. These models have to make significant simplifying assumptions, which will add considerably greater error over the time frames involved in climate change than they already do for shorter-term policymaking purposes.

The steps that go into producing economic impacts of climate change are not the only hazard identified. Most economic models used in scenario design are not able to include many of the most severe impacts likely from climate change, including tipping points and the resulting societal disruption, which would certainly have economic consequences. The biggest risk comes not only with the models themselves, but how they are applied and the degree to which their shortfalls, judgements and assumptions as well as limitations are properly understood.

The IFoA summarizes its conclusions:

“Time is too short to wait for models that are perfect. Development is needed, including looking beyond the commonly used general equilibrium economic models that underpin many approaches today, to find solutions that can realistically capture risk drivers and the interaction between policy, technology, the real economy and markets. A practical fix is to use qualitative scenarios that reflect the emerging reality of climate change to complement models, as well as out-of-model adjustments and margins to reflect uncertainty.”

One of the examples shared by the IFoA comes from forthcoming research by Carbon Tracker that outlines alternative damage functions that financial institutions could use to supplement the regulatory models. These simpler models posit a temperature beyond which humanity could adapt, whether that is 4°, 5° or 6° C, where economic activity will shrink 100% from current levels.

Between the economy of today’s 1.2° C of warming and the temperature rise limit of the economy, they make assumptions of logarithmic curves to generate an alternative input for scenarios. The simpler alternative of a logarithmic function for climate damage may not always be more accurate than econometric models for climate loss, but it adds diversity in the models used to reflect the deep uncertainty in climate risk modeling.

These quantitative models can be complemented with qualitative scenario analysis tracing likely channels through which climate impacts have a cascading impact that would be relevant for a financial institution. At the end of the day, the energy transition and its success or failure and the resulting climate impacts are uncharted territory.

Models are often based on historical data to guide assumptions, help influence judgments, and provide guidance for where a particular model’s limitations may exist. When looking at climate-related risks, “given the challenges of calibrating a model to a complex basket of never-before-experienced risks, users should beware of spurious accuracy – it is better to be roughly right than precisely wrong”.