Journal of Financial Stability 69 (2023) 101190

Contents lists available at ScienceDirect

Journal of Financial Stability

journal homepage: www.elsevier.com/locate/jfstabil

Climate risks and financial stability: Evidence from the European financial
system
Miia Chabot ∗, Jean-Louis Bertrand
ESSCA School of Management, 49003 Angers, France

ARTICLE INFO ABSTRACT

JEL classification: Climate-related risks have become a major concern for financial regulators and can pose a significant threat to
G21 financial stability. In this paper, we first propose a theoretical framework for the transmission of climate risks
E44 to financial institutions and the financial system. We then estimate the influence of physical and transition
E58
risks on the European financial system through bank-level and system-wide measures of financial stability. We
Q51
find that Scope 3 greenhouse gas emissions, chronic and acute climate risks negatively affect financial stability
Q54
at both the financial institution and system levels. Temperature anomalies, heat waves, wildfires and droughts
Keywords: are among the most significant risks. As Europe warms twice as fast as the rest of the world, our theoretical
Financial stability
and empirical results urge regulators to mandatorily require the assessment and disclosure of corporate climate
Physical risks
risks to allow banks to adjust their prudential capital requirements.
Transition risk
Climate change

1. Introduction financial impact on economic activity and financial assets of a changing

climate, which includes both gradual changes in weather patterns and
Climate change is one of the most important challenges facing the climate variability (termed ‘‘chronic’’), and extreme weather events
world’s economies in the coming decades (WEF, 2022; Ranger et al., (termed ‘‘acute’’). Liability risks are related to financial claims that
2022). The conclusions of the sixth Intergovernmental Panel on Climate could occur decades from now, provided that parties who have suffered
Change report (IPCC, 2021) are unequivocal: climate change is global, loss or damage from the effects of climate change are able to seek
rapid and intensifying. The speed at which the climate is changing is
compensation from those they hold legally responsible (Batten et al.,
unprecedented in thousands of years. Human activities and greenhouse
2016; Monasterolo, 2020; ECB, 2020; BIS, 2021; NGFS, 2020).
gas (GHG) emissions in particular are causing an increase in the Earth’s
A significant body of literature has been developed by both aca-
average surface temperature, which in turn leads to increased climate
variability and more frequent and intense extreme climate events. demics and regulators to investigate these risks and their consequences
Europe is particularly exposed as temperature is rising twice as fast as on the financial system. Research has initially focused on transition
in the rest of the world.1 risk (Bressan and Romagnoli, 2021). The measurement of transition
The transition away from fossil fuels, whether driven by compelling risk as it applies to financial institutions is largely based on the level
climate policies, technology, or consumer preferences, creates opportu- of GHG emissions associated to bank lending. Ehlers et al. (2022)
nities for some industries. Transition however poses significant risks for find that despite material risks faced by some borrowers, the overall
the many other industries that depend on fossil fuels, causing potential carbon footprint, including indirect emissions, does not seem to be
losses in profits and value for the companies involved and the banks fully priced. Kacperczyk and Peydró (2021) and Reghezza et al. (2021)
that finance them. The Financial Stability Board (FSB) was tasked observe that firms with high emissions receive less bank credit and
with clarifying the short- and long-term risks to financial stability and have to reduce their investments. In addition, a rapidly growing body
considering measures to mitigate them. Three risk categories were of literature based on stress tests and IPCC GHG emission scenarios
proposed by Carney (2015): transition, physical, and liability risks. highlights the potential materiality of transition risk for the financial
Transition risk refers to the financial risks posed by the implementation
system (Battiston et al., 2017; Vermeulen et al., 2019; Reinders et al.,
of climate policies and regulations, both directly on high-emitting com-
2020; NGFS, 2019; Allen et al., 2020; Roncoroni et al., 2021). However,
panies and indirectly on their lenders and investors. Physical risk is the

∗ Corresponding author.
E-mail addresses: miia.chabot@essca.fr (M. Chabot), jean-louis.bertrand@essca.fr (J.-L. Bertrand).
1
Temperatures in Europe increase more than twice global average. WMO (2022). Press release No. 02112022.

https://doi.org/10.1016/j.jfs.2023.101190
Received 26 April 2022; Received in revised form 23 October 2023; Accepted 7 November 2023
Available online 10 November 2023
1572-3089/© 2023 Elsevier B.V. All rights reserved.
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

the amount of GHGs already emitted will remain in the atmosphere national climate conditions and extreme events affect financial stability
for decades. Hence, the benefits of climate policies, technological im- measured at the individual financial institution and system levels. We
provements or consumer preferences in terms of risk reduction will not considered four different measures of financial stability, two that apply
have positive effects on climate for some time to come. Meanwhile, to individual financial institutions, and two to the financial system
companies must adapt to the consequences of rapidly increasing climate as a whole. These measures are based on Z-scores, the probability of
variability, and banks that finance them must plan to adapt their capital default, market conditions, and the volatility of the financial system.
adequacy. Thus, a second body of literature focused on the influence Our empirical analysis extends from January 2001 to December 2021
of physical risks is emerging (Caloia and Jansen, 2022). Klomp (2017) for physical risks and from 2016 to 2021 for transition risks. Transition
and Mallucci (2020) explore the influence of natural disasters on bank risk is measured by GHG emissions (Scope 1 and 2) and Scope 3.
lending. Huang et al. (2018) find that the probability of losses due to Physical risks, both chronic and acute, are measured using actual
major storms, floods, and heat waves is associated with lower and more historical climate data. Our empirical work follows the work initiated
volatile corporate profits and cash flows. by Dell et al. (2012, 2014) and Burke and Hsiang (2015).4 Chronic risks
Physical risks are progressively integrated into the methodologies are measured using temperature and precipitation anomalies rather
for analyzing the impact on financial stability (BIS, 2021; Bavandi et al., than average values, as unexpected deviations from climate normals
2022). In Europe, the European Central Bank (ECB) carried out a first are a more representative measure of risk associated with climate
EU economy-wide climate risk stress test on 4 millions corporates and variability (Bertrand et al., 2015).
1600 consolidated banking groups that includes three physical risks, We find that transition risk measured as Scope 1 and 2 has limited
namely floods, wildfires, and sea-level rise (Alogoskoufis et al., 2021). influence on financial stability, which is consistent with Ehlers et al.
The expected losses by hazard type projected to 2050 showed that (2022) who find that the price of transition risk is low. This result
wildfires and floods have the potential to destroy physical capital. The is likely explained by the fact that financial institutions are low GHG
study emphasizes wildfires as the most widespread events affecting emitters. Most of their transition risk depends on the GHG emissions of
a wider geographical area than floods. Yet, 41 years of empirical the companies they finance, which fall within the Scope 3 of banks.
extreme events evidence (1980–2020) show that hydrological events Focusing on European financial institutions that publish Scope 3
(floods) accounted for 44% of total economic losses, meteorological emissions, the results show that transition risk has a highly signifi-
events (storms) for 34%, and climatological events (heat waves, cold cant negative influence on the financial stability at both institution
waves, droughts and wildfires) for only 14% altogether (European and financial system levels. This result is important. Given the low
Environment Agency, 2022). This difference is likely due to a lack of proportion of companies and banks that disclose Scope 3, it implies
knowledge of the exact location of assets at risk. The approach used that the transition risk as currently estimated in the literature using
by the ECB relied on banks providing the location of corporate clients’ mandatory climate-related disclosures (i.e., Scope 1 and 2) is most
exposures, to which the ECB applied backward-looking and forward- likely underestimated. It also highlights the importance for companies
looking climate data. Information related to the location and value and banks to estimate and disclose Scope 3 emissions, and for regulators
of each asset at risk is not accessible to banks as it is not disclosed to publish legally binding methodological guidelines for estimating and
in annual reports. For lack of a better alternative, the location they reporting them.
provided to the ECB was in most cases the address of the head office With respect to physical risks, we show that temperature anomalies
associated to the loan documentation, not that of the production assets, have a very significant influence on the financial stability of institu-
warehouses, or key suppliers, and distribution networks exposed to tions and the system as a whole. This result demonstrates that the
climate risks. Hence, measurement of physical risks on the European chronic risk, which has long been known to influence the economic
financial system was likely biased. and financial activity of many firms, is transmitted to the banks that
In response to the lack of climate-related disclosures, the FSB estab- finance them, and to the financial system as a whole. The influence
lished the Task Force on Climate-related Financial Disclosures (TCFD). that we measure, although very significant, is once again likely to be
Its main mission was to develop a set of voluntary and consistent underestimated, given the annual time step used in this study, which
recommendations on the information that companies should provide prevents us from the ability to measure the additional effects of intra-
to investors, lenders and insurance underwriters on their climate- annual climate variability (Starr-McCluer, 2000). The physical risk we
related financial risks. The initial reporting framework was launched in identify at the individual level relates to heat waves, drought, and
2017 (TCFD, 2021). The 2021 status report shows that climate-related wildfires, consistent with the ECB EU economy-wide first stress-test.
financial reporting is building momentum, with 2,600 organizations Although often considered to have primarily a local influence, we show
having adopted the TCFD framework, including 1,069 financial insti- that extreme events such as storms and tornadoes, together with heat
tutions responsible for $194 trillion in assets. In Europe, climate risk waves, drought and forest fires, influence financial stability at the
reporting, in conformity with the Corporate Sustainability Reporting level of the eurozone system. The relative difference in significance
Directive (CSRD2 ), is not expected until 2025. It will initially apply to between the institution level (low) and the system level (high) may
companies with more than 500 employees, and the detailed content of be related to the amplification effects within the financial system,
the climate-related disclosures has yet to be written.3 mainly due to a high degree of interconnectedness between financial
In this paper, we empirically investigate the influence of climate institutions (Chabot et al., 2023). This again is an important result.
risks on financial stability using a dataset of 130 European financial in- Our results contribute to the growing literature on climate risk and
financial stability. Our approach accounts for the fact that climate
stitutions. To address some of the shortcomings of existing approaches,
risk is primarily local and specific, but its impact can be widespread
we developed a theoretical framework of the transmission of climate
and affect a wide range of other financial institutions. Our method
risks to financial institutions. With this framework, we estimate how
overcomes the current lack of geo-located data, and demonstrates that
climate risk materially affects individual financial institutions and the
2
The 2022/2464 Directive of 14 December 2022 can be found here: https: financial system as a whole. The speed at which the climate is changing,
//eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32022L2464. and the increasing intensity of extreme events, which in most European
3
The European Commission is meant to adopt the first set of standards countries has already reached the intensity predicted for the end of the
by mid-2023, based on the draft standards published by European Financial
Reporting Advisory Group (EFRAG) in November 2022. The first set of draft
4
of European Sustainability Reporting Standards (ESRS) is accessible at https: See Kolstad and Moore (2020) for a comprehensive review of empirical
//efrag.org/lab6. methods to measure the effects of climate on economic indicators.

2
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Fig. 1. Theoretical framework.

Chronic climate risks include temperature and precipitation anomalies; Acute climate risks include flooding, wild fires, cold and heat waves, droughts, storms, tornadoes, and
hurricanes. The categories and list of extreme events are displayed in Table 1.

century by IPCC scenarios, call for new methods of risk measurement. can affect the supply and operating costs, as in the case of the agri-
Our method provides an alternative measure of climate-related risks food industry (Kumar et al., 2022) or the construction industry (Schuldt
that can be used to plan adaptation strategies and implement short-term et al., 2021). Further, the increase in climate variability that turns
risk reduction mechanisms. Our empirical results on physical risks urge into weather shocks may lead to increased operating costs and lower
regulators to issue mandatory climate-related corporate disclosures and productivity, and drive some businesses and industries to financial
adapt prudential capital requirements to climate-related risks. distress (Agnew and Palutikof, 2006; Dell et al., 2012; Cachon et al.,
The remainder of the paper is organized as follows. In the next 2012; Bertrand and Parnaudeau, 2019). The approach of estimating
section, we provide an overview of the relevant literature. Section 3 the influence of weather shocks on economic sectors and corporates
presents data and describes the empirical methodology. Section 4 is summarized in a survey paper by Dell et al. (2014). The effects
presents the results and discusses them in relation to the existing of climate variability (chronic risks) on corporates as listed in the
literature. Section 5 concludes. ‘‘Households and corporates’’ box of Fig. 1 include consumption and
demand, productivity drop, operational losses, sales losses, earnings
drops and increased operating costs.
2. Theoretical framework
Extreme weather events (acute risks) can cause temporary or perma-
nent business interruptions, affecting companies’ production assets or
As part of the supervisory work on the transparency of banks’
affecting their partners in the supply and distribution chains (Pankratz
risk profiles, the European Central Bank assessed the level of climate-
and Schiller, 2022). An extensive body of literature and news articles
related risks disclosures of European financial institutions (ECB, 2022).
have documented the impact of natural disasters on global supply
The ECB concluded that virtually none of the financial institutions in the
chains (see for instance Bland and Kwong, 2011 or Abe and Ye, 2013).
scope of the assessment met the minimum level of disclosures expected by
Evidence also shows that acute physical risks reduce corporate prof-
the ECB. As a result, there is a lack of information on the nature,
itability and potentially increase credit risk. Studies based on historical
the location, and the value of bank assets exposed to climate risks. data find that natural disasters can result in short-term moderate de-
This does not currently allow a direct assessment of the influence of clines in corporate sales. For example, US corporates have been shown
climate conditions and events on bank risks and financial stability. To to experience an average drop of 2 to 3 percentage points in sales
overcome this issue, we designed a theoretical transmission model of growth, following a major natural disaster that affects their suppliers,
climate-related risks (Fig. 1) that builds on two streams of the literature: ultimately causing a 1% drop in corporates’ equity value (Barrot and
physical climate risks and their impact on the economy, and transition Sauvagnat, 2016). Weishi Gu and Hale (2023) and other related stud-
risk and its influence on corporates and financial institutions. ies (Schuldt et al., 2020; Nagar and Schoenfeld, 2022; Jia et al., 2022)
We start by considering how physical climate risks affect households show that extreme events influence corporate investments, negatively
and corporates. Climate variability affects demand and consumers’ pur- immediately after, and positively in the long-run. Noth and Schuwer
chasing decisions of a myriad of products. Climate variability globally (2018) find that extreme climate events increase the predicted prob-
affects 70% of corporates (Larsen, 2006). Retailers have long been ability of default as well as non-performing asset ratios, and reduce
aware that deviations from normal climate conditions affect sales. Early bank equity ratios. Extreme events also affect household wealth, and
work on the influence of climate variability and its impact on corporate credit capacity. There is empirical evidence of property value declines
revenues dates back to Steele (1951), Linden (1962), and Maunder resulting from damage caused by extreme climate events such as flood-
(1973). Starr-McCluer (2000) and Lazo et al. (2011) extend the analysis ing (Bin and Polaski, 2004; Ortega and Taspinar, 2018), although the
to US retail sales at the level of the U.S., while Parnaudeau and Bertrand magnitude of the effect, and how long it persists, vary. In Fig. 1,
(2018) show that the cumulative annual contribution of temperature these effects are listed as business disruption, supply chain disruption,
and precipitation can have a material effect on corporate sales and cash- physical damage, reinvestment and replacement, drop in investments,
flows in large variety of activity sectors. Similarly, climate variability real estate prices, and mortgage defaults.

3
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Transition risks stand on the concept of disorderly low-carbon tran- which link climate risks to the solvency and default probability of
sition. They refer to the economic and financial losses arising from a financial institutions, and to market volatility and financial conditions
sharp revaluation of carbon-intensive and low-carbon assets, induced in the financial system. The next section describes the data, variables,
by a sudden change in policy and/or regulation that cannot be fully an- and methodology.
ticipated by corporate risk managers and financial actors (NGFS, 2019;
Monasterolo, 2020). The climate policies needed to achieve a net-zero 3. Data and methodology
carbon economy by 2050 require a rapid and drastic transformation of
industrialized and developing economies (e.g., in their energy, produc-
Our study covers the period from January 2001 to December 2021
tion, and consumption systems). These policies may have significant
and is supported by several databases. Financial and accounting data as
economic consequences, negative or positive, on many economic sec-
well as GHG and Scope 3 data are sourced from a Bloomberg terminal.
tors, leading to adjustments in the value of financial assets held or
The climate data used for chronic physical risk are daily temperature
issued by companies and sovereign entities. Firms may face higher op-
and precipitation observations from national meteorological services
erating expenses because of a higher tax on GHG emissions (Bernardini
(see in Appendix Table 8). Temperature and precipitation climate
et al., 2019). The impact of this tax could reduce profits and the ability
data range from January 1991 to December 2021 to allow for the
to meet debt repayments, and affect the company’s creditworthiness.
In turn, this could limit the access to funding, and increase the cost of calculation of seasonal averages as defined by the World Meteorological
such funding. Increases in credit costs for corporates in certain sectors Organization (WMO, 2017). Extreme events for acute physical risk were
may curtail their ability to invest in new technologies and equipment retrieved from the BD CatNat from Ubyrisk.
required to transition their business. Empirical evidence suggests that
the 2015 Paris Agreement resulted in a higher cost of credit for corpo- 3.1. Sample
rates in polluting industries, for example fossil fuel companies (Seltzer
et al., 2022; Delis et al., 2019). In addition, corporates whose business Our analysis is exploratory by nature, and takes the approach of
and revenues depend on the production or use of fossil fuels may suffer extending the perspective to a sample of financial institutions with dif-
asset losses, resulting in so-called ‘‘stranded assets’’. These losses could ferent risk profiles and climate exposures, from systemically important
in turn negatively impact the value of these corporates, and reduce their financial institutions (SIFIs) to regional banks, provided that financial
ability to get access to credit (Battiston et al., 2021). In Fig. 1, these information is available over a relatively long period of time. We built
additional effects are listed in ‘‘the Households and corporates’’ box as a representative database of the European financial system by selecting
stranded assets and new capital expenditure. cooperative, mutual, and listed financial institutions of different sizes
We now examine how financial losses to households and businesses (see Chabot (2021) for a discussion of the features of the European
caused by climate-related risks can affect banks (box ‘‘Financial in- banking markets). The geographical breakdown of the 130 financial
stitutions’’ in Fig. 1). The increase in climate variability and in the institutions in the panel, for which we had a complete set of variables,
number and intensity of extreme events is causing more and more is provided in Appendix (Table 9). Table 1 at the end of this section
damage to real estate and other productive assets, not all of which summarizes the variables used in this paper.
are insured. Over the last two decades, the number of extreme climate
events in Europe has gone up by 35%, but the associated economic
3.2. Dependent variables: Financial stability measures
and financial losses have jumped by almost 60%, 40% of which are
not insured (Monasterolo, 2020; European Environment Agency, 2022).
These losses can in turn negatively impact the value of financial con- Financial stability is measured using successively four different
tracts and financial portfolios exposed to these assets, such as bank variables: Z-score (Z), Default Probability (DP), Bloomberg Financial
loans and pension fund stocks and bonds. Cortes and Strahan (2017) conditions (BFC), and Market volatility (V). Z and DP are bank-level
and Ivanov et al. (2022) show how extreme events affect bank lending measures of financial stability. BFC and V are financial system level
and credit conditions. Mallucci (2020) and Klomp (2017) highlight the measures. Z-scores are widespread measures of financial stability. We
effects of climate risks on the pricing of sovereign debt and capital follow Laeven and Levine (2009), Lepetit and Strobel (2015), Tonzer
markets. Pagliari (2023) analyzes the impact of floods on small Euro- (2015), and Noth and Schuwer (2023) and define Z for each bank 𝑖
pean banks, and shows that an adverse event leading to a decrease in over the time period 𝑡 as:
lending to households and businesses results in a decline in the return 𝑅𝑂𝐴𝑖𝑡 + 𝐸𝐴𝑖𝑡
on assets of these banks. In Fig. 1, we list these effects in the ‘‘Financial 𝑍𝑖𝑡 = (1)
𝜎𝑅𝑂𝐴
institutions’’ box as losses in residential loans, underwriting losses, asset
ROA is the Return on Assets, EA is Equity over total Assets, and 𝜎𝑅𝑂𝐴
price adjustments, decrease in earnings, and collateral depreciation.
the standard deviation of the Return on Assets.
Like non-financial companies, banks and other financial institutions are
exposed to transition risk. Although their direct emissions (Scope 1) are DP measures the 5-year probability of default. DP incorporates
structurally low, indirect emissions from lending and investment (Scope fundamental factors such as sector risk, market sentiment, and the
3) have the same potential impact on banks as those experienced by economic cycle to determine the probability of default. DP is estimated
businesses, i.e., asset price adjustments, decrease in earnings, stranded by Bloomberg following Merton (1974), Altman and Sabato (2005),
assets, increased cost of capital, and financial distress (Roncoroni et al., Altman (2010). It is based on a combination of measures such as prof-
2021; Kacperczyk and Peydró, 2021; Reghezza et al., 2021; Ehlers itability (ROA, ROE), capital structure, liquidity (liquid assets and size),
et al., 2022). market information (distance-to-default measures), and an estimate of
The theoretical framework we built (Fig. 1) highlights the channels insolvency based on the study of the institution’s equity.
through which climate-related risks can influence financial stability BFC index assesses the overall level of financial stress in the euro
across households and corporates. The lack of data on the location area (based on money, bond, and equity markets) to determine the
of assets exposed to climate risks makes it virtually impossible to availability and cost of credit. A positive value indicates accommoda-
measure directly the impact on financial institutions and the financial tive financial conditions, and a negative value indicates tight financial
system (horizontal arrows). However, the theoretical framework and conditions. Finally, V is the Vstoxx index of the Eurozone. Its US equiv-
the literature on which it is based allow us to consider indirect mea- alent is known as the VIX index. V is estimated using a methodology
sures of the impact of climate risks on financial institutions and the developed jointly by Goldman Sachs and Deutsche Börse. The higher
financial system. This is represented in Fig. 1 by the upper arrows, the index, the greater the volatility of the market (Osina, 2019).

4
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Table 1
Summary list and description of variables.
Abbrev. Variable Description
Z Z-score Measure of financial stability of each financial institution
DP Default Probability Measure of default probability of each financial institution
BFC Financial Conditions Measure of overall level of financial stress in the euro area
based on credit availability estimated by Bloomberg
V Volatility Measure of overall level of financial stress measured by Vstoxx
RGDP Real GDP Real Gross Domestic Product per country
3M Money market rate 3-month money market rate
PX Asset Price Daily closing price of each financial institution
ITBA Interbank Assets Interbank assets of each financial institution
NPA Non-Performing Assets Non-performing assets of each financial institution
T1 Tier-1 Tier 1 Capital of each financial institution
GHG Emissions 1 & 2 Scope 1 and 2 GHG emissions of each financial institution
Scope 3 Emissions 3 Scope 3 GHG emissions of each financial institution
ANOT Temperature Anomaly Average difference between observed temperatures and their
30-year average
ANOP Temperature Anomaly Average difference between observed precipitations and their
30-year average
HYD Hydrological Number of Flooding and mudflow events
CLI Climatological Number of Forest fires, coldwaves, heatwaves, and droughts
MET Meteorological Number of cyclones, storms, hail, tornadoes, waterspouts,
blizzards, avalanches, and freezing rain

Z,DP, BFC, and V are the dependent variables; GHG and Scope 3 are transition risk variables; ANOT and ANOP are chronic
climate risk variables; HYD, CLI, and MET are acute climate risk variables. They represent the number of material events
classified level 3 or above in each risk category; RGDP, 3M, PX, ITBA, NPA, T1 are economic and financial variables.

3.3. Independent variables The absence of a legally binding GHG measurement methodol-
ogy has some implications on the quality of the data, and its con-
3.3.1. Economic variables sistence across data providers. An emerging literature on GHG data
The economic variables we use are the determinants of finan- quality identified some inconsistencies, gaps, and omissions in the way
cial stability identified in the literature. Following Caccioli et al. emissions are disclosed in sustainability reports, and across different
(2013), Brunetti et al. (2015), Kanno (2015), Liu et al. (2015), Paltalidis channels (Dragomir, 2012; Talbot and Boiral, 2018; Papadopoulos,
et al. (2015) and Chabot (2021), we use asset price (PX), interbank 2022). Inconsistencies are found to be low in direct emissions, and
assets (ITBA), non performing assets (NPA), and Tier 1 (T1). We also progressively increase in indirect emissions (Busch et al., 2022).
use real GDP growth rate (RGDP) and the 3 month money market rate This is likely due to the fact that Scope 1 and 2 are mandatory in
(3M) (Battiston et al., 2012; Tonzer, 2015; Embree and Roberts, 2009). most countries but the disclosure of Scope 3 is voluntary. In addition,
the scope of disclosure for Scope 3 can vary over time for a single com-
pany and also between companies, as disclosures practices for Scope 3
3.3.2. Transition risk variables
can vary (Mandyck, 2022). Scope 3 data from Bloomberg include all
To test the influence of transition risk on financial stability, we use
indirect emissions other than Scope 2 that occur in a company’s value
both ‘‘Total GHG emissions’’ and ‘‘Scope 3’’ published by Bloomberg.
chain, such as purchased goods and services, business travel, employee
Total GHG emissions is the sum of Scope 1 GHG emissions + Scope 2
commuting, waste disposal, use of sold products, transportation and
emissions. Scope 1 covers direct emissions from owned or controlled
distribution (up- and downstream), investments, and leased assets. In
sources of the financial institution. Scope 2 covers indirect emissions
the case of financial institutions, Scope 3 also includes GHG emissions
from the generation of purchased electricity, steam, heating and cooling
related to loans and investments. Scope 3 is likely to be the most
consumed by the financial institution. The E.U. requires large com-
significant transition risk variable, and should allow to directly link
panies to report their environmental and social impact, as part as
financing and investments to emissions. Due to the fact that Scope 3
their non-financial reporting. This is covered under the Non-Financial
emissions have mainly started to be reported after the Paris agreements
Reporting Directive (NFRD), or Directive 2014/95/EU. It is important
in 2015, the analysis period considered for the transition risk is 2016–
to note that reporting scope 1 and 2 is not a requirement of the NFRD, 2021. This period also applies to Scope 1 and 2 data for reasons of
which simply requires companies to publish the non-financial report. comparability of results.
Companies can still exclude scope 1 and 2 from their report.
Financial institutions are typically low-emitting companies. Direct 3.3.3. Chronic risk variables
emissions are likely to be small relative to indirect emissions. While Climate variability can influence the activity of the financial institu-
companies that fall within NFRD may report Scope 1 and 2, the tion directly and indirectly through the influence on the companies to
reporting on scope 3 is scarcer than scope 1 and 2 (EFRAG, 2021). which it lends or in which it has invested. The construction of a relevant
In the absence of a legally binding regulation, the best practices in climate variable can be done in a bottom-up approach under the
the field of corporate emissions accounting and reporting have been conditions that (1) each financial institution publishes the geographic
developed through the cooperation of non-governmental organizations distribution of loans and investments related to its customers, and (2)
and non-profit institutions. Examples include the GHG Protocol (WRI their customers in turn publish the geographic distribution of their
and WBCCSD, 2021), the Global Reporting Initiative (GRI) and the activities. As this is not the case, we built on Fig. 1 to develop metrics
Sustainability Accounting Standards Board (SASB), among others. The aimed at capturing the combined direct and indirect influence on each
most widely used framework to account and report corporate GHG financial institution. These dependent and explanatory variables are
emissions in the EU is the GHG Protocol developed by the cooperation positioned on the theoretical framework at the end of this section
between the World Resources Institute (WRI) and the World Business (Fig. 2).
Council for Sustainable Development (WBCSD). This international stan- In the specific case of Europe, banks’ corporate and retail clients are
dard covers the accounting and reporting of seven greenhouse gases mainly domestic (Chabot, 2021). Alogoskoufis et al. (2021) find that
included in the Kyoto Protocol. across countries, banks have a strong home bias, giving loans predominantly

5
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Fig. 2. Explanatory and dependent variables in the theoretical framework.

Z, DP, BFC, and V are the dependent variables; GHG and SCOPE 3 are transition risk variables; ANOT and ANOP are chronic climate risk variables; HYD, CLI, MET are acute
climate risk variables; Data are annualized when used in the models.

to domestic rather than foreign firms (...). In all countries, with the exception to country, depending on the size and geography of the country. Due
of Ireland and Luxemburg, domestic firms make up at least 50% of bank to the granularity of the financial data, the chronic climate data used
portfolios and at euro area level 80% of banks’ exposures are to domestic in the models are annual. The daily data were converted to annual data
firms. The composition of bank portfolios in terms of domestic versus foreign using a calibration method.5
firms determines the extent to which the country-level climate risk of the
firms (...) translates into country-level climate risk for banks. Another 3.3.4. Acute risk variables
study from Duijm and Schoenmaker (2017) analyzes the proportion Access to data on extreme weather events is more challenging as
of domestic and European assets relative to total assets on a panel of most databases are not public and by construction contain some bias.
the 61 largest European banks. The information about the distribution Biases are reflected in damage thresholds, the nature of the events, the
of assets between domestic and non-domestic markets is not readily completeness of the description of the events, their localization, their
available, and data was hand-collected. The study confirms that a large consequences, and the countries covered. We analyzed four of the most
proportion of the assets are either domestic or European. In relative complete databases: BD Catnat, Munich Re, Swiss Re, and CRED.6 A
terms, a small proportion of the assets is outside of the combined comparison of the databases over the period 2001–2021 indicates that
‘‘domestic and rest of Europe’’ (see Appendix Table 15). the total estimated damages are similar between the 4 databases: 3797
In addition, annual temperature and precipitation anomalies are billion US dollars for BD Catnat, compared to 3752 for Munich Re, 3604
highly correlated between neighbor countries. For instance, if there
for Swiss Re, and 3066 for CRED. In terms of number of events and
is a warm temperature anomaly in France in a given year, the same
completeness, BD Catnat is the most exhaustive, with 18,080 events
anomaly is usually experienced in neighbor countries, such as Germany,
compared to 16,294, 3218 and 7278 respectively. Extreme events in
Italy, or Belgium (see Appendix Fig. 4). Temperature anomalies in
the BD Catnat database are grouped into three categories: hydrological
France can be considered a proxy of German, Italian, or Belgian tem-
(HYD), climatic (CLI) and meteorological (MET). They cover a wide
perature anomalies. The strong emphasis of banks’ assets on ‘‘domestic
range of climatic hazards, from floods and freezing rain to fires and
and rest of Europe’’ reinforces the relevance of building national indices
storms. In this analysis, we consider all the events of natural origin that
to explore the influence of chronic climate risks.
caused a material damage. The word material is to be understood in its
Following Quayle and Diaz (1980), Parsons (2001), and Dell et al.
accounting definition, i.e., significant enough to warrant reporting to
(2014), and following World Meteorological Organization’s guidelines
investors, insurers, or other stakeholders. The damages apply to both
for creating a set of national climate indices (WMO, 2017), the chronic
material goods and people. Each event is classified on a scale of 1 to
risk variables for each country are the aggregate of the climate data for
5 according to its degree of materiality. Level 1 and 2 events are local
all stations in the country, weighted by the corresponding population
(town, county) and have limited damage (no human consequence, and
data assigned to each station. The concept in creating a country climate
partial damage to buildings). They are excluded from our database.
index to be used in economic or financial models is to reproduce
the average climate conditions experienced by economic actors in the Level 3 and 4 events have a spatial extension to one or more regions,
considered area, not the average climate conditions experienced in the
area. The chronic risk variables we built in this paper are derived
5
The complete process to transform daily temperature and precipitation
from daily observations of temperature and precipitation measured
data into monthly, quarterly, or annual temperature and precipitation anomaly
across a wide range of ground weather stations in each European
indices is developed in Bertrand et al. (2015), Bertrand and Parnaudeau
country. The number of stations in each country is the one used by
(2017).
the national meteorological office to calculate temperature or precip- 6
Munich Re and Swiss Re are reinsurance companies, CRED is maintained
itation to produce national climate indices. This number depends on by the University of Leuven (Belgium), and BD Catnat is from Ubyrisk,
the number of climatically homogeneous geographical areas needed a private risk management consulting firm. Other providers include the
for the national index to be representative. The number of stations is Asian Disaster Reduction Center, EMA (Australia), the Federal Emergency
different for temperature and precipitation. It also differs from country Management Agency (USA), and Reliefweb.

6
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

or to a country, and have a significant human or material cost, or both. 0.32 ◦ C to 0.48 ◦ C, further confirming the rise in climate variability.
Level 5 extends to several countries. For the purpose of this analysis, we Average precipitation over the same periods however did not exhibit
included all events classified levels 3 to 5. In the database, each event any trend or significant change in the standard deviation. These statis-
is assigned to a country. A level 5 event is assigned to all countries that tics are in line with the IPCC reports which indicate that the annual
have suffered damage in connection with the event. cumulative precipitation has been globally constant in Europe. Its
distribution over the year however shows that the summer months
3.4. Methodology and models are increasingly dryer to the benefit of the winter months which are
wetter (IPCC, 2021).
The influence of climate risks on financial stability is carried out A focus on acute risk variables, i.e., extreme events, shows that
using panel data models. Kolstad and Moore’s (2020) recent literature extreme events increased both in terms of the number of events (+33%)
review on statistical approaches applied to climate and economics and their financial consequences (+62%), in line with the findings of
showed that the use of panel data is highly relevant when the research the latest (IPCC, 2021) report. In the database of European extreme
question aims to understand the response of a system as a whole to climate events, events categorized as meteorological were the most
climate change (Deschênes and Greenstone, 2007; Dell et al., 2012; frequent (54%), followed by climatic events (29%) and hydrological
Burke and Hsiang, 2015; Colacito et al., 2019). In particular, panel data events (17%).
allows us to assess the conceptual aspects over time by analyzing a A correlation analysis between the financial stability variables and
succession of years of observations of the same financial institutions the set of explanatory variables was carried out (see Table 11 in
and, for a given year, over the total observed population. We first Appendix). We verified the absence of significant correlations between
estimate the impact of economic and financial variables on financial climate risk variables. We verified that all series of variables are
stability measured by four different dependent variables (Z and DP at stationary. We also confirmed the absence of endogeneity between
bank-level, and BFC and V at system-level). We then test the influence the dependent variables and the climate variables, and the absence
of transition risk on financial stability by successively testing Total GHG of clustering (country or period), autocorrelations or outliers in the
and Scope 3 emissions. Finally, we test the influence of physical risks, residuals. We can thus proceed to the analysis of the estimation of the
both chronic and acute. In step 1, the model is as follows: models.
FS𝑖𝑡 = 𝛼𝑖 + 𝛽(EcoFin)𝑖𝑡 + 𝜖𝑖𝑡 (2)
4.2. Financial stability and control variables
where FS𝑖𝑡 is financial stability successively measured at the individual
level of each financial institution by Z and DP, and at the financial As a first step, we confirm the relationships between economic and
system level by BFC and V. EcoFin𝑖𝑡 gathers asset prices (PX), 3 month financial variables and financial stability (Table 2). Results are in line
money market rate (3M), real GDP (RGDP), interbank assets (ITBA), with the relevant literature (Chabot, 2021). At the individual level, Z
non performing assets (NPA), and Tier one (T1). In this model and the is positively influenced by an increase in stock market returns (PX),
next three models, α𝑖 is a constant and 𝜖𝑖𝑡 is the disturbance term. In short-term interest rates (3M), and Tier 1 (T1). The probability of
step 2, we add transition risk variables to Eq. (2) as follows: default (DP) decreases when PX and T1 increase. DP increases when 3M
and non-performing assets (NPA) increase. At system level, financial
FS𝑖𝑡 = 𝛼𝑖 + 𝛽(EcoFin)𝑖𝑡 + 𝛾(Transition)𝑖𝑡 + 𝜖𝑖𝑡 (3)
conditions (BFC) improve with an increase in stock returns and Tier
where Transition𝑖𝑡 is successively ‘‘GHG’’ and ‘‘Scope 3’’ emissions. In 1, and a decrease in short-term rates. Volatility (V) decreases with an
step 3, we add physical risk variables to Eq. (2) as follows: increase in PX, 3M, and real GDP (RGDP), and increases when NPA
rises.
FS𝑖𝑡 = 𝛼𝑖 + 𝛽(EcoFin)𝑖𝑡 + 𝛿(Physical)𝑖𝑡 + 𝜖𝑖𝑡 (4)

where Physical𝑖𝑡 gathers chronic risk variables, i.e., anomalies of tem- 4.3. Financial stability and transition risk
perature (ANOT) and anomalies of precipitation (ANOP), and acute risk
variables, namely the number of hydrological extreme events (HYD), We add total GHG emissions to the model and test their influence on
climatic extreme events (CLI), meteorological extreme events (MET). financial stability over the past five years. The relationships observed
previously are confirmed, except for non-performing assets for which
4. Results and discussion we observe a decrease in both the value of the coefficients and the level
of significance. Real GDP however is now significant in 3 models out
This section presents and discusses the results of the influence of of 4 (Table 3).
transition and physical climate risks on financial stability at bank and At bank level, we find that an increase in GHG emissions has a
system levels. The presentation of the results follows the steps of the negative effect on Z with a level of significance of only .1. At the system
methodology. The results are reported in the same format, displaying level, higher emissions are associated to lower financial conditions
the influence of each explanatory variable and its significance on (BFC) and higher volatility (V). Again, the level of significance is only
financial stability, at the individual financial institution level (columns .1. The fact that DP is not significant can be explained by the infor-
Z and DP in each table in this section), and at the system level (columns mation and statistical indicators which do not include environmental
BFC and V). In each table, significance at the .01, .05, and .10 level is issues. These results are nevertheless consistent with the fact that Scope
represented by ***, **, and * respectively. 1 and 2 of financial institutions is low. Therefore, the level of transition
risk exposure is low when transition risks are measured by the only
4.1. Descriptive statistics and correlation analysis two Scopes that banks are required to disclose. We next investigate
the influence of Scope 3. As scope 3 disclosure is done on a disparate
Descriptive statistics are summarized in Table 10 in Appendix. A and voluntary basis in Europe, the proportion of financial institutions
focus on chronic risk variables shows that the average temperature in our database that disclose Scope 3 is about 20%. This is consistent
anomaly of the last decade (2011–2020) across our panel of European with the fact that only 15% of the 109 European banks they supervise
countries is 0.39 ◦ C higher than the first decade (2001–2010). This publish data as part of Scope 3 on the emissions of the companies they
is consistent with WMO’s statement confirming that temperatures in finance (Arnold, 2022).
Europe have increased at more than twice the global average over the Financial institutions’ exposure to transition risk is mainly indirect
past 30 years. In addition, from one decade to the next, the standard and Scope 3 is likely to be much larger than Scope 1 and 2. Financial
deviation of temperature anomalies increased by almost 50%, from institutions cover virtually all activity sectors as they finance public

7
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Table 2 Table 4
Financial stability and economic and financial variables. Financial stability and Scope 3 emissions.
Z DP BFC V Z DP BFC V
PX .19*** −.05*** 2.25*** −7.006*** PX 1.24** −.004 17.81** −192.2**
(2.05) (−2.53) (3.78) (−3.22) (2.21) (−.05) (2.13) (−2.14)
3M .10*** −.02*** −8.54*** 1.79*** 3M .21*** .01 −.56 27.69***
(1.87) (−2.65) (−11.56) (6.64) (5.86) (.45) (−1.42) (6.51)
RGDP .01 .006 .78 −9.67*** RGDP .16*** −0.02 12.04*** 107.74***
(.37) (.26) (1.66) (−5.63) (2.73) (−.91) (6.87) (5.70)
ITBA .87 −.33 −4.401* 1.16 ITBA −1.05 .08 10.04*** 108.36***
(.95) (−1.30) (−1.56) (1.13) (−.82) (.28) (2.57) (2.57)
NPA −3.79*** 1.01*** −19.47* 6.96** NPA −7.09** .97*** −81.80*** 882.74***
(−6.09) (3.15) (−1.74) (1.71) (−2.89) (2.16) (−6.78) (6.78)
T1 .90*** −.41*** 3.07*** −4.39* T1 1.31** −.13* 17.10*** −184.59***
(1.93) (−2.10) (3.96) (−1.55) (1.72) (−1.58) (2.9) (−2.90)
c .03** .03*** .90*** 19.74*** Scope 3 −1.75*** 3.93 −3.09*** 3.34***
(1.67) (4.94) (3.17) (18.98) (−2.92) (.94) (−3.27) (3.27)
c −.006 .03*** −3.95*** 59.06***
F-statistic 3.13*** 2.91*** 2.36*** 1.34***
(−.16) (4.82) (−13.88) (19.22)
Cross-section F: fixed effects yes yes no no
Cross-section Chi-Square : fixed effects yes yes no no F-statistic 2.96*** 1.98** 20.35*** 21.51***
Hausman test: random effects no no no no Cross-section F: fixed effects no no yes yes
Estimation under the presence of f.e f.e f.e f.e Cross-section Chi-Square: fixed effects yes yes yes yes
SE Clust. No No No No Hausman test: random effects yes yes no no
Estimation under the presence of: r.e r.e f.e f.e
R Sq. .56 .55 .51 .53
SE Clust. No No No No
Adj. R Sq. .48 .47 .45 .45
R Sq. .39 .34 .9 .9
Z and DP are individual level measures of financial stability; BFC and V are Eurozone
Adj. R Sq. .29 .24 .83 .84
system level measures of financial stability; RGDP, 3M, PX, ITBA, NPA, and T1 are
economica and financial variables. Data are annualized when used in the models. All Z and DP are individual level measures of financial stability; BFC and V are Eurozone
variables are stationary. F-test is a test a joint relevance of the variables in the models. system level measures of financial stability; Scope 3 are Scope 3 (indirect) emissions
R-squ. is the goodness of fit. The Hausman test verifies the presence of random effects. of individual financial institutions that measure transition risk; RGDP, 3M, PX, ITBA,
SE Clust. verifies the absence of standard error clustering (period and cross-sections). NPA, and T1 are economic and financial variables. Data are annualized when used in
Results apply to 2001–2021. the models. All variables are stationary. F-test is a test a joint relevance of the variables
in the models. R-squ. is the goodness of fit. The Hausman test verifies the presence of
random effects. SE Clust. verifies the absence of standard error clustering (period and
Table 3 cross-sections). Results apply to 2016–2021.
Financial stability and GHG emissions.
Z DP BFC V
PX .91* −.33* 3.49 −6.01*
(1.56) (−1.49) (1.21) (−1.50) At system level, an increase in Scope 3 has a negative effect on financial
3M 3.47*** .24 −5.12*** 5.74*** conditions and increases market volatility. The results are significant at
(2.64) (.86) (−2.12) (1.95) the .01 level.
RGDP 1.04*** −.14 .31*** 1.82***
These results on Scope 3 and financial stability are insightful in
(2.28) (−1.37) (4.45) (2.31)
ITBA −1.52*** .07** 1.79* −3.63** a context where Blackrock, a founding member of Task Force on
(−1.82) (1.81) (1.48) (−1.70) Climate-Related Financial Disclosures (TCFD) and one of the largest
NPA −1.42 1.3*** −7.96* 2.61* asset managers in the world, advocates for a flexible approach to rule-
(−.91) (2.36) (−1.37) (1.61) making based on ‘‘comply or explain’’, rather than mandating complete
T1 1.51*** −.30*** 2.63* −5.96*
(3.52) (−2.90) (1.20) (−1.62)
Scope 3 disclosures in annual and quarterly reports. In the US, under
GHG −9.28* 3.80 −2.22* 3.71* the March 2022 climate-related rule proposal of the SEC, registrants
(−1.36) (1.22) (−1.61) (1.55) would be obligated to disclose their greenhouse gas emissions for their
c .93*** −.02 −16.79*** 197.69*** most recently completed fiscal year. All filers except smaller reporting
(2.86) (−.34) (−2.53) (2.55)
companies would also be required to disclose their Scope 3 emissions
F-statistic 3.93*** 2.67*** 1.09*** 6.36*** if the Scope 3 emissions are material, or if the registrant has a target
Cross-section F: fixed effects no no no no
or goal that includes Scope 3 emissions. Blackrock’s comments are not
Cross-section Chi-Square: fixed effects no no no no
Hausman test: random effects yes yes yes yes intended to allow companies and banks to escape disclosure of Scope 3
Estimation under the presence of r.e r.e r.e r.e emissions, especially since Blackrock, in a recent letter to CEOs, admits
SE Clust. No No No No that as investors, they use Scope 3 emissions as a proxy for the degree
R Sq. .58 .51 .38 .41 of exposure companies have to carbon-intensive business models and tech-
Adj. R Sq. .51 .43 .21 .32 nologies. The market criticisms of the proposed Scope 3 requirements
Z and DP are individual level measures of financial stability; BFC and V are Eurozone focus on ambiguities in data collection and measurement, calculation
system level measures of financial stability; GHG are Scope 1 and Scope 2 emissions methodologies, timing of deadlines, and materiality standards. Given
of individual financial institutions that measure transition risk; RGDP, 3M, PX, ITBA,
our results on Scope 3, it is critical that market regulators and the
NPA, and T1 are economic and financial variables. Data are annualized when used in
the models. All variables are stationary. F-test is a test a joint relevance of the variables bodies in charge of the main reporting frameworks (e.g., TCFD, IASB)
in the models. R-squ. is the goodness of fit. The Hausman test verifies the presence of align their objectives and resources to select a common methodology
random effects. SE Clust. verifies the absence of standard error clustering (period and for calculating and disclosing Scope 3. Our results also show that tran-
cross-sections). Results apply to 2016–2021. sition risk as it applies to financial institutions, when GHG emissions
are limited to Scope 1 and 2, is likely to be underestimated.

and private companies. If reporting carbon emissions of loans and 4.4. Financial stability and physical risks
investments, Scope 3 emissions should be significant and allow to
identify their influence on financial stability. This is what our results We now consider the influence of both chronic and acute climate
confirm (Table 4). At bank level, Scope 3 emissions negatively affect Z. events on financial stability (Table 5).

8
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

At bank level, temperature anomalies (ANOT) have a significant Table 5

Financial stability and physical climate risks.
influence on Z and DP. Both are significant at the .01 level. Academics
have long demonstrated the influence of temperature on economic Z DP BFC V

activity (Deschênes and Greenstone, 2007; Dell et al., 2014; Burke PX .93 −.009 1.32*** −1.08***
(1.43) (−.35) (1.88) (−2.11)
and Hsiang, 2015; Newell et al., 2021). Banks’ exposure to temper-
3M 3.85*** .03 −5.12*** 5.74***
ature anomalies is mainly indirect. Our results confirm that chronic (2.85) (.04) (−9.30) (3.39)
climate risks are indeed transmitted from the real economy to the RGDP .98*** −.04 .31*** 1.88***
financial system via banks and financial institutions. Climatic events (2.10) (−.23) (2.81) −4.97
ITBA −1.95* .1 1.39 8.51
(e.g., heatwaves, droughts, wildfires) is the only category of extreme
(−1.59) (.43) (1.22) (.87)
events for which we find a significant influence, but only at the .1 NPA −3.30* .41 9.31 1.51***
level. Climatic extreme events influence both Z and DP, again, at the (−1.61) (.58) (1.82) (3.75)
.1 level. This result is consistent with Alogoskoufis et al. (2021), who T1 1.56*** −.36*** 9.25*** −6.99***
found wildfires to be a significant risk to financial stability, although (3.13) (−2.11) (2.92) (−3.27)
ANOT −.005*** .003*** −3.12*** 1.13***
in our study we did not measure the specific influence of wildfires, (−3.14) (1.95) (−4.27) (3.66)
which are considered collectively with heat waves, cold waves, and ANOP .015 .05* 6.21 9.89
drought as a single physical risk variable (CLI). This result is also (.34) (1.46) (.44) (.21)
consistent with Acevedo et al. (2020) and Miller et al. (2021) on CLI −.0011* .0006* −7.16*** 1.83***
(−1.38) (1.34) (−2.62) (2.31)
weather shocks and heatwaves on output and GDP, and earlier related
HYD .002 .0024 −9.80 −1.18
findings from Nordhaus (2006), Burke et al. (2009), and Cachon et al. (.58) (.91) (−1.26) (−.41)
(2012). MET .0001 −1.87 −3.19*** 1.44***
At system level, temperature anomalies have a very significant influ- (.84) (−.12) (−2.84) (2.03)
c 1.03*** .11* −16.79*** 197.69***
ence on both BFC and V. In addition, two categories of extreme events
(3.08) (.50) (−1.11) (4.29)
have a very significant impact on financial stability, namely climatic
F-statistic 2.94*** 1.83** 1.49*** 2.10***
events and meteorological events (e.g., storms). Both are significant
Cross-section F: fixed effects yes yes no no
at the .01 level. These results are all the more remarkable as the Cross-section Chi-Square: fixed effects yes yes no no
level of aggregation of the data likely underestimates the effects of Hausman test: random effects yes yes yes yes
climate risks. Indeed, a cold anomaly can have negative consequences Estimation under the presence of: r.e r.e r.e r.e
SE Clust. No No No No
on a given activity sector in Q2, while a warm anomaly of the same
size in Q3 also has negative consequences. This is the case for the R Sq. .51 .6 .4 .43
Adj. R Sq. .43 .54 .24 .13
apparel industry for instance (Bertrand et al., 2015). In this example,
when aggregating the data annually, cold and warm anomalies offset Z and DP are individual level measures of financial stability; BFC and V are Eurozone
system level measures of financial stability; ANOT and ANOP are climate chronic risk
each other. There is no temperature anomaly, and therefore no ability
variables; CLI, HYD, and MET are climate acute risk variables; RGDP, 3M, PX, ITBA,
to measure the two quarterly losses yet due to temperature. This NPA, and T1 are economic and financial variables. Data are annualized when used in
phenomenon has been identified by Starr-McCluer (2000), who gave the models. All variables are stationary. F-test is a test a joint relevance of the variables
it the name of wash-out effect. The difference of significance level in the models. R-squ. is the goodness of fit. The Hausman test verifies the presence of
random effects. SE Clust. verifies the absence of standard error clustering (period and
between bank-level (low) and system-level (high) may be explained by
cross-sections). Results apply to 2001–2021.
the compounding and amplification effects identified in the literature,
which is mainly due to the interconnections between financial institu-
tions (Chabot et al., 2023). The high level of significance of the market
indicators (BFV and V) may also be explained by the fact that extreme ESG rating agencies offer investors a way to screen companies for
weather and climate events affect large areas and are associated with ESG performance in a similar way to how credit ratings allow investors
significant damage, both financial and human. This is especially true to screen companies for creditworthiness (Berg et al., 2020). ESG
in the case of storms and tornadoes. ratings are broken down into many indicators and hierarchies, the first
Overall, our results are in line with the findings of existing scenario level of which is Environment (E), Social (S), and Governance (G). Pillar
E is at times used in the literature as a proxy for measuring transition
and stress-test analyses (Battiston et al., 2017; Vermeulen et al., 2019;
risk when no other measures are available, due to the fact that GHG
Reinders et al., 2020; NGFS, 2019; Allen et al., 2020; Roncoroni et al.,
emissions and mitigation strategies are important components of the
2021): physical and transition risks significantly affect financial sta-
E-score. S&P Global ESG Scores are proprietary data that result from
bility at bank and system levels (Coelho and Restoy, 2022). They are
a combination of verified company disclosures, media and stakeholder
also in line with Noth and Schuwer (2023), who find that in the U.S.,
analysis, and in-depth engagement activities with companies. Entity-
weather-related natural disasters significantly weaken the stability of banks
specific scores are assessed on a relative basis, against sector peers.
with business activities in affected regions.
The Scores range from 0 to 100. Scores of 0 to 25 indicate poor ESG
performance and insufficient transparency in the public reporting of
4.5. Robustness material ESG data. Scores above 75 indicate excellent relative ESG
performance and a high degree of transparency. We limit the study
We used several tests to confirm the robustness and relevance of the period of the influence of ESG to the most recent years (2016–2021)
models. The F-test is a joint test that supplements the Student’s tests for comparability reasons. Table 12 in the Appendix summarizes the
and allows to validate the explanatory variables in the tested models. main elements that make up the different E, S, and G pillars.
If significant, the variables are collectively relevant in explaining the The first additional control variable we consider is the environ-
dependent variable. The Cross-section F and Cross-section Chi-square mental score (E). The E-score is primarily influenced by the level of
tests evaluate the presence of fixed effects. The Hausman (1978) test emissions and mitigation strategies that have been implemented, or
verifies whether the random effects (r.e) estimation is more relevant that are being considered, to reduce risk. The analysis of the individual
than the fixed effects (f.e) estimation method. In addition to these tests, pillars confirms that E has an influence on financial stability both at
we controlled for ESG scores. The ESG scores used here were provided bank and system levels (Table 6). As E increases, Z decreases, and V
by S&P Global Ratings. ESG scores from S&P Global ratings range from increases. Results are significant at .05 and .01 respectively. This is in
2016 to 2021. line with existing literature: environmentally-active institutions (high

9
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Table 6 Table 7
Financial stability and E Scores. Financial stability and ESG overall scores.
Z DP BFC V Z DP BFC V
PX .06 −1.45*** 4.3 −34.44*** PX 2.34 −.02 5.36* −32.17**
(.03) (−2.11) (1.37) (−3.40) (1.01) (−1.06) (1.56) (−1.81)
3M 2.732*** .1 −5.12*** 1.45 3M 2.386** .79*** −5.12 3.6
(2.43) (.24) (−2.33) (.50) (1.90) (8.97) (−2.40) (.22)
RGDP .81*** −.21*** .31*** 5.33*** RGDP .50* −.42*** .31*** 5.38***
(3.24) (−3.09) (6.15) (7.88) (1.50) (−10.71) (5.17) (2.41)
ITBA −3.18*** −.05 −1.35 4.56 ITBA −2.77 −.02 −4.02* −14.7*
(−4.01) (−.30) (−.56) (.43) (−1.23) (−.62) (−1.61) (−1.47)
NPA −1.25 .29 3.9 −28.02*** NPA −.85 .76** 1.29* −37.4*
(−.60) (1.31) (1.98) (−2.72) (−.27) (1.71) (1.62) (−1.42)
T1 5.48*** −.87*** −3.04 11.35** T1 8.15*** −.20** 4.92* 14.30***
(3.56) (−3.17) (−1.13) (1.72) (8.46) (−2.08) (1.63) (2.47)
E −2.26** .22*** −5.01 1.3** ESG .305*** 2.98 5.71* 4.9
(−2.08) (2.62) (−.42) (1.81) (3.06) (1.22) (1.60) (.30)
c .69** .07* −16.79*** 20.90*** c .39* −.11*** −16.79*** 20.63***
(2.80) (0.62) (−2.86) (18.73) (1.19) (−2.53) (−3.04) (2.57)
F-statistic 6.82*** 1.84*** 1.66* 7.03*** F-statistic 11.41*** 2.58*** 1.50*** 6.13***
Cross-section F: fixed effects yes yes no no Cross-section F: fixed effects yes no no no
Cross-section Chi-Square: fixed effects yes yes no no Cross-section Chi-Square: fixed effects yes no no no
Hausman test: random effects no no no yes Hausman test: random effects no yes yes yes
Estimation under the presence of: f.e f.e f.e r.e Estimation under the presence of: f.e r.e r.e r.e
R Sq. .83 .58 .46 .28 R Sq. .93 .42 .48 .27
Adj. R Sq. .71 .26 .18 .24 Adj. R Sq. .85 .33 .31 .23

Z and DP are individual level measures of financial stability; BFC and V are Eurozone Z and DP are individual level measures of financial stability; BFC and V are Eurozone
system level measures of financial stability; RGDP, 3M, PX, ITBA, NPA, and T1 are system level measures of financial stability; RGDP, 3M, PX, ITBA, NPA, and T1 are
economic and financial variables; E is the environmental score from S&P Global Ratings. economic and financial variables; ESG is the ESG rating from S&P Global Ratings .
Data are annualized when used in the models. All variables are stationary. F-test is a Data are annualized when used in the models. All variables are stationary. F-test is a
test a joint relevance of the variables in the models. R-squ. is the goodness of fit. The test a joint relevance of the variables in the models. R-squ. is the goodness of fit. The
Hausman test verifies the presence of random effects. SE Clust. verifies the absence of Hausman test verifies the presence of random effects. SE Clust. verifies the absence of
standard error clustering (period and cross-sections). Results apply to 2016–2021. standard error clustering (period and cross-sections). Results apply to 2016–2021.

E) are associated to a lower risk (Bouslah et al., 2018). With respect to the implications of climate-related risks for financial stability (Carney
the S score and the G score, we find that G scores are negatively associ- et al., 2019; Baranovic et al., 2021). Potential gaps may exist in the
ated to the probability of default (DP) at the .05 level (see Tables 13 and Pillar 1 framework for operational risks, since it does not cover losses
14 in Appendix). This is consistent with Pastor et al. (2021) and Bernile related to strategic and reputational risks, the latter of which may
et al. (2018). We also find that S scores negatively impact DP with a .01 be closely related to transition, liability, or environmental risks. Su-
level of significance. At system level, we find that G has an influence pervisory measures and enhanced climate-related disclosures represent
on V, but with a low level of significance. Better customer engagement, important tools for the financial sector, to better guard against climate
board oversight, controls, transparency, and values together (high S and risks in the short and medium term. Pillar 1 captures losses related
G) result in better management, higher profitability, and lower risks as from inadequate or failed internal purposes, people, and systems or
perceived by shareholders and markets (Friede et al., 2015; Chambers from external events. This includes, at least conceptually, some aspects
et al., 2018). of climate risks materializing in actual physical damages of the bank
ESG scores are often used as ‘‘financial soundness’’ proxy variables, assets, or liability risk reflected in legal or compliance risk. In contrast,
and as such, they should lower the risk. At financial institution-level, Pillar 2 covers risks which are underestimated or not covered by Pillar
Table 7 shows that ESG scores have a significant influence on Z at 1. Given the challenges of capturing the impact of climate-related
the.01 level. The higher the ESG score, the stronger the financial financial risks, some of the principles and methodologies underpinning
the Basel Pillar 1 framework might not hold. In particular, how and
stability. Again, the influence on DP is however not significant. At
when these risks will manifest over different time horizons creates
system-level, BFC is the only financial stability variable that is influ-
uncertainty in addition to existing uncertainties related to climate and
enced by ESG overall scores, only at the .1 level. The higher ESG scores,
associated economic long-term scenarios. Some parts of the current
the better financial conditions. The results are consistent with Chiara-
prudential framework are backward-looking, as they rely on consistent,
monte et al. (2015), who find that the ESG overall score, as well as
historical data to gauge the relationships between risk factors and
its sub-pillars, reduce bank fragility during periods of financial distress.
exposures to which our findings can contribute. Pillar 2 is a natural
Companies with high environmental, social and governance scores tend
candidate for ensuring that banks effectively report and manage phys-
to have higher excess returns, lower cost of capital, lower volatility,
ical climate risks, and have sufficient loss-absorbing capacity against
and therefore lower exposure to tail risks (Friede et al., 2015; La Torre
them.
et al., 2020; Giese et al., 2019). The process of modifying the macroprudential framework to take
account of short- and long-term climate risks, however, requires rein-
4.6. Recommendations forced measures to standardize, enforce, and monitor corporate climate-
related risk information. In turn, the ability of financial institutions to
Central banks as well as academia have highlighted that the risks assess their own climate risk is dependent on the quality of climate-
of inaction with respect to climate-related risks are far greater than the related disclosures provided by the companies they invest in or lend
risks of acting on the basis of partial data (NGFS, 2019; Coelho and to Feridun and Gungor (2020). The moves from the UK government
Restoy, 2021). Policy makers and supervisors are gradually agreeing and the SEC to make TCFD-aligned climate risk disclosure mandatory
on the need to revise the prudential framework to take full account of is a step in the right direction (Simpson, 2022).

10
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Fig. 3. Acute physical climate risk annual statistics*. *: France, Germany, Italy, Spain, and the United-Kingdom.

Table 8
Anomalies of temperature and precipitations: Data Sources.
Country Source
Austria Zentralanstalt für Meteorologie und Geodynamik
Belgium Royal Meteorological Institute
Finland Finnish Meteorological Insitute
France Meteo-France
Germany Deutscher Wetterdienst
Greece Hellenic National Meteorological Service
Irland Met éireann
Italy Servizio Meteorologico
Netherlands Koninklijk Nederlands Meteorologisch Instituut
Fig. 4. Correlation between annual temperature anomalies across major European Spain Agencia Estatal de Meteorologia
countries. Sweden Swedish Meteorological and Hydrological Insitute
Switzeland MeteoSwiss
UK Met Office

5. Concluding comments

In this paper, we analyze the influence of transition and physical climate-related risk exposures in their lending and investment portfo-
climate risks on financial stability on a panel of European financial lios. Finally, in a context of rising climate variability, our results suggest
institutions using a top-down approach. We find that both transition that climate risk should be factored into prudential regulations as early
and physical climate risks significantly influence financial stability. An as possible.
increase in Scope 3 emissions and climate variability, both chronic and
acute, has a negative effect on financial stability at the level of each
institution and at the level of the financial system. Data availability
Our results complement the growing literature on climate risk and
financial stability. One specificity of climate risk is that it is mostly Data used in this study are not publicly available as they are derived
local, but its impact can be widespread, depending on the nature of from proprietary databases. This is the case for climate and extreme
the risk and its intensity. The theoretical framework we propose builds event data, Bloomberg financial data, and ESG scores from S & P, which
on the transmission channels from the real economy, which has long the authors have access to through subscriptions and agreements be-
been known to be exposed to climate risks, to the financial system tween national meteorological agencies and the research organization
through financial institutions and their high level of interconnected- with which they are affiliated. Data are available from the correspond-
ness. In a context of accelerating climate change and climate variability,
ing author upon reasonable request, subject to the agreement of the
the method we expose should be considered in addition to existing
owner of the requested data.
methodologies to offer regulators and risk managers the opportunity
to improve short-term risk assessment and adjust capital requirements
to both short-term and long-term horizons. Our results stress the im- Appendix
portance of using climate variables and actual extreme events to assess
the short-term exposure of financial institutions based on actual climate
A.1. Tables and figures
variability and historical extreme events, to complement long-term
existing scenario-based models.
Our findings support the need for mandatory corporate climate See Figs. 3 and 4 and Tables 8–15.
risk disclosures so that financial institutions are able to value their

11
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Table 9 Table 12
Geographical distribution of financial institutions. S&P Global ESG Scores, (2016–2021)
Country Number of Financial Percentage in Score (Variable) Description
Institutions the sample
ESG (ESG) Overall ESG score
Austria 10 8%
Environmental score (E) Greenhouse Gas Emissions
Belgium 12 9%
Waste and Pollution
Finland 4 3%
Water Use
France 12 9%
Land Use
Germany 12 9%
Greece 9 7% Social (S) Workforce and Diversity
Ireland 3 2% Safety Management
Italy 12 9% Customer engagement
Netherlands 10 8% Communities
Spain 12 9% Governance (G) Structure and Oversight
Sweden 11 8% Code and Values
Switzerland 11 8% Transparency and Reporting
United Kingdom 12 9% Financial and Operational risks
Total 130 100%

Table 13
Financial stability and Social (S) Scores.
Z DP BFC V
PX .11 −1.48*** 2.90** −37.26***
Table 10
(.05) (−2.09) (2.009) (−2.17)
Descriptive statistics of variables.
3M 3.151*** .11 −4.17*** 1.45
Financial stability variables Mean Std Median Min Max (3.94) (.27) (−3.01) (.09)
Z .233 1.02 .09 −21.76 8.27 RGDP .90*** −.21*** .30* 5.30***
DP .0231 .037 .016 .0002 .763 (4.74) (−3.41) (1.47) (2.36)
BFC −.62 1.92 .03 −7.39 1.14 ITBA −3.11*** −.03 −.43 7.49
V 23.17 9.16 21.76 13.51 51.21 (−6.20) (−.26) (−.45) (.63)
Economic Variables NPA −1.15 .31 2.22 −27.41
(−.55) (1.38) (.99) (−1.05)
PX 118.75 313.02 28.46 .0004 4062
T1 5.74*** −.92*** −.86** 12.21***
3M 1.37 1.73 .85 −.57 4.86
(3.30) (−3.78) (−1.97) (2.34)
RGDP 1.08 2.51 1.90 −10.80 11.40
S 1.14* −.22*** 9.9 1.4*
ITBA 38.030 75.504 4878 18.35 787,295
(1.52) (−3.90) (1.07) (1.48)
NPA 9221 18.35 4127 .15 297,725
c .71*** .07* −1.86*** 21.00***
T1 24,408 33,770 9700 −2070 383,783
(3.20) (.62) (−2.61) (2.71)
Transition risk variables
F-statistic 6.52*** 1.84*** 17.88*** 7.11***
GHG 2058 36,755 39.55 .06 693,606
Cross-section F: fixed effects yes yes no no
Scope 3 151.5 970.9 7.80 .02 9,912.1
Cross-section Chi-Square: fixed effects yes yes no no
Physical climate risk variables
Hausman test: random effects no no yes yes
ANOT .85 1.10 .70 −.54 9.26 Estimation under the presence of: f.e f.e r.e r.e
ANOP .98 .09 .98 .74 1.26
R Sq. .83 .58 .50 .28
HYD 2.21 2.47 1.00 .00 8.00
Adj. R Sq. .70 .26 .47 .24
CLI 4.59 9.20 1.00 .00 40.00
MET 10.90 16.43 6.00 .00 66.00
ESG variables
Table 14
ESG Overall 60.86 26.80 65.00 2.00 99.00 Financial stability and Governance (G) Scores.
E 64.40 23.98 67.00 10.00 99.00 Z DP BFC V
S 60.58 26.29 62.00 .00 98.00
G 57.63 28.94 63.00 3.00 99.00 PX .47 −1.36** 2.90** −37.30***
(.22) (−1.92) (2.04) (−2.24)
3M 3.682*** .11 −4.15*** 1.14
(3.70) (−.24) (−3.03) (.07)
RGDP 1.005*** −.22*** .39* 5.30***
(4.36) (−3.19) (1.47) (2.36)
Table 11 ITBA −3.18*** −.10 −.60 10.02
Correlation matrix between financial stability and physical risk variables. (−4.42) (−.55) (−.54) (.75)
NPA −1.21 .23 2.09 −25.48
Z DP BFC V ANOT ANOP HYD CLI MET
(−.53) (1.29) (.96) (−1.01)
Z 1 T1 6.51*** −.77*** −.90** 12.81***
DP −.4354 1 (4.62) (−2.65) (−1.91) (2.26)
BFC .1810 −.3278 1 G .03 −.10** 1.06 −1.61*
V −.1747 .0409 −.4767 1 (.09) (−1.82) (1.03) (−1.42)
ANOT .0766 −.0053 −.2369 −.2862 1 c .75*** .06* −1.53*** 20.88***
ANOP .0482 −.0055 .1815 .1372 −.2043 1 (2.54) (.52) (−2.64) (2.73)
HYD .1331 .0007 −.0368 −.0216 −.1189 .1618 1
F-statistic 6.36*** 1.84*** 17.84*** 7.14***
CLI .1020 −.0144 .0853 .0641 −.0316 −.1157 .2139 1
Cross-section F: fixed effects yes yes no no
MET .1381 −.0064 −.0098 −.0278 .0614 −.0421 .2217 .2533 1
Cross-section Chi-Square: fixed effects yes yes no no
Hausman test: random effects no no yes yes
Estimation under the presence of: f.e f.e r.e r.e
R Sq. .82 .58 .52 .27
Adj. R Sq. .69 .26 .46 .21

12
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Table 15 Bressan, G., Romagnoli, S., 2021. Climate risks and weather derivatives: A copula-based
Geographical distribution of assets per country as a percentage of total assets. pricing model. J. Financial Stab. 54 (100877).
Source: Adapted from : Duijm and Schoenmaker, 2017. Brunetti, C., Harris, J., Mankad, S., Michailidis, G., 2015. Interconnectedness in the
Domestic Rest of Europe Total Interbank Market. FED Working Paper, October, 2015–090.
Burke, M., Hsiang, S., 2015. Global non-linear effect of temperature on economic
France 66.4 20.4 86.8
production. Nature 527, 235–239.
UK 55.0 8.8 63.8
Burke, M., Miguel, E., Satyanath, S., Dykema, J., Lobell, D., 2009. Warming increases
Germany 50.2 24.7 74.9
the risk of civil war in Africa. Proc. Natl. Acad. Sci. 106, 20670–20674.
Spain 47.4 26.9 74.3
Busch, T., Johnson, M., Pioch, T., 2022. Corporate carbon performance data: Quo
Netherlands 61.2 23.5 84.7
vadis? J. Ind. Ecol. 26 (1), 350–363.
Italy 67.7 29.3 97.0
Caccioli, F., Farmer, J., Nick Foti, D.R., 2013. How interbank lending amplifies
Sweden 48.8 46.3 95.1
overlapping portfolio contagion: A case study of the Austrian banking network.
Belgium 59.2 37.1 96.3
SSRN.
Cachon, G., Gallino, S., Olivares, M., 2012. Weather and automobile assembly
productivity. Columbia Business School Research Paper No. 12/37.
Caloia, F., Jansen, D., 2022. Flood Risk and Financial Stability: Evidence from a Stress
References Test for the Netherlands. DNB Working Paper, No. 730.
Carney, M., 2015. Breaking the tragedy of the horizon – climate change and financial
Abe, M., Ye, L., 2013. Building resilient supply chains against natural disasters: The stability. Speech at Lloyd’s of London, 29 September 2015.
Carney, M., de Galhau, F.V., Elderson, F., 2019. Open Letter on Climate-Related
cases of Japan and Thailand. Glob. Bus. Rev. 14, 567–586.
Financial Risks. Bank of England report. 17 April.
Acevedo, S., Mrkaic, M., Novta, N., Pugacheva, E., Topalova, P., 2020. The effects
Chabot, M., 2021. Financial stability indices and financial network dynamics in europe.
of weather shocks on economic activity: What are the channels of impact? J.
Revue Economique 72 (4), 591–631.
Macroecon. 65, 103207.
Chabot, M., Bertrand, J., Courquin, V., 2023. Climate interconnectedness and financial
Agnew, M., Palutikof, J., 2006. Impacts of short-term climate variability in the UK on
stability. Finance 2023/0.
demand for domestic and international tourism. Clim. Res. 31 (1), 109–120.
Chambers, D., Black, K., Lacey, N., 2018. Alternative Investments: A Primer for
Allen, T., Dees, S., Boissinot, J., Caicedo, G., Chouard, V., Clerc, L., De Gaye, A.,
Investment Professionals. CFA Institute Research Foundation. Research Foundation
Devulder, A., Diot, S., Lisack, N., Pegoraro, F., Rabaté, M., Svartzman, R., Vernet, L.,
Books.
2020. Climate-related scenarios for financial stability assessment: An application to
Chiaramonte, L., Croci, E., Poli, F., 2015. Shoud we trust the Z-score? Evidence from
France. Banque de France, Working Paper Series, no 774.
the European banking industry. Glob. Finance J. 28, 111–131.
Alogoskoufis, S., Dunz, N., Emambakhsh, T., Munoz, M., Parisi, L., Salleo, C., 2021. Coelho, R., Restoy, F., 2021. Keynote speech: Patchy data is a good start: from Kuznets
ECB economy-wide climate stress test. Methodology and results. Occasional Paper and Clark to supervisors and climate. In: ECB-EBRD Joint Conference: Emerging
Series. Nà281. September 2021. climate-related risk supervision and implications for financial institutions.
Altman, E., 2010. The Z-metrics methodology for estimating company credit ratings Coelho, R., Restoy, F., 2022. The Regulatory Response to Climate Risks: Some
and default risk probability. Risk Metrics Group. Challenges. Financial Stability Institute, FSI Brief No.16.
Altman, E., Sabato, G., 2005. Modeling credit risk for SMEs: Evidence for the US market. Colacito, R., Hoffmann, B., Phan, T., 2019. Temperature and Growth: a panel analysis
SSRN, 872336. of the United States. J. Money Credit Bank. 51 (2–3), 313–368.
Arnold, M., 2022. ECB accuses eurozone banks of ‘white noise’ on climate risks. Cortes, K., Strahan, P., 2017. Tracing out capital flows: How financially integrated
Financial Times. 14 March 2022. banks respond to natural disasters. J. Financ. Econ. 125 (1), 182–199.
Baranovic, I., Busies, I., Coussens, W., Coussens, M., Hempell, H., 2021. The Challenge Delis, M., de Greiff, K., Losifidi, M., Ongena, S., 2019. Being Stranded the Carbon
of Capturing Climate Risks in the Banking Regulatory Framework: Is There a Need Bubble? Climate Policy Risk and the Pricing of Bank Loans. Swiss Finance Institute
for a Macroprudential Response? ECB Publications. Research Paper No. 18-10.
Barrot, J., Sauvagnat, J., 2016. Input specificity and the propagation of idiosyncratic Dell, M., Jones, B., Olken, B., 2012. Temperature shocks and economic growth:
shocks in production networks. Q. J. Econ. 131, 3. Evidence from the last half century. Am. Econ. J. 4 (3), 66–95.
Batten, S., Sowerbutts, R., Tanak, M., 2016. Let’s Talk About the Weather: the Impact Dell, M., Jones, B., Olken, B., 2014. What do we learn from the weather? The new
of Climate Change on Central Banks. Bank of England. Staff Working Paper No. climate−economy literature. J. Econ. Lit. 52, 740–798.
603. Deschênes, O., Greenstone, M., 2007. The economic impacts of climate change:
Battiston, S., Dafermos, Y., Monasterolo, I., 2021. Climate risks and financial stability. Evidence from agricultural output and random fluctuations in weather. Amer. Econ.
J. Financial Stab. 54, 100867. Rev. 97 (1), 354–385.
Battiston, S., Mandel, A., Monasterolo, I., Schützen, F., Visentin, G., 2017. A climate Dragomir, V., 2012. The disclosure of industrial greenhouse gas emissions: a critical
stress-test of the financial system. Nature Clim. Change 7 (4), 283–288. assessment of corporate sustainability reports. J. Clean. Prod. 29, 222–237.
Battiston, S., Puliga, M., Kaushik, R., Tasca, P., Caldarelli, G., 2012. Debtrank: Too Duijm, P., Schoenmaker, D., 2017. European banks straddling borders: Risky or
central to fail? Financial networks, the FED and systemic risk. Sci. Rep. 2–6. rewarding? SSRN.
Bavandi, A., Berrais, D., Dolk, M., Mahul, O., 2022. Physical climate risk assessment: ECB, 2020. Guide on climate-related and environmental risks. Supervisory expectations
Practical lessons for the development of climate scenarios with extreme weather relating to risk management and disclosures. European Central Bank. Banking
events from emerging markets and developing economies. Network for Greening Supervision.
the Financial System. Technical Document. September 2022. ECB, 2022. Supervisory Assessment of institutions’ climate-related and environmental
disclosures. European Central Bank. Eurosystem.
Berg, F., Kolbel, J., Rigobon, R., 2020. Aggregate Confusion: The Divergence of ESG
EFRAG, 2021. Current Non-Financial Reporting Formats and Practices. Appendix 4.6:
Ratings. MIT Sloan School Working Paper 5822-19.
Stream A6 Assessment Report. Technical Report. European Financial Reporting
Bernardini, E., Di Giampaolo, J., Faiella, I., Poli, R., 2019. The impact of carbon risk
Advisory Group. February 2021.
on stock returns: evidence from the European electric utilities. J. Sustain. Finance
Ehlers, T., Packer, F., de Greiff, K., 2022. The pricing of carbon risk in syndicated
Invest. 11 (1), 1–26.
loans: Which risks are priced and why? J. Bank. Financ. 136, 106180.
Bernile, G., Bhagwat, V., Yonker, S., 2018. Corporate governance, board diversity, and
Embree, L., Roberts, T., 2009. Network Analysis and Canada’s Large Value Transfer
firm value. J. Financ. Econ. 127 (3), 588–612.
System. Bank of Canada Discussion Paper, 2009–13.
Bertrand, J.-L., Brusset, X., Fortin, M., 2015. Assessing and hedging the cost of
European Environment Agency, 2022. Economic Losses and Fatalities from Weather-
unseasonal weather: Case of the apparel sector. European J. Oper. Res. 244,
and Climate-Related Events in Europe. Technical Report. February.
261–276.
Feridun, M., Gungor, H., 2020. Climate-related prudential risks in the banking sector:
Bertrand, J.-L., Parnaudeau, M., 2017. No more blaming the weather: A retailer’s A review of the emerging regulatory and supervisory practices. Sustainability 12,
approach to measuring and managing climate variability. Int. J. Ret. Distrib. Manag. 5325.
45 (7/8), 730–761. Friede, G., Busch, T., Bassen, A., 2015. ESG and financial performance: Aggregated
Bertrand, J., Parnaudeau, M., 2019. Understanding the economic effects of abnormal evidence from more than 2000 empirical studies. J. Sustain. Finance Invest. 5 (4),
weather to mitigate the risk of business failures. J. Bus. Res. 98, 391–402. 210–233.
Bin, O., Polaski, S., 2004. Effects of flood hazards on property values: Evidence before Giese, G., Lee, L., Melas, D., Nagy, Z., Nishikawa, L., 2019. How ESG affects equity
and after hurricane Floyd. Land Econom. 80 (4), 490–500. valuation, risk, and performance. J. Portfolio Manag. 45 (5), 69–83.
BIS, 2021. Climate-related risk drivers and their transmission channels. Basel Committee Hausman, J., 1978. Specification tests in econometrics. Econometrica 46, 1251–1271.
on Banking Supervision. Huang, H., Kerstein, J., Wang, C., 2018. The impact of climate risk on firm performance
Bland, B., Kwong, R., 2011. Supply chain disruption: Sunken ambitions. Financial and financing choices: An international comparison. J. Int. Bus. Stud. 49, 633–656.
Times. 3 november 2011. IPCC, 2021. AR6 Climate Change 2021: The Physical Science Basis. Contribution of
Bouslah, K., Kryzanowski, L., M’Zali, B., 2018. Social performance and firm risk: Impact Working Group I to the Sixth Assessment Report of the Intergovernmental Panel
of the financial crisis. J. Bus. Ethics 149 (3), 643–669. on Climate Change. Cambridge University Press.

13
M. Chabot and J.-L. Bertrand Journal of Financial Stability 69 (2023) 101190

Ivanov, I., Machiavelli, M., Santos, J., 2022. Bank lending networks and the propagation Pagliari, M., 2023. LSIs’ exposures to climate-change-related risks: An approach to assess
of natural disasters. Financial Manag. 51 (3), 903–927. physical risks. Int. J. Central Bank. 19 (1).
Jia, X., Ma, X., Xie, V., 2022. Expecting Floods: Firm Entry, Employment, and Aggregate Paltalidis, N., Gounopoulos, D., Kizys, R., Koutelidakis, Y., 2015. Transmission channels
Implications. National Bureau of Economic Research. Working Paper No. 30250. of systemic risk and contagion in the European financial network. J. Bank. Financ.
Kacperczyk, M., Peydró, J.-L., 2021. Carbon emissions and the bank-lending channel. 61 (Supplement 1), S36–S52, Global Trends in Banking, Regulations, and Financial
Center for Economic Policy Research. Discussion Paper DP16778. Markets.
Kanno, M., 2015. Assessing systemic risk using interbank exposures in the global Pankratz, N., Schiller, C., 2022. Climate Change and Adaptation in Global Supply-
banking system. J. Financial Stab. 20, 105–130. Chain Networks. The Federal Reserve. Finance and Economics Discussion Series.
Klomp, J., 2017. Flooded with debt. J. Int. Money Finance 73, 93–103. No. 2022-056. August 2022.
Kolstad, C., Moore, F., 2020. Estimating the economic impacts of climate change using Papadopoulos, G., 2022. Discrepancies in Corporate GHG Emissions Data and
weather observations. Rev. Environ. Econ. Policy 14 (1). their Impact on Firm Assessment Performance. JRC Technical Report. European
Kumar, L., Chhogyel, N., Gopalakrishnan, T., Hasan, K., Jayasinghe, S., Commission.
Kariyawasam, C., Kogo, B., Ratnayake, S., 2022. Climate change and future Parnaudeau, M., Bertrand, J., 2018. The contribution of weather variability to economic
of agri-food production. In: Bhat, R. (Ed.), Future Foods. Academic Press, pp. sectors. J. Bus. Res. 50 (43), 4632–4649.
49–79. Parsons, A.G., 2001. The association between daily weather and daily shopping
La Torre, M., Mango, F., Cafaro, A., Leo, S., 2020. Does the ESG index affect stock patterns. Australasian Mark. J. 9, 78–84.
return? Evidence from the Eurostoxx50. Sustainability 12 (6387). Pastor, L., Stambaugh, R.F., Taylor, L.A., 2021. Dissecting Green Returns. NBER
Laeven, L., Levine, R., 2009. Bank governance, regulation and risk taking. J. Financ. Working Paper No. 28940.
Econ. 93 (2), 259–275. Quayle, R., Diaz, H.F., 1980. Heating degree day data applied to residential heating
Larsen, P., 2006. Estimating the Sensitivity of U.S. Economic Sectors to Weather. Cornel energy consumption. J. Appl. Meteorol. 19, 241–246.
University, Ithaca, NY, USA, Orking Paper Cornel University, SSRN 900901. Ranger, N., Mahul, O., Monasterolo, I., 2022. Assessing Financial Risks from Physical
Lazo, J., Larsen, M.L.P., Waidmann, D., 2011. U.S. economic sensitivity to weather Climate Shocks: A Framework for Scenario Generation. IBRD/the World Bank
variability. Bull. Am. Meteorol. Soc. 92, 709–720. Technical Report.
Lepetit, L., Strobel, F., 2015. Bank insolvency risk and Z-score measures: A refinement. Reghezza, A., Altunbas, Y., Marques-Ibanez, D., Rodriguez d’Acri, C., Spaggiari, M.,
Finance Res. Lett. 13, 214–224. 2021. Do Banks Fuel Climate Change? ECB Working Paper 2021/2550.
Linden, F., 1962. Consumer markets: Merchandising weather. In: The Conference Board Reinders, H., Schoenmaker, D., Van Dijk, M., 2020. A finance approach to climate stress
Business Record. Vol. 19. No. 6. pp. 15–16. testing. SSRN 3573107.
Liu, Z., Quiet, S., Roth, B., 2015. Banking sector interconnectedness: what is it, how Roncoroni, A., Battiston, S., Escobar Farfan, L., Jaramillo, S.M., 2021. Climate risk and
can we measure it and why does it matter? Bank of England Publications Q. Bull. financial stability in the network of banks and investment funds. J. Financial Stab.
(Q2), 1–10. 54 (100870).
Mallucci, E., 2020. Natural disasters, climate change, and sovereign risk. FRB Schuldt, S., Nicholson, M., Adams, Y., Delorit, J., 2020. Upgrading destruction? How do
International Finance Discussion Paper. No 1291. climate-related and geophysical natural disasters impact sectoral FDI. Int. J. Clim.
Mandyck, J., 2022. What if banks had to disclose the climate impact of their Change Strategies Manag. 12 (2), 182–200.
investments? Harvard Bus. Rev. 20, 2022. Schuldt, S., Nicholson, M., Adams, Y., Delorit, J., 2021. Weather-related construction
Maunder, W., 1973. Weekly weather and economic activities on a national scale: An delays in a changing climate: A systematic state-of-the-art review. Sustainability
example using United States retail trade data. Weather 28 (1), 2–19. 13, 2861.
Merton, R., 1974. On the pricing of corporate debt: The risk structure of interest rates. Seltzer, L., Starks, L., Zhu, Q., 2022. Climate regulatory risks and corporate bonds.
J. Finance 29, 449–470. Federal reserve Bank of New York. No 1014.
Miller, S., Chua, K., Coggins, J., Mohtadi, H., 2021. Heat waves, climate change, and Simpson, A., 2022. Why the SEC is right to make climate risk disclosure mandatory.
economic output. J. Eur. Econom. Assoc. 19 (5), 237–247. Financial Times. 30 March 2022.
Monasterolo, I., 2020. Climate change and the financial system. Ann. Rev. Resour. Econ. Starr-McCluer, M., 2000. The Effect of Weather on Retail Sales. Technical Report,
12 (1), 299–320. Federal Reserve board of Governors, Washington, DC, SSRN: 221728.
Nagar, V., Schoenfeld, J., 2022. Measuring weather exposure with annual reports. Rev. Steele, A., 1951. Weathers effect on sales of a department store. J. Market. 15, 436–443.
Account. Stud.. Talbot, D., Boiral, O., 2018. The disclosure of industrial greenhouse gas emissions:
Newell, R., Prest, B., Sexton, S., 2021. The GDP-Temperature relationship: Implications A critical assessment of corporate sustainability reports. J. Bus. Ethics 147 (2),
for climate change damages. J. Environ. Econ. Manag. 108, 102445. 367–383.
NGFS, 2019. A Call for Action. Climate Change as a Source of Financial Risk. Technical TCFD, 2021. Task Force on Climate-Related Financial Disclosures. Guidance on Metrics,
report. Targets, and Transition Plans. TCFD Technical Report. October 2021.
NGFS, 2020. Network for Greening the Financial System. Guide to Climate Scenario Tonzer, L., 2015. Cross-border interbank networks, banking risk and contagion. J.
Analysis for Central Banks and Supervisors 2020. Technical report. Financial Stab. 18, 19–32.
Nordhaus, W., 2006. Geography and macroeconomics: New data and new findings. Vermeulen, R., Schets, E., Lohuis, M., Kölbl, B., Jansen, D., Heeringa, W., 2019. The
Proc. Natl. Acad. Sci. United States Am. 103, 3510–3517. Heat Is On: a Framework for Measuring Financial Stress Under Disruptive Energy
Noth, F., Schuwer, U., 2018. Natural disaster and bank stability: Evidence from the Transition Scenarios. DNB Working Paper, no 625.
U.S. financial system. SAFE Working Paper, No. 167. WEF, 2022. The Global Risks Report 2022 17th Edition. World Economic Forum
Noth, F., Schuwer, U., 2023. Natural disasters and bank stability: Evidence from the Technical Report.
U.S. financial system. J. Environ. Econ. Manag. 119. Weishi Gu, G., Hale, G., 2023. Climate risks and FDI. J. Int. Econ. 103731, (In Press,
Ortega, F., Taspinar, S., 2018. Rising sea levels and sinking property values: Hurricane Corrected Proof).
Sandy and New York’s housing market. J. Urban Econ. 106, 81–100. WMO, 2017. WMO guidelines on generating a defined set of national climate
Osina, N., 2019. Global liquidity, market sentiment, and financial stability indices. J. monitoring products. WMO-No. 1204.
Multinat. Financial Manag. 52–53. WRI, WBCCSD, 2021. The greenhouse gas protocol: A corporate accounting and
reporting standard. Revised version. Technical Report by the World Resources
Institute and the World Business Council for Sustainable Development.

14