Over decades, the discussions about climate change have been topics of both wide debate and scientific research. Questions have ranged from “What is climate change?” to “How can climate change be any different from natural climate variations?” to “Are we heading to a climate apocalypse?” The accumulated scientific research found that globally, atmospheric temperatures are rising above natural variations with potentially devastating consequences to people, communities, and impacting financial stability.
The tragedy of the horizon, as Mark Carney put it in his ground-breaking speech in 2015,1 is also a tragedy of inequality where countries who are the bigger greenhouse gas emitters are also likely more resilient to the physical and transition risks of climate change.
Many industries have taken steps to better understand, quantify, and report on climate-related risks, and to devise strategies to mitigate the impacts of climate change. It is important to note that these initiatives are in their infancy. In the financial services industry, central banks, and regulators are raising awareness about the financial risks and opportunities of climate change. Recently, in 2021, more than 450 financial institutions representing $130 trillion of total assets committed to net-zero targets by 2050, as part of the Glasgow Financial Alliance for Net Zero (GFANZ).
It highlighted that the financial services industry has a key role to play in tackling climate change, in:
Many central banks, regulators, and industry bodies have published guidelines and policies to address financial risks posed by climate change. While we do observe convergence in some areas of guidance across jurisdictions—for example, the use of forward-looking scenario analysis and standardized reporting—for the Board and the risk management function, a profusion of divergent regulatory frameworks, guidance, and standards exist. What is becoming clearer is that financial institutions are focusing on a broad range of topics covering climate-related calculations, controls, conduct, communication, culture, and their customers. These considerations include:
As we find ourselves in the decisive decade in transitioning to a lower carbon economy, there is an urgency for financial institutions to help manage both the financial and nonfinancial risks. In the next section, we will be discussing more details.
Firstly, it is important to understand the main issues, the science, and the data before discussing possible solutions. In the financial services industry, as in other industries, climate risk is becoming mainstream in corporate decision-making: the magnitude of climate instability will impact most aspects of financial decision-making. To assess and address climate change, a scientific approach can be borrowed—where a scientific mindset is used to investigate historic data and observational studies, build questions, and then formulate a hypothesis. However, this approach has its own limitations, as there is no historical precedent for climate change.
Discussions on climate change, and whether it is indeed an emergency or real, are not new and have been around for decades. However, at the time of writing, 97% of climate scientists agree that climate change exists.3 Research on the impacts of climate change shifted to forward-looking scenarios based on projected emissions grouped as greenhouse gases. Greenhouse gas emissions are a byproduct of both the extraction and burning of fossil fuels. It has an impact on the climate, that is, increasing the atmospheric temperature, but it also impacts human health.4
More specifically, polluting facilities that process and distribute fossil fuels increase the levels of greenhouse gases in the atmosphere, leading to a warming effect. The main culprits are CO2 and other pollutants such as methane, but also include nitrous oxide and fluorinated gases.5 These facilities create 80% of CO2 and 30% of methane pollution.6 Accumulated evidence has suggested that other contributors to climate change are human activities such as natural gas drilling, transportation, farming, deforestation, and fertilizers.7
Research indicates that human activities contribute to the association between greenhouse gas emissions and the average rise in the planet's temperature observed in the last century.8 More dramatically, in recent years, since 1975, the planet's temperature has risen by 2.3 to 2.7 degrees Fahrenheit or 1.5 to 1.8 degrees Celsius on average.9
However, climate change is more than global warming: it is defined as “a significant variation of average weather conditions”: becoming warmer, wetter, or drier—over decades or more.10 There are also natural causes of climate variation such as volcanic eruptions, variation in solar radiation, the movement of crustal plates, and oscillation of the ocean and atmospheric system under El Nino–Southern Oscillation (ENSO),11 however, it is anticipated that the frequency and severity of variations in weather patterns are increasing.
In this case, the longer-term trend differentiates climate change from natural weather variability.
Other impacts of human activity on the climate range from:
We need to appreciate that human civilization flourishes in a stable climate and that the planet is potentially accelerating toward instability based on the increased frequency and severity of extreme weather events. The socioeconomic impacts and hazards associated with extreme weather events are dramatic in terms of loss of human life, exemplified by the 2010 Russian heatwave that killed 55,000 people. Monetary losses of Hurricane Harvey equated to $125 billion (about $380 per person in the United States lost in 2017). As mentioned, climate change has worsened the hazards of extreme weather events; for instance, the Russian heatwave was estimated to be three times more likely due to global climate change.13
A global group of 11,000 scientists have endorsed 40 years of research on a range of measures that supports the hypothesis that the world is now facing a climate emergency; the research suggests that the following scenarios are plausible:14
Since 1992, the United Nations has been recognizing that changes to global climate patterns, and those at regional levels, pose serious issues to the world.15 The recognition has created notable accords in negotiation with participating countries. Table 9.1 provides a summarized list of the talks and accords to date.
Scientific evidence on climate change and its impending impacts over the past 28 years has directed the accords. The Paris Agreement is no exception.
Table 9.1 History of United Nations Climate Accords
Year | United Nations talk and associated accords |
---|---|
1992 | UN Framework Convention on Climate Change (UNFCCC) |
1995 | Berlin Mandate |
1997 | Kyoto Protocol |
2001 | Negotiations on the Kyoto Protocol that focused on emissions trading and how to account for carbon sinks in Bonn, Germany |
2005 | Kyoto Protocol takes effect excluding the United States |
2007 | Negotiations begin on Kyoto Protocol 2.0 |
2009 | Copenhagen Accord |
2010 | Cancun Agreements |
2011 | Draft of a new legally binding agreement to take place in 2015. Extension of Kyoto 2.0 to 2017 |
2013 | Warsaw International Mechanism for loss and damage with climate change impacts. Agreement also on initiative to end deforestation known as REDD+. |
2015 | Paris Agreement |
2018 | Rules for Paris Agreement decided |
2019 | UN Climate Action Summit for countries to submit nationally determined contribution plans as per the Paris Agreement |
2020 | Talks postponed due to the Covid19 pandemic. Emissions fall worldwide due to the pandemic, but reductions are not set to last. |
2021 | COP 26 Summit in Glasgow |
In 2015, the Paris Agreement set a target for the net increase in global temperatures, to be limited to between 1.5 and 2 degrees Celsius. In general, there are two ways for participating countries to reach their agreed emissions targets. The first is to reduce emissions gradually over time, and the second, and more drastic approach, is to eliminate emissions by shutting down polluting facilities. The latter can be achieved by replacing fossil fuel production with renewable alternatives. Before we look at these options, let us look at the scientific reasons behind why emissions targets should be met.
Even a temperature increase of less than 2 degrees Celsius will affect lives and livelihood: it will likely lead to flooding caused by an increase in frequency of extreme rainfall, an increase in sea levels, and an increased extent of wildfires. In addition, to meet the Paris targets, participating countries will need to make and sustain changes in industry sectors and energy usage to reduce emissions by at least 6% annually until 2030.16 It has been reported that a new record in daily average carbon levels in the atmosphere was reached in 2021: it spiked to 421.21 parts per million, for the first time exceeding 420 parts per million. Such an increase, never before recorded in human history, confirmed that, according to sources, the planet's temperature increased by more than 2 degrees Celsius, compared to the Industrial Revolution.17 Keep in mind, even if greenhouse gases were to reduce significantly overnight, the planet's temperature is expected to continue to increase due to the historical accumulation of CO2 in the atmosphere.
Some research studies on climate change are suggesting that the Paris targets are not reasonably achievable, and that a 3 to 4 degrees Celsius warming is a more likely outcome.18
In its Fifth Assessment Report, the Intergovernmental Panel on Climate Change (IPCC) identified a set of representative concentration pathways (RCPs) that depict the greenhouse gas concentration (not emissions) trajectories under a range of warming scenarios.19 Thus, RCP is a way to show how greenhouse gas concentration may develop over time. However, they do not account for the social and economic drivers. The Shared Socioeconomic Pathways (SSPs) were created to complete this picture.20 Importantly, RCP and SSPs allow parallel approaches to better understand climate change scenarios, and the associated likely output warming. The IPCC has since published a Sixth Assessment Report that documents the individual contributions of three working groups focusing on physical sciences, climate change impacts (including adaptions and vulnerability), and mitigation. The report is a stock-take of current progress and provides the best- and worst-case scenarios of impacts on the environment and human society.21
Following the Paris Agreement and COP26, there is broad awareness that transitioning to a low carbon economy poses significant risks, new challenges, but also investment opportunities for financial services. Several leading industry bodies provided well-defined guidance with global applicability such as:
The policy frameworks, guidelines, and principles are all supported by key assessments and reported data.
For organizations to effectively assess and address the risks associated with climate change, it is helpful to evaluate the channels by which climate change impacts the financial sector. These include:
Each of these climate risk categories transmits through the economy at both macroeconomic and microeconomic levels, to financial risks such as credit risk, market risk, liquidity risk, operational risk, and reputational risk.
For risk practitioners, to assess the impacts of climate change on portfolios and incorporate climate risk assessments in decision-making, questions about the accuracy and reliability of climate risk models prevail. The following considerations are highlighted:23
According to the Basel Committee, “A bank's ability to assess its overall exposure to climate risks across all of its significant operations will be heavily dependent upon the quality of its IT systems and its ability to aggregate and manage large amounts of data.”
The above have influenced a growing amount of work24,25 that suggests traditional models will not be adequate for modeling climate risk. Even if traditional approaches are used by risk modelers or climate scientists at least initially, these will require retuning or redevelopment at some stage. So, the larger question for risk managers becomes: “How do we use AI and machine learning to create accurate models that can reconstruct climate physical and transition risk drivers, retrain regularly, and capture the interconnections between the climate, the macroeconomy, and balance sheets?”
The scale and complexity of the climate change problem demand new thinking and new technologies, including the use of AI and machine learning. AI and machine learning are very effective at modeling complex relationships at scale, and that is exactly what is required in terms of climate risk. There are various auxiliary functions that are leveraging AI and machine learning to better assess and address climate change, including its use to better understand the extent of the problem, to monitor progress, and to address data quality issues.26
Firstly, analytics can help to better understand and define the climate risk problem. There is a lot of climate data available, like geospatial data, sensor data, and satellite images, but innovation is needed to analyze and combine it with traditional risk data and processes. New methods are needed to analyze new and unique types of data, including unstructured data (such as sentiment).
For example, with climate change it is well known that there is an increased risk of flooding—more frequent flooding will not only impact collateral values, but also supply chains and the long-term creditworthiness of a region. With the use of “smart city” analytics, sensors and rain gauges, residential areas can be better protected from flooding. By using newer technologies, monitoring water levels in real-time to predict flood risk, lives and livelihood can be better protected.
Secondly, to reduce greenhouse gas emissions, energy intensive and high-carbon industries will likely turn to buying carbon offsets. There are still open questions on how carbon offset markets are governed and how to guarantee the integrity of such a market, but computer vision and sensors can help monitor the greenhouse gas emissions of carbon offset projects.
And lastly, the use of AI and machine learning can assess data quality (see Chapter 2) and address data gaps through synthetic data creation.
But for now, we want to further explain the measures taken by the United Nations that have set targets to limit the rise of global temperatures to between 1.5 and 2 degrees Celsius, and how the most current targets are being used as the basis of climate risk scenarios for climate risk stress testing.
Financial institutions have started strengthening their climate risk assessment capacity. Now that the definition of climate risk is beginning to take shape, banks are in the planning stages of building the right governance frameworks.
Financial institutions will need well-considered environmental risk management expertise, but it is important that the banking industry develop their own capacity and not solely rely on regulators for guidance. Ideally, financial services industry will collaborate with regulators to further develop and refine regulations. There will also need to be local adjustments as each country has a unique set of social, economic, regional, and political conditions. Thus, the approaches to climate risk management will evolve over time.
For example, the Swedish regulator put a new policy in place: an aggressive target for its country to become the world's first net-zero fossil fuel emissions nation, as part of its Roadmap 2050.27 Sweden has been on a greening journey to phase out its dependency on fossil fuels for over two decades. In addition, it has closed its last coal-fired power station two years ahead of schedule.28 Adopting a similar approach of net-zero fossil-fuel emissions would simply be catastrophic in countries that rely heavily on fossil fuel. In this case, a disruptive, disorderly approach will likely have detrimental consequences:
This is not to say that complacency is welcome. Climate risk has a direct impact on the economic environment that the financial services industry operates in, their business operations, and counterparties. Organizations will need to be proactive and disciplined to make progress toward responding to climate risks. A fine balance between mitigating climate risk and ensuring financial stability will need to be established, which will be influenced by the vulnerabilities to climate risk and how it evolves, government policy responses, and customer expectations and needs. This is a journey!
In most cases, financial institutions are not waiting for governments to drive the large-scale action needed to mitigate financial risks. They are responding to signals from institutional investors (such as the IIGCC29) and the public to decarbonize their portfolios, announcing net-zero targets, and diversifying away from thermal coal lending.30 In addition, they are creating sustainable finance and managing the financial and nonfinancial risks associated with climate instability.
Central banks and regulators are releasing forward-looking climate stress tests and disclosure expectations, mostly based on the recommendations by the Task Force on Climate-Related Financial Disclosures (TCFD).31 The Financial Stability Board formed the TCFD to help companies understand what types of disclosures financial markets need on climate-related risks. Without this information, investors may incorrectly evaluate or price assets that can lead to a misallocation of capital. While this will support convergence and standardization, the climate stress tests and TCFD recommendations need calibration and adjustment by countries to be relevant in each jurisdiction.
To assess the impacts of climate change, many organizations are extending their scenario analysis and stress testing frameworks to include climate risk scenarios. This is also the approach adopted by many central banks. Scenario analysis is a helpful tool to, firstly, comply with climate-related regulatory stress testing and secondly, allow various stakeholders to better understand the shorter- and longer-term implications under a range of forward-looking scenarios, and plan accordingly.
Running climate-related stress tests is a good starting point. Keep in mind that, for financial institutions, in general, the narratives of regulatory stress testing may not be sufficient to capture the unique impacts of climate change. Regulatory stress tests rely on only a handful of scenarios and broad-brush model assumptions that are not necessarily specific to a firm's risk profile. Responses are heavily reliant on human judgment.
Although there are plans to make regulatory and industry scenarios more granular, the scenarios currently available do not capture the full range of potential pathways. In addition, these do not account for policy changes or mitigating actions: for example, mitigating events such as technological improvements in renewable energy to avoid the physical risks before financial losses are realized.32
In most cases, the stress testing models in use today do not account for propagation channels through micro- and macroeconomic drivers, nor the nonlinear relationships between risk factors, mostly due to the popular use of linear modeling techniques.33
In addition to facilitating a more granular level of analysis of complex, nonlinear relationships, i.e., help ascertain the direct and indirect influences of physical and transition risks on the micro- and macroeconomy, and subsequently on financial balance sheets, at the individual and interrelated level,34 the use of AI and machine learning can help to better quantify the uncertainty of climate-based events.
Once the fundamentals are in place, the simplest way for a bank to build their climate risk management capacity is to extend their stress testing framework to incorporate climate risk scenarios and simulations. The extension will require to run scenarios with much longer time horizons (to 2050 and even 2100). As mentioned earlier, regulators and other industry bodies are releasing forward-looking climate stress tests to help ease the burden on financial institutions having to design their own scenarios. Of these, the Bank of England's Prudential Regulatory Authority (PRA) has published one of the most comprehensive climate stress tests to date.35 The developed 2021 biennial exploratory scenario (BES) has drawn from the lessons learned from climate scenarios in their 2019 Insurance Stress Tests. In addition, the NGFS has collaborated with climate scientists and published a set of climate scenarios. We will next describe both the BES and NGFS climate scenarios.
The Bank of England developed a 2021 biennial exploratory scenario (BES) that draws on the lessons learned from climate scenarios in the 2019 Insurance Stress Tests. The aim is to:
The key features of the BES are summarized in Figure 9.1 Key aspects include:
As part of sizing the risks on their assets of the climate scenarios, participants modeled the impact based on three levels of granularity, namely: corporate exposures, household exposures, and government exposures. For these exposures, participants were encouraged to assess individual corporate counterparty-risk using climate disclosures like those released by the TCFD. A challenge was how to assess corporate counterparties at the counterparty level, given that there is not a standardized set of information on emissions data available. The regulator used the findings from the exploratory exercise to assess the safety and soundness of the financial system under the different scenarios, and to determine whether additional capital to absorb climate risk losses is necessary. Based on the responses from participants and the analysis of the regulator, a key finding from the CBES exercise is that climate risk will place a drag on the profitability of banks, making the system more vulnerable to future potential shocks. At this stage, the regulator did not introduce a climate-specific capital add-on.36
The Network for Greening the Financial System (NGFS) published climate scenarios in 2020 and their updated version in 2021. A set of six scenarios were provided based on these dimensions:
Despite the challenges associated with climate stress testing, scenario analysis and simulations, and heavy reliance on expert judgment, several financial institutions in jurisdictions such as those across Europe, the Middle East, and Africa (EMEA) and Asia Pacific (APAC) are building capacity to improve quantifying climate risk by collecting climate data, integrating that with their incumbent data management processes, and performing early-stage stress testing and scenario analysis that are broadly aligned to those of the NGFS. Financial institutions are either utilizing their existing risk models by augmenting with climate-based variables, or creating new climate risk models.
With all this in mind, it is no wonder that organizations are not sure where to start.
Although most central banks have issued guidelines on how regulated entities, including banks, can apply prudent practices for climate risk management, in many jurisdictions a lack of clarity remains on:
To not be caught off guard, for organizations a good starting point is to review their capability to see if it meets the fundamentals of a robust analytic ecosystem (Figure 9.2):
Organizations will need granular physical and transition risk data, and that data will require use of advanced data-quality procedures to assure data quality. High-performance, parallel-processing engines can accommodate the granular data requirements and frequency of forward-looking risk simulation cycles.
As we know, data is the foundation of any analytics-derived outcome. Without the quality, validation, and assurance measures of data, the outcome is likely to be inferior. Provided that climate risk modeling is a relatively new topic, climate risk data for measuring financial impacts is not that readily available, especially loan-level information, and not yet of adequate quality to support large-scale, robust model development and simulation.
To better address the limitations, organizations will need to focus on the following areas to achieve robustness in their climate risk management efforts:
AI and machine learning are very effective at modeling complex relationships at scale and that is exactly what is required in terms of climate risk. The broad set of applications of AI and machine learning makes it particularly effective as a tool in the fight against climate change. Included in this section are a few tangible examples where AI and machine learning are effectively employed to assist climate risk management.
With the increased adoption of cloud-based computing, an example of that is a cloud computing providers, namely Google, who initiated the development of a machine learning application, called Deepmind, to improve energy efficiency in the cooling of their data centers. The Deepmind project employs ensembles of neural networks to predict energy efficiency. The models capture the dynamic interactions between the temperature observed from sensors, power pump speeds, and other parameters within their data centers. The project led to a significant reduction in energy consumption of 40%.37
Another example is in the increased risk of flooding. The increased risk of flooding is a major physical risk due the increased risk of natural disasters. Physical hazards, such as flooding, have potential to impact not only the collateral values associated with banks' exposures, but also the cost of capital and affordability of insurance. It can have a knock-on impact on businesses, supply chains, and the long-term creditworthiness of a region. Recently, an American town planner employed AI and machine learning on a smart city project to better protect residents from river flooding. By installing sensors and rain gauges to monitor water levels in real-time and using that data and analytics to predict flood risk, residents can be warned proactively and traffic is rerouted. In this case, the use of AI and real-time data mitigates the socioeconomic impacts of the increased risk of flooding.
Recognizing the need for an enterprise climate risk solution, a Tier 1 bank in Asia Pacific extended their stress testing solution to quantify the impacts of climate risk following a bottom-up analytical approach. They employed scenario analysis with supporting internal and external reports. A climate risk model assesses the impacts of each industry or subindustry's decline in asset values and quantifies the climate risk at exposure level. It links the bank's portfolio strategy with its finance and auditing processes for business planning and policy decision-making.
For financial institutions, assessing the carbon emissions of counterparties presents practical challenges in data collection. In this case, the firm employed AI and machine learning to upload and process sustainability reports and counterparty information on a cloud-based platform. Geo-location data, scenario, and other risk data from third-party providers are uploaded through APIs. By modeling the real estate data, disaster prediction data, and financial data of the counterparty, the degree of financial impact and the impact on collateral value are projected. The information is then used to assess the impacts of climate risk on the counterparty's balance sheet and P&L.
An analytics company in the Nordics created a sustainability investment screening solution to help their clients. By utilizing its web interface, financial investors can select companies that are more sustainable than others. The solution performs investment screening and monitoring based on the use of unstructured data, machine learning algorithms, and an intelligent triage process.
Specifically, the solution incorporates the corporate entities of existing investments. The application downloads the annual reports of each company that the client's financial investors are considering for investing, via the Danish registry, in the form of unstructured textual information. In addition, it uses sustainability reports from the UN Global Database via Robotic Process Automation (RPA). In this case, RPA was needed because an API was unavailable. The RPA performs screen-scraping, extracts the information, and stores it in a structured dataset. The sustainability reports include sustainability development goals (SDGs), as well as human rights, labor rights, the environment, and anti-corruption information.
AI and machine learning are used to improve the process by identifying climate-specific information using new types of data, generating SDG scores, and analyzing trends and variations.
For financial institutions, there is a pressing need to incorporate climate risk in their financial decision-making and risk management processes. As we find ourselves in the decisive decade with regards to transitioning to a lower carbon economy, there is an urgency for financial institutions to lead the way in sustainable finance and help manage both the financial and nonfinancial risks of the transition. The scale and complexity of the problem demand new thinking and new technologies. With that in mind, it's imperative for organizations to build their climate risk management capacity early: in forward-looking climate risk modeling approaches and employing AI and machine learning responsibly.