NGFS climate scenarios underestimate the impact of climate change


The Network for the Greening of the Financial Sector (NGFS), the ‘green club’ of central banks, has recently published the latest version of its climate scenarios. These are intended to be used in the financial sector to model the impact of climate change, as well as mitigation efforts to limit its worst effects. However, at the current state of these scenarios, there is a real danger that they give financial institutions a false sense of comfort, leading to too little climate mitigation action. At the Sustainable Finance Lab, we recently responded to the NGFS’ call for feedback on the last vintage of the scenarios. This blog summarizes our feedback and suggestions for improvements.

A good start

NGFS climate scenarios are a tool that enables a common starting point for understanding and integrating climate risks across the financial sector. They can be used to conduct scenario analyses to better understand various climate mitigation pathways and key economic and financial variables.

In its latest vintage, NGFS offers three clusters of six scenarios in total.[1] These clusters and scenarios differ in parameters such as temperature rise mitigation ambition, severity of the policy response, and optimism regarding the technological progress. For instance, the most ambitious Orderly Net Zero 2050 scenario assumes a temperature increase of 1.4oC above pre-industrial levels, quick policy reaction, and high level of technological innovation, including the use of CO2 removal. In contrast, the most pessimistic Current Policies scenario projects temperature increase above 3oC, with no significant policy reaction and slow technological response.

The added value of NGFS’ work is enabling standardization and comparison of results across the financial sector. This effort largely attempts to prevent a multitude of different tools used by different institutions and reduce the chance of greenwashing. While these scenarios offer a laudable first step, they still have room for improvement.

Dubious assumptions, questionable choices

NGFS scenarios are generated using Integrated Assessment Models (IAMs). This is a family of models that integrate natural processes with human economic activities and model their interactions. In the NGFS work, IAMs are used to make projections about greenhouse gas emissions pathways based on economic processes, and the corresponding temperature increases. A damage function – relationship between temperature increase and GDP loss – is applied to estimate the extent of the physical risk. Transition risk is proxied by modelling the dynamics of carbon price introduction and its evolution over time.

Each of these steps come with their own problematic assumptions. Firstly, IAMs belong to the family of neoclassical, equilibrium models. This means that their starting assumptions include having a representative, utility-maximizing rational actor, with perfect foresight and markets that always clear. These features have been heavily criticized in the literature as unrealistic and oversimplifying. Secondly, these models cannot accurately represent dis-equilibrium state, which is what large-scale climate change could cause. Thirdly, none of these models include tipping points. These are the so-called ‘points of no return’ that geoscientists believe we might encounter if the temperature of the earth keeps increasing. Lastly, IAMs do not model the financial sector explicitly. This last point is striking, given the NGFS’s target audience for their scenarios.

Aside from problematic assumptions of IAMs, the NGFS’s choice of the damage function is not without its issues. The Technical documentation chooses the damage function of Kalkuhl and Wenz due to its completeness in accounting for past damages. However, the predicted outputs of this function are rather low, showing less than 20% of GDP loss for a temperature increase above 4oC. This finding is not aligned with what we know about climate science and the realistic impact such temperature increase might have. For example, the last time the world was 3oC hotter than in the pre-industrial period was three million years ago, when sea levels were 10-20 meters higher than today. At the same time, by the year 2050 one billion people could live on coastlines at elevations of 10 meters. It is thus hard to believe that the magnitude of damages caused by sea level rise and flooding, including migrations and resulting broader social instability, could be equivalent to relatively small damages predicted by the damage function chosen by the NGFS.

A more general issue with the methodology used by Kalkuhl and Wenz is trying to infer conclusions about the future effects of climate change based on the past data. One of the reasons for this are the aforementioned tipping points and a large-scale uncertainty about the future states of the world given large temperature increases.

False sense of security

Money is the lifeblood of the economy, and it is the role of the financial sector to steer the money flows towards the most productive and sustainable parts of the economy. Having an inaccurate set of data to work with gives the financial actors a skewed basis for action. For instance, as a result of the last vintage of NGFS scenarios, predicted damages of climate change account for only between 15% and 20% of GDP loss in the year 2100. This number is almost certainly a lower bound and might be difficult to reconcile with everything we know about climate change. In addition, it might signal to the financial sector a lack of urgency and give room for delayed action. This is difficult to justify, given the key role of the financial sector in the green transition.

The NGFS’s feedback process is thus crucial to underscore main issues with their scenarios and offer better alternatives. Firstly, different models should be used to complement or replace IAMs. Various non-equilibrium models, such as Stock-Flow Consistent and Agent-Based models, could more accurately represent the economy and the financial sector. Secondly, more realistic damage functions, more aligned with climate science, should be used to give a better idea of potential damages. Thirdly, more climate risks, such as tipping points, should be included in scenario design. This is also true of social and political risks: wars and migrations caused by climate change could cause large-scale social dislocations.

Mitigating climate change and speeding up the green transition require action of all societal groups. The financial sector has one of the key roles to play here. The work done by central banks to help inform them on the dangers of climate change is a good beginning, but should be steered in a more realistic, sober direction. The future of the rest of us might depend on them.


[1] The scenarios are: Net Zero 2050, Below 2°C, Divergent Net Zero, Delayed Transition, Nationally
Determined Contributions (NDCs), Current Policies, clustered in groups Orderly, Disorderly and Hothouse world, two each and respectively.