Working paper

Climate Risk Stress Testing

A Conceptual Review

Henk Jan Reinders

Dirk Schoenmaker
Mathijs van Dijk
 1

Working paper

Climate Risk Stress Testing:

A Conceptual Review

March 2023

Henk Jan Reinders

Rotterdam School of Management, Erasmus University
International Monetary Fund

Dirk Schoenmaker
Rotterdam School of Management, Erasmus University
CEPR

Mathijs van Dijk

Rotterdam School of Management, Erasmus University
 2
Non-technical summary
This working paper summarizes the academic literature on Climate Risk Stress Testing (CRST) and
provides a framework for comparing types of climate-related shocks and CRST approaches. CRST
aims to predict the impact of climate-related shocks on the economic and financial system, by
modeling its effect on (macro)economic variables, financial sector variables, and firm-specific
variables such as balance sheets and P&L accounts. Considering recent predictions of increasing risks
of environmental tipping points, CRST is a crucial instrument for financial practitioners to understand
the impact on the stability of the financial system in (extremely) adverse circumstances. Through
understanding the consequences of climate-related shocks on the financial system, institutions can
more clearly identify actions which promote the stability of the system.

Climate-related shocks are defined in the financial world through risk, and can be either "transition
risks", which focus on climate-related changes within industries due to decarbonization or other
climate-related regulation, or "physical risks", which focus more on the direct effects of the changing
climate through natural disasters and other climate-related catastrophes. The authors define 6 types
of climate-related shocks in terms of the abruptness and severity of climate change, but also the
abruptness of reassessment or repricing of real and/or financial assets.

Based on the reviewed literature on CRST, the authors conclude that there is a wide variety of
approaches and substantial differences in investigated climate shocks and methodology.
Furthermore, several identified climate shocks have not yet been applied in CRST exercises. Similarly,
models have not included feedback effects between systems and primarily focus on specific asset
classes (equity, bonds, or loans). In conclusion, there is room to develop in terms of climate-financial
modeling by studying different scenarios and creating modeling approaches which integrate climate,
economic, and financial variables and associated feedback effects.
 3
Abstract
We conceptually review Climate Risk Stress Testing (CRST) approaches to assess the impact of
climate-related shocks on financial system stability. We distinguish between climate, economic, and
financial modeling steps, and identify six types of climate shocks and four different approaches
(macro-financial, micro-financial, non-structural, and disaster risk). Our review identifies several key
limitations in current CRST approaches: (i) neglect of certain climate shock types (Green Swan and
Minsky-type events); (ii) overreliance on macro models (with low sectoral and spatial granularity);
(iii) incomplete modeling (lack of feedback effects); and (iv) limited scope (subset of causal channels
and asset classes). We argue that these limitations may lead to significant underestimation of
potential system-wide financial losses and offer suggestions for improving CRST approaches.

E-mail addresses: reinders@rsm.nl, schoenmaker@rsm.nl, and madijk@rsm.nl. The authors are

grateful to Jean Boissinot, Dion Bongaerts, Wouter Botzen, Jean-Stéphane Mésonnier, Romain
Svartzman, colleagues at the World Bank and the International Monetary Fund, and participants at
the 2021 Stress Testing Research conference of the Federal Reserve for stimulating discussions and
useful suggestions. The opinions in this paper are those of the authors and do not necessarily coincide
with those of the International Monetary Fund.
 4

1. Introduction
Climate change and the associated policy measures to mitigate Greenhouse Gas (GHG) emissions
pose a novel challenge for central banks and financial supervisors and their traditional ways of
gauging potential losses from severe adverse events. While financial risk methodologies are
typically assuming that the future will be similar to the past, climate change is likely to lead to
fundamental and often detrimental changes over time in a broad set of regions and economic
sectors (Intergovernmental Panel on Climate Change, 2022). The economic costs related to climatic
change are potentially very high, for example due to the increasing frequency and severity of
natural disasters and sea-level rise in many regions of the world (e.g., Tol, 2002). Moreover,
reducing GHG emissions to limit climate change is expected to come at an economic cost in many
economic sectors, at least in the short run (Acemoglu et al., 2012; Nordhaus, 1992). This means
that, one way or another, it is likely that a broad range of economic and financial assets will face
changes in their value.
From the perspective of regulators concerned with the health of the financial system, a key
question is what the potential impacts of climate change and its mitigation are on the profitability,
solvency, and liquidity of banks (Campiglio et al., 2018). Efforts in recent years by central banks
and financial supervisors have focused on understanding and gauging climate-related financial
risks, including “transition risks”, which capture structural changes in the economy due to GHG
emission reduction, and “physical risks”, which capture the effects of a changing climate (Batten et
al., 2016; Nieto, 2019). Of specific interest are transition and physical risk scenarios that could
cause large economic and financial losses, impeding financial stability.
To assess risk scenarios that could cause large losses, financial sector stress testing is an
often-used technique that measures the vulnerability of a portfolio, a financial institution, or an
entire financial system under different hypothetical events or scenarios (Ong and Jobst, 2020).
Stress tests are designed to estimate what would happen to financial sector variables under adverse
(i.e., severe but plausible) circumstances, that have not materialized yet. Stress testing per
definition looks at extreme scenarios in the spectrum of all possible scenarios. Risks investigated
are hence in the tail end of the scenario probability distribution (Slijkerman et al., 2013; Bolton et
al., 2020). Traditional macro-financial stress testing approaches based on estimated GDP impacts
may, however, underestimate losses in adverse scenarios, amongst others due to limitations in
 5

neoclassical economic modeling (Keen, 2021) and the potential for unknown and highly non-linear
“tipping points” occurring (Lenton, 2008; Armstrong McKay et al., 2022). It is hence important to
develop new climate risk stress testing approaches.
Stress tests usually involve a form of modeling, i.e., a function that maps input variables to
the financial system variables of interest. Typically, a financial system stress test model consists of
at least four elements (Borio et al., 2014). This includes one or more (climate) shock scenarios, a
model to translate (climate) shock variables to (macro)economic variables, a model to translate
(macro)economic variables to financial sector variables, and a stress test model to apply shocked
financial sector variables to financial institutions’ balance sheets as well as profit and loss accounts
(see Figure 1). Models to translate (macro)economic variables to financial sector variables are
sometimes also referred to as “satellite models” (Oura and Schumacher, 2012). Furthermore,
financial system stress tests may include one or several feedback loops between key variables, such
as amplification within the banking sector and “credit crunches” whereby the supply of credit to
the economy declines sharply (e.g., Acemoglu et al., 2015; Silva et al., 2018). These feedback loops
may cause additional financial losses.
A specific strand of stress testing research has emerged that investigates the effects of
adverse climate and climate policy shocks on financial portfolios and the financial system as a
whole (e.g., Battiston et al., 2017; Allen et al., 2020; Jung et al, 2021; Reinders et al., 2023). This
paper contributes to this emerging literature in three ways. First, we survey the stress testing
methods that have been developed to date. Second, we conceptually classify scenarios and
approaches to climate risk stress testing. And third, we discuss the models that constitute climate
risk stress tests.
Our results are relevant for both academics and practitioners that design climate risk stress
tests. Stress test outcomes should feed into policymaking and risk management to take action to
avoid or reduce the impact of adverse scenarios. This is in particular the case for central banks and
financial sector prudential supervisors, whose main aim is to ensure the stability of the financial
system and to make sure that financial commitments (including to retail depositors and
policyholders) can be met even in highly adverse circumstances. Financial institutions should also
feed stress test outcomes into their risk management policies. Actions by public authorities and
 6

private institutions to reduce the impact of adverse scenarios can speed up the transition (e.g.
financial institutions rapidly reducing exposures to carbon-intensive industries).
This paper is organized as follows. In Section 2, we first discuss the stress testing of climate-
related financial risks and the features that make it different from traditional stress testing. Section
3 then proceeds by reviewing the “building block” models that are employed to understand the
climate-financial relationship. In Section 4, we identify and discuss four common typologies that
have emerged (traditional macro-financial, micro-financial, non-structural, and disaster risk).
Section 5 classifies and reviews the main top-down exercises that have been carried out so far while
the last section concludes and provides avenues for further research.

2. Climate Risk Stress Testing (CRST)

We define Climate Risk Stress Testing (CRST) as a technique that measures the vulnerability of a
portfolio, a financial institution, or an entire financial system to adverse climate-related hazards
and scenarios.1 A typical example of a climate related hazard would be an isolated extreme weather
event, such as a typhoon or flood occurring, while climate scenarios usually focus on a wider set of
variables changing over time, such as changing weather patterns, changing economic structures or
changing climate policies. In the remainder of this paper, we use the term “climate shocks” to refer
to both acute climate hazards and chronic climate change scenarios.
Climate-related risks have several unique characteristics that set them apart from
traditional financial sector risks. First, climate change is characterized by deep uncertainties and
complex non-linear effects that materialize over an extended period of time (Weitzman, 2009;
Monasterolo, 2020). Second, most climate change variables are thought to change gradually over a
relatively long time horizon, which implies that the most severe effects occur decades or even
centuries from now. Third, the level of climate change mitigation is the outcome of a complex socio-
economic process where potential financial losses by investors and lenders may, in turn, affect the
level of ambition of climate policies, such as putting a price on carbon. Fourth, climate change does

1
Hazards usually referring to a single risk factor while scenarios usually involve multiple risk factors evolving over time. It is
common practice in financial sector stress testing to measure financial impact as the differential in the outcome variable (e.g.,
capital adequacy) compared to the situation with either no hazard or a baseline scenario occurring. Hence, besides the shock
scenario a baseline scenario is needed.
 7

not only affect the economy but also vice versa since virtually all economic activity is associated
with at least some GHG emissions (Battiston et al., 2017).
To assess climate-related financial risks, CRST needs to be able to translate initial
parameters (i.e., a climate shock) into variables that are relevant to assess the impact on financial
system (e.g., solvency and liquidity ratios). Since climate-change scenarios usually have no
precedent in history there is a need for more fundamental (structural) models to estimate the
impact of climate scenarios on financial institutions. This in contrast to traditional stress testing
exercises in which the shock parameters are mostly calibrated on historical events, such as GDP
impacts in past financial crises (Allen et al., 2020). Since both transition and physical risks have
asymmetrical sectoral and spatial impacts, this has increased the need to use approaches to risk
assessment that are more granular (e.g., approaches at a sector, firm and asset level) than the
traditional macro-level approaches. Figure 1 provides a conceptual model for CRST depicting the
main modeling steps, generalizing the modeling steps presented in Borio et al. (2014) by labelling
the intermediary steps “economic state variables” and “financial state variables”. 2 There are also
relevant feedback effects between the different dependent, independent, and moderating variables.
These include systemic feedback effects within the financial sector, feedback effects from the
financial sector to the economy, and feedback effects from the economy to climate variables. A
combination of these feedback loops could cause feedback loops all the way from the financial
system back to climate shocks (e.g., when a financial crisis leads to an economic downturn that
reduces mitigation efforts).

2
We generalize these steps since for CRST we find that economic modeling at the micro (e.g., firm) level is becoming more
common – contrasting the traditional approach through macroeconomic variables.
 8

Figure 1 – Conceptual model from climate shocks to financial system impact

Besides the modeling through economic and financial variables, which is arguably the most
important overall channel, there are conceivably more direct channels that are relevant within the
CRST context. This includes climate-related shocks that have a direct effect on financial state
variables or financial institutions (i.e., not through economic state variables). First, this could occur
due to pricing effects based on green consumer preferences, also referred to as a “greenium” (Alessi
et al., 2021). Second, there may be direct effects of climate-related shocks on the pricing and
valuation of weather and disaster linked financial instruments, such as catastrophe bonds
(Lakdawalla and Zanjani, 2012).3 Third, direct effects on financial institution solvency could occur
when prudential regulation would incorporate climate-related elements in determining capital
requirements, such as a green supportive factor or a brown penalizing factor (Boot et al., 2022).

2.1 Climate shocks

Climate change can lead to several types of adverse shocks to the financial system. We identify six
types of shocks, summarized in Figure 2. The first three shocks (abrupt transition, gradual
transition, and “hot house world” are well established in the climate stress testing literature and
distinguish themselves along two dimensions: either an orderly or disorderly transition and either
low or high global warming – the high warming scenario referred to as a “hot house world” (Steffen

3
As far as we are aware, direct effects have to date not been incorporated in CRST exercises. The effect of pro-environmental
preferences on bond prices is found to be limited to date (e.g., Zerbib, 2019).
 9

et al., 2018; Network for Greening the Financial System, 2022). We omit the scenario with an
orderly transition and low global warming as this scenario does not fall in the adverse category that
is suitable for climate stress testing. A fourth type of shock (climate-related disaster) is based on an
emerging literature that investigates scenarios related to the occurrence of one or more climate-
related natural disasters on financial institutions (e.g., Klomp, 2014; Schüwer, Lambert and Noth,
2019; Hallegatte et al., 2022). These shocks are relevant in a climate change context, since climate
change is expected to affect the frequency and intensity of natural disasters (Intergovernmental
Panel on Climate Change, 2022), including tropical cyclones (e.g., Knutson et al., 2010) and floods
(e.g., Hirabayashi et al., 2013; Arnell and Gosling, 2016). This is in particular the case for the most
extreme disasters, whose frequency is expected to increase substantially due to climate change
(Coronese et al., 2019).
Furthermore, we argue that financial sector losses can occur when financial sector agents
suddenly change their perception of current and future risks, which would be rapidly reflected in
today’s market prices of financial instruments. In the climate context, this could chiefly be for two
reasons. First, a shock could emanate directly from our changing perception of the state of the
global climate system. This could include the unexpected occurrence of climate tipping points
(Lenton, 2008; Bolton et al., 2020; Armstrong McKay et al., 2022) or changing insights from climate
science – for example when research would find that sea-level rise occurs more quickly than
previously thought. Bolton et al. (2020) label these tipping points and changing insights as a “Green
Swan” event. Second, a shock could emanate from the financial sector if it fails to continuously
incorporate the latest climate science and financial sector agents suddenly do so at some point in
time – for example due to increased awareness, a large natural disaster event, or strongly improved
climate risk data. We label the latter as a “Minsky-type” shock (Minsky, 1992). The main difference
between the two types of shocks is whether it is economic or financial state variables that change
suddenly. A Green Swan shock represents a sudden change in understanding of economic
fundamentals, while a Minsky-type shock emanates from a disconnect between economic
fundamentals and financial asset values that is suddenly corrected. While economic fundamentals
can change gradually under a Minsky-type scenario, financial shocks could be sudden when
investor sentiment changes. Such scenarios could especially occur in environments where there is
a lack of adequate data, understanding, and transparency to price and assess financial risks.
 10

Figure 2 – Classification of climate shocks

Abrupt transition Gradual transition

Sudden change in economic conditions due to climate Gradual changes in economic conditions due to
policies, technological or societal change. climate policies, technological or societal change.

Hot house world Climate-related disaster

Gradual changes in economic conditions due to Sudden economic damages resulting from climate-
gradual albeit severe changes in weather-patterns related natural disasters (e.g., typhoons, floods,
and sea-level rise. droughts).

Green Swan Minsky-type

Sudden reassessment of the value of real assets due Sudden repricing of financial assets based on climate-
to changing insights from climate science (e.g., on related factors due to changing sentiments and/or
sea-level rise) or tipping points occurring. better data availability.

3. CRST building blocks

In this section, we review the main underlying models used in CRST exercises, distinguishing
between economic vulnerability, financial vulnerability and stress test modeling steps in line with
the conceptual model provided in Figure 1. We specifically focus on the level of analysis and main
moderating variables that are relevant as part of the modeling. Table 1 provides an overview of the
models, which modeling step they are mainly used for (i.e., economic vulnerability, financial
vulnerability, or financial system impact), and their level of analysis, and the CRST exercises that
they are applied in. We discuss each model in turn in the next sections.
 11

Table 1 – Models used in CRST

Model Modeling step Level of analysis Applied in

Economic
1. Computable General Equilibrium (CGE) Macro or sector • World Bank Group (2021)
vulnerability
Economic • Vermeulen et al. (2018)
2. Macro-structural models Macro
vulnerability • Allen et al. (2020)
• Allen et al. (2020)
Economic
3. Integrated Assessment Models (IAMs) Macro • European Central Bank
vulnerability
(2022)
Economic
4. Firm valuation models Firm • Reinders et al. (2023)
vulnerability
Economic
5. Disaster risk models Location / asset • Hallegatte et al. (2022)
vulnerability
• Vermeulen et al. (2018)
• Allen et al. (2020)
6. Macro-financial (satellite) risk models Financial vulnerability Asset class • European Central Bank
(2022)
• Grippa and Mann (2020)
7. Structural credit risk models Financial vulnerability Firm / asset • Reinders et al. (2023)

8. Non-structural empirical models Financial vulnerability Firm / asset • Jung et al. (2021)
• Battiston et al. (2017)
Financial system • Vermeulen et al. (2018)
9. Financial impact models Bank
impact • World Bank Group (2021)
• Hallegatte et al. (2022)

3.1 Computable General Equilibrium (CGE) models

Computable General Equilibrium (CGE) models are economy-wide models that focus on the long-
term effects of policy changes. The focus of these models is on the longer-term effects of large-scale
reforms of the economy, such as subsidy reforms and carbon taxation (Burns et al., 2009). CGE
models typically integrate Input-Output (IO) tables or Social Accounting Matrices (SAMs) as a result
of which CGE models tend to have a high sectoral detail compared with macrostructural models.4
CGE models provide a richer analysis than using merely IO tables or SAMs, as they account for price
and behavioral changes endogenously while applying macroeconomic constraints such as the
supply of labor (Anvari et al., 2022). This makes CGE models especially suitable to analyze the
economic effects of carbon taxation across economic sectors in an economy.

4
Input-output (IO) models focus on interactions between different sectors in the economy, thereby providing insight in the
flow of goods as intermediate inputs for final consumption. They provide a static picture of the economy by using fixed-
coefficients for the cost structure of firms (e.g., Leontief, 1951, Ghosh, 1958).
 12

Firm and consumer behavior is usually assumed to be fully flexible, with each reacting to
changes in incentives in a way that is consistent with economic theory. Due to this assumption,
outcomes of CGE models tend to be most valid for longer time horizons. For shorter time horizons,
non-modelled frictions in the economy (e.g., in the labor market) may cause CGE models to
underestimate economic damages from a policy shock (such as a change in carbon tax regime). This
could also be the case due to supply-side constraints, such as having enough suitable areas to
deploy wind and solar power. Moreover, assumptions are needed on structural changes in the
economy due to product and process innovation, including those driven by technological progress.
This is especially relevant when modeling longer run transition scenarios, since economic costs and
gains related to decarbonization depend strongly on the availability and cost of low-carbon
alternatives, such as renewable energy sources.

3.2 Macrostructural models

Macrostructural models use econometrically estimated parameters to establish relations between
key economic variables, such as relative prices, changes in employment, unemployment, and
inflation. The underlying structure of macrostructural models is based on economic theory and
usually focuses on short-term disequilibrium behavior in the economy following an initial shock.
Usually, these models have a less rigid theoretical foundation than DSGE or CGE models (Blanchard,
2018). In comparison to CGE models, macrostructural models are better able to reflect frictions
such as delayed responses by households and industries to changing relative prices (Burns et al.,
2009). This addresses non-modelled frictions and associated potential underestimations of
economic impact by CGE models and makes macrostructural models especially suited to investigate
shock scenarios.
A key example of a macrostructural model is the National Institute Global Econometric
Model (NiGEM) model, which has been used by the Network for Greening the Financial System
(NGFS) to determine the evolution of economic variables that are not present within IAM
frameworks. NiGEM consists of linked individual country models. Country models are economy-
wide systems of dynamic equations. It is New Keynesian in structure in that it assumes agents with
rational expectations and there are nominal rigidities that slow adjustment process to external
shocks (Hantzsche et al., 2018). A main drawback of macrostructural models is that they usually do
 13

not have the same level of sectoral detail as CGE models, making them less suitable when
considering micro or sector level financial exposure data. Another limitation of these models is that
they are reduced-form models and not structural. They are based on historical relations which may
become invalid in the presence of large shocks such as structural changes in climate policies (Lucas,
1976; Anvari et al. 2022).

3.3 Integrated assessment models

Integrated Assessment Models (IAMs) encompass approaches that integrate knowledge from two
or more domains into a single framework. IAMs have been extensively used to model the interplay
between GHG emissions, changes in climate and sea level, and economic variables (e.g., Nordhaus,
1992; Tol, 2002). Many are developed to analyze the expected costs and benefits of climate policies
and determine optimal transition paths to reduce GHG emissions (Weyant, 2020). From a climate
risk stress testing perspective, IAMs provide a link between climate and economic outcomes,
including feedback loops from economic outcomes to climate variables over time (chiefly through
GHG emissions). This allows for the construction of decarbonization trajectories as well as expected
damages from climate change over time.
IAMs differ in their complexity and interconnections that they consider. This includes the
degree of technological detail, the degree of sectoral detail, the availability of mitigation
technologies and options, and the method by which they reach a solution for each time period
(Gambhir et al., 2019).5 One of their important uses has been within the context of the IPCC
assessment reports, to quantify the effect of climate mitigation in different shared socio-economic
pathways (SSPs) (Intergovernmental Panel on Climate Change, 2022). The focus of IAMs is on long
term horizons, up until 2050 and beyond (e.g., Rogelj, 2018). This makes them less suitable to
assess more adverse scenarios from a stress testing point of view, such as the sudden introduction
of climate policies or the occurrence of a natural disaster. IAMs have been criticized for using
unrealistic and unchallenged assumptions which could lead to an underestimation of the economic
costs due to climate change. These critiques concern the use of financial discount rates instead of

5
Six widely used models in global mitigation scenario analysis are IMAE, MESSAGE-GLOBIOM, AIM/CGE, GCAM,
REMIND-MAgPIE, and WITCH-GLOBIOM. Three of these models have also been used to develop the scenarios for the
NGFS (Network for Greening the Financial System, 2022).
 14

typically lower social discount rates (e.g., Dasgupta, 2008; Stern, 2008) and a systemic and gross
underestimation of damages from unmanaged climate change (Stern, 2018), including classifying a
large fraction (up to 90 percent) of GDP as unaffected by climate change (Keen, 2021).

3.4 Firm valuation models

Discounted Cash Flow (DCF) models are used in finance to value firms and other types of assets
(e.g., real-estate) by estimating the cash flows that they produce over their lifetime (e.g., Berk and
DeMarzo, 2020). Cash flows are then adjusted using a time dependent discount factor to account
for investor preferences (such as risk aversion and the time value of money). In this setting, the
impact of carbon pricing policies on firm valuation can be investigated by including a cost of carbon
as a negative cash flow. The estimated impact of this negative cash flow on firm valuation can then
be linked to a Merton structural credit risk model to estimate market value changes in the firm’s
equity and debt (Reinders et al., 2023). To estimate the effect of a carbon tax on firm value it is
necessary to include parameters on cost pass-through and firm-level emission abatement potential
(e.g., Fabra and Reguant, 2013; Smale et al., 2006). This methodology can be used to estimate effects
at asset level, hence increasing the precision of the analysis, but is limited to the direct effect of
carbon taxation hence potentially leading to an underestimation of total losses in financial
portfolios.

3.5 Disaster risk models

Catastrophe (CAT) modeling has been developed since the 1980s, mostly from a practical
perspective to manage property insurance risks. According to Mitchell-Wallace et al. (2017), most
catastrophe models adopt a modular approach that consists of a hazard module (simulating a set
of potential adverse events, such as hurricanes or floods), a vulnerability module (providing
damages as a function of the intensity of the hazard), an exposure module (providing exposure data
such as location, characteristics and value of objects), and a financial loss module (calculating for
example insured losses). This fits to our conceptual model in that catastrophe modeling combines
the full chain between climate events and financial loss for insurers.
 15

Typical CAT models do not consider the changes in disaster frequencies in the future due to
climate change since property insurance contracts are usually annually renewed. However, for
climate stress testing purposes, the hazard module of catastrophe risk models can be recalibrated
to represent a future climate condition (Ranger and Niehörster, 2012). The future path of climatic
hazards can be estimated using global or regional climate models, for example for hurricanes
(Bender et al., 2010) and floods (Winsemius et al., 2013). Furthermore, the standard outputs of CAT
models for insurance companies (e.g., estimated damages) can be combined with a macroeconomic
model to estimate a broader range of relevant economic state variables such as GDP and
unemployment over time (Hallegatte et al., 2022) making them suitable for stress testing financial
institutions other than insurers as well.
A main drawback of CAT models is the limited historical data that is available for calibration
of the model and making future projections, especially for the tail ends of the distribution. Although
it is possible to use Extreme Value Theory (EVT) to estimate the probability and magnitude of
disasters outside of the observed historical sample (e.g., Embrechts and Resnick, 1999), deep
uncertainty remains in projecting disaster risks into the future (Weitzman, 2009; Ranger and
Niehörster, 2012). Furthermore, most financial institutions other than insurance (for whom CAT
modeling is traditionally developed) do not have data available on the precise geographical
locations of the assets that they finance, which may include firms that have productive assets in
different locations.

3.6 Macro-financial (satellite) models

Since macroeconomic models usually do not include measures of financial risks, stress testing
involves the estimation of auxiliary or satellite models that link a measure of financial risk to
macroeconomic variables. Specifically, satellite models are used to determine credit losses and
various components of profit, including funding costs, under various scenarios (Jobst et al., 2013).
Those auxiliary or satellite models often take the form of (panel) regression models with commonly
employed methods being (i) structural econometric models, (ii) vector autoregressive (VAR)
models, and (iii) pure statistical approaches (Foglia, 2009).
An early example of such approach is given by Blanschke et al. (2001) who regress the
nonperforming loan (NPL) ratio against a set of macroeconomic variables including the nominal
 16

interest rate, the inflation rate, the change in real GDP, and the change in the terms of trade. A recent
example of a VAR approach is provided by Gamba et al. (2017) who estimate the proportion of loans
that is in a certain credit rating category for different asset classes. They include GDP growth, the
interest rate on mortgage and corporate loans, the house price index, and the unemployment rate
as independent variables. Since satellite models treat macroeconomic variables as exogenous this
implies that any feedback effects from a distressed financial sector to the economy are not taken
into account. Another drawback of this approach is that it implicitly assumes that climate-related
economic shocks have similar effects as the economic shocks that occurred in the past and were
used to perform the regression modeling. It hence has little specificity to climate-related shocks.

3.7 Structural credit risk models

Structural credit risk models have been developed to value financial instruments, including options
and debt positions (e.g., Scholes and Black, 1973; Merton, 1974). Structural factors that determine
the market value of debt include leverage, the interest rate, and asset value volatility. Merton’s key
insight is that equity can be viewed as a residual claim on assets after the debt has been repaid
(Merton, 1974). This implies that the equity holder has a call option on the value of the firm’s assets,
where the payoff is the maximum of (since a corporation has limited liability) and the value of firm’s
assets minus the face value of the debt. Conversely, the debt holder has a risk-free bond and is short
a put option of the firm’s assets. Overall, Merton’s contingent claims approach implies that a
negative asset valuation shock will affect the value of both equity and debt in a non-linear manner.
This is especially important when considering asymmetric shocks that affect some debt claims to a
higher extent than others, which is characteristic for both physical risk, where especially assets in
some geographic regions could be affected, and transition risk, where especially assets in certain
transition-sensitive industries could be affected (Battiston, 2017).
Structural credit risk models can be used to distribute firm-level asset valuation shocks to
the holders of equity and debt and can thereby be used for stress tests with highly granular (e.g.,
firm level) data. A main drawback of structural credit risk models is that their simplified structural
form is not always applicable to the portfolios under investigation and its assumptions (e.g., firm
asset value follows a random walk, firms have only one type of plain vanilla debt) do not always
hold, resulting in potentially biased estimates of credit risk. Specifically for secured loans, such as
 17

most European mortgages, the standard Merton model may overestimate potential losses and
adjustments to the model are needed (Reinders et al., 2023).

3.8 Non-structural empirical models

Besides the more traditional satellite models that are mainly used for credit risks there is a broader
class of empirical approaches that can inform financial sector stress testing. This usually involves
investigating a more “direct” relation between scenario variables and financial sector variables,
such as financial asset value or financial institution solvency. For example, Klomp (2014)
investigates to effect of natural disasters on the distance-to-default of banks and Scholtens and van
der Goot (2014) investigate the effect of carbon prices on stock prices of firms in different
industries. There are also studies investigating the effect of natural disasters on sovereign credit
ratings, which allows the inclusion of a sovereign bond channel in the stress testing (e.g., Standard
& Poor’s, 2019).
In general, these approaches allow for the estimation of elasticities between the
independent and dependent variables, which both may have several time lags. Empirical models
are mostly suitable to assess scenarios that have already occurred in the past and hence historical
events are available to estimate a direct relation. This is challenging for adverse transition risk
scenarios, where historically carbon price shocks are found to be incremental and relatively benign
in their impact on financial asset values (e.g., Scholtens and van der Goot, 2014).

3.9 Financial impact models

As a final step in most stress testing approaches, the estimated changes in financial sector state
variables are linked to financial institution specific variables to obtain a measure of overall impact.
According to Oura and Schumacher (2012), two main approaches are used: models that build on a
detailed analysis of balance sheets of individual institutions (sometimes called “fundamental
approaches”) and models based on summary default measures for individual assets, institutions,
or entire financial systems as embedded in market prices (such as stocks, bonds, and derivatives).
The fundamental approach consists mostly of accounting-based translation of the estimated impact
on financial sector state variables to metrics such as capital and liquidity ratios (e.g., Cihak, 2007;
Schmieder et al., 2011). Depending on the granularity of the balance sheet data, this assessment can
 18

be at the sectoral or individual financial institution level. The market-price based approach derives
implicit default probabilities using option price models (Gray, 2007). It can amongst others be used
to translate estimated impacts on financial sector state variables into expected or unexpected
losses at market value at the portfolio, institution, or sector level.
Besides the first-round (direct) impacts of climate scenarios, financial sector dynamics and
feedback loops to the real economy may lead to asset price revaluations and shocks to financial
institutions that are much larger than the initial shock. Three effects that have been identified in
the literature are amplification within financial networks (e.g., Acemoglu et al. 2015; Allen et al.,
2012; Battiston and Martinez-Jaramillo, 2018) also referred to as “financial contagion”, feedback
loops to the economy due to a reduction in available finance (credit crunch; see, for example, Silva
et al., 2018), and interactions between financial losses and sovereign credit spreads due to
sovereign contingent liabilities in the case of the default of financial institutions (in particular,
banks) also referred to as “doom loops” (e.g., Farhi and Tirole, 2018).

4. CRST modeling approaches

In this section, we identify and discuss four typical climate-financial modeling approaches, based
on the currently existing CRST exercises and available economic and financial models. One key
observation in the development of CRST is the emergence of more granular modeling approaches
next to the traditional macro-financial approaches, leveraging on the sector specific and/or location
specific impacts of most climate shocks. Taking a more micro approach at sector, firm, or asset level
allows for more precise allocation of risks to financial institutions based on their sectoral and
spatial asset exposures, which helps to identify those institutions that are most at risk. This micro
approach allows central banks and supervisory authorities to assess microprudential risks next to
the more macroprudential (systemic) ones. Furthermore, specific approaches have been developed
that are less structural in nature. For example, Battiston et al. (2017) largely set aside the need for
economic and financial modeling by assuming a full loss of financial value in selected climate-policy
relevant sectors. Finally, specifically for disaster risk shocks, stress test have been developed that
build on the disaster risk models traditionally employed in the (re)insurance industry. We describe
each modeling approach in turn.
 19

4.1. Macro-financial (traditional) approach

The macro-financial approach builds on a well-established approach to do climate stress testing
(e.g., Jobst et al., 2013; Ong and Jobst, 2020) and is characterized by the translation of climate shock
variables into macro-economic variables such as GDP, unemployment, and inflation. Differences in
those variables compared to a baseline scenario are then used to estimate (i) accounting based
financial risk measures, such as changes in nonperforming loans (NPLs) and/or (ii) market based
financial losses, such as changes in expected losses on loan portfolios or market value of tradable
securities. This second step is usually mostly non-structural in nature, using regression models to
estimate relations between economic and financial variables. The main models used in this
approach are macroeconomic models such as CGE models, IAMs, and macro-structural models.
Since macroeconomic variables usually cannot be structurally linked to financial outcomes, the
financial models used are typically empirical in nature (i.e., macro-financial satellite models).
Figure 3 depicts a typical macro-financial approach, with satellite models for bank profitability,
NPLs, and credit spreads.6

Figure 3 – Typical macro-financial approach

4.2. Micro-financial approach

The micro-financial approach is characterized by the translation of climate shock variables into
micro-economic state variables at firm or asset level, such as earnings, enterprise value, and
leverage. The main economic models used in this approach are valuation models, often based on

6
The depicted satellite models are not exhaustive and other mediating financial variables have been used as well (e.g., interest
rates, provisions, and stock prices).
 20

discounted cash flow (DCF) analysis. Differences in variables compared to the baseline scenario are
then used to estimate (i) accounting based financial risk measures, such as changes in
nonperforming loans (NPLs) and/or (ii) market based financial losses, such as changes in expected
losses on loan portfolios or market value of tradable securities. This second step can be both
structural (e.g., based on option pricing models) or non-structural in nature, using regression
models to predict financial risk parameters such as asset level probability of defaults. Figure 4
depicts a typical micro-financial approach.

Figure 4 – Typical micro-financial approach

4.3. Non-structural (direct/empirical) approach

This approach is characterized by treating the economic effects of a shock as a black box and
directly modeling the relation between the climate shock and the financial outcomes. Usually, these
approaches are empirical in nature (e.g., Jung et al., 2021, Klomp, 2014). However some studies
investigate worst-case scenarios where it is assumed that assets in certain sectors lose their value
completely (e.g., Battiston, 2017). Since very few mediating and moderating variables are used, this
approach is mostly suitable to stress test events that have already occurred in the past, such as
large natural disasters. In the non-structural approach, it is not clear which economic and/or
financial channels are affected under the climate scenario. Some studies have attempted to estimate
the effect of carbon taxation on equity prices, however the investigated shock size has been limited
so far (e.g., Scholtens and van der Goot, 2014). Figure 5 depicts a typical non-structural approach,
with dependent variables being either financial asset prices or financial institution viability
indicators (such as the capital adequacy ratio).
 21

Figure 5 – Typical non-structural approach

4.4. Disaster risk approach

The disaster risk approach is characterized by the linking of disaster risk models used in the
(re)insurance industry to financial sector outcomes, through variables such as economic damage
and total factor productivity (TFP). These variables can in turn be linked to insurance liabilities, or
linked to non-insurance financial variables such as NPLs and provisions. Since large natural
disasters have occurred in the past, the relation between economic and financial variables can be
estimated empirically, and there are some studies that have used theoretical and computational
modeling to assess future damages under climate change scenarios (SFC, 2020; Zhou, Endendijk,
and Botzen, 2023). It is potentially also possible to link damage estimates at an asset level in a more
structural way to the value of financial assets, although we are not aware of any existing published
CRST study that employs such an approach. Figure 6 depicts a typical disaster risk approach, with
the top channel through insurance liabilities reflecting traditional insurance catastrophe stress
tests (e.g., EIOPA, 2022).

Figure 6. Typical disaster risk approach.

4.5 The structural form of the climate-financial relation

Part of the complexity of CRST lies in getting the functional form right regarding the propagation of
climate shocks to financial sector outcomes. As shown by the complexity of the models presented
in the previous sections, the shape of this relation is dependent on a range of moderating variables,
which could make outcomes both context and time dependent (i.e., empirical estimates of
 22

elasticities between climate variables and financial sector variables may not be universal). The
structural models that we reviewed point to several relevant moderating variables, which include
firm level characteristics (such as adaptability and the potential to abate GHG emissions), market
structure, financial asset characteristics (such as duration, leverage, and seniority of the
instrument), and business model characteristics of financial institutions (such as asset composition
and the adaptability of business models over time). Figure 7 provides a (non-exhaustive) overview.
A crucial component of the climate-financial relation are feedback loops. Figure 7 highlights
the feedback loops within the financial system (intra-financial), from the financial system to the
economy (macro-financial) and from the economy to climate risk (climate-economic). These
feedback loops are endogenous and may amplify the initial shock, as happened during the Global
Financial Crisis of 2008-2009. Current CSRT modelling approaches do not include these important
feedback loops.

Figure 7 – Moderating variables in the climate-financial relation

5. CRST exercises to date

In this section we review the main system-wide CRST exercises that have been conducted to date.
Most of these are conducted by central banks, financial supervisors, international organizations,
and academics (e.g., Battiston et al., 2017; Vermeulen et al., 2018; Allen et al., 2020; World Bank
 23

Group, 2021; European Central bank, 2022; Hallegatte et al., 2022; Reinders et al., 2023). CRST
exercises can be either bottom-up, when financial institutions make their own assessments (often
using different methodologies, but similar scenarios and assumptions) after which results are
aggregated, or top-down, when the assessment methodology is harmonized and data is obtained
from public sources or from participating financial institutions (Ong and Jobst, 2020). Since
methodologies are only harmonized for top-down exercises, we focus on this subset of all system-
wide CRST exercises. We collect and compare data on scope (i.e., financial institutions and asset
classes that are covered), the type(s) of climate shock assessed, modeling variables, and main
outcomes. We furthermore classify each exercise according to the four modeling approaches
identified in section 4 (macro-financial, micro-financial, non-structural, disaster risk) and the type
of model used.
Results are provided in Table 2. All CRST exercises include banks in the analysis, with two
including other financial institutions such as insurers and pension funds. More than half of the
investigated studies exclusively focus either on only the loan portfolio or only the equity portfolio
of financial institutions. Only Vermeulen et al. (2018) and Allen et al. (2020) assess a complete set
of equity, bond, and loan portfolios using structural modeling. Jung et al. (2021) investigate the full
impact of climate shocks on banks’ expected capital shortfall but only use historical shocks (e.g., the
fossil-fuel price collapse in 2020). By not assessing all asset classes, several studies provide partial
results and hence are likely to underestimate system-wide losses in adverse climate shock
scenarios. This is in particular the case for Battiston et al. (2017) who solely assess the impact on
equity exposures for European banks, while more than 90 percent of assets of European banks
consists of loans (Reinders et al., 2023). All top-down CRST exercises to date have assumed that
balance sheet exposures remain constant during the stress test horizon.
In terms of shocks investigated, six exercises examine abrupt or sudden transition scenarios.
Usually, these scenarios are operationalized by defining a carbon price path, with sharply
increasing carbon prices during the assessed time horizon (for example, during the next five to ten
years). Two studies investigate the impact of climate-related disasters on the financial sector
(World Bank Group, 2021; Hallegatte, 2022). Furthermore, two studies investigate more gradual
scenarios that unfold over a longer-term horizon, including gradual transitions and hot house world
scenarios (Allen et al., 2020; European Central Bank, 2022). The financial system impact of the
 24

latter two studies is however small, chiefly because traditional financial sector stress testing
typically has a time horizon of 3 to 5 years – in line with the limited duration of most debt
instruments (e.g., Bolton et al., 2020). However, while climate change may be gradual, the Green
Swan and Minsky-type of climate shocks would likely lead to abrupt changes in financial markets
in the short term (e.g., a severe price decline of coastal assets due to news on accelerated melting
of the Greenland ice sheet). To our knowledge, there are nevertheless to date no top-down CRST
exercises that have assessed such scenarios. We suspect that this could be due to a lack of suitable
models and difficulties in assigning probabilities to severe scenarios occurring (in order to credibly
claim that these are “severe but plausible” scenarios).

Table 2 – Overview of top-down CRST exercises

CRST Scope Climate Economic Financial Financial Classification Type of models
exercise shock state state system used
variables variables impact
variable

Battiston • Banks Abrupt N/A N/A Fraction of Non- Financial impact

et al. (primarily) transition portfolio structural models (for
(2017) • Equity (full loss of exposed second round
value) effects)

Vermeulen • Banks, Abrupt GDP Credit risk Financial Macro- Macro-

et al. insurers, transition Unemployment spread losses as financial structural model
(2018) pension (100 USD Risk-free (bonds) percentage of Macro-financial
funds carbon tax) interest rate Expected total assets (satellite) risk
• Equity, Equity indices losses (between 1 model
bonds, (level) (loans) and 11 Financial impact
loans percent) model

World • Banks Abrupt Sectoral value Non- Capital Macro- Computable

Bank • Loans, transition added performing adequacy ratio financial General
Group sovereign Climate- Economic loans (between – Disaster risk Equilibrium
(2021) bonds related damage and 5 Financial impact
disaster percentage model
(flood) points
decrease)
 25

Table 2 (cont.)– Overview of top-down CRST exercises

CRST Scope Climate Economic Financial Financial Classification Type of models
exercise shock state variables state system used
variables impact
variable

Allen et al. • Banks, Abrupt GDP, inflation, Probability N/A Macro- Macro-structural
(2020) insurers transition unemployment of default, financial model
• Equity, Gradual (12 total) market Integrated
debt, loans transition value of Assessment
equity Model
Macro-financial
(satellite) risk
models

Reinders • Banks Abrupt Firm value Market Market value Micro- Firm valuation
et al. • Equity, transition value of losses financial model
(2023) corporate (100-200 equity and Structural credit
loans EUR carbon debt risk model
tax)

European • Banks Gradual GDP, firm Probability Expected Micro- Integrated

Central • Bonds, transition earnings and of default, losses financial Assessment
Bank corporate (disorderly) leverage loss given Macro- Model
(2022) loans Hot house default, financial
world bond credit
spread

Hallegatte • Banks Climate- Economic Earnings Capital Disaster risk Disaster risk
et al. • Loans related damage Market adequacy Macro- model
(2022) disaster Total Factor value of ratio financial Financial
(typhoon) Productivity equity and (between 1 impact model
(TFP) debt and 10
NPLs percentage
points
decrease)

Grippa • Banks Abrupt Interest Loan losses N/A Micro- Macro-financial

and Mann • Corporate transition Coverage Ratio (increase of financial (satellite) risk
(2020) debt (ICR) 0.8-0.9 ppt) Non- models
Oil revenues structural
(SVAR)

Jung et al. • Large Abrupt N/A N/A CRISK Non- Non-structural

(2021) global transition (expected structural empirical model
banks (2020 fossil- capital
fuel price shortfall in
collapse) climate stress
scenario)
 26

In terms of modelling, we find that there is no single comprehensive model used for CRST
and, as a result, most findings are partial in nature (i.e., not covering all relevant transmission
channels). In all but one CRST exercise the economic, financial, and financial institution variables
are modelled in separate steps and then linked through economic and financial state variables.
Given the limited number of state variables typically included in CRST exercises, it is highly likely
that relevant transmission channels are omitted (for example, changes in risk-free interest rates).
More integrated structural models are conceivable but, to our knowledge, have not been developed
and used yet for CRST. For instance, it has been argued that Dynamic Stochastic General
Equilibrium (DSGE) models can be modified to include climate change and economic and financial
sector dynamics all in one model, but these models do to date not exist (Arndt et al., 2020, Anvari
et al., 2022). Stacking of multiple models may furthermore increase model-error while using
different models leads to lower comparability between the outcomes of different CRST exercises.

5.1 Shortcomings
Our review points to several shortcomings that need to be addressed with the further development
of CRST. Taken together, these shortcomings could lead to underestimation of the impact of climate
chocks on financial institution viability and hence also the potential of climate shocks to lead to
financial instability. First, not all relevant climate shocks are assessed within a CRST context. For
example, recent climate models indicate increasing risks of tipping points (e.g., Armstrong McKay
et al., 2022). The latest science on climate shocks should be used in CRST, which implies the
development of more adequate severe but plausible “green swan” scenarios beyond those provided
by the traditional IAMs. Furthermore, increased attention should go to potential Minsky-type
shocks emanating in the financial sector itself, if climate sentiment among investors and lenders
changes suddenly. Second, CRST is highly reliant on traditional macro-financial stress testing.
However, due to the asymmetric and non-linear relation between climate shocks and financial
institution outcomes there is a high risk of model misspecification. Most of the traditional models
discussed in this paper make strong assumptions, that are compounded when individual models
are linked to each other. More granular modeling approaches should be developed that refine
assumptions and provide a complementary angle to the outcome of macro-financial stress tests.
Third, a fundamental shortcoming of current CSRT modeling approach is the lack of modeling
 27

feedback loops, which may amplify the impact of climate shocks. This calls for much more
integrated modeling approaches, which captures better the non-linear relationship between
climate, economic, and financial variables. Fourth, nearly all existing CRST exercises are partial in
nature. Specifically, this relates to the limited availability of data and models to cover all asset
classes (e.g., loans, bonds, and equity), all relevant risk channels (e.g., changes in risk-free interest
rates and/or risk premiums), and all relevant financial institutions (e.g., banks, insurers, pension
funds). This is an issue that, among others, should be addressed by improving data availability.

Table 3 – CRST shortcomings

Climate shock Explanation

• No Green Swan and Minsky • Lack of understanding and assessment of “Green Swan” and
shocks Minsky-type scenarios

Vulnerability modeling Explanation

• Overreliance on macro models • Current climate stress testing is highly reliant on IAMs and
traditional macroeconomic stress testing

• Ignoring feedback loops • Most studies ignore feedback effects on the financial sector,
economy, and climate

• Partial set-up • Exercises are partial in nature, covering only limited sets of
causal channels and asset classes

6. Conclusion and discussion

CRST is a developing field with, so far, a wide variety of approaches to model the relation between
climate shocks and financial sector outcomes. Conceptually, this relation runs primarily through
economic and financial state variables. We find that almost all CRST exercises follow this approach.
There are however substantial differences in the type of climate shocks that are investigated (or
potentially could be investigated, but have not been yet) and the way that climate, economic, and
financial models are connected. We identify four types of modeling approaches: (a) traditional
macro-financial; (b) micro-financial; (c) non-structural; and (d) disaster risk. We also classify
 28

climate shocks into six types: (i) abrupt transition; (ii) gradual transition; (iii) hot house world; (iv)
climate-related disaster; (v) green swan; and (vi) Minsky-type. The latter two have not been
investigated in any CRST exercises that we are aware of, however could be especially relevant from
a financial stability perspective. Further shortcomings of CRST exercises to date include a high
reliance on the traditional macro-financial approach and IAMs, the lack of modeled feedback effects,
and the partial nature of the assessments (i.e., not covering all causal channels and asset classes).
We see several avenues for the future development of climate-financial modeling
approaches, summarized in Table 4. First, we think it is important to develop CRST exercises that
investigate the potentially most damaging scenarios (e.g., a “green swan” event or rapid repricing
of financial assets). Especially from a financial stability perspective it is important to better
understand unlikely but severe outcomes. Second, the next generation of climate-financial models
should include feedback loops, which can amplify the impact of initial climate shocks within the
economic and financial systems. Third, we suggest to further develop micro-based approaches that
allow for better sectoral and spatial disaggregation. This is especially relevant for micro-prudential
supervision and adequate pricing of climate-related financial risks. Fourth, disaster risk
approaches could be expanded to connect catastrophe models to financial outcomes other than
those for (re)insurance liabilities. This would allow disaster risk scenarios to be applied to banks
and other institutional investors.

Table 4 – Avenues for future research

Climate shock

• Improve understanding of tail risks related to a changing climate (e.g. tipping points)
• Assess plausible but severe “green swan” and Minsky-type scenarios on the economy
and financial sector

Vulnerability modeling

• Develop integrated modeling approaches that capture a comprehensive set of

feedback loops within the financial sector, and from the financial sector to the
economy and climate
 29

• Develop microeconomic approaches to climate stress testing (to assess impacts on

specific economic sectors and regions)
• Develop disaster risk stress tests for financial institutions other than insurers
(building on existing disaster risk models)
 30

References
Acemoglu, B. D., Aghion, P., Bursztyn, L., & Hemous, D. (2012). The environment and directed
technical change. American Economic Review, 102(1), 131–166.
Acemoglu, D., Ozdaglar, A., & Tahbaz-Salehi, A. (2015). Systemic risk and stability in
financial networks. American Economic Review, 105(2), 564-608.
Advisory Scientific Committee (ASC) (2016), ‘Too Late, Too Sudden: Transition to a Low-Carbon
Economy and Systemic Risk’, Report No. 6 of the Advisory Scientific Committee of the
European Systemic Risk Board, Frankfurt.
Alessi, L., Ossola, E., & Panzica, R. (2021). What greenium matters in the stock market? The
role of greenhouse gas emissions and environmental disclosures. Journal of Financial
Stability, 54, 100869.
Allen, F., Babus, A., & Carletti, E. (2012). Asset commonality, debt maturity and systemic risk. Journal
of Financial Economics, 104(3), 519-534.
Allen, T., Dees, S., Caicedo Graciano, C. M., Chouard, V., Clerc, L., de Gaye, A., ... & Vernet, L. (2020).
Climate-related scenarios for financial stability assessment: an application to France.
Banque de France Working Paper. No. 774.
Alogoskoufis, S., Dunz, N., Emambakhsh, T., Hennig, T., Kaijser, M., Kouratzoglou, C., and Salleo, C.
(2021). ECB economy-wide climate stress test: Methodology and results ECB Occasional
Paper. (No. 281).
Anvari, V., Arndt, C., Hartley, F., Makrelov, K., Strezepek, K., Thomas, T., Gabriel, S., & Merven, B.
(2022). A climate change modeling framework for financial stress testing in Southern Africa.
South African Reserve Bank Working Paper Series, WP/22/09.
Autorité de contrôl prudential et de resolution (2021). A first assessment of financial risks
stemming from climate change: The main results of the 2020 climate pilot exercise. ACPR
Analyzes et Synthéses No. 122-2021.
Arndt, C., Loewald, C., & Makrelov, K. (2020). Climate change and its implications for central banks
in emerging and developing economies. Economic Research and Statistics Department, South
African Reserve Bank.
Arnell, N. W., & Gosling, S. N. (2016). The impacts of climate change on river flood risk at the global
scale. Climatic Change, 134(3), 387-401.
 31

Armstrong McKay, D. I., Staal, A., Abrams, J. F., Winkelmann, R., Sakschewski, B., Loriani, S., ... &
Lenton, T. M. (2022). Exceeding 1.5° C global warming could trigger multiple climate tipping
points. Science, 377(6611), eabn7950.
Babiker, M. H. (2005). Climate change policy, market structure, and carbon leakage. Journal of
International Economics, 65(2), 421-445.
Bank of England (2022). Results of the 2021 Climate Biennial Exploratory Scenario (CBES). 24 May
2022. London.
Barth, J. R., Sun, Y., and Zhang, S. (2019). Banks and natural disasters. SSRN Electronic Journal.
Batten, S., Sowerbutts, R., & Tanaka, M. (2016). Let’s talk about the weather: The impact of climate
change on central banks. Bank of England, 603.
Battiston, S., Mandel, A., Monasterolo, I., Schütze, F., & Visentin, G. (2017). A climate stress-
test of the financial system. Nature Climate Change, 7(4), 283–288.
Battiston, S., & Martinez-Jaramillo, S. (2018). Financial networks and stress testing: Challenges and
new research avenues for systemic risk analysis and financial stability implications.
Journal of Financial Stability, 35, 6-16.
Baudino, P., & Svoronos, J. P. (2021). Stress-Testing Banks for Climate Change: a Comparison of
Practices. Bank for International Settlements, Financial Stability Institute.
Bender, M. A., Knutson, T. R., Tuleya, R. E., Sirutis, J. J., Vecchi, G. A., Garner, S. T., & Held, I. M. (2010).
Modeled impact of anthropogenic warming on the frequency of intense Atlantic hurricanes.
Science, 327(5964), 454-458.
Berk, J. and P. DeMarzo (2020), Corporate Finance, 5th edition, Pearson Education, Boston.
Blanchard, O. (2018). On the future of macroeconomic models. Oxford Review of Economic Policy,
34(1-2), 43-54.
Bolton, P., Després, M., Pereira da Silva, L. A., Samama, F., & Svartzman, R. (2020). The Green Swan.
Bank for International Settlements, Basel.
Boot, A., Grünewald, S., Schoenmaker, D., & van Tilburg, R. (2022). Climate risks are real and need
to become part of bank capital regulation. VoxEU column, 7 December.
Borio, C., Drehmann, M., & Tsatsaronis, K. (2014). Stress-testing macro stress testing: does it live up
to expectations?. Journal of Financial Stability, 12, 3-15.
 32

Bovenberg, A. L., Goulder, L. H., & Gurney, D. J. (2005). Efficiency costs of meeting industry-
distributional constraints under environmental permits and taxes. RAND Journal of
Economics, 36(4), 951-971.
Breeden S, & Hauser A (2019) A climate Minsky moment. In: Global Public Investor Report by the
Official Monetary and Financial Institutions Forum, OMFIF.
Cahen-Fourot, L., Campiglio, E., Godin, A., Kemp-Benedict, E., & Trsek, S. (2021). Capital stranding
cascades: The impact of decarbonisation on productive asset utilisation. Energy Economics,
103, 105581.
Campiglio, E., Dafermos, Y., Monnin, P., Ryan-Collins, J., Schotten, G., & Tanaka, M. (2018). Climate
change challenges for central banks and financial regulators. Nature Climate Change, 8(6),
1–13.
Cihák, M. M. (2007). Introduction to applied stress testing (Issues 7–59). International Monetary
Fund, Washington, D.C.
Coronese, M., Lamperti, F., Keller, K., Chiaromonte, F., & Roventini, A. (2019). Evidence for sharp
increase in the economic damages of extreme natural disasters. Proceedings of the National
Academy of Sciences, 116(43), 21450-21455.
Dasgupta, P. (2008). Discounting climate change. Journal of Risk and Uncertainty, 37(2), 141-
169.
Embrechts, P., Resnick, S. I., & Samorodnitsky, G. (1999). Extreme value theory as a risk
management tool. North American Actuarial Journal, 3(2), 30-41.
European Central Bank (2022). 2022 Climate Risk Stress Test. July 2022, Frankfurt.
European Insurance and Occupational Pensions Authority (2022). Methodological Principles of
Insurance Stress Testing – Climate Change Component. EIOPA-BOS-21/579. Frankfurt.
Fabra, N., & Reguant, M. (2013). Pass-through of emission costs in electricity markets.
American Economic Review, 104(9), 2872–2899.
Farhi, E., & Tirole, J. (2018). Deadly embrace: Sovereign and financial balance sheets doom loops.
Review of Economic Studies, 85(3), 1781-1823.
Foglia, A. (2009). Stress Testing Credit Risk: A Survey of Authorities’ Approaches. International
Journal of Central Banking. September 2009.
 33

Gamba, S., Jaulín, O., Lizarazo, A., Mendoza, J. C., Morales, P., Osorio, D., & Yanquen, E. (2017). SYSMO
I: a systemic stress model for the Colombian financial system. Borradores de Economía,
(1028).
Gambhir, A., Butnar, I., Li, P. H., Smith, P., & Strachan, N. (2019). A review of criticisms of integrated
assessment models and proposed approaches to address these, through the lens of BECCS.
Energies, 12(9), 1747.
Ghosh, A. (1958). Input-output approach in an allocation system. Economica, 25(97), 58-64.
Gray, D. F., Bodie, Z., & Merton, R. C. (2007). New Framework for Measuring and Managing
Macrofinancial Risk and Financial Stability. NBER Working Paper, w13607.
Grippa, P., & Mann, S. (2020). Climate-Related Stress Testing: Transition Risks in Norway. IMF
Working Paper. No. 2020/232. International Monetary Fund, Washington, D.C.
Hallegatte, S., Lipinsky, F., Morales, P., Oura, H., Ranger, N., Regelink, M. G. J., & Reinders, H. J.
(2022). Bank Stress Testing of Physical Risks under Climate Change Macro Scenarios:
Typhoon Risks to the Philippines. IMF Working Paper. WP/22/163.
Hantzsche, A., Lopresto, M., & Young, G. (2018). Using NiGEM in uncertain times: Introduction and
overview of NiGEM. National Institute Economic Review, 244.
Henry, J., Kok, C., Amzallag, A., Baudino, P., Cabral, I., Grodzicki, M., ... & Żochowski, D. (2013). A
macro stress testing framework for assessing systemic risks in the banking sector. ECB
Occasional Paper, No. 152.
Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., ... & Kanae, S.
(2013). Global flood risk under climate change. Nature Climate Change, 3(9), 816-821.
Intergovernmental Panel on Climate Change (2022). Climate Change 2022: Impacts, Adaptation and
Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK and
New York, USA.
Jobst, M. A. A., Ong, M. L. L., & Schmieder, M. C. (2013). A framework for macroprudential bank
solvency stress testing: Application to S-25 and other G-20 country FSAPs. International
Monetary Fund, Washington, D.C.
Jung, H., Engle, R. F., & Berner, R. (2021). Climate stress testing. FRB of New York Staff Report, (977).
 34

Klomp, J. (2014). Financial fragility and natural disasters: An empirical analysis. Journal of Financial
Stability, 13, 180-192.
Knutson, T. R., McBride, J. L., Chan, J., Emanuel, K., Holland, G., Landsea, C., ... & Sugi, M. (2010).
Tropical cyclones and climate change. Nature Geoscience, 3(3), 157-163.
Koch, N., & Bassen, A. (2013). Valuing the carbon exposure of European utilities. The role of
fuel mix, permit allocation and replacement investments. Energy Economics, 36, 431-443.
Lakdawalla, D., & Zanjani, G. (2012). Catastrophe bonds, reinsurance, and the optimal
collateralization of risk transfer. Journal of Risk and Insurance, 79(2), 449-476.
Lenton, T., H. Held, E. Kriegler, J. W. Hall, W. Lucht, S. Rahmstorf, and J. Schellenhuber. 2008. Tipping
elements in the Earth’s climate system. Nature 105(6): 1786–93.
Leontief, W. W. (1951). The structure of American economy, 1919-1939: an empirical
application of equilibrium analysis (No. HC106. 3 L3945 1951).
Merton, R. C. (1974). On the pricing of corporate debt: The risk structure of interest rates. Journal
of Finance, 29(2), 449–470.
Miller, R. E., & Blair, P. D. (2009). Input-output analysis: foundations and extensions. Cambridge
University Press, Cambridge.
Minsky H (1992) The Financial Instability Hypothesis, The Jerome Levy Economics Institute of Bard
College Working Paper. No. 74, May.
Mitchell-Wallace, K., M. Jones, J. Hillier and M. Foote (2017), Natural Catastrophe Risk Management
and Modeling, John Wiley & Sons, New Jersey.
Monasterolo, I. (2020). Climate change and the financial system. Annual Review of Resource
Economics, 12, 299-320.
Morana, C. and Sbrana, G., (2019). Climate change implications for the catastrophe bonds market:
An empirical analysis. Economic Modeling, 81, pp.274-294.
Network for Greening the Financial System (2020). Overview of Environmental Risk Analysis
by Financial Institutions. Paris.
Network for Greening the Financial System (2022). NGFS Climate Scenarios for central banks
and supervisors. Paris.
Nieto, M. J. (2019). Banks, climate risk and financial stability. Journal of Financial Regulation and
Compliance, 27(2), 243–262.
 35

Nordhaus, W. D. (1992). An optimal transition path for controlling greenhouse gases. Science,
258(5086), 1315–1319.
Ong, L. L., & Jobst, A. (2020). Stress Testing: Principles, Concepts, and Frameworks. International
Monetary Fund, Washington, D.C.
Oura, H., & Schumacher, L. B. (2012). Macrofinancial stress testing-principles and practices.
International Monetary Fund Policy Paper.
Ranger, N., & Niehörster, F. (2012). Deep uncertainty in long-term hurricane risk: scenario
generation and implications for future climate experiments. Global Environmental Change,
22(3), 703-712.
Reinders, H. J., Schoenmaker, D., & van Dijk, M. A. (2023). A finance approach to climate stress
testing. Journal of International Money and Finance,131, 102797.
Rogelj, J., Popp, A., Calvin, K. V, Luderer, G., Emmerling, J., Gernaat, D., Fujimori, S., Strefler, J.,
Hasegawa, T., & Marangoni, G. (2018). Scenarios towards limiting global mean temperature
increase below 1.5 C. Nature Climate Change, 8(4), 325.
Standard & Poor’s. 2019, “Credit Trends: The Cost of a Notch.” March 16, S&P Global Ratings.
Schmieder, M. C., Puhr, M. C., & Hasan, M. (2011). Next generation balance sheet stress testing.
International Monetary Fund.
Scholes, M., & Black, F. (1973). The pricing of options and corporate liabilities. Journal of Political
Economy, 81(3), 637-654.
Scholtens, B., & van der Goot, F. (2014). Carbon prices and firms' financial performance: an industry
perspective. Carbon Management, 5(5-6), 491-505.
Schüwer, U., Lambert, C., & Noth, F. (2019). How do banks react to catastrophic events? Evidence
from Hurricane Katrina. Review of Finance, 23(1), 75-116.
Silva, T. C., da Silva Alexandre, M., & Tabak, B. M. (2018). Bank lending and systemic risk: A financial
real sector network approach with feedback. Journal of Financial Stability, 38, 98-118.
Slijkerman, J.F., de Vries, C., & Schoenmaker, D. (2013). Systemic Risk and Diversification across
European Banks and Insurers. Journal of Banking & Finance, 37(3), 773-785.
Smale, R., Hartley, M., Hepburn, C., Ward, J., & Grubb, M. (2006). The impact of CO2emissions trading
on firm profits and market prices. Climate Policy, 6(1), 31–48.
 36

Steffen, W., Rockström, J., Richardson, K., Lenton, T. M., Folke, C., Liverman, D., ... & Schellnhuber, H.
J. (2018). Trajectories of the Earth System in the Anthropocene. Proceedings of the National
Academy of Sciences, 115(33), 8252-8259.
Stern, N. (2008). The economics of climate change. American Economic Review, 98(2), 1-37.
Stern, N. (2018). Public economics as if time matters: Climate change and the dynamics of policy.
Journal of Public Economics, 162, 4-17.
Tol, R. S. J. (2002). Estimates of the Damage Costs of Climate Change - Part I: Benchmark Estimates.
Environmental and Resource Economics 211: 47–73.
Vermeulen, R., Schets, E., Lohuis M., Kölbl, B., Jansen, D-J., & Heeringa, W. (2018). An energy
transition risk stress test for the financial system of the Netherlands. De Nederlandsche Bank
Occasional Studies, Volume 16 – 7.
Weitzman, M. L. (2009). On modeling and interpreting the economics of catastrophic climate
change. Review of Economics and Statistics, 91(1), 1-19.
Weyant, J. (2017). Some Contributions of Integrated Assessment Models of Global Climate Change.
Review of Environmental Economics and Policy, 11(1), 115-137.
Winsemius, H. C., Van Beek, L. P. H., Jongman, B., Ward, P. J., & Bouwman, A. (2013). A framework
for global river flood risk assessments. Hydrology and Earth System Sciences, 17(5), 1871-
1892.
World Bank Group (2021). Not-So-Magical Realism: A Climate Stress Test of the Colombian
Banking System (English). Equitable Growth, Finance and Institutions Insight. Washington,
D.C.
Zerbib, O. D. (2019). The effect of pro-environmental preferences on bond prices: Evidence from
green bonds. Journal of Banking & Finance, 98, 39-60.
Zhou, F., Endendijk, T., Botzen, W.J.W. (2023) A review of the financial sector impacts of risks
associated with climate change. Annual Review of Resource Economics. In Press.