Climate stress testing

Pushing the modelling frontier

The COVID-crisis and climate risk are instructive of a broader challenge: many emerging risks require an understanding of the linkages between the financial system, the real economy, and the broader environment in which society operates. Accordingly, system-wide stress tests need to be extended so that rather than considering the financial system in isolation, they consider the real economy as well as the environment. This will require a much more ambitious conception of what a ‘system-wide stress test’ looks like. We need to understand, and these models should capture, the interlinkages among climate, economic, financial, and financial stability risks. Different economies, geographies, sectors, and even firms are likely to face distinct risks from climate change, with each taking actions that affect the others.

A non-equilibrium model of the green energy transition

Stress Testing the Financial Macrocosm
Authors: J. Doyne Farmer, Alissa M. Kleinnijenhuis, Thom Wetzer
From: University of Oxford, Stanford University

A new stress testing methodology

In this paper, Jung et al. develop a climate stress testing methodology to test the resilience of the financial system. They construct a measure called CRISK, which indicates the expected capital shortfall of a financial institution. Their stress testing methodology involves three steps:

• The measurement of the climate risk factor using stranded asset portfolio return as a proxy measure for transition risk.
• The estimation of a time-varying climate beta of financial institutions using the Dynamic Conditional Beta (DCB) model, an approach to estimating regressions with time varying parameters.
• The computation of CRISK, which is a function of a given financial firm’s size, leverage, and expected equity loss conditional on climate stress.

We apply the methodology to measure the climate risk of 27 large global banks, whose aggregate oil and gas loan market share exceeds 80%. The stress scenario that we consider is a 50% drop in the return on stranded asset portfolio over six months. This corresponds to the first percentile of historical return on stranded asset portfolio. We find that, first, climate beta varies over time, highlighting the importance of dynamic estimation. Second, climate betas of banks move together over time, and there was a common spike in climate betas as well as in CRISKs when energy prices collapsed in 2020. The measured CRISKs for some of the banks were economically substantial. For instance, Citigroup’s CRISK increased by 73 billion US dollars during the year 2020. In other words, the expected amount of capital that Citigroup would need to raise under the climate stress scenario to restore a prudential capital ratio increased by 73 billion US dollars in 2020.

Climate Stress Testing
Authors: Hyeyoon Jung, Robert Engle, Richard Berner
From: Federal Reserve Bank of New York, New York University

A comparison of three recent climate stress tests

This is an interesting review of the different methodologies used in climate stress tests of banks conducted by the Dutch and French authorities in 2018 and 2021 respectively, and in the exercise currently underway in the United Kingdom. In particular, Baudino and Svoronos examine how these three stress tests dealt with: i) data availability and reliability; (ii) the adoption of very long time horizons; (iii) uncertainty around future pathways of key reference variables covering physical risks (eg floods, temperature increases and rising sea levels); and (iv) uncertainty relating to transition risks.
Stress-testing banks for climate change – a comparison of practices
Authors: Patrizia Baudino, Jean-Philippe Svoronos
From: BIS