In this study, researchers use machine learning to simulate the effects of rising emissions on sovereign creditworthiness and produce credible decision-ready green financial indicators. The findings show that climate-induced sovereign downgrades will emerge by the end of the decade, but honouring the Paris agreement would “nearly eliminate the effects”.
According to the authors, the results give “direct and immediate incentive” to invest in climate-smart investments as a precautionary strategy that will reduce debt servicing costs.
There is growing demand for ESG disclosures yet, authors say, climate science has failed to translate macroeconomic costs into credible metrics for conveying risks to financial decision makers. They highlight the need for metrics that are scientifically credible and real-world applicable.
The paper’s authors – Patrycja Klusak, Matthew Agarwala, Matt Burke, Moritz Kraemer, Kamiar Mohaddes – produce scientifically rigorous, climate-adjusted sovereign credit ratings and estimate the impact of this on the costs of corporate and sovereign debt.
The study explores the impact of physical climate risk on 109 countries, covering a wide range of income, growth and other economic variables. The results show the greatest impact on credit ratings are seen in the highest-rated economies.
Under the representative concentration pathway (RCP) 8.5 emissions scenario (or “business as usual”), climate-induced credit downgrades are expected for 81 nations by an average 2.18 notches by 2100. In contrast, under an RCP 2.6 scenario in which stringent climate policies are observed, 62 countries are likely to experience downgrades by an average 0.94 notches.
The authors evaluate the impact of these downgrades by considering two key indicators: corporate and sovereign debt costs. The findings show that climate-induced downgrades under an RCP 8.5 scenario result in US$135-203bn additional annual sovereign borrowing costs by 2100. Whereas under the 2.6 scenario, additional debt costs are approximately $44-67bn.
Additionally, in the RCP 8.5 scenario, the magnitude of corporate interest outlays is estimated between $35-66bn, while under RCP 2.6 additional costs range between $10-17bn.
As part of their research, the authors trained artificial intelligence models on S&P Global ratings from 2015-2020, combining this with economic modelling and S&P natural disaster risk assessment data. They used S&P data to stay as close as possible to credit rating practitioner approaches and to optimise predictive accuracy.
A significant limitation of the study is that it does not include climate-related political instability or transition and litigation risk. As the authors note, however, the model can be “readily extended” to include these when sufficiently robust quantitative estimates are available.
This page was last updated August 31, 2023
Share this article