Bringing climate scenarios to central bankers (and beyond)
For many years there has been a gulf between central banking and climate finance but today central bankers are important players in the fight against climate change. The Network for Greening the Financial System (NGFS) played a pivotal role in this shift. Giacomo Bressan and Maria Nieto argue that it is now time to push one step further: to make the NGFS long-term scenarios usable in the day-to-day operations of financial institutions.
What are the NGFS long-term climate scenarios?
Founded in 2017, the NGFS is “a group of Central Banks and Supervisors willing […] to […] contribute to the development of environment and climate risk management in the financial sector and to mobilize mainstream finance to support the transition toward a sustainable economy”. Climate scenarios are probably the most well-known NGFS initiative. Its long-term scenarios provide a window into different plausible futures for physical climate and transition risks, distinguishing between an ‘orderly’ and ‘disorderly’ transition, a ‘hot house world’, and ‘too little, too late’. Multiple models and their outputs are combined into a suite of models approach, to translate the scenario narratives into physical and transition risks on the economy.
Transition risk modelling is supported by three processed-based Integrated Assessment Models (PB-IAMs) developed by an academic consortium: the Global Change Analysis Model GCAM 6.0, MESSAGEix-GLOBIOM 1.1-M-R12, and REMIND-MAgPIE 3.2-4.6. These models differ in terms of their behaviour, proposed solutions and how they treat technological change. They are used to derive impacts of policy decisions on systems such as the energy or agricultural sectors. Then, the MAGICC model is used to translate greenhouse gas emissions into temperature pathways.
The approach to physical risks distinguishes between chronic and acute risks. The former covers average annual temperature, daily temperature variability, total annual precipitation, number of wet days and extreme daily rainfall. Acute risks include droughts, heatwaves, floods and cyclones. Lastly, the climate module of the macroeconomic model NiGEM is used to understand the implications of physical and transition risks for macroeconomic fundamental variables.
Scenarios from the International Energy Agency (IEA) and the Intergovernmental Panel on Climate Change (IPCC) are also often used by financial institutions. There are notable differences between the scenarios. For example, the NGFS has more scenarios, models, variables and sectors represented than the IEA. Ultimately, choosing between them should be based on purpose: whether this is for climate projection (IPCC), energy systems (IEA), or macroeconomic/financial risk analysis (NGFS).
The NGFS has meaningfully contributed to mainstreaming climate risk analysis. However, its current approach has limitations beyond those of PB-IAMS that are well documented in the literature. So where does room for improvement lie?
Improving the usability of NGFS data
Standardising transition risk analysis
A key issue is the starting point: calibration. We focus on two main greenhouse gases, namely CH4 (methane) and CO2 (carbon dioxide), and compare the NGFS results for the year 2020 with historical data from ClimateWatch. We observe meaningful deviations between the NGFS data and observed emissions, for all regions, which is relevant because PB-IAMs account for the carbon budget for each temperature objective. As shown in Figure 1, this calibration issue is widespread.
Regional comparison across models is impaired because different models have different regional structures. For example, in one of the PB-IAMs (MESSAGEix-GLOBIOM), results are available for the world aggregate and 12 regions. Country-level results are downscaled. Another PB-IAM (GCAM) uses 32 regions, and a third PB-IAM (REMIND-MAgPIE) uses 12 regions, which are different from those used in MESSAGEix-GLOBIOM. In the NiGEM climate module, many countries are modelled directly and yet differ from the downscaled ones.
This serves to weaken data availability. Practical use of the scenarios requires consistent availability of variables at the country level, which is not currently guaranteed. For example, CH4 emissions are available only from REMIND-MAgPIE for the EU28, available for all models in Latin America, and missing for REMIND-MAgPIE in South Asia. This greatly harms usability, as it reduces the possibility for uncertainty analysis and impairs regional comparability.
Overall, a question of cross-model comparability emerges. Beyond the inconsistencies we have highlighted, model assumptions also differ meaningfully and accumulation of apparently minor issues leads to counterintuitive results. First, there is the occurrence of values out of range with respect to model range variation. Second, there are extremely large and potentially uninformative uncertainty ranges across models (e.g. for the estimate of the carbon price where a range of more than 100% can be observed). Third, the lack of uncertainty ranges due to lack of variables availability across models.
Figure 1. Deviations in emissions data for CO2 and CH4 between NGFS (2023) and Climate Data Watch (CAITS)
Source: Authors’ analysis based on NGFS (2023) and CAITS
Harmonising physical risk analysis
Chronic risk analysis misses a key variable, namely sea-level rise, which matters in terms of direct climate damages, adaptation costs, and non-GDP negative effects such as land loss and migration. Furthermore, the calibration window for the relationship between temperature, precipitation and GDP is the period 1979–2019, assuming a stationarity that is contested in the literature due to the presence of non-linearities and tipping points, the inadequacy of historical data, and cascading and compound risks. Transmission channels from temperature to GDP include only productivity losses and capital depreciation, missing other effects such as increased mortality, migration and agricultural yields. Ultimately, several approximations are combined to reach the chronic risk estimates. For example, the calibration of the temperature–GDP relationship is performed at the subnational level, while country-level temperature pathways are downscaled. Lastly, recent research has highlighted data anomalies in the quantification of chronic damages, which may have led to their overestimation.
Acute-risk modelling employs multiple approaches, which also greatly impacts the results. Modelling of flood risk focuses on average losses and neglects tail risks. This, combined with other limitations such as mixed spatial resolutions, neglect of coastal and pluvial floods, and calibration issues (e.g. the usage of 2005 data for exposure), largely explains the limited impact of floods on global GDP, according to the NGFS.
Droughts and heatwaves follow a different approach. Drought impacts on agricultural crop yields are considered at a 50x50km scale. This neglects non-agricultural areas, cities and cascading effects.
Overall, inconsistencies accumulate for acute risks too, for example, differences in the reference Representative Concentration Pathway (RCP) scenario (e.g. using RCP 8.5 for heatwaves, but 4.5 and 6.0 for cyclones) and exposure data (e.g. using 2005 population for heatwaves, but LitPop for cyclones).
Conclusions
The work of the NGFS has strengthened understanding of climate risks in finance. However, as our research shows, the usability of its long-term scenarios is still hampered by limitations in the modelling approach, inconsistencies in the results and a lack of regional comparability. The areas for improvements we have highlighted boil down to, first and foremost, standardisation of assumptions, including regions, exposure and scenarios; and second, better calibration, including using more recent data and incorporating more hazards. These actions would make the NGFS scenarios more fit for purpose in the day-to-day decision-making of central banks and financial institutions.
Giacomo Bressan is an Associate Director at Morningstar Analytics. Maria Nieto is a Senior Visiting Fellow at CETEx.
The views in this commentary are those of the authors and do not represent the views of Morningstar Sustainalytics, CETEx senior management or its funders. Any errors or omissions remain those of the authors.
Photo: Yves Bernardi, Pixabay