(AGENPARL) – Tue 08 July 2025 Mercati, infrastrutture, sistemi di pagamento
(Markets, Infrastructures, Payment Systems)
The Rise of Climate Risks:
Evidence from Expected Default Frequencies for Firms
Number
July 2025
by Matilde Faralli and Francesco Ruggiero
Mercati, infrastrutture, sistemi di pagamento
(Markets, Infrastructures, Payment Systems)
The Rise of Climate Risks:
Evidence from Expected Default Frequencies for Firms
by Matilde Faralli and Francesco Ruggiero
Number 62 – July 2025
The papers published in the ‘Markets, Infrastructures, Payment Systems’ series provide
information and analysis on aspects regarding the institutional duties of the Bank of
Italy in relation to the monitoring of financial markets and payment systems and the
development and management of the corresponding infrastructures in order to foster
a better understanding of these issues and stimulate discussion among institutions,
economic actors and citizens.
The views expressed in the papers are those of the authors and do not necessarily reflect
those of the Bank of Italy.
The series is available online at http://www.bancaditalia.it.
Printed copies can be requested from the Paolo Baffi Library:
Editorial Board: Stefano Siviero, Paolo Del Giovane, Massimo Doria,
Giuseppe Zingrillo, Paolo Libri, Guerino Ardizzi, Paolo Bramini, Francesco Columba,
Luca Filidi, Tiziana Pietraforte, Alfonso Puorro, Antonio Sparacino.
Secretariat: Yi Teresa Wu.
ISSN 2724-6418 (online)
ISSN 2724-640X (print)
Banca d’Italia
Via Nazionale, 91 – 00184 Rome – Italy
Designed and printing by the Printing and Publishing Division of the Bank of Italy
THE RISE OF CLIMATE RISKS: EVIDENCE FROM EXPECTED
DEFAULT FREQUENCIES FOR FIRMS
by Matilde Faralli* and Francesco Ruggiero**
Abstract
The paper investigates the relationship between climate transition risk and credit risk by analysing
firms’ carbon emissions and Moody’s Expected Default Frequencies (EDFs). The results suggest that the
Paris Agreement was a turning point in the relationship between emissions and credit risk: following
the Agreement, the correlation between emission levels and EDFs became positive and statistically
significant. By decomposing the EDFs into their core components, increased asset volatility is found
to be the main channel through which transition risk affects credit risk for high-emissions companies.
The analysis sheds light on the mechanisms linking climate transition risk to financial risk. The
results are robust across different model specifications, control variables and geographic areas, and
indicate that climate-related financial risks have become increasingly important for credit markets.
JEL Classification: G30, G32, C13, H23.
Keywords: Climate Change, Credit Risk, EDF, Carbon Emissions, Transition Risk.
Sintesi
Il lavoro analizza la relazione tra rischio di transizione climatica e rischio di credito esaminando le
emissioni di carbonio delle imprese e le Expected Default Frequencies (EDFs) stimate da Moody’s. I
risultati suggeriscono che l’Accordo di Parigi ha rappresentato un punto di svolta nella relazione tra
emissioni e rischio di credito: successivamente a quell’accordo la correlazione statisticamente tra
livelli delle emissioni ed EDFs è divenuta positiva e statisticamente significativa. Scomponendo le
EDFs nelle loro componenti fondamentali, la maggiore volatilità degli attivi viene identificato come
il principale canale attraverso cui il rischio di transizione incide sul rischio di credito per le imprese
ad alte emissioni. L’analisi contribuisce a chiarire i meccanismi che collegano il rischio climatico di
transizione al rischio finanziario. I risultati si mostrano robusti rispetto a diverse specificazioni del
modello, variabili di controllo e aree geografiche, e indicano che i rischi finanziari legati al clima
hanno assunto una crescente rilevanza per i mercati del credito.
* Imperial College Business School.
** Banca d’Italia, Financial Risk Management Directorate.
CONTENTS
1. Introduction
2. Data and Methodology
2.1 Data
2.2 Methodology
3. Descriptive Statistics
4. Results
4.1 Carbon Intensity and Credit Risk
4.2 Other Measures of Credit Risk
4.3 Time Dynamics of Transition and Credit Risk
4.4 Mechanisms: EDF Components
4.5 Geographic Areas
5. Robustness
6. Conclusions
References
Appendix A
Appendix B
1. Introduction
With increasing attention devoted to climate issues, scholars have studied the effect of
climate change on the economy from several perspectives.1 A fast-growing strand of the literature investigates the relationship between climate-transition risk (i.e. the risks stemming
from implementing policy mitigating strategies) and firms’ credit risk. Measuring this link
is complicated because both variables suffer from endogeneity problems and measurement
errors. Using ESG ratings as a proxy for firms’ sustainability, earlier studies show that an
increase in ESG scores leads to lower CDS spreads (Barth et al., 2022), better credit ratings
(Devalle et al., 2017) and lower bond risk premia (Kotró and Márkus, 2020).
Other studies use carbon emissions to investigate the effect of transition risk on bond
ratings and yield spreads (Seltzer et al., 2022), option implied volatility slope (Ilhan et al.,
2021), CDS spreads (Blasberg et al., 2021) and market-implied distance-to-default (Bouchet
and Le Guenedal, 2020; Capasso et al., 2020; Carbone et al., 2021; Kabir et al., 2021). These
latter studies construct a measure of default risk by estimating Merton’s distance to default
(Merton, 1974). Capasso et al. (2020) and Kabir et al. (2021) find a positive correlation
between the probability of default and firms’ carbon footprints. Carbone et al. (2021) find
some evidence only when using relative carbon emissions (i.e. carbon intensity), but not
when using emissions in level.
Our paper builds on this latter stream of research. We study the relationship between
climate transition risk and credit risk by collecting data on emissions and Moody’s Expected Default Frequencies (EDFs) for 1,308 firms from 2008 to 2022. Moody’s EDF, a
market-implied probability that a firm will default, has a number of desirable features for
Acknowledgments: We thank the participants of the JRC Summer School on Sustainable Finance,
the PhD Informal Seminars at the University of Naples Federico II, and our colleagues at Banca d’Italia for
their valuable comments. The views expressed here are those of the authors and do not necessarily reflect
those of Banca d’Italia. Any remaining errors are our own.
For a summary of the related literature on climate finance, see Giglio et al. (2021)
our analysis. First, compared to studies that use computed distances to default, Moody’s uses
proprietary actual default data to obtain physical probabilities of default.2 Starting from the
risk-neutral distance to default obtained with an improved version of Merton (1974) model,
they project the computed distances to default onto actual default data, hence obtaining a
linear mapping from the risk-neutral probabilities into physical default probabilities. This
approach has the advantage of excluding risk-aversion adjustment components that complicate the inference on the data. Second, using the EDF we study the time dimension of credit
risk (i.e. whether climate risks affect the probability of a firm defaulting in 1, 5, or 10 years)
as well as the term structure of climate change risk. This provides us with information on
when and how transition risks are expected to materialize and how they will affect firms
in different sectors. Third, the EDF can be broken down into its main components: asset
volatility, the market value of assets, and the default point. This allows us to disentangle
the effect of carbon emissions on the different drivers of credit risk and thus indirectly on
the default probability. By construction, the EDF drivers fully explain the EDF variability.
Hence, by regressing EDF components on emission levels, we can show how carbon emissions
individually affect each EDF driver.
In examining how climate-related risks affect credit markets, we distinguish between
absolute emissions (measured as total Scope 1 emissions) and relative emissions (measured
as carbon intensity), and assess how the market perception of these measures changes over
time. Specifically, we regress Moody’s 1-, 5-, and 10-year EDFs on absolute direct emissions
and emission intensity. Our results indicate that, after accounting for firm fundamentals
and sectoral variations, absolute emissions have no effect on credit risk. Instead, we find
a positive relationship between carbon intensity and credit risk, consistent with previous
Risk-neutral probabilities are useful tools for pricing derivatives, as they assume market participants
are indifferent to risk and only care about expected returns. However, they fail to incorporate real-world
risks, such as climate risks, that can significantly impact firms’ financial performance. Hence, by taking into
account the uncertainty and potential impacts of climate risks, physical probabilities (i.e. historically-based)
allow for a more accurate assessment of a firm’s credit risk.
studies suggesting that more polluting firms within a sector face higher default risks (Blasberg
et al., 2021; Capasso et al., 2020; Carbone et al., 2021; Kabir et al., 2021). However, this
relationship is sensitive to specification choices and driven largely by upper-tail observations.
We further explore this issue using a quintile regression approach, which confirms that the
baseline patterns persist.
We hypothesize that absolute emissions became increasingly relevant for credit risk following the adoption of the Paris Climate Accords (also referred to as the Paris Agreement).
To test this, we examine whether the signing and ratification of the agreement marked a
structural break in policies, particularly from a corporate perspective. In other words, we
explore the relevance of the Paris Agreement as a potential determinant of increased credit
risk through its amplifying effect on climate transition risk. Our underlying hypothesis is
that for signatory countries guided by the agreement’s pledges, the probability of increased
transition risk should rise in response to stricter climate regulations, thereby increasing uncertainty for domestic firms.3 This effect should be more pronounced for firms with higher
carbon emissions (top quintiles), which may face greater challenges in complying with stricter
carbon regulations. This uncertainty would then directly translate into higher asset volatility,
consequently leading to a higher EDF.
In the aftermath of the Paris Agreement, we find that higher direct emissions lead to an
increase in the probability of default for firms within the same credit rating class. This finding
suggests that policies aimed at reducing carbon emissions are effective in reducing transition
risk and improving creditworthiness, highlighting the potential benefits of environmentally
responsible practices for firms.
We also provide evidence of the channels through which carbon emissions affect credit
risk by breaking-down EDF into its core components: asset volatility, the market value
Although the agreement does not legally bind signatory countries to enforce its pledges, several authors
have shown that investors and financial markets react to the signaled increased global commitment on climate
action (see Monasterolo and De Angelis, 2020; Kruse et al., 2020; Seltzer et al., 2022).
of assets, and default point. First, we document how each component enters linearly into
the EDF. We show how, by construction, higher asset volatility is associated with a higher
probability of default, while higher default points and market value of assets reduce default
risk. Then, we analyze the effect of carbon emissions on each EDF component. We find that
the increased default risk of high emitters is driven by an increase in asset volatility after
the Paris Agreement. Finally, we corroborate our hypothesis that firms with high credit
risks are penalized only when there are regulations in place that internalize the costs of
polluting through an analysis focused on geographical differences. Different countries and
regions have varying regulations and policies to address climate change, which can influence
the speed and scale of the transition to a low-carbon economy. For instance, the US has
been less stringent in implementing carbon reduction policies compared to other countries,
while European carbon policies are more stringent and accurately tailored for various sectors.
Consistent with our intuition that looser regulation on carbon emissions does not penalize
polluting firms, we do not find a clear relationship between emissions and credit risk for US
firms. However, when examining the EU sample, we observe that large emitters face a higher
credit risk premium compared to their peers. This result is also confirmed when looking at
the effect of carbon price surprises connected to the EU Emission Trading System (EU ETS)
on credit risk.
To test the robustness of the findings, we conduct a series of supplementary analysis
to ensure that our results are not driven by the choice of credit or transition risk proxies.
We show that our findings are consistent when using alternative winsorization thresholds,
and also different measures of credit risk, including CDS spreads, CDS-implied ratings, and
Moody’s ratings. In all cases, we observe a positive correlation between credit risk and carbon intensity, but no significant relationship with absolute emissions. We also acknowledge
that the EDF, given its sensitivity to market expectations and future credit risk related to the
transition to a low-carbon economy, may reflect forward-looking transition risk rather than
the long-term, backward-looking impact of carbon emissions, which is typically represented
by absolute emissions (Bolton and Kacperczyk, 2021b). To address the forward-looking
dimension of transition risk, we use Trucost’s carbon earnings-at-risk—which captures the
potential financial cost firms may incur under future carbon pricing—and the announcement
of science-based targets, which are voluntary corporate commitments to reduce greenhouse
gas emissions in line with the goals of the Paris Agreement. We perform additional robustness checks to test the consistency of the relationship between carbon emissions and
the probability of default. Our results indicate that forward-looking transition risks —measured by carbon earnings-at-risk relative to EBITDA— are positively associated with higher
default probabilities. We also find that medium- and long-term EDFs temporarily decline
around the announcement of science-based targets for high emitters, suggesting a short-lived
reduction in perceived credit risk. Furthermore, the findings remain robust to the inclusion
of lagged emission levels, although we observe no significant effect from the short-term rate
of change in carbon emissions.
To the best of our knowledge, this is the first paper in the related literature that using
Moody’s EDFs to proxy the probability of default provides insights into the drivers of climate
transition risks on firm-level credit risk before and after the Paris Agreement.4 This study
complements the existing literature in two key aspects. First, we highlight that the signing of
the Paris Agreement and the subsequent rise in investors’ awareness of climate risks marked a
structural break in policies, particularly from a corporate perspective. The Paris Agreement
appears to be a significant determinant of increased credit risk through its amplifying effect
on climate transition risk. Second, we identify the channels through which carbon emissions
affect firms’ probability of default, along with the relative weights of each. Our analysis
reveals that this influence primarily occurs through increased asset volatility, especially for
high-emission firms. Finally, we provide additional evidence that climate regulation strongly
In a different research setting, Acharya et al. (2022) use Moody’s EDFs to show that heat stress exposure
increases credit risk of municipal as well as corporate bonds.
impacts the integration of climate risks into credit risk assessments by offering a comparative
analysis of firms in the US and the EU, which are subject to different sets of rules and
requirements concerning climate risk.
Related Literature. In the remaining part of this section, we provide a survey of how
previous studies dealt with quantifying firms’ climate-related risks and their main findings.
Earlier studies addressed this issue using corporate social responsibility metrics (e.g.
Stellner et al., 2015) and environmental scores (Höck et al., 2020). For example, Höck et al.
(2020) show that a higher environmental score leads to lower CDS spreads for firms with
ex-ante high creditworthiness, low leverage, and high market capitalization. In contrast,
Stellner et al. (2015), investigate the effect of higher corporate social responsibility (CSR)
scores on credit ratings and zero-volatility spreads (z-spreads); they find that stronger results
are driven by countries’ ESG performance, suggesting that the regulatory environment allows
a larger reduction in credit risk when companies display a higher CSR score.
Other studies used ESG ratings to show how an increase in ESG scores leads to lower
CDS spreads (Barth et al., 2022), better ratings by Moody’s (Devalle et al., 2017), and lower
bonds’ risk premia (Kotró and Márkus, 2020). Similarly to Höck et al. (2020), also Barth
et al. (2022) conclude that higher ESG ratings correlate with lower credit risk (proxy by
CDS spreads), with a stronger effect for European firms and for firms with medium ESG
ratings.5
One of the criticisms often raised against the use of ESG data is that they are unstandardized, non-compulsory and not fully transparent in the way they are constructed (see,
for example, Berg et al., 2022), thus making it hard to disentangle the drivers of their effect
on credit risk.
Another relevant paper belonging to this strand of literature is Henisz and McGlinch (2019). They show
that previous years’ higher ESG ratings, using RavenPack’s reported news, are strongly correlated with lower
future assets volatility.
To respond to these criticisms, scholars explored different paths. Seltzer et al. (2022)
study the effect of poor environmental profiles or high carbon footprints on credit rating
scores and bond spreads for firms around the 2015 Paris Agreement. They find that firms
with high pre-existing emissions present worse scores and higher spreads, with more pronounced effects in strictly regulated US states. Blasberg et al. (2021) study the correlation
between CDS spreads and transition risk, proxied by carbon intensity and emissions. They
find that climate risk has a heterogeneous effect across sectors and on the term structure of
firms’ credit risk. Ilhan et al. (2021) provide evidence that an increase in carbon intensity
leads to a larger option implied volatility slope, in particular for left tail regions.
Our paper relies on previous studies in choosing carbon disclosures as a proxy of climate
risk, more precisely transition risk. However, a recent strand of literature relies on the
construction of climate-corrected ratings. Kölbel et al. (2024) train an AI algorithm for
languages to see whether regulatory risk disclosures affect CDS spread. They find that
while disclosing transition risks increases CDS spreads, especially after the Paris Climate
Agreement of 2015, disclosing physical risks decreases CDS spreads. Other related papers
such as Klusak et al. (2023) construct a model similar to S&P’s such as to incorporate
climate physical risks into sovereign ratings for possible future climate scenarios. Sautner
et al. (2023) use quarterly earnings calls to construct an annual firm-level measure of firms’
exposure to climate. Further work might use one of these novel firms’ climate exposure
variables to corroborate our findings further.
Finally, it is important to note that our paper specifically addresses the risks associated with the potential implementation of mitigation strategies. However, there is an everexpanding body of literature that delves into the consequences of directly implementing
carbon taxes on firms’ default risks. For instance, Di Virgilio et al. (2023) conducted a study
investigating the impact of different levels of carbon taxes on energy prices and revealed
that the probability of default for 200,000 non-financial Italian firms was only minimally
affected. Similarly, Aiello and Angelico (2023) observed that the imposition of a carbon tax
had a modest impact on default rates for Italian banks, which remained below historical
averages. These findings align with the growing consensus, prompting several central banks
to undertake stress tests to comprehensively evaluate the quantifiable impact of climate risks
on financial institutions.
Our paper is structured as follows. Section 2 presents the data and methodology. Section 3 provides descriptive evidence of the relationship between carbon emissions and the
probability of default. Section 4 discusses the empirical findings, and Section 5 reports the
results of a battery of robustness checks. Finally, Section 6 concludes the paper.
2. Data and Methodology
2.1. Data
The dataset is constructed by gathering data from four main databases: carbon emissions
are retrieved from MSCI, EDFs and ratings from Moody’s CreditEdge, CDS spreads and
stock prices are obtained from Refinitiv, and balance sheet information is sourced from CRSP
and Compustat. We also collect carbon-earnings-at-risk data from Trucost to enhance our
analysis with a forward-looking measure of transition risk..
Our initial sample comprises all firms with yearly carbon emissions listed in the MSCI
database for three advanced economies: the United States, United Kingdom, and the European Union, covering the period from 2008 to the end of 2022. We match the data with
Moody’s CreditEdge to obtain monthly Expected Default Frequencies (EDFs), EDF components (asset volatility, market value of assets and default point) and Moody’s credit ratings.
Additionally, we incorporate quarterly balance sheet data from Compustat Global and North
America via WRDS.6
We also collect 5-year single-name CDS spreads for unsecured debt with the “Modified
Modified Restructuring” clause (MM14) from 2008 to 2023. All CDS spreads are in US
When possible we impute missing values using previous quarter values.
dollars. Following Gao et al. (2021) the data are aggregated at the monthly level by taking
the mean over the month within each entity. We exclude all CDSs with a spread higher than
4000 basis points (Zhang et al., 2009) and illiquid CDS (Blasberg et al., 2021).7
As part of our strategy to identify the effect of climate on firms’ credit risk, we apply
several filters to the sample analyzed. We winsorize EDF, EDF components, CDS Spread,
absolute emission and emission intensity at the 5% level to avoid results driven by a few
distressed companies, extremely high carbon emitters, or incorrectly reporting zero emissions.
We then discard firms with less than 7 years of complete data. Finally, we exclude firms in the
financial sector, public administration and other services sectors to avoid misinterpretation
of the outcomes driven by these entities’ significantly different financial behavior.
The final dataset includes 1,308 firms from 2008 to 2022, containing monthly EDFs,
quarterly balance sheet information, and yearly carbon emissions. For the analysis, we
additionally use two sub-samples: one consisting of 205 firms with monthly CDS spreads,
allowing comparisons with previous studies, and another comprising 615 firms with Moody’s
ratings to control for credit risk.
2.2. Methodology
To study how transition risks influence firms’ credit risks, our primary outcome variables
are Expected Default Frequency (EDF) over 1-year, 5-year, and 10-year horizons. EDF
represents the likelihood of a firm’s default within a specified period. Our independent variables of interest are absolute emissions and emissions intensity as proxies for transition risks.
Absolute emissions reflect alignment with zero-carbon emissions goals outlined by carbon
policies (Bolton and Kacperczyk, 2021a), while emissions intensity is included to address
potential reliance on intensity measures in market valuation as well to ensure comparability
across firms (Hartzmark and Shue, 2022; Zhang, 2025).
The results are consistent when using median and end-of-the-month CDS spreads.
We estimate the baseline regression model using a high-dimensional fixed effects methodology, specified as follows:
EDFi,t “ α ` β1 ˚ Emissionsi,t ` δXi,t´1 ` F E ` ϵi,t
where EDF is the dependent variable measured at monthly frequency, and the key independent variables are yearly emissions—measured as LogpScope1q and Carbon Intensity.
The vector Xi,t´1 contains control variables at in the previous quarter, specifically firm size
(logarithm of total assets), debt ratio, operating margin ratio, and capital intensity. Additionally, we include quarterly intangible capital to control for firm-level innovation and
efficiency. Given that default probability is influenced by several firm-specific factors, these
control variables help isolate the effect of the climate variable on each firm’s default probability.
To address the potential concern that contemporaneous absolute emissions might primarily reflect a firm’s sales activity (Zhang, 2025), we also include current log sales as a control
variable. Since emissions data are recorded annually, we use yearly sales data for consistency.
Other control variables, however, are measured quarterly and lagged by one quarter.8
Lastly, we include a fixed effects (F E) matrix at the country, sector, and calendar-year
levels. Country-fixed effects control for variations in economic, regulatory, and institutional
factors across different nations, while sector-fixed effects account for sector-specific dynamics,
such as varying levels of carbon intensity or default risk inherent to certain industries. Yearfixed effects capture significant events, such as the 2007–2008 financial crisis or the COVID-19
pandemic, both of which globally increased default probabilities. To account for within-firm
correlation, which arises because EDF is measured monthly while emissions are reported
yearly, we follow standard practice in the literature and cluster standard errors at the firm
Another concern is that emissions are released with a 10-12 months lag (Zhang, 2025). For robustness,
table B10 replicates the baseline analysis with lagged variables.
level.
3. Descriptive Statistics
The final dataset encompasses 1308 firms, offering a broad coverage across geographical
regions and sectors. Geographically, the sample includes 58% of firms from the United States,
30% from the Euro Area, and 12% from the United Kingdom. Sector-wise, around 46% of
the companies are from the manufacturing sector, followed by 9% in the information sector,
with the fewest firms in agriculture, management, and education sectors.
Table 1. Summary Statistics
Variables
Median
1 Year EDF (%)
5 Year EDF (%)
10 Year EDF (%)
Asset Volatility
Log(Market Value Assets)
Log(Default Point)
Mean CDS Spreads
Moody’s Ratings
Derived CDS Ratings
Log(Scope 1)
Carbon Intensity
Log(Current Sales)
Debt Ratio
Operating Margin Ratio
Capital Intensity
Intangible Assets
21.83
112.98
10.10
10.49
-0.49
20.29
77.51
10.25
96.17
28.28
40.89
11.40
10.14
27.25
394.43
21.00
21.00
15.94
13.32
13.65
-4189.50 36.05
Observation
213,555
213,555
213,555
213,555
213,555
213,555
34,159
82,848
43,205
17,909
17,909
17,909
71,246
71,246
71,246
71,246
71,246
Note: 1-Year EDF, 5-Year EDF, 10-Year EDF, 5-Year CDS Spreads, Log(Scope 1) and Carbon Intensity are
winsorized at the bottom and top 5%. The number of observations varies depending on the data frequency:
EDF and EDF components (asset volatility, market value of assets, and default point), CDS spreads, and
ratings are monthly variables. Log(Scope 1), Carbon Intensity and Log(Current Sales) are reported at
the firm-year level (17,909 observations), and fundamentals are reported at the firm-quarter level (71,246
observations). See Table B1 for variables’ description and sources.
Figure 1. EDF Trends by Absolute Emissions and Emission Intensity: Quintile and Time Series Comparisons
1-Year EDF (%)
EDF (%)
Mean 5-Year EDF(%)
Quintile of absolute emissions
Mean 1-Year EDF(%)
Mean 10-Year EDF(%)
1st Quintile Log(Scope 1)
(a) EDF by absolute emission
5th Quintile Log(Scope 1)
(b) 1-Year EDF by quintiles of absolute emission
1-Year EDF (%)
EDF (%)
Mean 5-Year EDF(%)
Quintile of emission intensity
Mean 1-Year EDF(%)
Mean 10-Year EDF(%)
1st Quintile Emission Intensity
(c) EDF by emission intensity
5th Quintile Emission Intensity
(d) 1-Year EDF by quintiles of emission intensity
The figures present the average 1-Year EDF, 5-Year EDF, and 10-Year EDF by quintiles of absolute emissions
(a) and emission intensity (c), as well as the time series of the 1-Year EDF for the lowest and highest quintiles
of absolute emissions (b) and emission intensity (d). Emissions are measured as the logarithm of Scope 1
emissions, and emission intensity is defined as Scope 1 emissions divided by sales. Quintiles of emissions are
estimated within each year. Higher quintiles of emissions represent larger emitters. The EDF and emissions
are winsorized at the top and bottom 5%.
Table 1 reports the summary statistics for the main variables in our analysis (see Table
B1 for an in-depth description of the variables used in the analysis). The average firm in our
sample has a 1-year probability of default of 0.37% and a debt ratio of 27%. As anticipated,
we observe a positive term structure of the EDF, where 10-year EDF has a higher mean and
lower standard deviation (1% and 0.79%) compared to the 5-year EDF (0.73% and 0.75%)
and the 1-year EDF (0.37% and 0.61%).
To explore how emissions relate to credit risk, Figures 1 (a) and (c) show the average
1-year, 5-year, and 10-year EDFs divided by quintiles of total direct emissions and emission
intensity, respectively. Figures 1 (b) and (d) display the time series of the 1-year EDF from
2008 to 2022 for the first and fifth quintiles.
In panel (a), we observe that firms in the upper quintiles (high emitters) tend to have
a lower probability of default than firms in the first quintile. This result is striking, as
it challenges the common expectation that the market penalizes firms with higher carbon
footprints. While a size effect could contribute to this descriptive evidence, the fact that this
effect persists over longer horizons, even when using carbon intensity in panel (c) —where
emissions are scaled by sales, inherently accounting for firm size— adds robustness to this
finding.
Turning to the time series, panel (b) reveals that the average 1-year EDF for firms in the
top (bottom) quintile significantly increased (decreased) after 2015. This aligns with previous
studies highlighting the substantial impact of the 2015 Paris Agreement on shaping investors’
and policymakers’ perceptions of firms’ risks (Bolton and Kacperczyk, 2021b; Carbone et al.,
2021; Capasso et al., 2020; Seltzer et al., 2022; Barth et al., 2022; Kölbel et al., 2024). A
two-sample t-test confirms that the average 1-year EDF for firms in the bottom quintile is
not statistically different from those in the top quintile between 2015 and 2021. However,
the difference becomes statistically significant again in 2022.
Panel (d) shows that for emission intensity, the EDFs of the first and fifth quintiles
become statistically indistinguishable as early as 2011. However, firms in the top quintile
exhibit significantly higher risk in 2015, 2016, 2017, and 2020. The appendix (Figure A1)
provides a detailed view of all quintiles for both total direct emissions and emission intensity,
confirming similar trends over longer horizons.
4. Results
4.1. Carbon Intensity and Credit Risk
We begin by examining the relationship between carbon emissions and expected default
frequencies (EDFs) at 1-year, 5-year, and 10-year horizons. Our main variables of interest are
log(Scope 1) emissions and carbon intensity. Log(Scope 1) measures direct carbon emissions
from sources owned or controlled by the firm, while carbon intensity, calculated as Scope
1 carbon emissions over total sales, reflects a firm’s efficiency in managing carbon output
relative to its economic activity.
Table 2 summarizes the baseline relationship between emissions metrics and EDFs. In
detail, Columns (1) and (2) of Table 2 suggest a negative correlation between both absolute
emissions and emission intensity with default probability across all horizons. For every 1%
increase in absolute emissions, the 1-year EDF decreases by 0.0180 percentage points, while
the 5-year and 10-year EDFs decrease by 0.0550 and 0.103 percentage points, respectively.
Similarly, higher carbon intensity is associated with lower default probability in the medium
(5-year) and long term (10-year), by 0.23 and 0.049 percentage points respectively.
However, once firm-level controls and sector fixed effects are introduced in columns (3)
and (4) to account for firm fundamentals and within-sector variation, the results change
notably. In the full specification, the effect of absolute emissions becomes statistically insignificant, while carbon intensity shows a positive and significant relationship with credit
risk. Specifically, a 1% increase in carbon intensity corresponds to increases of 0.016, 0.023,
Table 2. Analysis using Emission Levels and Intensity
Panel A:
1-Year EDF
Log(Scope 1)
-0.018***
(0.004)
-0.005
(0.005)
Panel B:
-0.055***
(0.005)
Carbon Intensity
-0.023***
(0.008)
-0.103***
(0.006)
Carbon Intensity
Year FE
Country FE
Industry FE
Controls
0.023**
(0.009)
10-Year EDF
0.014
(0.009)
Panel C:
Log(Scope 1)
0.016**
(0.007)
5-Year EDF
Log(Scope 1)
0.008
(0.007)
Carbon Intensity
0.004
(0.009)
-0.049***
(0.008)
0.019**
(0.009)
213,555
213,555
213,555
213,555
Note: The table reports the regression of current log absolute emission and carbon intensity on EDF. The
controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and current-year
log sales. EDF and emissions are winsorized at the bottom and top 5%. Standard errors in parentheses are
clustered at the firm level. Statistical significance is reported as * for p ă 0.10, ** for p ă 0.05 and *** for
p ă 0.01.
and 0.019 percentage points in the 1-year, 5-year, and 10-year EDF, respectively.9 10
Taken together, these findings suggest that relative emissions, rather than absolute levels,
may have informational value in explaining credit risk—but this result is highly sensitive to
the treatment of extreme values. In particular, the apparent positive relationship between
carbon intensity and EDFs is not robust to alternative winsorization choices, raising caution
about interpreting it as evidence of a systematic link. To further investigate heterogeneity in
results, we perform a quantile regression analysis, as detailed in Table B3 in the appendix.
Columns (1) and (2) —without controls and fixed effects—show a negative correlation between emissions and EDFs. Once controls are included, the relationship turns positive. Figure A2 plots the coefficients across EDF quantiles (5th to 95th percentile) and shows that the
positive association becomes statistically significant only above the 60th percentile—i.e., for
firms with higher transition risk. This effect is more pronounced for emission intensity, while
the impact of total emissions remains generally weak and statistically insignificant across the
distribution. These results suggest that transition risk, as proxied by carbon intensity, may
matter more for firms with higher credit risk. Although a positive relationship is frequently
documented in the literature, we find that it holds predominantly for firms at the extreme
end of the emission intensity distribution, questioning the robustness of this finding across
the full sample.
Several factors could explain the observed non-significant or even negative relationship
between emissions and EDFs. First, high-emission firms often operate in carbon-intensive
industries with substantial entry barriers, economies of scale, and market power, potentially
All variables are winsorized at the 5th and 95th percentiles in the baseline specification. Table B2 reports
results without winsorization. We observe that when sector and control variables are included, the positive
effect of carbon intensity disappears in the non-winsorized version suggesting that coefficients estimated in
Table 2 are largely driven by a small number of extreme observations in the upper tail of the carbon intensity
distribution.
To address concerns that the observed results may be overstated due to the differing temporal dimensions
of EDF and Emissions, Table B13 presents the baseline estimates using EDF aggregated at the yearly level.
The results remain consistent under this alternative specification.
reducing competitive pressure and default risk. Second, these firms might be more aware
of climate risks and take proactive mitigation measures, such as investing in low-carbon
technologies or diversifying activities. Third, the results in the full sample might be driven, at
least partially, by the presence of firms subject to different climate-related regulations, which
could affect the degree of integration of climate risks into credit risk for those firms. Fourth,
firms might benefit from implicit or explicit government subsidies or consumer tolerance
for carbon emissions, enhancing their cash flows and reducing default risk. We extend our
analysis to account for some of these explanations in detail later in the paper.
In the following section, we validate our findings using alternative measures of creditworthiness, including Mean CDS Spreads, Moody’s Ratings, and CDS implicit ratings.
4.2. Other Measures of Credit Risk
To ensure comparability with previous studies, we replicate our initial analysis using
alternative measures of credit risk. We begin by examining the correlations between our
primary measure, the Expected Default Frequency (EDF), and other commonly used credit
risk variables. Table 3 presents the pairwise correlations among 1-year, 5-year, and 10-year
EDF; Mean CDS Spreads; Moody’s Ratings; and CDS Implied Ratings. All correlation coefficients are statistically significant at the 1% level. Notably, CDS Spreads exhibit strong
correlations with all other credit risk measures. Moody’s Ratings and CDS Implied Ratings show particularly high correlations with the 10-year EDF (0.68 and 0.77, respectively),
although their correlations with the 1-year EDF are notably lower (0.51 for both).
Table 4 replicates our initial baseline specification with year, country, and sector fixed
effects, using alternative measures of credit risk. These measures include CDS spreads and
ratings from Moody’s and CDS-implied sources, where lower values of the rating variable
indicate better creditworthiness. It is important to highlight that this analysis is based on a
smaller sample size, comprising 205 firms with CDS spreads, 615 firms with Moody’s ratings,
and 343 firms with CDS-implied ratings.
Table 3. Pairwise correlation of credit risk variables
Variables
(1) CDS Spreads
(1) CDS Spreads
(2) Moody’s Ratings
0.68***
(3) CDS Implied Ratings
0.77***
(4) Y1-EDF
0.65***
(5) Y5-EDF
0.68***
(6) Y10-EDF
0.65***
(2) Moody’s Ratings
(3) CDS Implied Ratings
(4) Y1-EDF
(5) Y5-EDF
(6) Y10-EDF
0.74***
0.51***
0.62***
0.66***
0.51***
0.60***
0.62***
0.90***
0.75***
0.94***
Note: The table reports the pairwise correlation across 1, 5 and 10-year EDF, Mean CDS Spreads, Moody’s
Ratings and CDS Implied Ratings. Statistical significance are reported such as * for p ă 0.10, ** for p ă 0.05
and *** for p ă 0.01.
Table 4 shows that while absolute emissions exhibit no significant relationship with CDS
spreads or Moody’s ratings (though they correlate with CDS-implied ratings), carbon intensity remains positively associated with all credit risk measures. This pattern suggests that
credit markets place greater emphasis on firms’ emission efficiency relative to their economic
output than on their absolute emission levels.
Table 4. Other measures of credit risk
Mean CDS
Log(Scope 1)
3.913
(3.325)
Carbon Intensity
Year, Country & Industry FE
34,159
8.840***
(2.909)
34,159
Moody’s Ratings
CDS implied Ratings
0.074
(0.067)
0.232**
(0.113)
82,848
0.140***
(0.051)
82,848
43,205
0.288***
(0.077)
43,205
Note: The controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and
current-year log sales. The Mean CDS, Log(Scope 1) and Carbon Intensity are winsorized at the bottom
and top 5%. The sample spans the years from 2008 to 2022, including 205 firms with CDS spreads, 615 firms
with Moody’s ratings, and 343 firms with CDS-implied ratings. Standard errors in parentheses are clustered
at the firm level. Statistical significance are reported such as * for p ă 0.10, ** for p ă 0.05 and *** for
p ă 0.01.
Interestingly, the positive relationship with credit risk only appears when emissions are
considered in relation to economic activity (carbon intensity) rather than in absolute terms,
similar to Carbone et al. (2021) and Blasberg et al. (2021). This observation supports our
earlier findings, but also highlights a potential issue with market incentives. The main goal
should be to reduce absolute emissions, not just improve relative efficiency.
To investigate whether the Paris Agreement influenced this paradigm by prioritizing absolute emission reduction for both governments and companies, the following section examines
how the relationship between emissions and expected default frequency evolves annually
throughout our sample period.
4.3. Time Dynamics of Transition and Credit Risk
The 2015 Paris Agreement marks a significant turning point in collective awareness of
climate risks. We investigate whether the effect of emissions on credit risk changes around
this period, motivated by Figures 1 (b) and (d), which show a structural change in 1-year
EDFs between top and bottom emission quintiles after 2015. Following the methodology of
Acharya et al. (2022), we estimate:
EDFi,t “ γi `γt `
Iy rβy Emissioni,t `θy Ratingi,t s`βEmissioni,t `θRatingi,t `θXi,t `ϵit (2)
y“2009
The dependent variables are the 1-year, 5-year, and 10-year EDFs for firm i at time t. The
coefficients of interest, βy, capture the year-by-year sensitivity of EDF to Emissions—both
absolute and relative (intensity)— compared to the base year 2008. We control for credit
quality by including Moody’s ratings interacted with year indicators and add firm characteristics, including size, debt ratio, operating margin ratio, capital intensity, intangible assets,
and log current-sales. The specification includes firm and year fixed effects, with standard
errors clustered at the firm level.
Figure 2 shows the interaction coefficients between log(Scope 1) emissions, carbon intensity, and time (using 2008 as the baseline year), along with their 95% confidence intervals.
For all EDF horizons, the coefficients are generally insignificant before 2015, with the exception of the 1-year EDF and emission intensity (Figure b), which displays a positive trend
Figure 2. EDF change around the Paris Agreement
1-Year EDF
1-Year EDF
(a) 1-Year EDF & Absolute Emission
(b) 1-Year EDF & Carbon Intensity
5-Year EDF
10-Year EDF
(d) 5-Year EDF & Carbon Intensity
(c) 5-Year EDF & Absolute Emission
10-Year EDF
5-Year EDF
(e) 10-Year EDF & Absolute Emission
(f ) 10-Year EDF & Carbon Intensity
Note: the figure presents yearly interaction coefficients of absolute emissions (left column) and carbon
intensity (right column) with 1-year, 5-year, and 10-year EDFs. The base year is 2008. Coefficients are
estimated using firm and year-fixed effects, with 95% confidence intervals displayed.
beginning in 2011. However, we observe a marked shift after 2015, with coefficients becoming
strongly positive and significant across all horizons before declining in the last three years
of the sample.11
The evidence suggests that economic actors internalized implicit climate risk costs following the Paris Agreement, either through direct emission reduction commitments or higher
pollution costs imposed by signatory countries. Tables B4 and B5 in the Appendix show the
coefficients for absolute emissions and emission intensity increase substantially in magnitude
and significance after 2015, nearly doubling in 2016 across all EDF horizons. Although these
effects remain significant in subsequent years, they gradually decrease, possibly reflecting
uncertainty about the agreement’s implementation.
We propose two possible non-mutually exclusive interpretations of these findings. First,
the Paris Agreement signaled heightened expectations of future carbon regulation and taxation, potentially increasing costs and reducing revenues for high-emission firms, thereby
raising their default risk. Second, the Agreement may have shifted stakeholder preferences,
reducing demand for high-emission firms while boosting support for low-carbon alternatives,
creating a divergence in default risk profiles.
To further investigate the relationship between carbon emissions and credit risk, the
next section decomposes the Expected Default Frequency (EDF) into its key structural
components and examines the impact of emissions on each. This component-level analysis
enables us to identify the channels through which carbon emissions influence firms’ credit
risk, thereby shedding light on the underlying drivers of the observed aggregate effect.
Since the regression includes Moody’s ratings to control for credit risk across firms, the sample size is
smaller than that used in section 4.1, as only 615 firms have ratings. Figures A3 (which do not include
ratings) reveal an even larger effect, showing that higher emitters exhibit greater credit risk compared to
lower emitters as early as 2011, relative to 2008. This difference continues to grow until 2015, after which it
stabilizes.
4.4. Mechanisms: EDF Components
The EDF measures the probability of a firm defaulting over a certain time horizon. It
is computed as the probability that the value of the firm falls below a certain threshold (its
liabilities payable, also defined as the default point) within a certain time, using an extended
version of the Merton (1974) model. Standard EDFs incorporate balance sheet data and
market data, thus they tend to express a “market based” default frequency over a given
horizon. Three primary components determine a firm’s EDF: asset volatility, market value
of assets, and the default point. The model employs an iterative approach to simultaneously
estimate the value of the assets and their volatility. Once these are determined, the distance
to default (DTD) is calculated as the number of standard deviations separating the current
value of the assets from the default point; the default point is an estimate of the level of the
market value of a company’s assets below which the firm would fail to make scheduled debt
payments. Finally, the DTD is converted into a PD using a cumulative normal distribution
and then calibrated using Moody’s historical default data to obtain the EDF.
To understand how transition risks influence EDF, we begin by examining how each EDF
component contributes to the overall default probability. Using a series of OLS regressions,
we assess the role of these components in explaining variations in EDF. While this approach
simplifies the complex relationships underlying EDF dynamics, it provides valuable insights
into the channels through which carbon emissions may affect credit risk. The resulting
coefficients can be interpreted as the relative weights of each component in explaining EDF
variability.
Table 5 presents the estimated coefficients from a series of regressions analyzing the
individual effects of each EDF component. Columns (1) to (3) include each component
separately, revealing that all coefficients are highly significant. Asset volatility exhibits
a strong positive association with EDF. Conversely, higher market values of assets and
default points are associated with lower EDF. While the sign of the market value of assets
aligns with theoretical expectations—since larger asset values act as a buffer against financial
Table 5. The weights of EDF’s Components
1-Year EDF
Asset Volatility
0.015***
(0.001)
Log(Market Value)
5-Year EDF
10-Year EDF
0.044***
(0.002)
0.062***
(0.002)
0.067***
(0.002)
-0.628***
(0.022)
-0.864***
(0.026)
-0.836***
(0.022)
-0.016***
(0.006)
0.612***
(0.022)
0.787***
(0.026)
0.688***
(0.022)
0.479***
(0.043)
213,555
0.465***
(0.066)
213,555
1.213***
(0.086)
213,555
1.878***
(0.093)
213,555
-0.111***
(0.007)
Log(Default Point)
Constant
1-Year EDF
0.032
(0.030)
213,555
1.320***
(0.062)
213,555
Note: The dependent variable is the 1-Year EDF in columns 1 to 4, the 5-Year EDF in column 5, and the
10-Year EDF in column 6. All variables are winsorized at the top and bottom 5%. The regressions do not
include fixed effects or additional controls. Standard errors in parentheses are clustered at the firm level and
statistical significance is reported as * for p ă 0.10, ** for p ă 0.05 and *** for p ă 0.01.
distress—the negative coefficient on the default point is more difficult to interpret. A higher
default point should, in principle, imply a higher likelihood of default, and thus the negative
coefficient in column (3) may reflect omitted variable bias due to the exclusion of asset
volatility and market value.
We therefore focus on columns (4) to (6), where all EDF components are included jointly.
Across all specifications, asset volatility is positively and significantly associated with EDF,
with coefficient magnitudes increasing with the time horizon. For example, the coefficient
on asset volatility rises from 0.044 in column (4) for the 1-Year EDF to 0.067 in column
(6) for the 10-Year EDF. The market value of assets remains negatively and significantly
associated with EDF, capturing the role of firm size and financial strength in reducing default
probability. The magnitude of this effect is relatively stable across maturities. The default
point, which proxies for a firm’s debt obligations or leverage, displays a horizon-dependent
relationship with EDF. In the 1-Year EDF specification (column 4), the coefficient is 0.612,
while for longer horizons (columns 5 and 6), the coefficient increases substantially, indicating
that as the default point rises, so does the likelihood of default—consistent with theoretical
expectations.
Table 6. Pre and Post Paris Agreement for EDF Component
Asset Volatility
Log(Scope 1)
PA*Log(Scope 1)
-0.577***
(0.076)
0.100***
(0.038)
Log(Market Value of Assets)
Log(Default Point)
-0.040***
(0.008)
0.010*
(0.005)
-0.010***
(0.004)
0.010***
(0.003)
Carbon Intensity
-0.163**
(0.065)
-0.013**
(0.006)
0.006
(0.004)
PA*Carbon Intensity
0.051
(0.044)
213,555
-0.013***
(0.004)
213,555
0.009***
(0.003)
213,555
Year, Country and Industry FE
Controls
213,555
213,555
213,555
Note: the dependent variables are Asset Volatility, Market Value of Assets, and Default point. The table
reports coefficients for Log(Scope 1) and Carbon Intensity, and their interaction with the Paris Agreement
(PA) indicator (i.e. taking value 1 after 2015). Emissions and EDF components are winsorized at the top
and bottom 5%. The controls included are size, debt ratio, operating margin, capital intensity, intangible
assets, and current-year log sales. The fixed effects (FE) included are Year, Country and Sector FE. Standard
errors in parentheses are clustered at the firm level. Statistical significance is reported as * for p ă 0.10, **
for p ă 0.05 and *** for p ă 0.01.
We then use the EDF components to identify the channels through which carbon emissions affect EDF. Table 6 presents regression estimates where each EDF component is regressed on carbon emissions and carbon intensity variables, with interactions for the Paris
Agreement (PA) indicator to capture potential structural shifts post-PA (i.e. this indicator
takes value one after 2015).
The results show that before 2015 higher emissions correlate with lower asset volatility. Following the Paris Agreement, high emitting firms experience a significant increase
in asset volatility. This shift likely reflects heightened uncertainty and market repricing of
carbon-intensive firms due to the anticipated regulatory, market, and operational adjustments needed to align with climate targets.
For the other two EDF components —market asset value and default point- the results
indicate that higher emissions are associated with lower asset values and higher default
points. These effects are amplified post-Paris Agreement, where high emitters experience
relatively lower asset values and even higher default points compared to before 2015. The
effect on firm value is consistent with theoretical expectations, particularly under scenarios
where carbon-intensive firms face an increased risk of stranded assets (Bolton et al., 2020).
To quantify these effects, we compute the marginal contribution of emissions to EDF
through each component. The results suggest that asset volatility is the most influential
channel. For a one-standard-deviation increase in asset volatility, the associated rise in EDF
attributable to emissions is 0.0374 percentage points. By comparison, the corresponding
marginal effects through market value and the default point are 0.0092 and 0.0105 percentage
points, respectively.12
A similar pattern holds when carbon intensity is used instead of absolute emissions.
However, the relationship between carbon intensity and EDF components—particularly asset
volatility—appears weaker in the post-2015 period. This is consistent with the interpretation
that the Paris Agreement shifted investor and regulatory attention toward firms’ absolute
emissions rather than their emissions efficiency.
To gain a deeper understanding of the drivers behind the increased EDF for high emitters,
we replicate the specification from Acharya et al. (2022) to control for variations in firm credit
risk and track the evolution of this effect over time. In this context, we plot the difference
These effects are calculated as the product of the estimated coefficient capturing the impact of emissions
on each structural component of the EDF —namely, asset volatility, market value, and the default point—
and the corresponding coefficient of each component on the EDF. The resulting values represent the marginal
contribution of emissions to EDF through each channel, evaluated at a one-standard-deviation increase in
the respective component.
Figure 3. Difference in EDF Components Between Top and Bottom Emission Quintiles
Asset Volatility
Asset Volatility
(a) Asset Volatility & Absolute Emission
(b) Asset Volatility & Carbon Intensity
Market Value Assets
Market Value Assets
(c) Log(Market Value Assets) & Absolute
Emission
(d) Log(Market Value Assets) & Carbon
Intensity
Default Point
Default Point
(e) Log(Default Point) & Absolute Emission
(f ) Log(Default Point) & Carbon Intensity
Note: the figure presents differences in EDF components (asset volatility, market value of assets, and default
point) between firms in the top and bottom quintiles of emissions. Panels compare absolute emissions (left
column) and carbon intensity (right column), with estimates including country, sector and year-fixed effects.
The 95% confidence intervals are also displayed.
between the top quintile and the bottom quintile of emissions, both for absolute emissions
and intensity.
EDF Componenti,t “ γi ` γt `
Iy βy 1pTop Quintile Emissionqi,t ` θy Ratingi,t
y“2009
`θRatingi,t ` θXi,t ` ϵi,t
Figure 3 illustrates changes in EDF components by comparing firms in the highest and
lowest quintiles of emissions. The findings show that asset volatility increased for firms in
the highest quintile of absolute emissions between 2016 and 2018, though this effect appears
to dissipate in subsequent years. For carbon intensity, a sharp increase is observed as early
as 2011, which helps explain the results presented in Table 6 and aligns with the idea that
emissions efficiency was more relevant before the Paris Agreement. In terms of market value,
a notable decline is observed in 2015 and after 2019 for firms with high absolute emissions.
For default points, neither total direct emissions nor emission intensity exhibit meaningful
changes.13
Overall, we document the impact of emissions on firms’ default probability by isolating
their effects on each EDF component. The next section examines how emissions affect EDFs
differently across jurisdictions —specifically, the United States, European Union, and Great
Britain— leveraging cross-country variation in policy and regulatory implementation.
Figure A4 presents the effects on market value of assets and default point in levels. The contrasting
results relative to the log specification arises from the strong skewness in the distribution of both variables.
High emitters tend to have higher absolute levels. Consequently, level regressions may mask the relative
underperformance of high emitters in percentage terms after the Paris Agreement —a pattern consistent
with increased market penalization of emissions.
4.5. Geographic Areas
In this section, we examine whether a firm’s geographical location influences the relationship between climate and credit risk. Regional differences in the impact of transition
risk on credit risk are well-documented. Evidence discussed in the literature suggests that
jurisdictions with stricter regulations tend to experience heightened effects of transition risks
on credit outcomes (see, for example, (Seltzer et al., 2022) for U.S. firms).
Table 7 presents estimated coefficients for the United States, Euro Area, and Great
Britain across three EDF horizons. All specifications include firm-level controls, as well as
year and sector fixed effects. Country fixed effects are included only for firms within the
European Union, where multiple countries are represented.
The results reveal distinct regional patterns across all horizons. For the United States,
no statistically significant correlation is found between emissions and EDF. In contrast, European Union firms exhibit a significant positive correlation between both absolute emissions
and emission intensity and EDF, consistent with the findings of Capasso et al. (2020). In the
United Kingdom, absolute emissions are positively and significantly associated with EDF,
but no significant relationship is observed for emission intensity.
Regional heterogeneity in the credit impact of climate risk likely stems from policy differences, such as the EU’s established Emission Trading Scheme compared to the historically
fragmented approach in the US, highlighting differences in the regulatory and market landscapes faced by firms across these regions. For instance, the EU ETS trading system, established in 2005, imposes heightened credit risks on large polluters by capping total emissions
and penalizing excess pollution. By comparison, the United States operates only limited
cap-and-trade systems, restricted to a few states. The disparate results likely reflect varying
levels of regulatory commitment and enforcement, as well as heterogeneous firm responses
to climate-related pressures and market signals.
In the appendix, we test how policy stringency affects credit risk. Table B8 reports results
for the interaction of monthly carbon price surprises measured as the euro change in carbon
Table 7. Results by Geographical Location
Panel A:
1-Year EDF
Country
Log(Scope 1)
-0.009
(0.010)
Carbon Intensity
125,075
0.008
(0.011)
125,075
-0.012
(0.013)
125,075
0.010
(0.014)
125,075
Panel C:
-0.021
(0.014)
Carbon Intensity
Year, Country and Sector FE
Control
0.023*
(0.014)
0.025***
(0.009)
62,875
25,605
0.013
(0.015)
25,605
0.032**
(0.015)
62,875
0.038***
(0.013)
62,875
0.036*
(0.020)
25,605
0.013
(0.022)
25,605
10-Year EDF
Log(Scope 1)
62,875
5-Year EDF
Carbon Intensity
0.020*
(0.011)
Panel B:
Log(Scope 1)
125,075
0.023
(0.015)
0.005
(0.015)
125,075
62,875
0.017
(0.019)
0.034***
(0.012)
62,875
25,605
0.009
(0.022)
25,605
Note: The controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and
current-year log sales. The sample spans the years from 2008 to 2022. The dependent and independent
variables are winsorized at the bottom and top 5%. Standard errors in parentheses are clustered at the firm
level. Statistical significance is reported as * for p ă 0.10, ** for p ă 0.05 and *** for p ă 0.01.
price relative to prevailing wholesale electricity prices (Känzig, 2023). We set the carbon
price surprise to one if there is a positive carbon price shock in a given month and zero
otherwise. Since the EU ETS trading scheme operates exclusively in Europe, this analysis
is restricted to that geographical area.14
Column (1) shows that companies with higher absolute emissions face increased credit risk
during months with carbon price surprises. No statistically significant effects are observed for
other horizons, which is expected as the variable is measured daily and aggregated monthly,
making the effect short-term in nature. Furthermore, the results appear specific to absolute
emissions rather than emission intensity, aligning with the structure of the EU ETS that
enforces a fixed cap on total emissions, primarily penalizing absolute emissions rather than
emission intensity.
5. Robustness
We begin by assessing the sensitivity of our results to data preprocessing choices by
varying the winsorization threshold. We then conduct a broader set of robustness checks
to evaluate the stability of our findings. Specifically, we examine the relationship between
credit risk and multiple proxies for transition risk, incorporating Trucost’s carbon earningsat-risk metric and the announcement of science-based targets to better capture the forwardlooking dimension of climate transition exposure. As part of our robustness strategy, we
also re-estimate the EDF decomposition using an alternative methodology based on Structural Equation Modeling (SEM). Furthermore, we replicate the analysis across several subsamples, segmented by credit ratings, sectoral greenness, and public ownership status.
Note that we use the carbon price surprise as a shock variable because the implementation of the EU
ETS in 2005 predates our dataset, the second phase in 2008 coincides with our first year, and the third phase
in 2013 is very close to the Paris Agreement.
Varying winsorization threshold. To ensure that our results are not unduly influenced by
large issuers or extreme observations that could distort the relationship between emissions
and credit risk, we apply a 5% winsorization in the main analysis. This approach limits
the impact of outliers and helps to reveal underlying patterns in the data. To verify the
robustness of our findings to this choice, we replicate the analysis using a stricter 1% winsorization. The core results remain broadly consistent across both thresholds, confirming
that our conclusions are not driven by a small set of extreme values. The only meaningful
difference appears in the exploratory analysis of carbon intensity, where statistical significance weakens under the 1% threshold—suggesting that the previously observed effect may
be attributable to a handful of extreme observations. Other key results, including those
related to EDF levels, ratings, and regional variation, remain stable under both winsorization choices, reinforcing the credibility of our empirical findings. All tables underlying this
robustness check are available from the authors upon request, but are not reported in the
paper to conserve space.
Forward Looking Risk. we replicate the analysis using carbon earnings-at-risks from Trucost.
Carbon earnings-at-risk measure the additional financial cost that a company could face
due to possible future carbon pricing. This is calculated for each firm based on its sector,
operations, and a given price policy scenario (low, medium, and high). 15 For our analysis,
we use the firm’s carbon earnings-at-risks as a percentage of EBITDA, forecasted for the
year 2030. Given data availability, we only have yearly data from 2017 to 2022. We choose
the 2030 earnings-at-risk horizon because it allows us to exploit the 10-year EDF, both
representing long-term risk. This allows us to investigate whether the EDF incorporates
forward-looking transition risks. Given the sample period (2017 to 2022) and a shift in
behavior after 2015, we expect carbon earnings-at-risk to be reflected in EDFs.
See https://www.spglobal.com/en/Perspectives/IIF-2019/Trucost-Carbon-Earnings-at-Risk.pdf
more details.
Table 8. Forward Looking Transition Risk
1-Year EDF
Log(Scope 1)
FWR Low
0.004
(0.007)
0.042**
(0.019)
0.005
(0.007)
0.004
(0.007)
0.012
(0.010)
0.057**
(0.024)
0.013
(0.010)
0.013
(0.010)
0.001
(0.010)
0.051**
(0.022)
0.002
(0.010)
0.002
(0.010)
0.006***
(0.001)
FWR High
0.008***
(0.002)
0.004***
(0.001)
1-Year EDF
Carbon Intensity
FWR Low
0.014*
(0.008)
0.039***
(0.015)
FWR Medium
0.015*
(0.008)
54,319
54,319
0.007***
(0.002)
0.005***
(0.001)
5-Year EDF
0.015*
(0.008)
0.023**
(0.011)
0.052***
(0.018)
0.005***
(0.001)
FWR High
Year, Country and Industry FE
Controls
10-Year EDF
FWR Medium
5-Year EDF
0.024**
(0.011)
10-Year EDF
0.024**
(0.011)
0.020*
(0.011)
0.045***
(0.015)
0.007***
(0.002)
0.004***
(0.001)
54,319
54,319
54,319
0.005***
(0.001)
0.021*
(0.011)
0.021*
(0.011)
0.007***
(0.001)
0.005***
(0.001)
54,319
54,319
54,319
0.004***
(0.001)
54,319
Note: The controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and
current-year log sales. The sample spans the years from 2017 to 2022. FLR stands for Forward-Looking Risk
calculated as the additional financial cost that a company could face due to possible future carbon pricing.
The log(Scope 1) and 10-Year EDF variables are winsorized at the bottom and top 5%. Standard errors in
parentheses are clustered at the firm level * for p ă 0.10, ** for p ă 0.05 and *** for p ă 0.01.
Table 8 shows a positive correlation between EDF and forward-looking transition risks
(FLR) across all policy scenarios (low, medium, and high). In other words, firms facing
higher future carbon pricing or regulatory pressures are associated with increased default
risk. In Panel A, which uses Scope 1 emissions, the results show no statistically significant relationship with EDF, suggesting that absolute emissions alone do not fully capture
forward-looking transition risks. In contrast, Panel B, which examines carbon intensity, reveals a consistently though weakly significant correlation with EDF. This effect holds across
all FLR scenarios, suggesting that firms with higher carbon intensity are less efficient in
their operations and more exposed to potential regulatory costs and market shifts aimed
at reducing carbon footprints. In other words, the significant relationship we find between
forward-looking earnings-at-risk and EDFs indicates that markets, at least partially, internalize future carbon-related earnings shocks into credit risk evaluations, providing direct
evidence of investors’ forward-looking climate risk pricing.
Science-based targets. A crucial aspect of transition risk involves firms’ voluntary commitments to reduce their emissions. Science-based targets provide companies with a clear
roadmap for cutting greenhouse gas emissions, aligned with the latest climate science and
the goals of the Paris Agreement, while supporting sustainable business growth. The Science
Based Targets Initiative (SBTi) includes 5,246 firms that have disclosed targets between 2014
and 2023; however, ISINs are available for only 2,084 of these firms.16 After matching on
ISIN and restricting the sample to firms with data available for the three months before and
after the target publication, we retain 336 firms, 77% of which disclosed their targets after
2020. Geographically, 43% are based in the EU, 35% in the US, and 21% in the UK.
We exploit the timing of the actual publication of the targets to assess the effect on credit
risk. To this end, we estimate a version of equation 2, now using firm and month-year fixed
effects to absorb time shocks, and define the month prior to publication as the reference
https://sciencebasedtargets.org/target-dashboard
period. We do not include credit ratings’ controls due to the already limited sample size.
Table 9 presents the estimated coefficients of the interaction of emission, either absolute
emissions (columns 1–3) or carbon intensity (columns 4–6), with time indicators. The estimates suggest that large emitters who set science-based targets experience a temporary
reduction in EDF following the announcement. This effect is statistically significant for
the 5-year and 10-year EDF, but fades within one to two months—indicating a short-lived
response.
EDF decomposition: SEM. As a further robustness check, we explore the decomposition of
EDF using Structural Equation Modeling (SEM). The variables employed in our decomposition are, by construction, key determinants of EDF and are inherently interrelated.
Our baseline approach allows us to examine how carbon emissions influence EDF indirectly
through its components. However, SEM offers a complementary framework by estimating
all structural equations simultaneously and explicitly accounting for the correlations among
them. This approach enables a more integrated understanding of how carbon-related variables propagate through the determinants of credit risk. The results from this alternative
methodology confirm the main findings presented in the paper. Specifically, the indirect
effects of emissions and carbon intensity on EDF remain statistically significant and directionally consistent with our baseline results. For transparency and completeness, we report
the key outputs from the SEM analysis in Appendix Appendix B (Tables B6-B7), which
present the estimated effects of carbon variables on EDF through its components. Full estimation tables are available upon request. These results reinforce the robustness of our
conclusions to the choice of decomposition method.
Other measures of transition risk. Our analysis considers absolute emission levels (i.e. the
long-term effect of carbon emissions) and carbon intensity (i.e. emissions relative to sales).
For robustness, we replicate our initial analysis by incorporating the rate of change in carbon
emissions and intensity (i.e. the short-term effect of emissions) in Table B9 and lagged values
Table 9. Changes in EDF Around Science-Based Target Announcement
Log(Scope 1)
Carbon Intensity
1 Year EDF
0.028˚
(0.015)
5 Year EDF
0.004
(0.014)
10 Year EDF
-0.021˚
(0.013)
1 Year EDF
0.001
(0.013)
5 Year EDF
0.008
(0.011)
10 Year EDF
0.007
(0.008)
Emission × 3M Before
-0.004
(0.004)
-0.002
(0.003)
-0.000
(0.003)
0.001
(0.005)
-0.001
(0.003)
0.000
(0.002)
Emission × 2M Before
-0.006˚
(0.003)
-0.004
(0.003)
-0.002
(0.002)
-0.007˚˚
(0.003)
-0.004
(0.003)
-0.003
(0.002)
Emission × Event Month
-0.005
(0.003)
-0.006˚˚
(0.002)
-0.005˚˚
(0.002)
-0.003
(0.006)
-0.005˚˚˚
(0.002)
-0.006˚˚˚
(0.002)
Emission × 1M After
-0.007
(0.004)
-0.006˚
(0.003)
-0.005˚
(0.003)
-0.006
(0.007)
-0.006˚˚
(0.003)
-0.006˚˚
(0.003)
Emission × 2M After
-0.007
(0.005)
-0.006
(0.004)
-0.004
(0.003)
-0.005
(0.007)
-0.004
(0.003)
-0.004˚
(0.002)
Emission × 3M After
-0.006
(0.006)
2,349
-0.003
(0.005)
2,349
-0.002
(0.004)
2,349
-0.003
(0.010)
2,349
-0.003
(0.007)
2,349
-0.004
(0.006)
2,349
Emission
Month*Year FE
Firm FE
Note: The controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and
current-year log sales. The sample covers the period from 2017 to 2022. Emissions are measured as log(Scope
1) in columns 1–3 and as Carbon Intensity in columns 4–6. The reference month is the month before the
firm publicly announces a science-based target. The log(Scope 1) and 10-Year EDF variables are winsorized
at the 5th and 95th percentiles. Standard errors, clustered at the firm level, are reported in parentheses.
Standard errors in parentheses are clustered at the firm level * for p ă 0.10, ** for p ă 0.05 and *** for
p ă 0.01.
in Table B10 (i.e. the temporal effect of emissions).
Table B9 in the appendix indicates that in the short-term, the carbon intensity coefficient
remains positive but its effect on EDFs is not significant, as observed in the initial analysis.
On the other hand, an increase in the rate of change of absolute carbon emissions today
leads to a lower EDF in the short term, particularly for the 1-year horizon. This finding is
consistent with expectations, as short-term changes in absolute emissions are closely linked
to short-term credit risk and serve as a proxy for shifts in economic activity, which typically
correspond to a lower risk of bankruptcy. In contrast, emission efficiency, as measured by
carbon intensity, does not affect significantly short-term credit risk.
In Table B10, we replicate the baseline model using one-year lagged values for Scope 1
emissions and carbon intensity to deal with lagged information of emission (Zhang, 2025).
The results show that carbon intensity coefficient is positive and statistically significant
across all three horizons. Conversely, while the coefficients for absolute emissions are positive,
we do not find a significant relationship between absolute emissions and EDFs for the 1-year
and 10-year horizons. Nonetheless, there is a weakly significant positive effect on 5-year
EDFs.
Breakdown by ratings, sector and public ownership. Tables B11 and B12 look at possible
effects driven by Moody’s credit ratings, the ”greenness” of the sector, and whether a Government entity holds a majority stake in the company. We categorize firms with investmentgrade ratings (from Aaa to Baa3) as “good rating” firms, and those with ratings below
investment grade (Baa1 to Caa3) as “bad rating” firms.
The results reveal a negative association between absolute emissions and EDFs across all
three horizons for investment-grade firms, where higher emissions is significantly associated
to lower EDFs at the 1% level. In contrast, no statistically significant relationship is observed
between credit risks and a firm’s classification as belonging to a “green” or “brown” sector
except for a weak positive effect of emissions on EDFs for firms in the most polluting sectors.
This indicates that sector-wide environmental attributes do not substantially influence EDF
in this context, possibly reflecting heterogeneity in regulatory pressures or operational efficiency.17 Finally, among firms with majority government ownership, emissions are positively
associated with credit risk. This finding suggests that public ownership introduces dynamics
that amplify the perceived risks of emissions. Potential explanations include inefficiencies
in publicly managed operations, heightened regulatory exposure, or market expectations of
government accountability for environmental performance.
Interestingly, when focusing on emissions intensity, we observe a similar effect for public
ownership, which indeed reflect the fact that public ownership amplifies the financial risks
associated to higher emissions, regardless of whether the metric is measured in absolute or
intensity terms. We do not find any effect for rating, whether investment-grade or noninvestment-grade. Finally, in the case of carbon emissions, sectoral differences in emissions
intensity present a more differentiated picture. Firms in “brown” industries are associated
with a higher probability of default for all EDF horizons. Conversely, firms in green industries
remain mostly insulated from such risks.
6. Conclusions
We employ a comprehensive yet straightforward approach to estimate the effect of carbon
emissions on credit risk. In our initial analysis, we test the relationship between absolute
emissions and carbon intensity with EDFs. We find some evidence that carbon intensity is
positively associated with EDFs, consistent with previous studies, although this relationship
is sensitive to specification choices and driven largely by upper-tail observations. Nevertheless, our analysis identifies the Paris Agreement as a pivotal structural break, significantly
increasing the sensitivity of firms’ default risks to absolute emissions. Specifically, post17
Green industries include consumer discretionary, consumer staples, healthcare, and telecommunications;
brown industries include utilities, energy, and materials sectors.
2015, firms with high total direct emissions became riskier, primarily due to increased asset
volatility.This finding highlights the importance of international climate agreements as catalysts for market perception shifts and evolving market expectations related to climate policy
developments
We also provide evidence that a firm’s geographical location influences the relationship
between climate risk and credit risk. Regional differences in the impact of transition risk
on credit outcomes are well-documented, and our findings align with prior expectations.
US firms exhibit different relationships compared to EU firms, reflecting their divergent
approaches to climate mitigation policies and carbon emissions regulation.
Finally, we document how firm-level heterogeneity affects the estimated relationship. We
find that the impact of emissions on credit risk is stronger for high emitters, firms in “brown”
sectors, and those with substantial public ownership.
In summary, our study provides empirical evidence that climate transition risks have become increasingly influential in shaping corporate default probabilities, particularly following
the Paris Agreement. We show that absolute emissions became significantly correlated with
EDFs after the Agreement, primarily through increased asset volatility. While our reducedform approach does not allow precise quantification of the magnitude of these effects, we
believe it offers important insights into how transition risks impact financial stability. Future
work could quantify these impacts more precisely by employing structural models. Despite
its simplifying assumptions, our analysis provides novel and insightful implications for both
academics and policymakers.
References
Acharya, V. V., Johnson, T., Sundaresan, S. and Tomunen, T. (2022), ‘Is physical climate
risk priced? evidence from regional variation in exposure to heat stress’, Available at SSRN
Aiello, M. A. and Angelico, C. (2023), ‘Climate change and credit risk: The effect of carbon
tax on italian banks’ business loan default rates’, Journal of Policy Modeling 45(1), 187–
Barth, F., Hübel, B. and Scholz, H. (2022), ‘Esg and corporate credit spreads’, The Journal
of Risk Finance 23(2), 169–190.
Berg, F., Koelbel, J. F. and Rigobon, R. (2022), ‘Aggregate confusion: The divergence of
esg ratings’, Review of Finance 26(6), 1315–1344.
Blasberg, A., Kiesel, R. and Taschini, L. (2021), ‘Carbon default swap -disentangling the
Bolton, P., Despres, M., Da Silva, L. A. P., Samama, F. and Svartzman, R. (2020), ‘The
green swan’, BIS Books .
Bolton, P. and Kacperczyk, M. (2021a), ‘Carbon disclosure and the cost of capital’, Available
Bolton, P. and Kacperczyk, M. (2021b), ‘Do investors care about carbon risk?’, Journal of
financial economics 142(2), 517–549.
Bouchet, V. and Le Guenedal, T. (2020), ‘Credit risk sensitivity to carbon price’, Available
Capasso, G., Gianfrate, G. and Spinelli, M. (2020), ‘Climate change and credit risk’, Journal
of Cleaner Production 266, 121634.
Carbone, S., Giuzio, M., Kapadia, S., Krämer, J. S., Nyholm, K. and Vozian, K. (2021),
‘The low-carbon transition, climate commitments and firm credit risk’, ECB Working
Paper No. 2631 .
Devalle, A., Fiandrino, S. and Cantino, V. (2017), ‘The linkage between esg performance
and credit ratings: A firm-level perspective analysis’, International Journal of Business
and Management 12(9), 1–53.
Di Virgilio, S., Faiella, I., Mistretta, A. and Narizzano, S. (2023), ‘Assessing credit risk
sensitivity to climate and energy shocks’, Banca d’Italia Working Paper .
Gao, F., Li, Y., Wang, X. and Zhong, Z. K. (2021), ‘Corporate social responsibility and the
term structure of cds spreads’, Journal of International Financial Markets, Institutions
and Money 74, 101406.
Giglio, S., Kelly, B. and Stroebel, J. (2021), ‘Climate finance’, Annual Review of Financial
Economics 13, 15–36.
Hartzmark, S. M. and Shue, K. (2022), ‘Counterproductive sustainable investing: The impact
Henisz, W. J. and McGlinch, J. (2019), ‘Esg, material credit events, and credit risk’, Journal
of Applied Corporate Finance 31(2), 105–117.
Höck, A., Klein, C., Landau, A. and Zwergel, B. (2020), ‘The effect of environmental sustainability on credit risk’, Journal of Asset Management 21(2), 85–93.
Ilhan, E., Sautner, Z. and Vilkov, G. (2021), ‘Carbon tail risk’, The Review of Financial
Studies 34(3), 1540–1571.
Kabir, M. N., Rahman, S., Rahman, M. A. and Anwar, M. (2021), ‘Carbon emissions and default risk: International evidence from firm-level data’, Economic Modelling 103, 105617.
Känzig, D. R. (2023), ‘The unequal economic consequences of carbon pricing’, Available at
Klusak, P., Agarwala, M., Burke, M., Kraemer, M. and Mohaddes, K. (2023), ‘Rising temperatures, falling ratings: The effect of climate change on sovereign creditworthiness’,
Management Science 69(12), 7468–7491.
Kölbel, J. F., Leippold, M., Rillaerts, J. and Wang, Q. (2024), ‘Ask bert: How regulatory
disclosure of transition and physical climate risks affects the cds term structure’, Journal
of Financial Econometrics 22(1), 30–69.
Kotró, B. and Márkus, M. (2020), ‘The impact of esg rating on corporate bond yields’,
Economy and Finance: English-Language Edition of Gazdaság és Pénzügy 7(4), 474–488.
Kruse, T., Mohnen, M. and Sato, M. (2020), ‘Are financial markets aligned with climate
action?: new evidence from the paris agreement’, Grantham Research Institute on Climate
Change and the Environment .
Merton, R. C. (1974), ‘On the pricing of corporate debt: The risk structure of interest rates’,
The Journal of Finance 29(2), 449–470.
Monasterolo, I. and De Angelis, L. (2020), ‘Blind to carbon risk? an analysis of stock market
reaction to the paris agreement’, Ecological Economics 170, 106571.
Sautner, Z., Van Lent, L., Vilkov, G. and Zhang, R. (2023), ‘Firm-level climate change
exposure’, The Journal of Finance 78(3), 1449–1498.
Seltzer, L., Starks, L. T. and Zhu, Q. (2022), ‘Climate regulatory risks and corporate bonds’,
FRB of New York Staff Report No. 1014 .
Stellner, C., Klein, C. and Zwergel, B. (2015), ‘Corporate social responsibility and eurozone
corporate bonds: The moderating role of country sustainability’, Journal of Banking &
Finance 59, 538–549.
Zhang, B. Y., Zhou, H. and Zhu, H. (2009), ‘Explaining credit default swap spreads with
the equity volatility and jump risks of individual firms’, The Review of Financial Studies
22(12), 5099–5131.
Zhang, S. (2025), ‘Carbon returns across the globe’, The Journal of Finance 80(1), 615–645.
Appendix A. Appendix Figures
Figure A1. Time-series EDF by quintiles of total emissions and emission intensity
1-Year EDF (%)
1-Year EDF (%)
1st Quintile
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
1st Quintile
(a) Total Emission & 1-Year EDF
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
(b) Carbon Intensity & 1-Year EDF
5-Year EDF (%)
5-Year EDF (%)
1st Quintile
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
1st Quintile
(c) Total Emission & 5-Year EDF
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
(d) Carbon Intensity & 5-Year EDF
1-Year EDF (%)
1st Quintile
2nd Quintile
3rd Quintile
10-Year EDF (%)
4th Quintile
5th Quintile
1st Quintile
(e) Total Emission & 10-Year EDF
2nd Quintile
3rd Quintile
4th Quintile
5th Quintile
(f ) Carbon Intensity & 10-Year EDF
Figure A2. Analysis using Quantile Regression
Estimated Effect on EDF
Estimated Effect on EDF
Total Emission Quintiles
Emission Intensity Quintiles
(a) 1-Year EDF & Total Emission
(b) 1-Year EDF & Carbon Intensity
Estimated Effect on EDF
Estimated Effect on EDF
Total Emission Quintiles
Emission Intensity Quintiles
(c) 5-Year EDF & Total Emission
(d) 5-Year EDF & Carbon Intensity
Estimated Effect on EDF
Estimated Effect on EDF
Total Emission Quintiles
Emission Intensity Quintiles
(e) 10-Year EDF & Total Emission
(f ) 10-Year EDF & Carbon Intensity
Note: the figure presents the quintile regression of absolute emissions (left column) and carbon intensity
(right column) on 1-year, 5-year, and 10-year EDFs. Coefficients are estimated running the specification (1)
with all controls and year, country, and sector fixed effects. Standard errors are clustered at the firm level
with 95% confidence intervals displayed.
(a) Total Emission & 1-Year EDF
(b) Carbon Intensity & 1-Year EDF
5-Year EDF
5-Year EDF
1-Year EDF
1-Year EDF
Figure A3. EDF change around Paris Agreement (without controlling for Moody’s ratings)
(c) Total Emission & 5-Year EDF
(d) Carbon Intensity & 5-Year EDF
10-Year EDF
10-Year EDF
(e) Total Emission & 10-Year EDF
(f ) Carbon Intensity & 10-Year EDF
Note: the figure presents yearly interaction coefficients of absolute emissions (left column) and carbon intensity (right column)
with 1-year, 5-year, and 10-year EDFs. The base year is 2008. Coefficients are estimated using firm and year-fixed effects, with
95% confidence intervals displayed.
15000
10000
10000
Market Value of Assets
Market Value of Assets
Figure A4. Difference in EDF Components Between Top and Bottom Emission Quintiles (in Levels)
-5000
-5000
-1000
-2000
Default Point
(b) Market Value Assets & Carbon Intensity
(a) Market Value Assets & Absolute Emission
Default Point
-10000
(c) Default Point & Absolute Emission
(d) Default Point & Carbon Intensity
Note: the figure presents differences in market value of assets and default point (both in levels) between
firms in the top and bottom quintiles of emissions. Panels compare absolute emissions (left column) and
carbon intensity (right column), with estimates including country, sector and year-fixed effects. The 95%
confidence intervals are also displayed.
Appendix B. Appendix Tables
Table B1. Source and Description Variables
Variable Name
Description
Source
1-Year EDF
5-Year EDF
10-Year EDF
Mean CDS Spreads
Carbon Intensity
Ln(scope 1)
Debt Ratio
Operating Margin Ratio
Country
Sector
Year EDF
Asset Volatility
Market Value of Assets
Default Point
Moody’s Ratings
Derived CDS Ratings
Capital Intensity
Intangible Assets
Public Ownership
1-Yr EDF (%)
5-Yr EDF (%)
10-Yr EDF (%)
Mean monthly CDS spreads
Carbon Intensity Scope 1+2 (t/ USD in million sales)
ln(Scope 1 Emissions)
ln(Total Assets)
(current liabilities + long-term debt)/Total assets
Operating income/Sales
Country of the firms
Sector from 2-digits NAICS code of the firms
Asset Volatility (EDF) (%)
Market Value of Assets (EDF)
Default Point (EDF)
Clean Moody’s Ratings: encoded from 1 for AAA to 21 for C
Clean derived CDS Ratings: encoded from 1 for AAA to 21 for C
Property, Plant and Equipment divided by Total Assets
Intangible Assets over Total Assets
The indicator takes value 1 if the ultimate owner is a public entity
CreditEdge
CreditEdge
CreditEdge
Refinitiv
CRSP/Compustat
CRSP/Compustat
CRSP/Compustat
EDF-MSCI
EDF-MSCI
CreditEdge
CreditEdge
CreditEdge
CreditEdge
CreditEdge
CRSP/Compustat
CRSP/Compustat
Orbis
Note: All CRSP/Compustat variables are expressed in USD millions.
Table B2. Analysis using Emission Levels and Intensity (not winsorized)
Panel A
1-Year EDF
Log(Scope 1)
-0.039***
(0.010)
-0.003
(0.004)
Panel B
-0.072***
(0.009)
Carbon Intensity
-0.008***
(0.003)
-0.118***
(0.008)
Carbon Intensity
Year FE
Country FE
Sector FE
Controls
0.004
(0.004)
10-Year EDF
-0.022
(0.017)
Panel C
Log(Scope 1)
0.004
(0.005)
5-Year EDF
Log(Scope 1)
-0.009
(0.021)
Carbon Intensity
-0.007
(0.014)
-0.014***
(0.003)
0.000
(0.003)
213,555
213,555
213,555
213,555
Note: the controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and
log sales in the current year. The dependent and independent variables are not winsorized. Standard errors
in parentheses are clustered at the firm level, * for p ă 0.10, ** for p ă 0.05 and for *** p ă 0.01.
Table B3. Quantile Regression
Panel A
1-Year EDF
Log(Scope 1)
-0.008***
(0.001)
Carbon Intensity
-0.003*
(0.002)
-0.051***
(0.004)
Carbon Intensity
-0.026***
(0.006)
-0.108***
(0.005)
Carbon Intensity
Year FE
Country FE
Sector FE
Controls
0.013**
(0.006)
10-Year EDF
0.009
(0.007)
Panel C
Log(Scope 1)
0.005***
(0.002)
5-Year EDF
0.005***
(0.002)
Panel B
Log(Scope 1)
-0.003
(0.009)
-0.054***
(0.008)
0.014
(0.009)
213,555
213,555
213,555
213,555
Note: the controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and
log sales in the current year. The dependent and independent variables are winsorized at the top and bottom
5%. Standard errors in parentheses are clustered at the firm level, * for p ă 0.10, ** for p ă 0.05 and for
*** p ă 0.01.
Table B4. Pre and Post Paris Agreement for Total Emission
Log(Scope 1)
Year 2009 ˆ Log(Scope 1)
Year 2010 ˆ Log(Scope 1)
Year 2011 ˆ Log(Scope 1)
Year 2012 ˆ Log(Scope 1)
Year 2013 ˆ Log(Scope 1)
Year 2014 ˆ Log(Scope 1)
Year 2015 ˆ Log(Scope 1)
Year 2016 ˆ Log(Scope 1)
Year 2017 ˆ Log(Scope 1)
Year 2018 ˆ Log(Scope 1)
Year 2019 ˆ Log(Scope 1)
Year 2020 ˆ Log(Scope 1)
Year 2021 ˆ Log(Scope 1)
Year 2022 ˆ Log(Scope 1)
Year, Country and Sector FE
Controls
1 Year EDF
-0.063***
(0.013)
-0.004
(0.010)
0.009
(0.011)
0.024**
(0.011)
0.039***
(0.012)
0.040***
(0.011)
0.037***
(0.011)
0.050***
(0.012)
0.074***
(0.013)
0.064***
(0.013)
0.057***
(0.012)
0.066***
(0.013)
0.058***
(0.013)
0.046***
(0.012)
0.041***
82,848
5 Year EDF
-0.072***
(0.015)
-0.002
(0.008)
0.002
(0.009)
0.017
(0.011)
0.029***
(0.011)
0.031***
(0.011)
0.033***
(0.011)
0.050***
(0.012)
0.074***
(0.013)
0.077***
(0.013)
0.066***
(0.013)
0.073***
(0.014)
0.066***
(0.015)
0.055***
(0.014)
0.048***
82,848
10 Year EDF
-0.076***
(0.014)
-0.001
(0.006)
-0.003
(0.008)
0.009
(0.009)
0.019**
(0.009)
0.018*
(0.009)
0.022**
(0.010)
0.037***
(0.011)
0.059***
(0.011)
0.063***
(0.012)
0.054***
(0.012)
0.062***
(0.013)
0.059***
(0.013)
0.048***
(0.013)
0.045***
82,848
Asset Volatility
-0.396**
(0.166)
-0.038
(0.042)
0.067
(0.058)
0.232***
(0.069)
0.327***
(0.076)
0.395***
(0.082)
0.487***
(0.085)
0.582***
(0.090)
0.736***
(0.091)
0.815***
(0.092)
0.735***
(0.093)
0.523***
(0.092)
0.204**
(0.092)
0.095
(0.095)
0.143
82,847
Market Value of Assets
0.026*
(0.014)
0.005
(0.004)
-0.006
(0.005)
-0.012**
(0.005)
-0.015**
(0.006)
-0.023***
(0.006)
-0.022***
(0.007)
-0.025***
(0.008)
-0.017**
(0.008)
-0.019**
(0.009)
-0.025***
(0.009)
-0.034***
(0.009)
-0.031***
(0.010)
-0.039***
(0.010)
-0.030***
82,847
Default Point
0.024***
(0.009)
0.005*
(0.003)
0.008**
(0.004)
0.002
(0.004)
0.006
(0.004)
0.005
(0.004)
0.005
(0.004)
0.004
(0.005)
0.005
(0.005)
0.000
(0.005)
-0.004
(0.005)
-0.001
(0.005)
-0.004
(0.005)
-0.008
(0.005)
-0.013**
82,847
Note: The coefficients of interest in the table are those for Log(Scope 1), the interaction between the year indicators (with 2008
as the base year) and Log(Scope 1). The controls included are size, debt ratio, operating margin, capital intensity, intangible
assets, and log sales in the current year. We do not report the coefficients of the interaction between absolute emissions and
years’ dummies and between Moody’s rating and years’ dummies. The specification includes year, country and sector fixed
effects. The dependent and independent variables are winsorized at the bottom and top 5%. Standard errors in parentheses
are clustered at the firm level, * for p ă 0.10, ** for p ă 0.05 and *** for p ă 0.01.
Table B5. Pre and Post Paris Agreement for Emission Intensity
Carbon Intensity
Year 2009 ˆ Carbon Intensity
Year 2010 ˆ Carbon Intensity
Year 2011 ˆ Carbon Intensity
Year 2012 ˆ Carbon Intensity
Year 2013 ˆ Carbon Intensity
Year 2014 ˆ Carbon Intensity
Year 2015 ˆ Carbon Intensity
Year 2016 ˆ Carbon Intensity
Year 2017 ˆ Carbon Intensity
Year 2018 ˆ Carbon Intensity
Year 2019 ˆ Carbon Intensity
Year 2020 ˆ Carbon Intensity
Year 2021 ˆ Carbon Intensity
Year 2022 ˆ Carbon Intensity
Year, Country and Sector FE
Controls
1 Year EDF
-0.039***
(0.009)
-0.004
(0.006)
0.013**
(0.006)
0.025***
(0.007)
0.025***
(0.008)
0.034***
(0.008)
0.036***
(0.008)
0.043***
(0.008)
0.055***
(0.010)
0.058***
(0.010)
0.045***
(0.008)
0.048***
(0.009)
0.044***
(0.010)
0.035***
(0.009)
0.036***
(0.013)
82,848
5 Year EDF
-0.043***
(0.010)
-0.000
(0.005)
0.009
(0.006)
0.020***
(0.007)
0.021***
(0.008)
0.025***
(0.008)
0.031***
(0.008)
0.041***
(0.009)
0.052***
(0.010)
0.061***
(0.011)
0.049***
(0.009)
0.050***
(0.011)
0.046***
(0.011)
0.039***
(0.011)
0.038***
(0.013)
82,848
10 Year EDF
-0.042***
(0.010)
0.000
(0.004)
0.006
(0.005)
0.015**
(0.006)
0.014**
(0.006)
0.016**
(0.007)
0.023***
(0.007)
0.031***
(0.008)
0.041***
(0.009)
0.049***
(0.009)
0.039***
(0.009)
0.041***
(0.010)
0.039***
(0.010)
0.032***
(0.010)
0.033***
(0.012)
82,848
Asset Volatility
-0.106
(0.092)
0.010
(0.029)
0.097***
(0.038)
0.201***
(0.044)
0.215***
(0.052)
0.250***
(0.058)
0.317***
(0.057)
0.371***
(0.061)
0.502***
(0.061)
0.602***
(0.065)
0.541***
(0.066)
0.354***
(0.061)
0.104*
(0.062)
0.062
(0.067)
0.139*
(0.075)
82,847
Market Value of Assets
0.024***
(0.008)
-0.002
(0.002)
-0.009***
(0.003)
-0.010***
(0.003)
-0.008*
(0.004)
-0.014***
(0.004)
-0.016***
(0.005)
-0.019***
(0.006)
-0.010*
(0.005)
-0.012**
(0.006)
-0.009
(0.006)
-0.011*
(0.007)
-0.011
(0.007)
-0.012*
(0.007)
-0.006
(0.007)
82,847
Default Point
0.014**
(0.006)
-0.000
(0.002)
0.001
(0.003)
-0.002
(0.003)
-0.000
(0.003)
-0.001
(0.003)
0.000
(0.003)
0.000
(0.004)
0.002
(0.004)
-0.001
(0.004)
0.001
(0.004)
0.003
(0.004)
0.004
(0.005)
0.003
(0.005)
-0.001
(0.005)
82,847
Note: The coefficients of interest in the table are those for Carbon Intensity, the interaction between the year indicators (with
2008 as the base year) and Carbon Intensity. The controls included are size, debt ratio, operating margin, capital intensity,
intangible assets, and log sales in the current year. We do not report the coefficients of the interaction between emissions and
years’ dummies and between Moody’s rating and years’ dummies. The specification includes year, country and sector fixed
effects. The dependent and independent variables are winsorized at the bottom and top 5%. Standard errors in parentheses
are clustered at the firm level, * for p ă 0.10, ** for p ă 0.05 and *** for p ă 0.01.
Table B6. Effect of Log(Scope 1) on EDF through its components
Indirect effect
1-Year EDF
5-Year EDF
10-Year EDF
Asset volatility
0.007026*** 0.008357***
(0.000452)
(0.000538)
0.004564*** 0.005918***
(0.000542)
(0.000703)
0.005826*** 0.007597***
(0.000346)
(0.000451)
0.008037***
(0.000517)
0.005607***
(0.000666)
0.006884***
(0.000409)
0.004972***
(0.000675)
0.008689***
(0.000742)
Log(Market value)
Log(Default point)
Direct effect
Log(Scope1)
0.009455***
(0.000738)
Note: This table reports the direct and indirect effects of logpScope1q on EDF, using a structural equation model (SEM) that
decomposes the overall effect into three EDF components (asset volatility, log(market value of assets), and log(default point)).
The indirect effect is obtained by multiplying (i) the coefficient of the variable of interest (asset volatility, log(market value of
assets), and log(default point)) from the main regression by (ii) the coefficient on the interaction term logpScope1qP ost ´ P aris
in the corresponding EDF-component regression. Standard errors are shown in parentheses * for p ă 0.10, ** for p ă 0.05 and
*** for p ă 0.01.
Table B7. Effect of Carbon intensity on EDF through its components
Indirect effect
1-Year EDF
5-Year EDF
10-Year EDF
Asset volatility
0.003983*** 0.004758***
(0.000597)
(0.000713)
0.007917*** 0.010242***
(0.000708)
(0.000916)
0.004169*** 0.005444***
(0.000449)
(0.000586)
0.004619***
(0.000692)
0.009639***
(0.000862)
0.004941***
(0.000532)
0.008586***
(0.000876)
0.013866***
(0.000969)
Log(Market value)
Log(Default point)
Direct effect
Carbon intensity
0.014839***
(0.000959)
Note: This table reports the direct and indirect effects of carbon intensity on EDF, using a structural equation model
(SEM) that decomposes the overall effect into three EDF components (asset volatility, log(market value of assets), and
log(default point)). The indirect effect is obtained by multiplying (i) the coefficient of the variable of interest (asset volatility, log(market value of assets), or log(default point)) from the main regression by (ii) the coefficient on the interaction term
carbon intensityP ost ´ P aris in the corresponding EDF-component regression. Standard errors are shown in parentheses * for
p ă 0.10, ** for p ă 0.05 and *** for p ă 0.01.
Table B8. Carbon Surprise and Credit Risk in Europe
1-Year EDF
Log(Scope 1)
CPSurprise ˆ Log(Scope 1)
0.017
(0.012)
0.004**
(0.002)
Carbon Intensity
CPSurprise ˆ Carbon Intensity
Year, Country and Sector FE
Controls
49,955
0.025***
(0.009)
0.003
(0.002)
49,955
5-Year EDF
0.030*
(0.016)
0.002
(0.002)
49,955
0.037***
(0.013)
0.002
(0.002)
49,955
10-Year EDF
0.023
(0.015)
-0.001
(0.001)
49,955
0.035***
(0.012)
0.000
(0.001)
49,955
Note: The coefficients of interest in the table are those for Carbon Intensity and Log(Scope 1) interacted with CPSurprise.
CPSurprise is a binary variable that takes value one if there is a positive carbon policy surprise and 0 otherwise. The carbon
policy surprises are measured as euro change in carbon price, relative to prevailing wholesale electricity price (Känzig, 2023).
The analysis is conducted for the Europe area only. The controls included are size, debt ratio, operating margin, capital
intensity, intangible assets, and current-year log sales. The specification includes year, country and sector fixed effects. The
dependent and independent variables are winsorized at the bottom and top 5%. Standard errors in parentheses are clustered
at the firm level, * for p ă 0.10, ** for p ă 0.05 and *** for p ă 0.01.
Table B9. Rate of Change in Emissions
1-Year EDF
∆t´pt´1q Log(Scope 1)
-0.023***
(0.009)
∆t´pt´1q Carbon Intensity
Year, Country and Sector FE
Controls
197,582
0.008
(0.008)
197,582
5-Year EDF
-0.016
(0.011)
197,582
0.005
(0.009)
197,582
10-Year EDF
-0.000
(0.011)
197,582
0.001
(0.009)
197,582
Note: The dependent variable is ∆logpScope1q constructed as logpScope1qt ´ logpScope1qt´1 . The controls
included are size, debt ratio, operating margin, capital intensity, intangible assets, and current-year log
sales.The Fixed Effects (FE) included are Year, Country, and Sector FE. The dependent and independent
variables are winsorized at the bottom and top 5%. Standard errors in parentheses are clustered at the firm
level, and statistical significance is * for p ă 0.10, ** for p ă 0.05 and for *** p ă 0.01.
Table B10. Lagged Emissions
1-Year EDF
Log(Scope 1)t´1
0.010
(0.007)
Carbon Intensityt´1
Year, Country and Sector FE
Controls
197,582
5-Year EDF
0.016*
(0.009)
0.017**
(0.007)
197,582 197,582
10-Year EDF
0.004
(0.009)
0.024***
(0.009)
197,582 197,582
0.020**
(0.009)
197,582
Note: The table reports the analysis using 1-year lagged emission variables. The controls included are size,
debt ratio, operating margin, capital intensity, intangible assets, and current-year log sales.The Fixed Effects
(FE) included are Year, Country, and Sector FE. The dependent and independent variables are winsorized
at the bottom and top 5%. Standard errors in parentheses are clustered at the firm level, and statistical
significance is reported as * for p ă 0.05, ** for p ă 0.01 and for *** p ă 0.001.
Table B11. Ratings, Sector and Public Ownership (Total Emissions)
Firms breakdown
Ratings
Sector
Green
Brown
Ownership
Public
Private
Panel (1) – 1-Year EDF
Log(Scope 1)
-0.016**
(0.006)
0.004
(0.022)
0.012
(0.019)
0.013 0.073***
(0.011) (0.024)
0.002
(0.007)
Panel (2) – 5-Year EDF
Log(Scope 1)
-0.022***
(0.008)
0.007
(0.026)
0.031
(0.026)
0.025*
(0.015)
0.087***
(0.032)
0.008
(0.010)
Panel (3) – 10-Year EDF
Log(Scope 1)
-0.030***
(0.009)
-0.007
(0.025)
0.027
(0.026)
0.016
(0.017)
0.061**
(0.026)
0.001
(0.010)
Year, Country and Sector FE
Controls
52,043
30,402
34,251
26,386
6,757
180,817
Note: The table reports the estimated coefficients of the baseline regression divided by ratings, sector and
public ownership with Year, Sector and Country FE. The controls included are size, debt ratio, operating
margin, capital intensity, intangible assets, and current-year log sales. The dependent and independent
variables are winsorized at the bottom and top 5%. Standard errors in parentheses are clustered at the firm
level, * for p ă 0.05, ** for p ă 0.01 and for *** p ă 0.001.
Table B12. Ratings, Sector and Public Ownership (Carbon Intensity)
Firms breakdown
Ratings
Sector
Green
Brown
Ownership
Public
Private
Panel (1) – 1-Year EDF
Carbon Intensity
-0.004
(0.004)
0.026
(0.024)
-0.083
(0.100)
0.021***
(0.007)
0.056***
(0.016)
0.011
(0.008)
Panel (2) – 5-Year EDF
Carbon Intensity
-0.005
(0.005)
0.030
(0.028)
-0.098
(0.135)
0.033***
(0.010)
0.075***
(0.020)
0.016
(0.011)
Panel (3) – 10-Year EDF
Carbon Intensity
-0.009
(0.006)
0.022
(0.025)
-0.169*
(0.092)
0.032*** 0.061***
(0.011)
(0.016)
0.012
(0.011)
Year, Country and Sector FE
Controls
52,043
30,402
34,251
26,386
180,817
6,757
Note: The table reports the estimated coefficients of the baseline regression divided by ratings, sector and
public ownership with Year, Sector and Country FE. The controls included are size, debt ratio, operating
margin, capital intensity, intangible assets, and current-year log sales. The dependent and independent
variables are winsorized at the bottom and top 5%. Standard errors in parentheses are clustered at the firm
level, * for p ă 0.05, ** for p ă 0.01 and for *** p ă 0.001.
Table B13. Yearly Analysis of Emission Levels and Intensity
Panel A
1-Year EDF
Log(Scope 1)
-0.018***
(0.004)
-0.005
(0.005)
Panel B
-0.055***
(0.005)
-0.023***
(0.008)
Panel C
-0.103***
(0.006)
0.022**
(0.009)
0.003
(0.010)
Carbon Intensity
Year FE
Country FE
Sector FE
Controls
10-Year EDF
Log(Scope 1)
0.014
(0.009)
Carbon Intensity
0.016**
(0.007)
5-Year EDF
Log(Scope 1)
0.008
(0.007)
Carbon Intensity
-0.049***
(0.008)
17,909
17,909
17,909
0.018*
(0.009)
17,909
Note: The controls included are size, debt ratio, operating margin, capital intensity, intangible assets, and
current-year log sales. EDF values are reported as the average EDF at the yearly level for each company.
The dependent and independent variables are winsorize at the bottom and top 5%. Standard errors in
parentheses are clustered at the firm level, * for p ă 0.10, ** for p ă 0.05 and for *** p ă 0.01.
Recently published papers in the ‘Markets, Infrastructures, Payment Systems’ series
n. 25
The bonfire of banknotes, by Michele Manna (Research Papers)
n. 26
Integrating DLTs with market infrastructures: analysis and proof-of-concept for secure DvP
between TIPS and DLT platforms, by Rosario La Rocca, Riccardo Mancini, Marco Benedetti,
Matteo Caruso, Stefano Cossu, Giuseppe Galano, Simone Mancini, Gabriele Marcelli, Piero
Martella, Matteo Nardelli and Ciro Oliviero (Research Papers)
n. 27
Statistical and forecasting use of electronic payment transactions: collaboration between
Bank of Italy and Istat, by Guerino Ardizzi and Alessandra Righi (Institutional Issues) (in
Italian)
n. 28
TIPS: a zero-downtime platform powered by automation, by Gianluca Caricato, Marco
Capotosto, Silvio Orsini and Pietro Tiberi (Research Papers)
n. 29
TARGET2 analytical tools for regulatory compliance, by Marc Glowka, Alexander Müller,
Livia Polo Friz, Sara Testi, Massimo Valentini and Stefano Vespucci (Institutional Issues)
n. 30
The security of retail payment instruments: evidence from supervisory data, by Massimiliano
Cologgi (Research Papers)
n. 31
Open Banking in the payment system: infrastructural evolution, innovation and security,
supervisory and oversight practices, by Roberto Pellitteri, Ravenio Parrini, Carlo Cafarotti
and Benedetto Andrea De Vendictis (Institutional Issues) (in Italian)
n. 32
Banks’ liquidity transformation rate: determinants and impact on lending, by Raffaele Lenzi,
Stefano Nobili, Filippo Perazzoli and Rosario Romeo (Research Papers)
n. 33
Investor behavior under market stress: evidence from the Italian sovereign bond market, by
Onofrio Panzarino (Research Papers)
n. 34
Siamese neural networks for detecting banknote printing defects, by Katia Boria, Andrea
Luciani, Sabina Marchetti and Marco Viticoli (Research Papers) (in Italian)
n. 35
Quantum safe payment systems, by Elena Bucciol and Pietro Tiberi
n. 36
Investigating the determinants of corporate bond credit spreads in the euro area, by Simone
Letta and Pasquale Mirante
n. 37
Smart Derivative Contracts in DatalogMTL, by Andrea Colombo, Luigi Bellomarini, Stefano
Ceri and Eleonora Laurenza
n. 38
Making it through the (crypto) winter: facts, figures and policy issues, by Guerino Ardizzi,
Marco Bevilacqua, Emanuela Cerrato and Alberto Di Iorio
n. 39
The Emissions Trading System of the European Union (EU ETS), by Mauro Bufano, Fabio
Capasso, Johnny Di Giampaolo and Nicola Pellegrini (in Italian)
n. 40
Banknote migration and the estimation of circulation in euro area countries: the italian
case, by Claudio Doria, Gianluca Maddaloni, Giuseppina Marocchi, Ferdinando Sasso,
Luca Serrai and Simonetta Zappa (in Italian)
n. 41
Assessing credit risk sensitivity to climate and energy shocks, by Stefano Di Virgilio, Ivan
Faiella, Alessandro Mistretta and Simone Narizzano
n. 42
Report on the payment attitudes of consumers in italy: results from the ecb space 2022
survey, by Gabriele Coletti, Alberto Di Iorio, Emanuele Pimpini and Giorgia Rocco
n. 43
A service architecture for an enhanced Cyber Threat Intelligence capability and its value
for the cyber resilience of Financial Market Infrastructures, by Giuseppe Amato, Simone
Ciccarone, Pasquale Digregorio and Giuseppe Natalucci
n. 44
Fine-tuning large language models for financial markets via ontological reasoning,
by Teodoro Baldazzi, Luigi Bellomarini, Stefano Ceri, Andrea Colombo, Andrea Gentili
and Emanuel Sallinger
n. 45
Sustainability at shareholder meetings in France, Germany and Italy, by Tiziana De Stefano,
Giuseppe Buscemi and Marco Fanari (in Italian)
n. 46
Money market rate stabilization systems over the last 20 years: the role of the minimum
reserve requirement, by Patrizia Ceccacci, Barbara Mazzetta, Stefano Nobili, Filippo
Perazzoli and Mattia Persico
n. 47
Technology providers in the payment sector: market and regulatory developments,
by Emanuela Cerrato, Enrica Detto, Daniele Natalizi, Federico Semorile and Fabio Zuffranieri
n. 48
The fundamental role of the repo market and central clearing, by Cristina Di Luigi, Antonio
Perrella and Alessio Ruggieri
n. 49
From Public to Internal Capital Markets: The Effects of Affiliated IPOs on Group Firms,
by Luana Zaccaria, Simone Narizzano, Francesco Savino and Antonio Scalia
n. 50
Byzantine Fault Tolerant consensus with confidential quorum certificate for a Central Bank
DLT, by Marco Benedetti, Francesco De Sclavis, Marco Favorito, Giuseppe Galano, Sara
Giammusso, Antonio Muci and Matteo Nardelli
n. 51
Environmental data and scores: lost in translation, by Enrico Bernardini, Marco Fanari,
Enrico Foscolo and Francesco Ruggiero
n. 52
How important are ESG factors for banks’ cost of debt? An empirical investigation,
by Stefano Nobili, Mattia Persico and Rosario Romeo
n. 53
The Bank of Italy’s statistical model for the credit assessment of non-financial firms,
by Simone Narizzano, Marco Orlandi, Antonio Scalia
n. 54
The revision of PSD2 and the interplay with MiCAR in the rules governing payment services:
evolution or revolution?, by Mattia Suardi
n. 55
Rating the Raters. A Central Bank Perspective, by Francesco Columba, Federica Orsini and
Stefano Tranquillo
n. 56
A general framework to assess the smooth implementation of monetary policy:
an application to the introduction of the digital euro, by Annalisa De Nicola and Michelina
Lo Russo
n. 57
The German and Italian Government Bond Markets: The Role of Banks versus Non-Banks.
A joint study by Banca d’Italia and Bundesbank, by Puriya Abbassi, Michele Leonardo
Bianchi, Daniela Della Gatta, Raffaele Gallo, Hanna Gohlke, Daniel Krause, Arianna
Miglietta, Luca Moller, Jens Orben, Onofrio Panzarino, Dario Ruzzi, Willy Scherrieble and
Michael Schmidt
n. 58
Chat Bankman-Fried? An Exploration of LLM Alignment in Finance, by Claudia Biancotti,
Carolina Camassa, Andrea Coletta, Oliver Giudice, Aldo Glielmo
n. 59
Modelling transition risk-adjusted probability of default, by Manuel Cugliari, Alessandra
Iannamorelli and Federica Vassalli
n. 60
The use of Banca d’Italia’s credit assessment system for Italian non-financial firms within
the Eurosystem’s collateral framework, by Stefano Di Virgilio, Alessandra Iannamorelli,
Francesco Monterisi and Simone Narizzano
n. 61
Fintech Classification Methodology, by Alessandro Lentini, Daniela Elena Munteanu and
Fabrizio Zennaro