Physical and transition risks

It is common to distinguish between physical and transition climate risks (Annex 2). The IPCC (2021) provided an updated analysis in 2021³ on the outlook for climate change and physical risk. In its report, it was noted that many changes in the climate system are becoming larger (increases in the frequency and intensity of hazards) in direct relation to increasing global warming giving rise to physical risks—risks arising from physical impact of climate change. In addition, low-likelihood, high-impact outcomes, such as ice-sheet collapse, abrupt ocean circulation changes, and global warming at rates substantially larger than previous model pathways cannot be ruled out. There is a high level of confidence that these low-likelihood outcomes would “worsen” throughout the 21st century, with a medium level of confidence of their exact magnitude. So-called general circulation models (GCMs) developed by climate scientists simulate the response of the global climate system and different indicators of climate change to increasing greenhouse gas concentrations under the reference scenarios. Specific hazards’ models (for example, hydrological models to analyze floods) build on GCMs⁴ to project the frequency and intensity of hazards (essentially damaging climate related events such as storms and floods, etc., Table 1), grouped in different categories. Furthermore, limiting human-induced global warming requires limiting cumulative CO₂ emissions. This gives rise to the transition risks, that is, those resulting from policy, technology, legal, and market changes that occur during the move to a low-carbon economy. Transition risks include assets becoming stranded, reputational damage, and financial distress of polluters. Underlying global climate change scenarios is that the larger are the policy steps taken to reduce the carbon intensity of economic activity, the lesser will be the rise in temperatures. As such, there are trade-offs between the economic and financial effects arising from: (1) the physical impacts of continued temperature rise versus (2) policies to mitigate temperature rise.

^{Table 1.}

Physical Risk Hazards, Indicators

Source: Adapted from EEA (2021).

^{Table 1.}

Physical Risk Hazards, Indicators

Heat and Cold	Wet and dry	Wind	Shown and Ice	Coastal	Open ocean
Mean air temperature Extreme heat Frost	Mean precipitation Heavy precipitation River floods Aridity Drought Fire	Severe wind strom Mean wind speed	Snow, glacier and ice sheet	Costal flood Sea level	Ocean chemistry Marine heatwave Mean ocean temperature

Source: Adapted from EEA (2021).

View Table

^{Table 1.}

Physical Risk Hazards, Indicators

Heat and Cold	Wet and dry	Wind	Shown and Ice	Coastal	Open ocean
Mean air temperature Extreme heat Frost	Mean precipitation Heavy precipitation River floods Aridity Drought Fire	Severe wind strom Mean wind speed	Snow, glacier and ice sheet	Costal flood Sea level	Ocean chemistry Marine heatwave Mean ocean temperature

Source: Adapted from EEA (2021).

Recognizing the complexity of this analysis and the importance of common starting points to illustrate the benefits of joint action, central banks around the world joined in 2019 to establish the Network for Greening the Financial System (NGFS). This network comprises 105 central banks and supervisors from advanced, emerging, and developing economies that are focusing on the need for (1) common and improved disclosure of climate risks, (2) developing reference scenarios and guidance on integration of climate risk analysis into macroeconomic and financial stability surveillance, (3) adapting supervisory and regulatory frameworks to address financial sector risks posed by climate change, and (4) broader collection of climate risk data. Our climate risk analysis framework is closely connected with the work of NGFS and its scenarios that build on IPCC’s scenarios, with efforts to look at shorter term horizons and country specificities of the IMF’s diverse membership.

Starting point: Standard FSAP risk analysis

FSAP risk analysis typically entails the development of scenario-based stress tests for assessing bank solvency and liquidity. FSAP stress tests are top-down exercises run by IMF teams, typically on bank level data with some degree of sectoral break-down. The most basic component of an FSAP stress test is top-down solvency analysis.⁵ The solvency stress testing methodology assesses credit, market, interest, and FX risks and their impact on bank profitability and capitalization over a 3–5 year stress testing horizon.

The process starts by using different macroeconomic modeling frameworks to design adverse macroeconomic scenarios. These scenarios are based on the narrative of risks included in a Risk Assessment Matrix (RAM), and typically characterized by negative GDP growth, rising unemployment, balance of payment and exchange rate shocks, and falling asset prices.

These adverse scenarios are then used as input into country-specific estimated relationships between macro drivers and risk factors (such as credit risk, interest income, etc.) based on their historical relationship, to generate impacts on bank income and capital. Uncertainty over the “true” model linking macro drivers and financial sector risks and variables can generally be controlled for (for example, using methods such as Bayesian Model Averaging (Raftery 1995, Hoeting and others 1999, and Viallefont and others 2001, Gross and Población, 2019). The assessment of bank resilience is then based on whether bank capital falls below certain regulatory thresholds (so-called “hurdle rates”) conditional on the macro-financial scenarios.

In some cases, granular data for individual corporate and household balance sheets are available and can be used in the solvency analysis. In this case, in addition to modelling risks at a more aggregated top-down level, the impact of adverse scenarios is assessed using corporate and households’ micro data (Gross and Población 2017, Gross and others forthcoming). Such a micro analysis provides a more granular perspective which can be important in identifying threshold effects.

Incorporating a climate risk analysis module

The first step in the climate risk analysis is to assess which climate risks and hazards are the most relevant for the country under consideration. This assessment is based on a climate risk assessment matrix (C-RAM) that IMF staff compile based on aggregate climate risk metrics that can be derived from a variety of sources for transition risk (carbon tax gap, share of the carbon-intense industry, etc.) and physical risk (change in hazards’ projections relative to history, damages due to different hazards, etc.). These metrics are used to identify which risk is relevant and material, following which a narrative of the risk and transmission channels is described. For a country where climate risks are important, two additional steps are added to the bank solvency stress testing framework to incorporate physical and transition risk:

(1) Temperature and emissions scenarios. These scenarios present the starting point of the analysis and are taken from the IPCC/NGFS.

(2) “Climate” scenarios. A design of climate scenarios entails mapping emissions and temperature scenarios into the evolution of physical risks (that is, projections of frequency and severity of relevant hazards) and transition risks (that is, pathways for carbon taxes).

To map climate scenarios into the resiliency of banks, two approaches could be used depending on the level of data granularity. A macro approach could be used to map climate scenarios into macro and sectoral scenarios and use the standard stress testing methodologies to assess the implications of climate risks for the banking system’s resiliency.⁶ If more granular data are available, a preferable, micro-macro approach focusing on borrower-level (corporates and households) assessments and their implications for banks can be considered. The additional micro granularity is valuable in the context of climate risk analysis, as discussed below.

Design and implementation challenges

There are several challenges in designing and implementing climate risk analysis.

• Climate change is a long-term phenomenon where the largest consequences of physical risk and the benefits of policy actions are typically judged to emerge over a 30- to 80-year horizon, well beyond the conventional 3- to 5-year horizon typically considered for risk analysis in FSAPs and other stress testing exercises. However, in conducting the stress tests, staff consider both the conventional and longer-term horizons given sizable uncertainty over modeling and policies. Among these are the increasing likelihood that the realization of long-term costs (including from stranded assets) could feed back into shorter-term horizons via a reassessment of market valuation of companies and thus banks. However, long time horizons come with high uncertainty about how policy and socio-economic factors might evolve. There is also a very wide range of climate models to choose from giving rise to sizable model uncertainty. Inherent uncertainty in modeling increases with the complexity of the system. This generates higher-than-typical uncertainty regarding projections of emissions and temperature. As such, there are many pathways for emissions and temperatures with high levels of uncertainty around them, though the general tendency is for a clear increase in temperatures without policy action to decarbonize.

• Modeling macroeconomic outcomes of physical and transition risk is complex. Challenges include (1) translating the impact of transition policies into macroeconomic and sectoral outcomes; this requires combining macro and sectoral models and modeling changes in the structure of the economy due to transition—something we do not observe in the historical data over the short term—while the timelines and magnitudes of transition policies may shift, including as unexpected physical risks emerge, new technologies are developed, or in response to other shocks. Moreover, the framework does not fully take into account adaptation to changing climate conditions, which is relevant in the long run as ongoing adaptation will reduce exposure and vulnerability to climate change.⁷ This endogenous reaction to climate change is very difficult to fully model and is one of the areas requiring further work; and (2) designing credible macro scenarios due to physical and transition shocks over a long horizon of up to 80 years.

• Bank risk analysis or stress testing often makes the simplifying assumption that the structure of bank balance sheets does not change over time (for example, the “static” balance sheet assumption). This assumption is particularly problematic in the case of assessing climate change considering the long horizons of the analysis over which it could reasonably be expected that there will be significant structural changes to economies and banking systems.

• Assessment of projections of physical risk models and their calibration requires special scientific expertise. History provides little guidance and extreme events cannot be ruled out over the near/medium term as the distribution of such events is likely shifting over time, with nonlinear threshold effects and tipping points of uncertain timing and possibly systemic impact.

• Climate risk analysis puts a premium on granular data, arguably even relative to standard stress testing, given the sectoral and geographical specificity of climate risks—physical risk exposure is highly location specific, while carbon exposures vary significantly at the firm level. But disclosure of transition risk exposures is limited, modeling projections of both risks is highly uncertain, with major work remaining on estimating damages from physical risks. And hazards data in a user-friendly format are expensive.

• Finally, physical and transition risks—currently analyzed separately due to complexities of analysis in each case—should in principle be jointly analyzed within one framework since the risks are interconnected. For example, higher carbon taxes and lower carbon emissions will increase transition risk but will have positive consequences on climate change (lower physical risk).

While scenario analysis offers a “what-if” framework that is suitable to deal with uncertainty about future climate and policy events over a long horizon, it is important to note that the methodologies presented here are still work in progress.

Temperature and emissions scenarios

FSAP teams are generally seeking to use, as a starting point for their analysis, the NGFS reference scenarios for emissions and temperature, and the concomitant pathways developed for both physical and transition risk (see NGFS 2021b for more details). The NGFS emissions and temperature scenarios (Figure 2) draw primarily on IPCC emissions pathways for the period up to 2100 and projections for consequent climate impacts—and combines them with pathways for GDP, population, and the urbanization rate--to help contextualize the setting of each emissions pathway. The scenarios also make a range of assumptions about how technology evolves.

View Full Size

CO₂ Emission by Scenario

Citation: Staff Climate Notes 2022, 005; 10.5089/9798400212895.066.A001

Source: IISAS NGFS Climate Scenarios Database, REMIND model.Note: NDCs = nationally determined contributions.

• Download Figure
• Download figure as PowerPoint slide

Currently, six global scenarios have been published by the NGFS, divided in three clusters (Figure 3). Each scenario is characterized by paths for emissions, temperatures, and carbon taxes. Physical and transition risk are seen as interconnected—strong and immediate policy action would raise near term transitions risk but limit physical risk; by contrast, delayed and weak action would lead to higher physical risk, lower transition risk today but even larger transition risk in the future. The scenario clusters include:

View Full Size

NGFS Scenarios Framework

Citation: Staff Climate Notes 2022, 005; 10.5089/9798400212895.066.A001

Source: NGFS Climate Scenarios for Central Banks and Supervisors, 2021.

• Download Figure
• Download figure as PowerPoint slide

• Orderly transition. Climate policies are introduced early and become gradually more stringent, consistent with the Paris Agreement. Both physical and transition risks are relatively subdued. This cluster includes two scenarios: (1) a scenario that limits global warming to 1.5°C by the end of the century relative to pre-industrial level through stringent climate policies and innovation, reaching global net zero CO2 emissions about 2050 and (2) a scenario that gradually increases the stringency of climate policies, giving a 67 percent chance of limiting global warming to below 2°C.

• Disorderly transition. These scenarios involve higher transition risk due to delayed or divergent policies (for example, carbon prices increase abruptly after a period of delay and not all countries follow their Paris commitments).

• Hot house world. These scenarios assume that current climate policies are maintained resulting in high levels of emissions such that the global temperature rise exceeds 3°C by the end of the century. There is no or limited transition risk, but physical risk is severe. The nationally determined contributions (NDCs) scenario, which falls under this cluster, includes all pledged policies even if not yet implemented.

Physical risk analysis

After having explained how general bank stress testing frameworks are designed and then adapted to the needs of climate change risk analysis, we now explain in greater detail the specifics of the approaches to assessing physical and transition risk.

Our physical risk framework seeks to assess the impact of climate change on the economy and the banking sector for countries where physical risk is relevant. A changing climate leads to changes in the frequency and intensity, of weather and climate extremes, and can result in unprecedented extremes. These hazards cause damages to physical assets, markets, and productivity that in turn can affect the resilience of the banking sector.

We focus on both long-term horizons spanning up to 2100, but also the standard FSAP horizon of 3 to 5 years. Both long- and near-term analysis are challenging. The long-term analysis uses distributions of projections of hazards for shock calibration and assumes static balance sheets. This latter assumption is particularly unsatisfactory in the face of structural change engendered by the response to climate change, but alternative approaches are not currently feasible. Shorter horizons are considered due to the possibility of extreme events happening over the near term and the fact that low-likelihood outcomes, and global warming at rates substantially larger than expected pathways cannot be ruled out. To account for this uncertainty over the magnitude of the shock, the projected distribution of some hazards over a long term could possibly be used to calibrate a severe climate-related shock over the near term. An alternative is to pick a point on the tail of the distribution of physical risk hazards from the recent history, though this could understate the effect of climate change and possible non-linearities.

We start with NGFS temperature and emissions scenarios, discussed in the previous section, that will affect projections of frequency and intensity of relevant hazards for a specific location. Data for hazards’ projections are obtained from private vendors or, in some cases, from country authorities. Typical hazards include precipitation, cyclones, floods, droughts, wildfires and heatwaves, and chronic physical risk (for example, sea level rise). These projections are aligned with NGFS scenarios. It is possible to obtain both country-level aggregate risk indicators that can be used to construct a heat map for a country-specific climate risk assessment matrix, and location-specific (that is, latitude and longitudes) granular data for the frequency and intensity of individual hazards, integrated with individual corporate data.

The next step is to estimate the impact of projected hazards in terms of their damages and productivity and assess, in turn, the effects on bank stability. Damages can be estimated for individual corporates or households and then linked to banks’ balance sheets when data on exposures is available. In many cases we are seeking to leverage World Bank staffs experience on catastrophe modeling to estimate damages. Damages can also be aggregated at country-level and used as an input (for example, capital and productivity shocks) in macro models to generate macrofinancial scenarios accounting for losses from climate risk. A macrofinancial scenario is then used to estimate the impact on bank solvency using the standard approach for banks’ stress tests.

Estimating damages

The damages generated by individual hazards (for example, the degree of capital stock destruction and/or productivity loss from a heat wave) are the key variables that link climate science with finance and economics. Damages from physical risks arise as the interaction of three components: the projections of individual hazards, the exposure of economic agents to these hazards, and their resulting vulnerability in the event the hazard materializes. There is no comprehensive public data set on damages and these granular data are only available, at a very high cost, from firms that specialize in catastrophe (CAT) modeling (CAT models were originally developed for property insurers to help manage rare disaster risks) or from the authorities, if they are available.

The approach to estimating damages from physical risks currently developed by the NGFS consortium is focused on chronic risk and based on top-down, country-level empirical relationships between GDP and temperature as a high-level approximation of the economic impacts of temperature scenarios that underlie hazard projections and their damages (Figure 4). GDP losses are calculated based on the methodology set out in Kalkuhl and Wenz (2020) at the country level for the change in average temperature in each RCP scenario. Estimates suggest a global GDP impact of up to 13 percent in the current policies scenario relative to a scenario with no additional warming. Resulting damages representing the impacts from chronic climate change are then exogenously introduced into the National Institute’s Global Econometric Model (NiGEM—proposed by the NGFS) as a shock to capacity to generate other macro and financial variables for use in stress testing.

View Full Size

Physical Risk—Global GDP Losses

Citation: Staff Climate Notes 2022, 005; 10.5089/9798400212895.066.A001

Source: IISAS NGFS Climate Scenarios Database, REMIND model.

• Download Figure
• Download figure as PowerPoint slide

The NGFS (2020) notes two main concerns with the current approach. First, historical trends may not hold in the future due to socio-economic changes, or because of potential structural breaks when a particular threshold is reached (for example, labor productivity drops off sharply above a given level of climate change). Second, macro-econometric approaches may still not capture all relevant transmission channels and risks such as low-probability high-impact events, sea-level rise, and extreme events. As a result, actual damages under these scenarios are expected to be larger than shown. As such, the NGFS is moving toward assessments of direct damages for acute risks only (currently available for tropical cyclone and floods), but the current approach remains in use as granular data on damages for all hazards are not publicly available.

Staff are also seeking to develop bottom-up methodologies to assess damages in a consistent manner for different hazards across countries which would then be used to calibrate macro models to assess the macro impact (Figure 5). This approach is similar to the NGFS approach on estimating damages due to cyclones and floods and approaches used by the Banque de France (ACPR 2021) and the Australian Prudential Regulation Authority (APRA 2021). Estimating damages entails linking projections of hazards to the exposures (that is, geolocation and value of assets) we are interested in, and their vulnerability (that is, the propensity or predisposition to be adversely affected when impacted by hazard events). The approach is set out below.

View Full Size

Physical Risk Analysis: Estimating

Citation: Staff Climate Notes 2022, 005; 10.5089/9798400212895.066.A001

Source: Geiger and others (2017).

• Download Figure
• Download figure as PowerPoint slide

Hazards Projection. We are currently exploring the use of data compiled by private vendors on the projections of frequency and intensity of hazards.

Exposures. Ideally, we would like to have geolocation data on infrastructure, insured properties, physical capital, and their valuation. Since there is no publicly available data set on exposures, we are currently relying on a proxy, namely, GDP measures that have been mapped from the national aggregate to the GDP in individual geographical units (so-called downscaling⁸). We can then proxy the capital stock at risk in these disaggregated geographical locations by using standard capital output ratios applied to the gridded GDP.

Vulnerability. We then would use hazard specific damage functions from existing studies and publicly available data sets (instead using the aggregate economywide damages studied by Kalkuhl and Wenze (2019)). Damage functions vary in shape and link the intensity of individual hazards to economic damages where larger intensity leads to larger damage in a nonlinear fashion (depending on the hazard, geography, and type of exposure).

Bank stability assessment

Once projections of damages of relevant hazards are estimated for different temperature and emissions scenarios, we apply these to the bank stability assessment. There are two approaches to conducting the stability analysis and data availability directs the choice of the approach.

Macro approach. This approach incorporates analysis of capital and productivity shocks due to hazard damages on macroeconomic and financial variables using macro models. Damages are aggregated at country-level and used as an input in macro models to generate macro financial scenarios due to climate shocks. Macro models used include: NiGEM, single-country Dynamic stochastic general equilibrium (DSGE) and global models developed by IMF staff. Once macro-financial scenarios are built, the standard FSAP stress testing approach for credit and market risks is applied to assess the risks and the impact on bank capital (see Annex 3 for an example from Philippines FSAP).

Micro-macro approach. This approach relies on an analysis of firms and households using micro models (integrated with, that is, connected to, the macro models) to estimate the impact of physical risk (and its impact on macro variables) on individual balance sheets. In this case, granular damages data are used to estimate damages to firms’ and households’ balance sheets in specific geographies, taking into account the mitigating effect due to insurance of assets if data are available.⁹ The financial risk exposure of the banking sector—including via collateral valuation—to specific geographies is used to link damages to the earnings of banks.

Transition risk analysis

Transition risk reflects the policy reaction to climate change. The proposed focus for FSAPs, in line with the NGFS scenarios, is on carbon taxation in countries that are vulnerable to transition risk. The logic of the framework is the following. The introduction/increase in carbon taxation will increase the cost of carbon emissions, which will weigh on the financial performance of carbon intensive firms, induce relative price shifts that disincentivize the non-renewables sector, and promote growth of renewables (especially if carbon tax proceeds are invested in renewables), all of which will have an impact on the wider economy and the banking sector.

The FSAP approach to transition risk analysis may start with NGFS scenarios on emissions and temperature. The next step is to derive a carbon price to achieve specified emissions in different scenarios.

The impact on banks’ health could be derived using two approaches. If granular bank exposure data are available, the impact on firms’ balance sheets is used to estimate the impact on banks. We assess the impact of carbon taxes directly on firms’ balance sheets—using national or vendor-provided, firm-level data. The assessment of firms’ health could be complemented by the analysis of macro and sectoral effects of carbon taxes using macro and computable general equilibrium (CGE) models (for example, ENV model (Chateau, Jaumotte, and Schwerhoff 2022), the G-CUBED model (McKibbin and Wilcoxen 2013¹⁰). In this context, we are leveraging CGE models developed by the IMF and World Bank staff (see for example the ENVISAGE model (Van der Mensbrugghe 2019)). The macro-financial and sectoral scenario is used to estimate the impact on bank solvency using the standard approach for banks’ stress tests, even without a granular assessment of firms’ health. Data and modeling availability determine the choice of the approach.

Staff plans to consider scenarios ranging from a standard 3- to 5-year horizon to longer-term scenarios (30 years) to be able to analyze the opportunities from carbon taxes. We are also considering upfront shocks to carbon prices for the purpose of sensitivity analysis.

Staff are also exploring the idea of a so-called “Climate Minsky moment,” (Annex 4) where current asset valuations could shift abruptly, driving upfront changes in credit risk from longer-term climate drivers. Such a sudden shift in valuation perceptions could arise if there is technological breakthrough or if consumers, firms, or financial markets suddenly change their expectations regarding how future policies, technologies, or physical risks may evolve, interact, and therefore impact asset valuations (for example reflecting a surprise regarding the true extent of exposure of firms to carbon).

Scenario design—where do carbon tax paths and other variables come from?

There are two approaches for deriving paths for carbon taxes consistent with achieving the benchmark emissions and temperature pathways examined by the IPCC (NGFS).

NGFS (2021) approach. Carbon prices are available from NGFS scenarios (Figure 6),¹¹ which are derived using Integrated Assessment Models (IAMs)¹² for a given GDP path (taken from SSPs or semi-endogenously derived in the case of some IAMs) as a proxy for energy demand. IAMs are used to model the interaction between energy, land, economy, climate systems, and policy needed to meet a particular temperature outcome13. Assumptions regarding policy and technologies (for example, carbon dioxide removal, the availability of solar, wind, and geothermal resources) play an important role in IAMs. Multiple models were used to produce the scenarios to capture a range of uncertainty in the results. The NGFS methodology then uses the NiGEM macro model to derive paths of other macro and financial variables consistent with carbon and GDP pathways from the IAMs. This approach, however, takes GDP as given and does not provide comprehensive sectoral impacts.

View Full Size

Carbon Price Development

($2010 per tonne CO₂)

Citation: Staff Climate Notes 2022, 005; 10.5089/9798400212895.066.A001

Source: IISAS NGFS Climate Scenarios Database, REMIND model.Note: NDCs = nationally determined contributions.

• Download Figure
• Download figure as PowerPoint slide

IAMs suggest that a carbon price of around $160 per tonne would be needed by the end of this decade to incentivize a transition towards net zero by 2050 (Figure 7). They suggest that world GDP impacts from transition risk are slightly positive in the Net Zero 2050 scenario as negative impacts on demand from higher carbon prices and energy costs are more than offset by the recycling of carbon revenues into government and private investment into new technologies, and lower employment taxes. GDP impacts are negative in the disorderly scenarios as the speed of the transition combined with investment uncertainty affects consumption and investment.

View Full Size

Transition Risk: GDP Impacts Relative to Prior Trend

Citation: Staff Climate Notes 2022, 005; 10.5089/9798400212895.066.A001

Source: IISAS NGFS Climate Scenarios Database, REMIND model.Note: NDCs = nationally determined contributions.

• Download Figure
• Download figure as PowerPoint slide

Although the NGFS approach offers a critical global perspective, there is scope for extensions. The NGFS scenarios offer a globally consistent modeling framework linked to alternative global climate pathways that include some tailoring to country-specific circumstances. However, not all IMF members are considered in all layers of the scenarios.¹⁴ The globally consistent approach has the benefit of motivating joint action and improving cross-country comparability among other factors. But such common approaches limit the flexibility of tailoring parameters of the scenario design to more country-specific features. These could include, for example, different schemes for reinvestment of carbon tax proceeds, applying versus not applying carbon taxes for different countries or groups of countries, or for devising carbon trade policies (such as carbon border adjustment mechanisms). The NGFS is working to increase industry segmentation in the scenarios, which is important for analyzing the effects of transition policies that have differential effects across sectors. Alternatives are to consider global CGE models, with significant global country coverage, comprehensive industry segmentation, and greater flexibility regarding reinvestment, carbon tax application, and related trade policies.

CGE approach. Dynamic CGE models can be used to derive carbon tax paths, their sectoral impact (substitution from non-renewables to renewables), and GDP and other macroeconomic variables consistent with IPCC scenarios on emissions and temperature. The advantage of this approach is that the macroeconomic variables and carbon taxes are fully endogenous and that CGE models provide a fuller sectoral decomposition of transition paths, by contrast with the more aggregate output of the IAMs employed for the NGFS scenarios. Moreover, CGE models (such as GTAP and ENVISAGE, as referenced earlier), allow obtaining detailed impacts on bilateral trade effects across countries. Our initial assessment is that fully endogenizing the determination of carbon prices and output consistent with target temperature/emission paths could well produce a larger impact on GDP over the near term comparing to IAMs for the same increase in carbon taxes. However, more analysis will be needed to assess the sensitivity of the models and scenario outcomes to such methodological differences. In any event, once the paths for aggregate and sectoral GDP, other macro variables and trade are determined by the CGE model, supplementary macro-financial models can be used to derive other financial variables required for financial system impact analysis (current modeling frameworks do not generate most such variables).

Bank stability assessment

The bank stability assessment requires a mapping from carbon taxes to impacts on the macroeconomy, firms’ balance sheets and bank capital. There are two approaches based on data availability:

Macro approach. Macro models are used to generate macro financial scenarios due to carbon tax shocks. Sectoral models are used to analyze cross-industry differences and opportunities from higher carbon taxes which are critical for assessing transition risks. Outputs from dynamic CGE models used for such sectorization (for example, carbon taxes, GDP) should be consistent with outputs of macro models/IAMs. Once macro-financial and sectoral scenarios are built, the standard FSAP stress testing approach for credit and market risks is applied to assess the risks and the impact on bank capital.

Micro-macro approach. This approach assesses the impact of carbon taxes on firms’ balance sheets considering firms’ carbon intensity. Staff are currently experimenting with data that projects carbon taxes and greenhouse gas emissions at the firm level and linking these to carbon price paths and macrofinancial and sectoral scenarios to generate estimates of firm level credit risk. The emissions are categorized as Scope 1, 2, or 3.¹⁵ Unlike scope 1 and 2 emissions, whose measurement requires keeping track of greenhouse gasses emitted directly by a company or embodied in the energy it buys, the measurement and disclosure of scope 3 emissions involve a number of challenges, including those related to data availability, use of estimates, calculation methodologies and other sources of uncertainty. Currently the availability and quality of scope 3 emission disclosures is very limited. Mapping firm-level credit risk into bank stress tests requires bank exposure-level data (Annex 5).

What else has been done so far in FSAPs?

Climate risk has been considered in other several FSAPs. Early analyses were undertaken in FSAPs for Chile, Colombia, Norway, and South Africa to test new methodological approaches and enrich discussions with the authorities in the context of the FSAP. Brief summaries of the approach are indicated below, with transition risks analysis mostly emphasizing the impact of carbon taxes on corporate performance thus providing sensitivity analysis to complement the bank stress test.

Norway: This FSAP pioneered a micro-level, debt-at-risk analysis of transition risk posed by carbon tax increases (the scenario modelled was a one-off increase in the domestic carbon price to $75 and $150 from current levels of about $45 per tonne of CO₂ equivalent) at the level of individual firms, though the impact was not mapped directly into credit risk and bank capital. The results suggest that debt at risk (the share of firms for which the interest coverage ratio drops below a threshold value following a carbon price shock) would amount to 2.2 percent for a carbon price increase to $75 per tonne CO2 and 4 percent for an increase of $150 per tonne CO₂.

Chile: In this FSAP for physical risk, a macro approach was taken to assess the impact of floods and droughts on capital stock, calibrated using historical data and analyzed using the IMF’s global macrofinancial models. The impact of historical droughts and floods on GDP was estimated to be around 2 percent of GDP, which had a small effect on banks (0.1 percentage point reduction in the aggregate capital ratio). For transition risk, a micro analysis was undertaken of the impact of higher carbon taxes (one-off increase to $100 per tonne CO₂) on credit risk facing individual firms. Firm default rates were linked to balance-sheet vulnerability indicators using firm level data on historical probabilities of default (PDs) and balance sheets. A micro-simulation was then performed on firm-specific balance sheets to establish “stressed” vulnerability indicators due to increased carbon prices; carbon price paths and firmspecific emission footprints were used as the anchoring mechanism for this simulation. The stressed indicators were used to produce firm-level stressed PDs and the cross-sector weighted average PDs were used to inform bank stress test results in terms of additional provisions and capital impact attributable to transition risk. The results suggest that the higher carbon taxes modeled would increase PDs by 0.7 percentage points after five years, which would reduce capital ratios up to 0.5 percentage points. The sectoral impact and the impact on GDP were not assessed.

South Africa: In this FSAP the physical risk analysis linked provincial nonperforming loans to historical droughts in the provinces while the transition risk analysis conducted micro-level studies of the effects different carbon tax scenarios on individual firm performance. In a scenario of a sudden and large rise in the price of carbon to a mid-point estimate needed to stabilize carbon equivalent emissions, corporate debt at risk among publicly listed firms would rise from a level of around 30 percent in the baseline to around 60 percent in the shock scenario.

Colombia: This FSAP conducted a transition risk micro analysis of the impact of carbon tax increase (a one-off increase in carbon price from $5 to $75/ per tonne CO2) on the financial performance of individual firms, similar to the Norway FSAP, but also assessed the impact on individual bank loans.

Climate risk analysis and Financial Policy

Staff’s proposed climate risk analysis can inform policy considerations in surveillance and capacity development by evaluating the magnitude of risk and potential pressure points for the financial system due to physical climate shocks and in the transition to a low-carbon economy. The resulting analysis can raise awareness of the risk, and adaptation needs and opportunities, including the need for banks to develop tools to manage climate risks and for financial sector supervisory authorities to adequately supervise this risk. This would potentially drive gradual early adjustment and help inform policies needed to enhance risk management and the resilience of the financial system.

As stated at the outset, while climate risk analysis is based on stress testing methodologies, climate risk analysis is not a standard stress test. As such, their uses, while complementary to policy discussions, are distinct.

The standard stress test is used for monitoring and assessing risks over the near term and to inform the design of microprudential and macroprudential tools. It is based on well-established methodology, extensive data disclosure, and a reasonable first-order understanding of the historical relationships between macro-financial variables and bank capital. For purposes of microprudential supervision, most supervisors use stress tests as a qualitative tool when judging banks’ overall resilience, risk exposures and controls. The results of the stress tests, in most of the cases, are not tied to additional capital requirements or other “binding supervisory tools” but often impact the “internal supervisory rating” of the bank and, consequently the content of the supervisory dialogue and intensity of the supervision process.

Climate risk analysis should in principle be informative of supervision and regulation, especially once methodologies have been further developed and validated. Their use, particularly for regulatory purposes, will require significant further work¹⁶ with challenges arising not just from the complexity of climate analysis relative to standard macro-financial modeling in stress testing, but also from its evolving nature, very long risk horizons, and poor quality of data disclosure¹⁷, reflecting also nonstandard nomenclatures¹⁸. This work also requires new expertise in climate science. NGFS (2022) suggests that, in the short term, in light of challenges posed by data gaps and methodological uncertainties, no jurisdiction as of yet envisage calibrating prudential policies such as capital requirements on the basis of their climate scenario analysis.

As the international community conducts further work in this area climate risk analysis can be very helpful to:

• Create awareness around the prudent management of climate risks and incentivize banks in improving their frameworks.

• Inform supervisors about the potential magnitude of climate-related risks in their jurisdictions and help deepen the understanding of the transmission channels from climate-related risks to the financial system. This is key for the sound development of supervisory plans and an informed decision on how much supervisory resources should be allocated to the issue.

• Identify data gaps and banks’ individual capacity to identify and manage climate-related risks. This allows supervisors to make more specific recommendations for banks to improve their risk management framework.

• Help supervisors to determine the extent to which the impact of climate risk is appropriately considered by banks in their management of credit, market, operational and other traditional risks.

• Compare the results with own evaluation of the solvency impact of climate risk. The incorporation of climate risk on ICAAP is nascent and likely to be an iterative process.

• Assess the long-term sustainability of individual banks business models.

• Build capacity to expand the Pillar 2 approaches.

• Develop metrics and indicators for ongoing monitoring of banks exposures to climate risk.

Currently, other institutions, including financial sector supervisors and central banks (for example, APRA, Bank of Canada, and Bank of England), use climate stress tests to measure the exposures of financial institutions to climate-related risks. This helps to understand the challenges to banks’ business models from these risks, the implications for the provision of financial services, and desired policy responses. Ultimately, climate risk analysis will be useful in directing financial institutions to enhance their disclosure and management of climate-related financial risks.

Collaboration opportunities on climate risk analysis

The complexity of climate risk analysis presents many opportunities for close collaboration to explore synergies, gain knowledge, and develop rigorous approaches. IMF staff have worked closely with the World Bank, and colleagues from the NGFS, among other interactions. There is also rising demand for capacity development on methodologies to assess the relevance and impact of different types of climate risks. The FSAP provides an important vehicle to test new approaches and scenario designs that are relevant for the IMF’s universal membership.

Collaboration with the World Bank. IMF staff cooperate closely with World Bank staff on climate risk work, leveraging their deep technical expertise to improve scenario design while maintaining the IMF’s primary role in undertaking stability assessments. Scenario design for physical risk has benefited from this collaboration given World Bank staff expertise in catastrophe modeling. This has been demonstrated in the fruitful collaboration on the Philippines FSAP. The potential impact of Carbon Border Adjustment tax policies adopted by some IMF members, especially for the IMF’s emerging market economy and developing country members, is also of interest in FSAPs. Detailed cross-border sectoral CGE modeling of associated scenarios, has been another collaboration opportunity.

NGFS. IMF staff participate as observers in the NGFS work stream on scenario design and analysis. The staff has also co-led the workstream on disclosure and are engaged in the workstream on supervision. The main objective of the NGFS workstream on scenario design and analysis is to undertake climate scenario analysis and promote its use within the regulatory community more broadly. To achieve this goal the workplan includes: (1) improving the NGFS climate scenarios developed previously developed by the NGFS workstream 2; (2) providing methodological guidance on scenario-based climate risk analysis for macroeconomic and financial stability surveillance; (3) updating the NGFS climate scenarios on a regular basis; and (4) promoting the use of the NGFS climate scenarios within the financial system. The staff have leveraged learning from the NGFS on climate and macro scenarios in the design of the approach proposed for FSAPs. As indicated previously, the staff intend to use different approaches to deriving carbon taxes including benchmarking scenario design, as feasible, of climate scenarios designed by NGFS. Indeed, transition risk scenarios designed by the work stream have already been used in FSAPs for the United Kingdom. Cross-border cooperation, at this early stage, is crucial to share experience in developing new methodologies.

What central banks and supervisors are doing

Of the 105 NGFS members, the NGFS (2021) reports that more than 30 central banks and regulators have adopted scenario analysis to better understand the macroeconomic and financial impacts of climate change (Annex 6). These range from shorter-term, top-down modelling exercises to exercises with a longer time horizon, in many cases with bottom-up participation of financial firms, with static and dynamic assumption on the evolution of their balance sheets. While it is too early to distill best practices in climate risk analysis, we present briefly six summary examples of such work (Table 2) with some comparison of the main elements of the exercises with the approach considered for FSAPs. The main takeaways are the following:

^{Table 2.}

Climate Scenario Design Approaches, IMF and Six Institutions

Source: IMF Staff Note: APRA = Australian Prudential Regulation Authority; BoC = Bank of Canada; BoE = Bank of England; BdF = Bank de France; ECB = European Central Bank; HKMA = Hong Kong Monetary Authority.

^{Table 2.}

Climate Scenario Design Approaches, IMF and Six Institutions

Source: IMF Staff Note: APRA = Australian Prudential Regulation Authority; BoC = Bank of Canada; BoE = Bank of England; BdF = Bank de France; ECB = European Central Bank; HKMA = Hong Kong Monetary Authority.

View Table

^{Table 2.}

Climate Scenario Design Approaches, IMF and Six Institutions

View Expanded

Source: IMF Staff Note: APRA = Australian Prudential Regulation Authority; BoC = Bank of Canada; BoE = Bank of England; BdF = Bank de France; ECB = European Central Bank; HKMA = Hong Kong Monetary Authority.

Transition risk scenarios. While all institutions rely on NGFS scenarios for transition and physical risk (temperature, emissions, macro impact in some cases), they take the scenarios as a starting point and build upon them to assess sectoral impacts (Bank of England 2021, Bank of Canada and Office of the Superintendent of Financial Institutions 2021) or derive carbon taxes (Bank of Canada and Office of the Superintendent of Financial Institutions as in the staff’s approach). The Banque de France also considers a sudden transition scenario.

Physical risk scenarios. Some institutions (Banque de France/Autorité de Contrôle Prudentiel et de Résolution (ACPR 2021), (APRA 2021)) follow a similar approach to that proposed by staff for considering granular hazards projections and estimation of their damages, relying on damages assessments from reinsurers or national meteorological institutions.

Macro scenarios for transition risk. Some institutions (Bank of Canada, Banque de France) combine macro models with CGE models to assess the macro and sectoral impact of carbon taxes or combine them with (Banque de France) or rely on NGFS macro simulations (ECB 2021).

Macro scenarios for physical risk. Most institutions leave to banks to conduct a bottom-up assessment of the macro implications of physical risks for their exposures. Some rely on the NGFS scenarios (ECB, Bank of England) which is based on the GDP-temperature empirical relationship.

The climate work by NGFS members, while at an early stage, is developing rapidly. At this stage only a few institutions have finalized the first iteration of their climate risk analysis with published results, but the literature from central banks on approaches and methodologies is expanding fast.