Global Finance Journal 52 (2022) 100571

Contents lists available at ScienceDirect

Global Finance Journal

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

It is not only what you say, but how you say it: ESG, corporate
news, and the impact on CDS spreads
Hans-Jörg Naumer *, Burcin Yurtoglu
WHU – Otto Beisheim School of Management, Burgplatz 2, 56179 Vallendar, Germany

A R T I C L E I N F O A B S T R A C T

JEL classifications: We assess the influence of corporate news on costs of financing for a sample of large European and
G12 U.S. firms from 2006 to 2016. Focusing on environmental, social, and governance (ESG) news
G30 items, we take into account volume (number of news items), tonality (positive, neutral, negative),
G34
and source (financial or mass media). We find that (1) the volume of ESG-related news is
K29
significantly associated with credit default swap (CDS) spreads and therefore matters for com­
Keywords:
panies’ refinancing costs; (2) news with positive (negative) tonality is associated with lower
Media
(higher) CDS spreads by about 4% (6%); (3) tonality matters even more for ESG-related news.
News
ESG These results hold for different subsamples and alternate specifications, and are relatively
CSR insensitive to omitted variables.
CDS spreads

1. Introduction

Corporate social responsibility (CSR) has gained increasing importance for firms’ operations over the past decade. According to
KPMG (2017), 78% of the world’s top companies include CSR information in their annual financial reports, indicating that CSR data
are relevant for their investors. The demand for sustainable and impact investing is growing—in the United States alone, investors
consider environmental, social, and governance (ESG) factors across $8.72 trillion of professionally managed assets, a 33% increase
since 2014 (US SIF Foundation, 2016).
Yet scholars disagree concerning the motivation for CSR and how it affects firm value and firm behavior (Margolis et al., 2009;
Margolis & Walsh, 2003; Orlitzky et al., 2003). There are two theoretical perspectives on CSR. The first posits that CSR activities aim to
maximize shareholder wealth by creating reputational capital among customers, employees, and other stakeholders (Bénabou &
Tirole, 2010; Deng et al., 2013; Ferrell et al., 2016). The second suggests that CSR activities do not aim to create value but divert cash
flows to investments that may benefit only some stakeholders (Cheng et al., 2014; Friedman, 1970; Hong et al., 2012; Jo et al., 2016;
Masulis & Reza, 2015, 2020). This view considers CSR as a reflection of managerial preferences and thus as an agency cost.
A large body of literature has studied the potential for value creation through CSR. On average, these studies suggest a small,
positive, and statistically significant relationship between CSR and various measures of corporate financial performance (Friede et al.,
2015; Margolis et al., 2009). However, there has been only limited research to examine the specific channels through which CSR-
related news influences firm value and firm performance. Here, we study one such channel by investigating how qualitative infor­
mation in media affects credit risk valuation in the credit default swap (CDS) market and thereby the cost of debt finance.

* Corresponding author.
E-mail addresses: [email protected] (H.-J. Naumer), [email protected] (B. Yurtoglu).

https://doi.org/10.1016/j.gfj.2020.100571
Received 27 December 2019; Received in revised form 18 August 2020; Accepted 20 August 2020
Available online 25 August 2020
1044-0283/© 2020 Elsevier Inc. All rights reserved.
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

To do so, we employ a unique source of media data provided by Media Tenor, an international company specialized in media
coverage and content analysis. Media Tenor evaluates the influence of news collected from mass media and financial data sources,
sorted into 24 categories reflecting both ESG and non-ESG activities and classified according to its neutral, positive, or negative to­
nality. We use 261,907 news stories provided by Media Tenor about 1703 companies that have been included in the S&P 500 and the
STOXX 600 indices during the period 2006–2016. To estimate the relation between CSR-related news stories and debt financing costs
reflected in CDS spreads, we use a firm fixed effects specification and control for various firm-, bond-, and market-level factors that are
known to correlate with credit risk valuations in the CDS market.
The remainder of the paper is organized as follows. The next part offers a review of the extant literature and formulates our
predictions. Part 3 describes our data, the construction of our dependent and independent variables, and the econometric model. Part 4
reports the empirical results. Part 5 reports the results of several robustness checks and tests of sensitivity to omitted variables. Part 6
concludes.

2. Literature review and hypotheses development

2.1. CSR and the cost of debt

There has been a large amount of research on the relationship between CSR and corporate financial performance (CFP). Systematic
surveys summarizing this relationship (Friede et al., 2015; Margolis et al., 2009; Margolis & Walsh, 2003; Orlitzky et al., 2003) report a
significantly positive, albeit small mean effect of corporate social performance (CSP) on CFP. A large subset of these studies uses
various proxies for firm value or profitability that matter principally to equity holders.
On the other hand, studies on the relationship between CSP and the cost of debt are relatively scarce, and they focus predominantly
on U.S. issuers. This paucity of research is surprising given the large and growing importance of debt finance for corporations.1 Bauer
and Hann (2010) focus on 582 U.S. issuers over the years 1995–2006 and report that environmental engagement is associated with
lower credit risk, which is reflected primarily in bond prices but not in credit ratings. Bauer et al. (2009) report that the quality of
employee relations predicts the cross-sectional variation in credit risk, and firms with stronger employee relations have lower cost of
debt on newly issued bonds. Oikonomou et al. (2014) use data from 3240 bonds issued by 742 US public corporations and find that
aspects of CSR such as support for local communities, higher product safety and quality, and avoidance of controversies about the
firm’s workforce decrease the cost of corporate debt. Using an international sample of 23,194 firm-year observations from 2003 to
2012, Chollet and Sandwidi (2018) analyze the relationship between firm-level CSR and financial risk. They report a “virtuous circle”
in which better CSR performance leads to lower financial risk, which in turn reinforces firms’ commitment to improved CSR practices.
Ge and Liu (2015) find that better CSP is associated with lower yield spreads on initial U.S. corporate bond offerings. This relationship
is stronger for issuers that are more likely to face greater agency problems and higher information asymmetries. Goss and Roberts
(2011) show that in the U.S. loan market also, better CSP is rewarded with lower risk premiums. Salvi et al. (2020) come to a similar
conclusion from evidence that more responsible firms benefit from lower bond spreads and improved bond ratings, while greater CSR-
related controversy penalizes firms in both respects. Amiraslani et al. (2019) study the relationship between CSR and bond spreads
during the 2008–2009 financial crisis, a period when the market and the economy faced a severe shock to overall trust. They report
that high-CSR firms had substantially lower bond spreads than low-CSR firms. These effects are more pronounced for firms with a
higher propensity to engage in asset substitution or diversion (i.e., firms with high-yield debt and/or lower asset tangibility, and firms
incorporated in states that provide less bondholder protection during insolvency).
On the other hand, there is only limited evidence on the relationship between CSP and cost of debt outside the United States. Menz
(2010) studies a European sample of 498 bond issues from 2004 to 2007 and finds no significant advantage in credit spreads of socially
responsible corporations. More recent evidence suggests that the strength of this relationship is amplified by country-level ESG factors
(Stellner et al., 2015). Superior CSP reduces risk and lowers corporate bond spreads (by approximately 10 basis points) if the firm’s
ESG efforts mirror the ESG performance of the country it is located in.
The study closest to our analysis is by Kölbel et al. (2017), who explore the relationship between corporate social irresponsibility
(CSI) and financial risk. They define CSI coverage as the number of news articles per quarter that blame a firm for CSI pertaining to
ecological issues, community relations, labor relations, or corporate governance. They use an international sample of 539 firms from
38 countries during 2008–2013 and report that firms receiving higher CSI coverage face higher financial risk reflected in higher CDS
spread. The reach of the reporting media outlet is critical for this relationship. Outlets with a higher reach are more likely to attract the
attention of a critical number of stakeholders to CSI and thereby generate financial risk. Kölbel et al. (2017) focus, however, only on
negative news and exclude television coverage.
Our study complements these findings in several ways. First, rather than focusing only on negative news, we pay attention to the
whole spectrum of news and differentiate among positive, neutral, and negative tonalities. Second, we distinguish between financial
and mass media. Third, we differentiate between two different sets of image factors, with and without relation to ESG. Our estimates of
the association between negative ESG reports and CDS spreads are close to the estimates presented by Kölbel et al. (2017). In addition,
we find a significant and sizeable relationship between positive news coverage and CDS spreads. We also compare the potential impact

1
SIFMA (2018) reports that corporate bonds outstanding totalled $9.0 trillion at the end of 2017, up 63.5% from $5.5 trillion in 2008. In 2017,
$1.65 trillion of corporate bonds were issued, more than double the $760.4 billion in 2008. In 2017, stock offerings raised $220.1 billion, down
14.1% from $256.3 billion in 2008.

2
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

of ESG-related and non-ESG-related news and find that the former has more value relevance for investors.

2.2. Media and financial markets

The general role of media in financial markets has been well documented (Tetlock, 2015). Public news announcements are a major
mechanism for disseminating information to investors, who use it to assess the future cash flows of financial assets and their riskiness.
Studies focus primarily on the potential impact of media coverage on stock returns and document that media coverage significantly
affects both trading volume and returns (Barber & Odean, 2008; Tetlock, 2007). Several studies report that even stale news, if widely
publicized, can dramatically raise short-term returns (Huberman & Regev, 2001) and influence prices of large firms and widely fol­
lowed stocks in the S&P 500 (Tetlock, 2010).
These studies posit at least three different roles for the media. First, media coverage is a principal channel for communicating value-
relevant information to the public and thus can reduce the informational asymmetries between corporate insiders and outside
shareholders (Diamond & Verrechia, 1991; Healy & Palepu, 2001). Second, media can play a critical role in shaping corporate policy
by pressuring corporate decision makers to behave in ways that are “socially acceptable” (Dyck et al., 2008) and can align the interests
of insiders and outside shareholders by reducing the private benefits of control. Third, media not only report events but can also shape
the public’s perception of these events. Gurun and Butler (2012) argue that “biased” local media report information in a favorable way
reflecting pressure from local advertisers. Supporting this view, Solomon (2012) finds that media coverage is more favorable for firms
that use an investor relations firm, raising the possibility that firms may have incentives to manage their media coverage (Ahern &
Sosyura, 2014; Cahan et al., 2015).

2.3. Hypotheses

The media influence the acquisition of reputational capital (Liu & McConnell, 2013) by collecting, selecting, certifying, and
presenting information (Dyck et al., 2008), thereby reducing the cost of obtaining relevant information for various stakeholders who
can affect firms’ future cash flows and cost of capital. This role is akin to improved disclosure in reducing information asymmetries (e.
g., Diamond and Verrechia (1991)), and can affect firm value through at least three channels. First, better and more information might
improve insiders’ investment decisions or give investors more confidence that insiders are not appropriating firm value in unseen ways
(e.g., Jensen & Meckling, 1976; Shleifer and Wolfenzon (2002)), lowering the rates at which bondholders are willing to hold the firm’s
debt. Second, better disclosure can improve liquidity, which should in turn increase share and bond prices—a channel proposed by
Amihud and Mendelson (1988). Third, greater media coverage can lower the cost of capital by attracting investor attention (Fang &
Peress, 2009; Merton, 1987). These arguments lead to our first hypothesis:
Hypothesis 1. Corporate news from various media channels helps lower uncertainty for investors and therefore affects the risk
premiums of corporate bonds.
CSR is one of the principal mechanisms through which firms can build consumer trust (PwC, 2016). A large number of studies show
that firms’ CSR records affect consumers’ broad assessment of firms, in addition to their evaluation of products, final consumption
decisions, and willingness to pay. Such findings are consistent with the notion that reputation is a valuable intangible asset that helps
firms to create value by affecting the economic choices of various stakeholders (Barney, 1991; Hall, 1992). On this line of reasoning,
publicized CSR activities can influence firm value by affecting firms’ reputational capital. While this relationship finds widespread
acceptance (Lange et al., 2011), there is limited research on the precise mechanisms underlying it. Indeed, most research focuses on the
stock of reputational perceptions measured at a given point in time; the flow of reputation acquisition (or loss) is much less studied
(Moser & Martin, 2012). If the value of CSR comes from stronger reputational capital, investors should have higher expectations about
a given firm’s future cash flow if they have more awareness of CSR.
Hypothesis 2. CSR-related corporate news from various media channels helps build reputational capital and thereby lower the risk
premiums of corporate bonds.
Starting with Niederhoffer (1971), several studies have quantified the linguistic content of news announcements and examined the
impact of positive and negative tonality on various outcomes, including share returns (Engelberg & Parsons, 2011; Tetlock, 2007),
earnings following news stories (Tetlock et al., 2008), and firm value (Gurun & Butler, 2012). (Dyck et al., 2008) document that stock
prices are more sensitive to earnings emphasized by the press and less so to earnings information provided by the firm. Solomon (2012)
examines how media coverage of stocks affects investors’ allocation of flows to mutual funds. We expect a similar influence of news
tonality on CDS spreads:
Hypothesis 3. Corporate news with positive (negative) tonality decreases (increases) the risk premiums of corporate bonds, whereas
news with neutral tonality has no impact.
News from financial media is addressed to an institutional audience (“Wall Street”) that can act quicker than private investors
(“Main Street”) and also with higher volumes (Ioannou & Serafeim, 2015). Moreover, the CDS segment is exclusively served by
professional and institutional traders, who are more likely to follow the financial press. In addition, ESG-related investments are driven
more by institutional investors (Friede et al., 2015) than by private investors, who invest in a more “attention-driven” fashion (Barber
& Odean, 2008). Nevertheless, according to Fang and Peress (2009, p. 2024) the breadth of information dissemination is important,
suggesting that the mass media have some influence on CDS spreads, as they contribute, for example, to reputational capital and

3
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

awareness. We therefore look separately at the potential influences of news flows from financial and mass media on CDS spreads and
propose that.
Hypothesis 4. ESG-related news from financial media has more impact on the risk premiums of corporate bonds than ESG-related
news from mass media.
A large body of literature in psychology and economics suggests that humans respond more to negative than to positive information
(Ito et al., 1998). Given a unit of positive information and a unit of negative information, actors are likely to react more to the latter
than to the former. Several studies in psychology (Baumeister et al., 2001; Cacioppo & Gardner, 1999; Rozin & Royzman, 2001)
document evidence of this negativity bias, and in economics, experimental work on loss aversion suggests that actors put more weight
on a loss in utility than they do on a gain of the same magnitude (Kahneman & Tversky; Tversky & Kahneman, 1991). Theories of loss
aversion are similar to “frequency-weight” accounts of impression formation in psychology in that they are a product of differential
reactions to negative and positive information. Loss aversion has been documented for individual decision-makers across various
laboratory and real-world environments including macroeconomic environments (Bowman et al., 1999; Rosenblatt-Wisch, 2008). It
follows that negative news should dominate agenda setting, so that.
Hypothesis 5. Negative news has more impact on the risk premiums of corporate bonds than positive news.

3. Data

3.1. Media data

Our data on media news coverage and news tonality come from Media Tenor, a media research institute that screens leading daily
newspapers and TV broadcasts. Media Tenor collects and classifies news with the help of analysts trained to capture the tonality of a
piece of news, following a standardized process based on human experience and not on an algorithm. The technique differs from a
simple word count in that humans extract and code the tonality of the news using a specified codebook. The analysts explicitly evaluate
and code the wording and content of all news articles on business and politics with more than five lines, and TV broadcasts lasting more
than 5 s. The coding covers (1) volume, that is, the frequency with which a topic is mentioned, and (2) tonality, which is explicit or
contextual, and categorized as positive, neutral, or negative. Media Tenor analysts use 115 thematic topics to aggregate and code the
content and tonality of a piece of news into a smaller set of 24 image factors. This dataset has not been employed in previous studies.
To capture the CSR-relevance of the news flow, out of these 24 image factors we construct 13 that are relevant to “Environmental”
(E), “Social” (S), or “Governance” (G) topics, and 11 that refer to non-ESG topics. In addition, Media Tenor data enable us to distinguish
(1) between financial media and mass media, and (2) among positive, neutral, and negative tonalities. Table 1 lists the 24 thematic
topics that are used to extract and code media news coverage.
The mass media sources covered by Media Tenor comprise six German newspapers and magazines (Bild Zeitung, Focus, Spiegel, Bild
am Sonntag, FAZ Sonntagszeitung, Welt am Sonntag) and 25 international TV stations, of which 13 are German and the remaining 12 are
international. Three of these TV stations are from the United States (CBS, FOX, NBC), two from the UK (BBC 2 Newsnight, Ten o’clock
News BBC 1), and the remaining ones from Austria, Switzerland, France, Italy, Spain, Canada, and China. The international financial
print media comprise include the Financial Times, the German Handelsblatt, and The Wall Street Journal (U.S. edition).2
Table 2 presents the country composition of our sample in terms of both companies and media sources. Our dataset comprises
15,444 firm-years from 1703 companies in 18 countries that are (or were) listed in either the S&P 500 or the STOXX 600 during the
period from January 2006 through October 2016. Roughly half of the sample is composed of U.S. companies and the remaining half are
European companies, with Great Britain, France, and Germany being the most important markets. Of the news articles, 40% are about
U.S. companies and 60% are about European companies, with Germany and Great Britain accounting for about 26% and 10% of all
news articles, respectively. Thus compared to Kölbel et al. (2017), who studied companies from 38 countries, we focus more on
European markets and the United States. Also, Media Tenor, with a total of 261,907 news articles, appears to screen roughly three
times more news articles than RepRisk, which Kölbel et al. used to construct their corporate social irresponsibility measure (CSI) -
because CSI news by definition includes only items with negative effects on stakeholders’ legitimate claims, whereas we look at news
with positive, negative, and neutral tonality.
Table 3 shows that slightly fewer than half of the news items cover ESG-related topics. Among these topics, the “Social” category is
the most important, accounting for about 70% of all ESG-related news. The share of “Governance” related news is 27% and that of
“Environment” news only 2.5%.
Table 4 shows that roughly half of the news has a neutral tonality, while the remaining half is split between positive and negative
news. Financial media are the dominant source, accounting for 71% of all items.
Untabulated statistics show that the distribution of news items is roughly equal over industries, with a slight overrepresentation of
consumer cyclicals and financials.

2
Our data source does not cover the international edition of the Wall Street Journal, which experienced steep decreases in sales and print
advertising revenue in the years before 2017, when it ceased operations.

4
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Table 1
Image factor categories used by Media Tenor for ESG and non-ESG news.
Environment Social Governance Non-ESG

Environment Customer relations/sales Management Strategy and M&A

History Company culture Reporting
Human resources Globalization and economic climate
Politics and regulation Finances, financial results
Pricing Market position
Production Investments
Products Ratings
Salaries and unions Stock
Society, CSR, and sponsoring R&D and innovation
Image, public relations, and scandals Own research
Other

Table 2
Sample description by country (2006–2016).
Number of observations Share (%) Number of news items Share (%)

USA 7834 50.73 105,644 40.34

Europe
Great Britain 2156 13.96 27,063 10.33
France 1903 12.32 17,419 6.65
Germany 1900 12.30 68,210 26.04
Switzerland 528 3.42 12,068 4.61
Netherlands 255 1.65 5172 1.97
Italy 213 1.38 7327 2.80
Spain 126 0.82 5034 1.92
Sweden 120 0.78 2439 0.93
Ireland 65 0.42 1868 0.71
Belgium 41 0.27 1837 0.70
Other 303 1.96 7826 2.99
All 15,444 100 261,907 100

Table 3
ESG and non-ESG-related news.
Number of news items Share in total news (%) Share in ESG news (%)

All ESG news 127,300 48.61

Environmental 3174 1.21 2.49
Social 89,794 34.28 70.54
Governance 34,332 13.11 26.97
Non-ESG news 134,607 51.39

3.2. CDS spreads

Our dependent variable is the natural logarithm of the five-year CDS spreads. CDS spreads have increasingly been used to gauge the
financial health of companies in commercial applications (Moody’s and Bloomberg’s CDS implied default probability). We use CDS
data from Bloomberg, which have been widely adopted in recent research related to credit derivatives (Amato & Remolona, 2003;
Fabozzi et al., 2016). We focus on single-name CDS spreads that refer to contracts of five-year maturity and senior unsecured debt,
because these have emerged as the benchmark in CDS trading.

3.3. Other covariates

• To minimize omitted variable concerns, we include a number of covariates that are known to correlate with CDS spreads and media
coverage (Ashbaugh-Skaife et al., 2006; Attig et al., 2013; Campbell & Taksler, 2003; Chen et al., 2007; Collin-Dufresne et al., 2001;
Fisher, 1959; Zhang et al., 2009). Firm-year specific covariates are ROA (return on assets), firm size (natural logarithm of total
assets), leverage (ratio of total debt to total assets), interest rate coverage (ratio of EBIT plus interest expense to interest expense),
and capital intensity (ratio of fixed assets to total assets). As news coverage and tonality may proxy for ESG activities, we also
control for the firm’s ESG performance, using ESG scores obtained from Thomson-Reuters ASSET 4.
• In addition, we control for the general state of the market, which may affect both the way journalists interpret firm-specific events
and the CDS spreads. To measure market-wide investor sentiment, we use the value of country-specific volatility indices and the
country-specific difference between the yield of 10-year and 2-year Treasuries. For European countries for which there was no

5
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Table 4
Summary statistics of media variables.
Number of news articles (2006–2016)

Total By company

Variable Abbreviation N Mean Std. Dev. Min. Max

Panel A. All media

Number of news items - ESG AM-ESG 127,300 232.8 393.0 0 3226
Number of news items - Non ESG AM-NESG 134,607 221.1 288.0 0 1993
Neutral news - ESG AM-ESG-Neu 70,542 126.2 196.4 0 1462
Neutral news - non ESG AM-NESG-Neu 66,202 108.2 137.4 0 932
Positive news - ESG AM-ESG-Pos 23,226 44.1 91.0 0 860
Positive news - non ESG AM-NESG-Pos 35,923 59.1 79.8 0 654
Negative news - ESG AM-ESG-Neg 33,532 62.5 113.4 0 904
Negative news - non ESG AM-NESG-Neg 32,342 53.8 76.5 0 484

Panel B. Mass media

Number of news items - ESG MM-ESG 44,759 85.8 169.5 0 1465
Number of news items - Non ESG MM-NESG 30,704 57.3 103.6 0 757
Neutral news - ESG MM-ESG-Neu 22,312 42.7 79.7 0 621
Neutral news - non ESG MM-NESG-Neu 15,309 28.6 49.2 0 318
Positive news - ESG MM-ESG-Pos 8274 16.2 41.1 0 406
Positive news - non ESG MM-NESG-Pos 7904 14.7 29.5 0 264
Negative news - ESG MM-ESG-Neg 14,173 26.9 53.3 0 438
Negative news - non ESG MM-NESG-Neg 7435 14.0 28.1 0 190

Panel C. Financial media

Number of news items - ESG FM-ESG 82,541 147.3 241.9 0 1761
Number of news items - non ESG FM-NESG 103,903 164.4 199.7 0 1491
Neutral news – ESG FM-ESG-Neu 48,230 83.9 127.7 0 999
Neutral news - non ESG FM-NESG-Neu 50,893 79.9 95.7 0 786
Positive news - ESG FM-ESG-Pos 14,952 28.1 55.7 0 454
Positive news - non ESG FM-NESG-Pos 28,019 44.6 55.4 0 390
Negative news – ESG FM-ESG-Neg 19,359 35.7 65.9 0 508
Negative news - non ESG FM-NESG-Neg 24,907 39.9 52.4 0 352

Notes: Sample consists of 261,907 news items on 1703 companies included in the S&P 500 or the STOXX 600 indices between January 2006 and
October 2016.

relevant volatility index for a certain time period, we substitute the value of VSTOXX (the STOXX Europe 600 Volatility Index). We
use a dummy variable (QE) that is set to one in years covering the quantitative easing programs (United States: December
2008–March 2010, November 2010–June 2011, and September 2012–December 2013; Europe: starting from March 2013). Finally,
we include the bid-ask spread on CDS contracts as a proxy for liquidity. Table 5 provides summary statistics, and Table 6 reports the
Pearson correlation coefficients.

The mean of the ln(CDS) is 4.23, with a standard deviation of almost 1. This mean value corresponds to 117.1 basis points (std.
dev = 232 bp). The ESG scores are expressed on a scale running from 0 to 100, and they are more widely spread, with a standard
deviation of about 17 around a mean value of 84.48. Some of the covariates, such as return on assets, leverage ratio, interest coverage
ratio, and bid-ask spread, contain some outliers. Hence, we winsorize these variables at their 1st and 99th percentile values. The
volatility index captures the volatility spikes prevalent after the U.S. subprime crises and the debt crisis in Europe, as well as the
interventions of the Federal Reserve Board and the European Central Banks with unconventional measures such as quantitative easing
(QE). The slope of the yield curve (the term spread of 10-year vs. 2-year Treasuries) ranges from a steepness close to 164 basis points
(maximum) to an inverse shape of minus 56 basis points.
The correlation matrix shows that all independent variables are correlated with the dependent variable (column 1) at a significance
level of at least 5% or higher. The media variables (without distinction between financial and mass media, to save space) are in most
cases negatively correlated. The correlations of the remaining variables suggest economically meaningful relationships with the
exception of the dummy variable for the ECB’s QE. All correlation coefficients are at moderate levels, while the correlations of the
media variables are generally higher. The correlations among covariates are below 0.5 with the exception of the bid-ask-spread and the
slope of the yield curve. With the CDS spreads these covariates correlate more strongly.

3.4. Empirical models

The natural logarithm of CDS spreads, ln(CDS), is our dependent variable. We take logs to reduce the influence of outlier firms with
high CDS spread. In our base specification, we winsorize variables with outliers (the bid-ask spread, the return-on-assets ratio, and the
interest rate coverage ratio) at the 1st and 99th percentiles.
We use the following econometric model for our monthly data:

6
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Table 5
Summary statistics of nonmedia variables.
Variable Definition Mean Std. Min Max
dev.

CDS Single name CDS spreads that refer to contracts of five-year maturity 117.06 231.72 3.41 6235.57
ln(CDS) Natural logarithm of CDS 4.23 0.92 1.22 8.73
ESG score Equally weighted ESG score 84.48 16.93 5.29 97.92
Leverage* The ratio of total debt to total assets 27.73 14.03 0 91.97
ROA* Return on assets (winsorized at 1/99%) 5.36 6.03 − 19.25 29.53
ICR* Interest coverage ratio (earnings before interest & taxes plus interest expenses divided by interest 11.57 31.09 − 30.19 660
expense) (winsorized at 1/99%)
Firm size Natural logarithm of total assets ($Mn.) 10.45 1.14 5.84 13.64
Capital The ratio of fixed assets to total assets 29.60 21.75 0 107.64
intensity
Bid-ask CDS* Bid-ask-spread on CDS contracts (winsorized at 1/99%) 8.054 8.58 1.75 65.68
Volatility The value of the volatility index, country specific 19.70 7.31 10.42 60.67
Slope-yield The yield of 10-year bonds by country minus the yield of 2-year bonds by country 1.64 0.90 − 0.56 3.78
QE-US Dummy variable for the Fed’s QE; 1 if the observation is in a “Quantitative Easing” period (12/ 0.10 0.30 0 1
2008–03/2010, 11/2010–06/2011, or 09/2012–12/2013)
QE-EMU Dummy variable for the ECBs’ QE; 1 if the observation is in the “Quantitative Easing” period 0.06 0.23 0 1
starting from 03/2013

Notes: Sample consists of 1703 companies that were included in the S&P 500 or the STOXX 600 indices between January 2006 and October 2016. The
variables Volatility and Slope-Yield make use of country-specific values of the volatility indices and bond yields. * indicates that the variable is
winsorized at the 1st and 99th percentiles of the distribution. Number of observations is 15,444 for all variables.

ln(CDS)i,t = β0 + β1 *Xi + β2 *Zi + β2 *NewsESG

i,t + β3 *Newsi,t
non− ESG
+ fi + gt + εi,t , (1)

where ln(CDS)i, t is the natural logarithm of CDS spread for company i at time t; Xi is a vector of characteristics measured at the firm
level, Zi is a vector of market characteristics, NewsESG non− ESG
i, t (Newsi, t ) represents ESG (non-ESG) related news flow, fi is a set of firm-
specific effects, gt is a set of time dummies, and εi, t is an error term. We also employ models in which we replace NewsESG non− ESG
i, t (Newsi, t )
with other variables that reflect the tonality of the news coverage.
We estimate both models where the firm effects, fi, are assumed to be random and models where they are assumed to be fixed. Our a
priori preference for the firm fixed effects model is confirmed by Hausman tests that reject the null hypothesis that firm-specific effects
are orthogonal to the regressors in all equations. Therefore, we present the coefficient estimates from the firm fixed effects specifi­
cations and not the firm random effects specifications.
We address the potential for correlated standard errors by clustering at the firm level, which allows for correlation within firms,
across time. We assess the sensitivity of our results to how we handle outliers in robustness checks, presented in Section 5.

4. Results

Table 7 reports the coefficients estimated from firm fixed effects regressions of the natural logarithm of the CDS spread, ln(CDS), on
variables reflecting total news flow from all media (AM) sources (column 1), ESG and non-ESG related news flows (column 2), and
news tonalities for all media sources (column 3). Column 4 captures the tonalities of ESG and non-ESG news (all media). Column 5
differentiates between financial and mass media. All equations include covariates that control for firm-, bond-, and market-level
variables known to correlate with CDS spreads. The t-statistics use standard errors that are clustered at the firm level.
Column 1 shows that total news flow stemming from all media (AM) sources has an insignificant association with risk premiums on
corporate bonds. It does not lend statistical support to Hypothesis 1. The coefficients on the covariates are economically meaningful,
and most of them are significant at the 5% level or better. Firms with higher ESG scores and higher profitability have lower risk
premiums, whereas larger firms and those with higher bid-ask spreads have higher risk premiums.
The coefficients estimated for QE dummies show a highly negative significant association with the CDS spread. The coefficients of
the QE dummies indicate a tightening of about 9% both for the US and Europe.
We include country × volatility and country × slope interactions, but do not tabulate their coefficients separately. The estimated
coefficients show that these interaction terms capture significant and positive associations with risk premiums.
In column 2 we report separate coefficients on ESG and non-ESG related news flows from all media for our monthly data. The
coefficients on these two variables, AM-ESG and AM-NESG, take on opposite signs: ESG-related news has a coefficient of − 0.004 (t-
value = − 0.47), while non-ESG-related (AM-NESG) news has a significant coefficient of 0.024 (t-value = 2.26). This result doesn’t
provide sufficient evidence in favor of Hypothesis 2, which predicts that ESG-related news is likely to help build reputational capital
and thereby lower the risk premiums on corporate bonds.
Hypothesis 3 predicts that corporate news with positive (negative) tonality decreases (increases) the risk premiums of corporate
bonds, whereas news with neutral tonality has no impact. We test this hypothesis in columns 3 and 4 of Table 7. We start by considering
all media news and distinguish among neutral, positive, and negative tonalities (column 3). We then separate ESG-related from non-
ESG-related news flows from all media sources, again distinguishing among the three tonalities (column 4). We observe that both ESG-
related and non-ESG-related news with neutral tonality have no association with risk premiums at the conventional levels of statistical

7
 H.-J. Naumer and B. Yurtoglu
Table 6
Correlation matrix.
Covariates

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

(15)

(16)

(17)

(18)
(1) ln(CDS) 1
(2) AM-ESG-Neu − 0.10* 1
(3) AM-ESG-Pos − 0.12* 0.56* 1
(4) AM-ESG-Neg − 0.03* 0.56* 0.49* 1
(5) AM-NESG-Neu − 0.08* 0.52* 0.43* 0.42* 1
(6) AM-NESG-Pos − 0.12* 0.43* 0.46* 0.37* 0.42* 1
(7) AM-NESG-Neg 0.05* 0.43* 0.38* 0.48* 0.40* 0.38* 1
8

(8) ESG Score − 0.18* 0.14* 0.09* 0.12* 0.13* 0.11* 0.12* 1
(9) Leverage 0.16* − 0.02* − 0.03* 0.03* 0.00 − 0.04* 0.03* − 0.04* 1
(10) ROA − 0.39* − 0.01* 0.01* − 0.07* − 0.03* 0.03* − 0.11* 0.04* − 0.22* 1
(11) ICR − 0.16* − 0.00* 0.01* − 0.02* − 0.02* 0.02* − 0.03* − 0.04* − 0.29* 0.34* 1
(12) Firm Size − 0.08* 0.36* 0.24* 0.32* 0.35* 0.23* 0.27* 0.34* 0.12* − 0.24* − 0.07* 1
(13) Capital Intensity 0.03* − 0.10* − 0.06* − 0.03* − 0.10* − 0.02* − 0.01* 0.07* 0.14* 0.00 − 0.08* − 0.17* 1
(14) Bid-Ask CDS 0.65* − 0.13* − 0.11* − 0.06* − 0.11* − 0.09* 0.01* − 0.21* 0.14* − 0.27* − 0.07* − 0.17* 0.05* 1
(15) Volatility 0.31* 0.02* − 0.01* 0.05* − 0.02* − 0.06* 0.02* 0.03* 0.01* − 0.04* − 0.02* 0.09* − 0.05* 0.15* 1
(16) Slope-Yield 0.42* − 0.03* − 0.05* 0.02* − 0.05* − 0.11* − 0.01* 0.08* 0.04* − 0.06* − 0.03* 0.08* − 0.01* 0.12* 0.37* 1
(17) QE-US 0.05* − 0.01* − 0.04* − 0.03* − 0.00 − 0.04* − 0.05* − 0.06* 0.01* 0.04* − 0.01* − 0.01* 0.04* − 0.01* − 0.10* 0.09* 1
(18) QE-EMU 0.01* 0.05* 0.06* 0.04* 0.08* 0.04* 0.05* − 0.01* − 0.00 − 0.06* − 0.01* − 0.03* − 0.01* 0.02* − 0.06* − 0.00 − 0.07* 1
*
Indicates statistical significance at the 5% level or higher.

Global Finance Journal 52 (2022) 100571

H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Table 7
Media coverage, tonality, and CDS spreads.
(1) (2) (3) (4) (5)

All news AM ESG & non-ESG All news tonality AM tonality FM & MM tonality

AM-All news 0.015

(1.22)
AM-ESG − 0.004
(− 0.47)
AM-NESG 0.024**
(2.26)
AM-Neu − 0.001
(− 0.09)
AM-Pos − 0.035***
(− 3.93)
AM-Neg 0.057***
(5.62)
AM-ESG-Neu − 0.012
(− 1.56)
AM-ESG-Pos − 0.052***
(− 4.84)
AM-ESG-Neg 0.022*
(1.95)
AM-NESG-Neu − 0.007
(− 0.72)
AM-NESG-Pos − 0.020*
(− 1.96)
AM-NESG-Neg 0.078***
(6.76)
FM-ESG-Neu − 0.004
(− 0.44)
FM-ESG-Pos − 0.053***
(− 4.17)
FM-ESG-Neg 0.015
(1.26)
FM-NESG-Neu − 0.002
(− 0.19)
FM-NESG-Pos − 0.015
(− 1.40)
FM-NESG-Neg 0.073***
(6.68)
MM-ESG-Neu − 0.033***
(− 2.60)
MM-ESG-Pos − 0.047***
(− 3.31)
MM-ESG-Neg 0.022
(1.44)
MM-NESG-Neu − 0.037***
(− 2.82)
MM-NESG-Pos 0.063***
(3.16)
MM-NESG-Neg − 0.037***
(− 2.82)
ESG score − 0.005*** − 0.005*** − 0.005*** − 0.005*** − 0.005***
(− 3.87) (− 3.85) (− 3.83) (− 3.83) (− 3.80)
Leverage 0.005** 0.005** 0.005** 0.005** 0.005**
(2.19) (2.19) (2.23) (2.23) (2.23)
ROA − 0.020*** − 0.020*** − 0.020*** − 0.020*** − 0.020***
(− 6.32) (− 6.34) (− 6.35) (− 6.40) (− 6.43)
ln(IRC) − 0.001** − 0.001** − 0.001** − 0.001** − 0.001**
(− 2.17) (− 2.16) (− 2.17) (− 2.16) (− 2.13)
Firm size 0.070 0.071 0.066 0.066 0.068
(1.15) (1.15) (1.09) (1.09) (1.13)
Capital intensity − 0.002 − 0.002 − 0.002 − 0.002 − 0.002
(− 0.70) (− 0.71) (− 0.73) (− 0.77) (− 0.79)
Bid-Ask CDS 0.035*** 0.035*** 0.035*** 0.035*** 0.035***
(15.06) (15.12) (14.93) (15.13) (15.19)
QE-US − 0.088*** − 0.089*** − 0.086*** − 0.085*** − 0.086***
(− 3.60) (− 3.63) (− 3.53) (− 3.50) (− 3.54)
QE-EMU − 0.087** − 0.086* − 0.082* − 0.079* − 0.081*
(− 1.97) (− 1.94) (− 1.87) (− 1.80) (− 1.87)
Constant 0.385 0.394 0.480 0.534 0.588
(continued on next page)

9
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Table 7 (continued )
(1) (2) (3) (4) (5)

All news AM ESG & non-ESG All news tonality AM tonality FM & MM tonality

(0.52) (0.53) (0.65) (0.72) (0.80)

Volatility index × country interactions Yes Yes Yes Yes Yes
Slope-yield × country interactions Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes
Firm fixed effects Yes Yes Yes Yes Yes
Observations 15,444 15,444 15,444 15,444 15,444
Within R2 0.545 0.546 0.549 0.552 0.554
F-statistic (p-value) 149.8 (0.000) 143.7 (0.000) 132.5 (0.000) 119.6 (0.000) 99.20 (0.000)

Notes: Table 7 reports the coefficients estimated from firm fixed effects regressions of the natural logarithm of the CDS spread, ln(CDS), on variables
reflecting total news flow from all media (AM) sources (column 1), ESG and non-ESG related news flows (column 2), and news tonality (column 3).
Column 4 estimates separate coefficients on the tonality of news flows from all media (AM). Column 5 reports coefficients on variables reflecting total
news flows from financial media (FM) and mass media (MM). All equations include covariates that control for firm-, bond-, and market-level variables
known to correlate with CDS spreads. We test for the joint significance of the media variables with an F-test and report the value of the F-statistic (p-
value). t-statistics are in parentheses and use standard errors clustered on firm. Asterisks denote statistical significance levels: *p < 0.1, **p < 0.05,
and ***p < 0.01. Coefficients significant at the 5% level or better are in bold.

significance. News with positive tonality, whether ESG-related or non-ESG-related, has a statistically significant negative sign. A one-
unit increase in ESG-related news with positive tonality is associated with a decrease in risk premiums on the order of 5.2% (column 4),
while non-ESG-related news with positive tonality is associated with a decrease on the order of 2% (albeit with weak significance [t-
value = − 1.96]). These are sizeable coefficients. Moreover, non-ESG-related news with negative tonality has a significantly positive
coefficient of 0.078 (t-value = 6.76), whereas the coefficient on ESG-related news with negative tonality is indistinguishable from zero.
Overall, these results lend support to Hypothesis 3.
Hypothesis 4 differentiates the effects of news from different types of media and predicts that ESG-related news from financial
media (FM) will affect risk premiums on corporate bonds more than will news from mass media (MM). We test this hypothesis in
column 5 of Table 7, by estimating separate coefficients on ESG-related news flows from FM and MM. Looking first at news with neutral
tonality, we observe that news from financial media has an insignificant coefficient, while that from mass media has a significantly
negative coefficient. For news with positive tonality, news from FM decreases risk premiums by 5.3%, while similar news from MM has
a smaller coefficient (4.7%), but the difference between the coefficients is not statistically significant. For news with negative tonality,
news flows from both FM and MM have coefficients with positive signs but, again, no statistical significance. These results lend some
support to the prediction of Hypothesis 4.
The last hypothesis predicts that news with negative tonality has more impact on the risk premiums of corporate bonds than news
with positive tonality. To test this hypothesis, we compare the absolute values of the coefficients reported in column 3, 4, and 5 in
Table 7. We observe a sizeable difference between the coefficients on ESG-related and non-ESG-related news. ESG-related news
variables with positive vs. negative tonality have coefficients whose magnitudes differ substantially in absolute value: 0.057 vs. 0.035
in column 3; 0.052 vs. 0.022 for ESG and 0.020 vs. 0.078 for non-ESG in column 4. In column 5, we find similar differences between
ESG-related news flows with positive and negative tonality: 0.053 vs. 0.015 for FM, and 0.047 vs. 0.022 for MM. In the case of non-
ESG-related news flows, positive news (0.015 in absolute value) seems to be dominated by negative news (0.073). All differences are
statistically significant at the 5% level.
However, these differences in some cases are opposite to the prediction of Hypothesis 5, since the effects of news with positive
tonality are not generally lower in absolute value than those of news with negative tonality, lending no support to Hypothesis 5.

5. Robustness

In this section, we explore the robustness of the results reported in Section 4 using different subsamples and slices of data, and also
using alternative specifications.
As our media sources are disproportionately in German, they might pay more attention to companies from German-speaking
countries and less attention to international or U.S.-based news or companies. To study the potential for such bias, we run our
basic regressions separately for four different subsamples: Germany, the full sample excluding Germany, the United States, and the
United Kingdom. To save space we report only the results obtained using news flows from all media (AM) and mass media (MM) with
different tonalities. We report these regressions in Panel A of Table 8.3
Column 1 in Panel A of Table 8 shows that the results for Germany share basically the same pattern of results reported in Table 7 for
the full sample. Both ESG-related and non-ESG related news with positive tonality from all media sources have significantly negative
associations with risk premiums of bonds issued by German companies. A similar result emerges when we exclude Germany from the
full sample (column 2), or when we study only U.S. or UK companies (columns 3 and 4). Columns 5–8 repeat the same exercise using
only the news flow from mass media (MM) and find a pattern highly similar to the one obtained in columns 1–4.

3
Other results are available upon request.

10
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Table 8
Regression results using subsamples and alternative specifications.
Panel A - Regression results using subsamples.

(1) (2) (3) (4) (5) (6) (7) (8)

All media Mass media

Germany Full sample excluding USA UK Germany Full sample excluding USA UK
Germany Germany

ESG-Neu − 0.015 − 0.013 − 0.012 − 0.028 − 0.036 − 0.040** − 0.053** − 0.073

(− 0.70) (− 1.25) (− 0.84) (− 1.15) (− 1.53) (− 2.14) (− 2.37) (− 1.24)
ESG-Pos − 0.111*** − 0.100*** − 0.074*** − 0.076** − 0.094*** − 0.086*** − 0.047* − 0.050
(− 4.37) (− 6.51) (− 4.57) (− 2.03) (− 3.19) (− 3.98) (− 1.91) (− 1.00)
ESG-Neg − 0.022 − 0.002 − 0.004 0.080** − 0.026 0.023 0.019 0.156***
(− 0.79) (− 0.16) (− 0.25) (2.29) (− 1.33) (1.09) (0.91) (3.83)
NESG-Neu 0.081* 0.008 0.019 − 0.007 0.076** − 0.032* 0.001 − 0.103***
(2.03) (0.71) (1.30) (− 0.34) (2.49) (− 1.94) (0.02) (− 3.64)
NESG-Pos − 0.129*** − 0.039*** − 0.058*** − 0.017 − 0.132*** − 0.081*** − 0.096*** − 0.075
(− 5.39) (− 3.15) (− 3.83) (− 0.68) (− 3.99) (− 4.10) (− 3.00) (− 1.41)
NESG-Neg 0.043 0.087*** 0.052*** 0.094*** 0.064* 0.091** 0.002 0.141**
(1.43) (4.88) (3.32) (3.53) (1.83) (2.46) (0.09) (2.37)
Covariates Yes Yes Yes Yes Yes Yes Yes Yes
Year Fixed Yes Yes Yes Yes Yes Yes Yes Yes
Effects
Firm Fixed Yes Yes Yes Yes Yes Yes Yes Yes
Effects
Constant Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1900 13,544 7834 2156 1900 13,544 7834 2156
Within-R2 0.565 0.555 0.576 0.617 0.559 0.553 0.573 0.619

Panel B - Regression results using alternative specifications.

(1) (2) (3)

Industry×year Country×year E/S/G

interactions interactions

AM-ESG-Neu − 0.012 − 0.015*

(− 1.24) (− 1.80)
AM-ESG-Pos − 0.104*** − 0.041***
(− 7.59) (− 3.95)
AM-ESG-Neg − 0.005 0.024**
(− 0.38) (2.36)
AM-NESG-Neu 0.014 − 0.003
(1.26) (− 0.34)
AM-NESG-Pos − 0.051*** − 0.024**
(− 4.42) (− 2.49)
AM-NESG-Neg 0.080*** 0.071***
(5.05) (7.04)
AM-environmental-Neu − 0.019
(− 1.13)
AM- environmental -Pos − 0.057
(− 1.46)
AM- environmental -Neg 0.105**
(2.36)
AM-social-Neu − 0.013
(− 1.46)
AM- social -Pos − 0.058***
(− 4.87)
AM- social -Neg 0.024**
(2.03)
AM- governance -Neu 0.009
(0.83)
AM- governance -Pos − 0.030**
(− 2.24)
AM- governance -Neg 0.042**
(2.48)
Covariates Yes Yes Yes
Year fixed effects Yes Yes Yes
Firm fixed effects Yes Yes Yes
Constant Yes Yes Yes
Observations 15,444 15,444 15,444
Within-R2 0.552 0.715 0.688

11
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Notes: Panel A, columns 1–4 report the coefficients estimated from firm fixed effects regressions of the natural logarithm of the CDS spread, ln(CDS),
on variables reflecting ESG and non-ESG related news flows with different tonalities from all media (AM) sources for different subsamples. Columns
5–8 repeat the same exercise for mass media (MM). Panel B, columns 1 and 2 analyze the potential impact of including industry-year and country-year
interactions; column 3 uses country-specific response surfaces for the nonmedia covariates; and column 4 reports separate estimates on Environ­
mental (E), social (S), and governance (G) related news flows. All equations include covariates that control for firm-, bond-, and market-level variables
known to correlate with CDS spreads. t-statistics are in parentheses and use standard errors clustered on firm. Asterisks denote statistical significance
levels: *p < 0.1, **p < 0.05, and ***p < 0.01. Coefficients significant at the 5% level or better are in bold.

Panel B of Table 8 reports three alternative specifications. In column 1 (2) we include industry×year (country×year) interactions
and cluster the standard errors on industry (year). None of these alternative specifications change the general pattern that we have
observed in previous results. Column 3 differentiates between the three different categories of ESG news and reports separate co­
efficients on environmental (E), social (S), and governance (G) categories with different tonalities. The regression results show that
negative-tonality news in all three ESG categories has significantly positive associations with CDS spreads. The coefficients for the
environmental category are larger in absolute value (0.105, t-value = 2.36) than those for social (0.024, t-value = 2.03) and gover­
nance (0.042, t-value = 2.48) related news. News with positive tonality in all three ESG categories has uniformly negative associations
with CDS spreads; however, here the social category exhibits the highest association (− 0.058, t-value = 4.87). None of these three
categories have meaningful associations with CDS spreads when the news has neutral tonality.
Black et al. (2014) emphasize that neither a firm fixed effects nor a firm random effects model with extensive covariates can fully
eliminate the potential for omitted variable bias. To assess the sensitivity of our results to unobserved and omitted variables, we follow
an approach introduced by Hosman et al. (2010). Their sensitivity analysis is based on the assumption that the—typically
unknown—omitted variable (partly) predicts the dependent variable as strongly (same t-statistic, or same F-statistic for multiple
variables) as the strongest (known) variable that is already included. Using this approach, Hosman et al. (2010) calculate a lower
bound for the variable of interest.
Table 9 presents the lower bounds on firm fixed effects estimates for ESG-related and non-ESG-related variables with different
tonalities. Columns 1 and 2 report the lower bound coefficients under the assumption that the single omitted covariate has the same
predictive power as the strongest observed predictors of ln(CDS) (the ones with the largest t-statistic) –that is, the volatility indices and
the bid-ask spread. In column 3, we assume that the omitted variable(s) have the same predictive power as the volatility indices and the
bid-ask spread variables combined (largest F-statistic). Column 4 assumes that the omitted variable(s) have the same predictive power
as all the firm characteristics combined.
The lower bounds obtained show that the coefficient for ESG-related news with positive tonality is robust and not sensitive to
omitted variables. The lower bounds obtained for the coefficients on ESG-related news with neutral and negative tonality are more
sensitive to different degrees of omitted variable bias, but have the predicted signs and remain predominantly significant, whereas the
sensitivity tests for non-ESG-related news show no sensitivity at all.

6. Conclusion

A rapidly growing literature studies the various ways through which media shape the information environment and suggests that
media reporting can affect financial market outcomes. The news content provided by media can also be a useful window into investors’
and managers’ beliefs and thereby shed light on financial markets (Tetlock, 2015). We contribute to this literature by studying the
relationship between CDS spreads and the amount and tonality of ESG-related and non-ESG-related news flow. While the overall news
flow does not have a significant association with CDS spreads, we find that the topic (ESG- vs. non-ESG-related), the tonality, and the
source (financial vs. mass media) matter for the cost of debt financing both statistically and economically.
Our study is subject to some limitations. Owing to data constraints, we did not distinguish among different ownership structures.
ESG-related news might have a weaker association with CDS spreads under high ownership concentration. For example, institutional
owners may be inclined to use their ownership rights to promote CSR (Dyck et al., 2019). Such ownership structures may render the
news flow less relevant to reputational capital.
As our dataset consists of companies that are or were part of two major indices, our sample does not contain many companies with
high-yield bonds. However, ESG-related news may help to overcome informational asymmetries even more for bonds of lower quality.
Future research that studies high-yield bonds may find stronger associations between CDS spreads and news flow.
Our panel data design, with firm and year fixed effects (FE) and extensive covariates, is not a true causal design, but can be seen as
the best research design that is realistically available for studying this research question across countries. And a bounds analysis
suggests no bias from omitted variables.
Our results suggest that ESG-related and non-ESG related news (from financial as well as from mass media) influence CDS spreads
and can therefore increase or decrease the cost of debt financing by as much as 3.5% per month. This influence seems to come primarily
from positive ESG news, but we also find evidence for an association between negative news and CDS spreads. Hence, it is not only
what (ESG or non-ESG) you say, but how (tonality) you say it.

CRediT authorship contribution statement

Both authors, Hans-Jörg Naumer and B. Burcin Yurtoglu are jointly responsible for Conceptualization, Methodology, Data Cura­
tion, Investigation, Formal analysis, Software, Validation, Writing, Writing - Review & Editing of the manuscript.

12
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Table 9
Sensitivity bounds on omitted variable bias.
(1) (2) (3) (4)

Omitted variable is assumed to have the same predictive power as

Volatility index Bid-ask spread Bid-ask spread and volatility index All firm-level characteristics

AM-ESG-Neu − 0.066*** − 0.014 − 0.036*** − 0.015*

(− 5.72) (− 1.39) (− 3.34) (− 1.67)
AM-ESG-Pos − 0.146*** − 0.111*** − 0.132*** − 0.149***
(− 8.75) (− 7.07) (− 8.52) (− 9.49)
AM-ESG-Neg 0.005 0.022 0.023 0.028**
(0.32) (1.40) (1.34) (2.10)
AM-NESG-Neu − 0.069*** 0.022* − 0.006 0.020*
(− 5.42) (1.76) (− 0.44) (1.84)
AM-NESG-Pos − 0.179*** − 0.055*** − 0.103*** − 0.063***
(− 13.40) (− 4.06) (− 7.09) (− 6.49)
AM-NESG-Neg 0.017 0.125*** 0.105*** 0.147***
(1.09) (6.11) (5.14) (11.91)

Notes: Table 9 presents lower bounds on firm fixed effects estimates for ESG- and non-ESG-related variables with different tonalities, using the
approach by Hosman et al. (2010). Columns 1 and 2 report the lower bound coefficients under the assumption that a single omitted covariate has the
same predictive power as the strongest observed predictors of ln(CDS) (largest t-statistic) as Volatility Index and the Bid-Ask spread. For this exercise
we use a specification that combines all volatility indices into a single vector and do not use country-specific response surfaces. In column 3, we
assume that the omitted variable(s) have the same predictive power as the bid-ask spread and volatility index combined (largest F-statistic). Column 4
assumes that the omitted variable(s) have the same predictive power as all of the firm characteristics combined. We use the standard errors from the
base regression to compute the t-values (in parentheses). *, **, and *** respectively indicate significance at the 10%, 5%, and 1% levels. Results
significant at the 5% level or better are in boldface.

Declaration of competing interest

None.

Acknowledgements

We are indebted to Dr. Roland Schatz and Matthias Vollbracht from Media Tenor: Roland for graciously supporting our research by
sharing a unique set of data, and to Matthias for preparing and customizing the data according to our research requirements. We thank
Dimitris Andriosopoulos for extensive comments and suggestions that improved this paper. This research did not receive any specific
grant from funding agencies in the public, commercial, or not-for-profit sectors.

References

Ahern, K. R., & Sosyura, D. (2014). Who writes the news? Corporate press releases during merger negotiations. The Journal of Finance, 69(1), 241–291.
Amato, J. D., & Remolona, E. M. (2003). The credit spread puzzle (SSRN Electronic Journal).
Amihud, Y., & Mendelson, H. (1988). Liquidity and asset prices: Financial management implications. Financial Management, 17(1), 5.
Amiraslani, H., Lins, K. V., Servaes, H., & Tamayo, A. (2019). The bond market benefits of corporate social capital. Retrieved from https://papers.ssrn.com/sol3/
papers.cfm?abstract_id=2978794.
Ashbaugh-Skaife, H., Collins, D. W., & LaFond, R. (2006). The effects of corporate governance on firms’ credit ratings. Journal of Accounting and Economics, 42(1–2),
203–243.
Attig, N., El Ghoul, S., Guedhami, O., & Suh, J. (2013). Corporate social responsibility and credit ratings. Journal of Business Ethics, 117(4), 679–694.
Barber, B. M., & Odean, T. (2008). All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors. Review of Financial
Studies, 21(2), 785–818.
Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of Management, 17(1), 99–120.
Bauer, R., Derwall, J., & Hann, D. (2009). Employee relations and credit risk (SSRN Electronic Journal).
Bauer, R., & Hann, D. (2010). Corporate environmental management and credit risk (SSRN Electronic Journal).
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of General Psychology, 5(4), 323–370.
Bénabou, R., & Tirole, J. (2010). Individual and corporate social responsibility. Economica, 77(305), 1–19.
Black, B., de Carvalho, A. G., Khanna, V., Kim, W., & Yurtoglu, B. (2014). Methods for multicountry studies of corporate governance: Evidence from the BRIKT
countries. Journal of Econometrics, 183(2), 230–240.
Bowman, D., Minehart, D., & Rabin, M. (1999). Loss aversion in a consumption–savings model. Journal of Economic Behavior & Organization, 38(2), 155–178.
Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review of Psychology, 50, 191–214.
Cahan, S. F., Chen, C., Chen, L.[. L.]., & Nguyen, N. H. (2015). Corporate social responsibility and media coverage. Journal of Banking & Finance, 59, 409–422.
Campbell, J. Y., & Taksler, G. B. (2003). Equity volatility and corporate bond yields. The Journal of Finance, 58(6), 2321–2350.
Chen, L.[. L.]., Lesmond, D. A., & Wei, J. (2007). Corporate yield spreads and bond liquidity. The Journal of Finance, 62(1), 119–149.
Cheng, B., Ioannou, I., & Serafeim, G. (2014). Corporate social responsibility and access to finance. Strategic Management Journal, 35(1), 1–23.
Chollet, P., & Sandwidi, B. W. (2018). CSR engagement and financial risk: A virtuous circle? International evidence. Global Finance Journal, 38, 65–81.
Collin-Dufresne, P., Goldstein, R. S., & Martin, J. S. (2001). The determinants of credit spread changes. The Journal of Finance, 56(6), 2177–2207.
Deng, X., Kang, J.-K., & Buen, S. L. (2013). Corporate social responsibility and stakeholder value maximization: Evidence from mergers. Journal of Financial Economics,
110(1), 87–109.
Diamond, D. W., & Verrechia, R. E. (1991). Disclosure, liquidity, and the cost of capital. The Journal of Finance, 46(4), 1325–1359.
Dyck, A., Lins, K. V., Roth, L., & Wagner, H. F. (2019). Do institutional investors drive corporate social responsibility? International evidence. Journal of Financial
Economics, 131(3), 693–714.

13
H.-J. Naumer and B. Yurtoglu Global Finance Journal 52 (2022) 100571

Dyck, A., Volchokova, N., & Zingales, L. (2008). The corporate governance role of the media: Evidence from Russia. The Journal of Finance, 63(3), 1093–1135.
Engelberg, J., & Parsons, C. A. (2011). The causal impact of media in financial markets. The Journal of Finance, 66(1), 67–97.
Fabozzi, F. J., Giacometti, R., & Tsuchida, N. (2016). Factor decomposition of the Eurozone sovereign CDS spreads. Journal of International Money and Finance, 65,
1–23.
Fang, L., & Peress, J. (2009). Media coverage and the cross-section of stock returns. The Journal of Finance, 64(5), 2023–2052.
Ferrell, A., Liang, H., & Renneboog, L. (2016). Socially responsible firms. Journal of Financial Economics, 122(3), 585–606.
Fisher, L. (1959). Determinants of risk premiums on corporate bonds. Journal of Political Economy, 67(3), 217–237.
Friede, G., Busch, T., & Bassen, A. (2015). ESG and financial performance: Aggregated evidence from more than 2000 empirical studies. Journal of Sustainable Finance
& Investment, 5(4), 210–233.
Friedman, M. (1970, September). 13. The New York Times Magazine: The social responsibility of business is to increase its profits. Retrieved from http://www.
colorado.edu/studentgroups/libertarians/issues/friedman-soc-resp-business.html.
Ge, W., & Liu, M. (2015). Corporate social responsibility and the cost of corporate bonds. Journal of Accounting and Public Policy, 34(6), 597–624.
Goss, A., & Roberts, G. S. (2011). The impact of corporate social responsibility on the cost of bank loans. Journal of Banking & Finance, 35(7), 1794–1810.
Gurun, U. G., & Butler, A. W. (2012). Don’t believe the hype: Local media slant, local advertising, and firm value. The Journal of Finance, 67(2), 561–598.
Hall, R. (1992). The strategic analysis of intangible resources. Strategic Management Journal, 13(2), 135–144.
Healy, P. M., & Palepu, K. G. (2001). Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature. Journal of
Accounting and Economics, 31(1–3), 405–440.
Hong, H., Kubik, J., & Scheinkman, J. (2012). Financial constraints on corporate goodness. Cambridge, MA: National Bureau of Economic Research.
Hosman, C. A., Hansen, B. B., & Holland, P. W. (2010). The sensitivity of linear regression coefficients’ confidence: Limits to the omission of a confounder. The Annals
of Applied Statistics, 4(2), 849–870.
Huberman, G., & Regev, T. (2001). Contagious speculation and a cure for cancer: A nonevent that made stock prices soar. The Journal of Finance, 56(1), 387–396.
Ioannou, I., & Serafeim, G. (2015). The impact of corporate social responsibility on investment recommendations: Analysts’ perceptions and shifting institutional
logics. Strategic Management Journal, 36(7), 1053–1081.
Ito, T. A., Larsen, J. T., Smith, N. K., & Cacioppo, J. T. (1998). Negative information weighs more heavily on the brain: The negativity bias in evaluative
categorizations. Journal of Personality and Social Psychology, 75(4), 887–900.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4),
305–360.
Jo, H., Song, M. H., & Tsang, A. (2016). Corporate social responsibility and stakeholder governance around the world. Global Finance Journal, 29, 42–69.
Kahneman, D., & Tversky, A. Choices, values, and frames, Vol. 4, pp. 269–278 (n.d.).
Kölbel, J. F., Busch, T., & Jancso, L. M. (2017). How media coverage of corporate social irresponsibility increases financial risk. Strategic Management Journal, 33(3),
520.
KPMG. (2017). The road head: The KPMG survey of corporate responsibility reporting 2017. Retrieved from https://www.google.de/url?
sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwim55n1wYDcAhXCLFAKHZf_BGQQFggvMAA&url=https%3A%2F%2F1.800.gay%3A443%2Fhttps%2Fassets.kpmg.com%
2Fcontent%2Fdam%2Fkpmg%2Fxx%2Fpdf%2F2017%2F10%2Fkpmg-survey-of-corporate-responsibility-reporting-2017.
pdf&usg=AOvVaw2oG1WuOUaYrAm3K3_-Bbfp.
Lange, D., Lee, P. M., & Dai, Y. (2011). Organizational reputation: A review. Journal of Management, 37(1), 153–184.
Liu, B., & McConnell, J. J. (2013). The role of the media in corporate governance: Do the media influence managers’ capital allocation decisions? Journal of Financial
Economics, 110(1), 1–17.
Margolis, J. D., Elfenbein, H. A., & Walsh, J. P. (2009). Does it pay to be good…and does it matter?: A meta-analysis of the relationship between corporate social and financial
performance (SSRN Electronic Journal).
Margolis, J. D., & Walsh, J. P. (2003). Misery loves companies: Rethinking social initiatives by business. Administrative Science Quarterly, 48(2), 268–305.
Masulis, R. W., & Reza, S. W. (2015). Agency problems of corporate philanthropy. Review of Financial Studies, 28(2), 592–636.
Masulis, R. W., & Reza, S. W. (2020). Private benefits and corporate investment and financing decisions: The case of corporate philanthropy. European Corporate
Governance Institute (ECGI)., 603.
Menz, K.-M. (2010). Corporate social responsibility: Is it rewarded by the corporate bond market? A critical note. Journal of Business Ethics, 96(1), 117–134.
Merton, R. C. (1987). A simple model of capital market equilibrium with incomplete information. The Journal of Finance, 42(3), 483–510.
Moser, D. V., & Martin, P. R. (2012). A broader perspective on corporate social responsibility research in accounting. The Accounting Review, 87(3), 797–806.
Niederhoffer, V. (1971). The analysis of world events and stock prices. The Journal of Business, 44(2), 193–219.
Oikonomou, I., Brooks, C., & Pavelin, S. (2014). The effects of corporate social performance on the cost of corporate debt and credit ratings. Financial Review, 49(1),
49–75.
Orlitzky, M., Schmidt, F. L., & Rynes, S. L. (2003). Corporate social and financial performance: A meta-analysis. Organization Studies, 24(3), 403–441.
PwC. (2016). Redefining business success in a changing world CEO survey: 19th Annual Global CEO Survey. Retrieved from https://www.pwc.com/gx/en/ceo-
survey/2016/landing-page/pwc-19th-annual-global-ceo-survey.pdf.
Rosenblatt-Wisch, R. (2008). Loss aversion in aggregate macroeconomic time series. European Economic Review, 52(7), 1140–1159.
Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity dominance, and contagion. Personality and Social Psychology Review, 5(4), 296–320.
Salvi, A., Giakoumelou, A., & Bertinetti, G. S. (2020). CSR in the bond market: Pricing stakeholders and the moderating role of the institutional context. Global Finance
Journal, Article 100522. article in press.
Shleifer, A., & Wolfenzon, D. (2002). Investor protection and equity markets. Journal of Financial Economics, 66(1), 3–27.
SIFMA. (2018). U.S. Capital Markets Deck: SIFMA Research. Retrieved from https://www.sifma.org/wp-content/uploads/2018/09/US-Capital-Markets-Deck-2018-
09-06-SIFMA.pdf.
Solomon, D. H. (2012). Selective publicity and stock prices. The Journal of Finance, 67(2), 599–638.
Stellner, C., Klein, C., & Zwergel, B. (2015). Corporate social responsibility and Eurozone corporate bonds: The moderating role of country sustainability. Journal of
Banking & Finance, 59, 538–549.
Tetlock, P. C. (2007). Giving content to investor sentiment: The role of media in the stock market. The Journal of Finance, 62(3), 1139–1168.
Tetlock, P. C. (2010). Does public financial news resolve asymmetric information? Review of Financial Studies, 23(9), 3520–3557.
Tetlock, P. C. (2015). The role of media in finance. In S. P. Anderson, J. Waldfogel, & D. Strömberg (Eds.), Vol. 1. Handbook of media economics (pp. 701–721).
Amsterdam: North Holland.
Tetlock, P. C., Saar-Tsechansky, M., & Macskassy, S. (2008). More than words: Quantifying language to measure firms’ fundamentals. The Journal of Finance, 63(3),
1437–1467.
Tversky, A., & Kahneman, D. (1991). Loss aversion in riskless choice: A reference-dependent model. The Quarterly Journal of Economics, 106(4), 1039–1061.
US SIF Foundation. (2016). 2016 Trends Report highlights. Retrieved from https://www.google.de/url?
sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwiN6rfVw4DcAhVFKlAKHeeZCHgQFggvMAA&url=https%3A%2F%2F1.800.gay%3A443%2Fhttps%2Fwww.ussif.org%2Ffiles%
2FTrends%2FUS%2520SIF%25202016%2520Trends%2520Overview.pdf&usg=AOvVaw1O7IrtOhojYuMJ98JMm3_x.
Zhang, B. Y., Zhou, H., & Zhu, H. (2009). Explaining credit default swap spreads with the equity volatility and jump risks of individual firms. Review of Financial
Studies, 22(12), 5099–5131.

14