SSRN Id4500037

The Effect of ESG Controversies on Firm Value: Evidence from Korea
ed
Tae-Wook Ahna, Junesuh Yib*
iew
aDeepSearch, Inc.,TWO IFC Building 15, Gukjegeumyung-ro, Yeongdeungpo-gu, Seoul,
Korea, Tel: +82-10-4554-0498; Email: [email protected]
b Dongguk Business School, Dongguk University, 30, Pildong-ro 1-gil, Jung-gu, Seoul, Korea;
Tel: +82-10-3478-3322; Email: [email protected]
v
re
*Corresponding author
er
pe
ot
tn
rin
ep
Pr
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4500037
ESG Controversies and Firm Value: Moderating Role of ESG Performance
ed
Abstract
This study investigates whether Environmental, Social, and Governance (ESG hereafter)
iew
controversies affect firm value in Korea. Moreover, based on a risk management perspective
on ESG performance, this study analyzes whether firms causing ESG controversies can
benefit from ESG engagement. Using unique research setting in Korea, this study reveals a
negative and significant relationship between ESG controversies and firm value. In this study,
a fixed-effect panel regression model is applied to examine the moderating effect of ESG
v
performance. As a result, the loss of firm value partially recovers from ex-ante ESG activities
before the ESG controversies. The findings are robust when conducting a subgroup analysis
re
indicates that ESG performance has a moderating effect in the face of negative ESG
controversial issues only for highly visible firms such as (1) a large conglomerate commonly
called ‘chaebol’, (2) more searched on the Internet and (3) high analyst coverage. In sum, this
study contributes to the literature on ESG and stakeholder theory and sheds light on the
er
negative consequences of controversies and the insurance-like effect of ESG performance on
the controversies. These findings encourage corporate executives to engage in ESG activities
for risk management purposes.
pe
JEL Classiﬁcation: M14, G39, G41, L82
Keywords: ESG controversies, Firm value, Moderating effect, ESG performance, Firm
visibility, Stakeholder theory, Insurance mechanism
ot
tn
rin
ep
Pr
1. INTRODUCTION
The issue of ESG (Environmental, Social, Governance) has become a priority for
ed
companies and investors. The supply of ESG-related mutual funds and exchange-traded funds
(ETF) have grown at a remarkable pace with net flows more than doubling to $405 billion
worldwide in 20201. This explosion of investors' interest in ESG is due to the growing
iew
perception that not only corporate financial factors but also non-financial factors are important
to increase sustainability. Accordingly, institutional investors such as pension funds are
incorporating the ESG performance of companies into investment research and decision-
making process. The recent expansion of the stewardship code by institutional investors is
v
further strengthening corporate ESG activities. This is the reason why institutional investors
re
must manage potential risk factors by considering not only financial performance but also non-
financial performance to properly perform the fiduciary duty of trustees in managing funds.
Hence, ESG is the key issue in fulfilling the trustee's responsibilities. Moreover, one of the
er
major global ESG initiatives addressed by the EU2 and OECD3 is reinforcing ESG disclosures,
and Korea plans to gradually mandate disclosures on environmental and social activities as
well as governance structures depending on the size of the company4.
pe
In particular, the COVID-19 pandemic greatly contributed to the spread of ESG
importance. The outbreak has raised concerns about the destruction of the environment and
ecosystem, income disparity, and social polarization. It generates public awareness of
companies' proper environmental (E) and social (S) activities and governance structure (G).
ot
Many governments such as the U.S. and the EU have announced Green New Deal (GND)
proposals5 related to the environment to overcome the economic recession caused by the
tn
COVID-19 pandemic, and the Ministry of Economy and Finance (MOEF) in Korea also signed
the agreement to launch the Korea Green New Deal Trust Fund (KGNDTF) with the Global
Green Growth Institute (GGGI) on March 17, 2022. With the signing of the Agreement, the
rin
1 Global Sustainable Investment Alliance(GSIA) and Broadridge. Morningstar announced that the flows into US sustainable
funds and ETF in 2020 reached $51.1 billion, a significant increase over $21.4 billion in 2019, and AUM of them in 2020
reached $236.4 billion, up more than 70% from 2019. (https://www.morningstar.com/articles/1019195/a-broken-record-
flows-for-us-sustainable-funds-again-reach-new-heights)
ep
2The revised EU Non-Financial Reporting Directive requires large entities and groups to disclose information on their
development, performance, and position, and the impact of their activities relating to ESG and other matters.
3 https://www.oecd.org/finance/Investment-Governance-Integration-ESG-Factors.pdf
4 Listed companies with assets of more than 2 trillion won are currently disclosing governance reports, and sustainability
Pr
reports, including environmental and social activities, must be disclosed in 2025 afterward. Depending on the size of the
company, the obligation to disclose governance, environment, and social activities will be gradually expanded, and all listed
companies on Korea Exchange(KRX) will have to disclose sustainability reports by 2030.
5 https://en.wikipedia.org/wiki/Green_New_Deal
Korean government plans to contribute 6 billion won to the Trust Fund from 2022 to 2026 to
support the development of overseas projects related to the Korea Green New Deal and
ed
implementation of policies and regulations, as well as capacity building in the developing
countries.
From the corporate management perspective, ESG engagement is becoming an essential
iew
factor to increase firm value and secure sustainability. However, when it comes to business
operations daily, anti-ESG incidents such as environmental pollution, industrial accidents,
fraudulent accounting, breach of trust, embezzlement, and unfair trade, that is, unexpected
controversies can arise. ESG controversies are negative ESG news stories such as suspicious
v
social behavior and product-harm scandals, lawsuits, and failure of corporate governance that
re
place a firm under the media spotlight (Aouadi & Marsat, 2018). In behavioral finance,
investors are more sensitive to corporate wrongdoings than ethical cases which could have a
greater impact on firm value. Many behavior finance literature6 claims that negativity bias
er
causes investors to place greater weight on bad news compared to good news (Kothari et al.,
2009; Reyes, 2019). Negative news exerts a greater impact on firms than positive news (Krüger,
2015a).
pe
Negative stock market performance attracts more attention from retail investors than
comparable positive performance (Reyes, 2019). In this regard, the National Pension Service
(NPS), Korea's largest institutional investor, has been evaluating ESG since last year in
consideration of unexpected controversial issues in the guidelines for exercising shareholder
ot
rights. MSCI, a global ESG rating agency, evaluates the severity of ESG controversy by rating
it as very severe, severe, moderate, and minor while Refinitiv and Morningstar are also
tn
applying a method of discounting existing ESG scores amid controversy.
If so, what will happen to firm value if controversies occur in firms that work hard on
rin
ESG activities?7? Once a controversial issue occurs, firm value is expected to decline as the
company's image or reputation is damaged. However, if the company usually performs well in
ESG activities, there would be a possibility that its firm value will not necessarily fall.
ep
6 Bird et al(2007), Flammer(2013), Klassen and McLaughlin(1996), Krüger(2015), Reyes(2019), Sabbaghi(2020)

7In the case of CSR, there are two theories where CSR and CSiR can coexist. CSR and CSiR are incompatible phenomena,
and a company cannot behave in a socially responsible and irresponsible manner at the same time, while the orthogonal
hypothesis proposes that CSR and CSiR can increase or decrease in the same period and even, in some cases, a positive
Pr
causal relationship is proposed: The higher the CSiR, the higher the CSR. CSR and CSiR occur simultaneously in external
groups of stakeholders and shareholders in multinational corporations located in different regions or different business areas
(Strike et al. (2006)), in the relationship between the environment and communities and suppliers (Keig et al. (2015)), and in
external groups of interest and shareholders (Cruz et al.(2014)).
Even if there are no extant studies about the relationship between ESG engagement
and ESG controversies, some research claims that corporate social responsibility (CSR)8,
ed
similar to ESG, can play an important role as a protection against negative events based on the
insurance mechanism (Aouadi & Marsat, 2018; Minor & Morgan, 2011; Ortiz‐de‐Mandojana
& Bansal, 2016). In other words, some studies focus on CSR performance to preserve the
iew
financial value of a company through its risk-reducing properties from the perspective of risk
management, so negative evaluation can be mitigated amid corporate social irresponsibility
(CSiR). This is the result of supporting the risk reduction hypothesis that corporate risk
decreases as CSR performance increases. The more CSR engagement of firms in controversial
v
industry sectors, the lower the firm risk (Jo & Na, 2012). On the other hand, investors need to
re
recognize controversial issues. Companies with high customer awareness of CSR performance
had a significant effect on firm value, while companies with low customer awareness had either
negative or insignificant relationships (Servaes & Tamayo, 2013). Moreover, in the case of a
er
negative event, the awareness effect is greater than that of a positive event. In the case of highly
recognized companies, CSR concerns are harmful to firm value (Groening & Kanuri, 2013).
Therefore, investors need to recognize the negative activities of companies, and the role of
pe
media as a main channel for delivering related information is also influential.
This study aims to examine the moderating effect of ESG performance of companies
with ESG controversies. In other words, we investigate whether the negative effect on firm
value is alleviated when a controversy occurs in companies that are active in ESG
ot
engagement. In the first step, news stories containing at least one controversial issue are
extracted from over 20 million Korean news articles over the 2013-2018 period. As investors'
tn
awareness of negative ESG events is important, 20 million articles are searched and analyzed
to derive keywords for ESG controversies. By collecting news stories searched with a list of
keywords related to ESG controversy, we calculate the score of controversy as the ratio of the
rin
number of negative articles over the total number of news articles on a particular firm and test
the relation between controversy score and firm value. We also examine the moderating
effect of ESG performance where firm value, which has been damaged due to the occurrence
ep
of controversial issues, is recovered with the joint effect of overall ESG performance and
ESG controversies. We also find that each E, S, and G performance is in conjunction with
ESG controversies. For the company's ESG performance, we further find the support for how
Pr
8 In this paper, we describe the difference between ESG and CSR. However, when we treat both terms as if they are
interchangeable, we will use the terminology ESG/CSR instead.
long the ESG moderating effect will last by using the period of the previous(t―1), current(t0),
and next fiscal year(t+1). We lastly attempt to check what firm-specific characteristic owns a
ed
significant moderating effect on ESG performance. Our sample splits are respectively based
on high vs. low ESG performance, membership of the large business group set by the Fair
Trade Commission (FTC), the company's ownership structure, the search volume of any
iew
query submitted to Google search engine, and analyst coverage. In addition, we test whether
the moderating effect of CSR performance exists. CSR is recognized as a similar concept to
ESG, but CSR is a more comprehensive concept including social issues, stakeholders, and
spontaneity, and is difficult to see as a direct consideration in the investment decision-making
v
process, so we figure out whether there is a difference between ESG moderating effect and
re
CSR moderating effect.
This study is important because we find support that the moderating effect of ESG exists,
not the moderating effect of CSR. Until now, there have been many studies on the moderating
er
effect of CSR performance, but studies on the moderating effect of ESG performance are
extremely limited. (Aouadi & Marsat, 2018) argue that they tested and found support that the
interactive effect of ESG controversy and ESG performance on firm value. However, they used
pe
CSR score9, not ESG ratings, as an indicator to measure ESG performance and a dummy
variable for ESG controversy: 1 for companies facing at least one controversy during the
previous year, 0 otherwise. The effects of ESG and CSR on firm value may be similar.
While CSR aims to make a business accountable, ESG criteria make its efforts measurable.
ot
With CSR performance varying massively between businesses and sectors, there is a lack
of comparable metrics available. ESG performance, on the other hand, is generally
tn
quantifiable to a far greater degree. The Socially Responsible Investments (SRI) funds also
target companies that are good at CSR but use the ESG framework for practical investment
decisions. Therefore, CSR and ESG performance may be different to the extent that they
rin
directly affect firm value. In addition, CSR focuses on social issues while ESG can be
divided into more sophisticated and detailed activities for sustainable management
including the environment and governance structure.
ep
Second, this study differs from previous studies in that we directly derive the
controversy score for the controversial issues published in the media. Most of the extant
Pr
9 (Aouadi & Marsat, 2018) use CSP score which measures the firm's established reputation related to CSR issues and it is
provided by Asset4-Thomson Reuters, one of the major ESG/CSR data providers. We use the ESG score rather than the CSR
rating or score as we focus on the relation between ESG controversy and ESG performance in this study.
studies on CSR or ESG controversies use existing data provided by ESG data providers such
as Thomson Reuters' Asset4 DB, MSCI, and Sustainalytics while this study uses a main
ed
keyword related to ESG controversies and a polarity measure driven by a deep learning
algorithm called the convolutional neural network (CNN) model from 20 million news articles
by
iew
more than 130 media sources and draws controversy scores for those companies which
committed negative events. It is quite useful to calculate a controversy score for unfavorable
ESG-related issues exposed to media due to investor sentiment that respond more sensitively
to negative information than positive information. In the case of Korea, it is more appropriate
v
to use media that react sensitively to social problems because social issue-related events such
re
as ‘gapjil’10 occur more frequently.
Third, it is also relevant to discover the firm-specific characteristics with the moderating
effect of ESG performance. It may be interesting to test the difference in the ESG moderating
er
effect depending on the investor awareness of firm visibility, such as the Fair Trade
Commission (FTC)'s ownership structure, Internet search frequency, and analyst coverage.
With the test results, investors should be able to distinguish one group with an interest in ESG
pe
activities. Those firms in the group will be able to recognize motives to perform ESG activities
more actively and help set directions for future ESG performance. The rest of the paper is
organized as follows. The next section reviews prior studies and develops three hypotheses.
Section 3 outlines data and research model. Section 4 presents and discusses the result of
ot
empirical studies. Section 5 contains the discussion of robust tests, and the last section
concludes the article and suggests potential avenues for future research.
tn
2. Theoretical Background and Hypotheses

Although many studies have been conducted on the effect of ESG performance on firm value,
rin
the results are mixed depending on ESG measurement variables, performance indicators,
measurement periods, measurement techniques, and measurement channels (Albuquerque et
al., 2019; Baron, 2007, 2008; Benabou & Tirole, 2010; Fatemi et al., 2015). On one hand,
ep
greater ESG performance can increase firm value through theoretical models. ESG activities
can affect firm value positively by conducting empirical academic studies (Borghesi et al.,
2014; Ferrell et al., 2016; Gao & Zhang, 2015; Gillan et al., 2010; Iliev & Roth, 2021). On
Pr
10According to The New York Times, Gapjil indicates “the abuse of underlings and subcontractors by executives who
behave like feudal lords .” It can represent the illegal or unethical, and immoral behaviors of a firm.
the other hand, it is empirically proved that the association between ESG performance and
firm value is negative (Buchanan et al., 2018; Di Giuli & Kostovetsky, 2014; Hong et al.,
ed
2012; Masulis & Reza, 2015). There are no relations that exist between them ((Hsu et al.,
2018; Humphrey et al., 2012). However, a meta-analysis of more than 2,000 related
empirical studies is conducted with the results that roughly 90% of studies find a nonnegative
iew
relation between ESG and financial performance (Friede et al., 2015). In conclusion, a
positive link between ESG performance and firm value exists. Therefore, most of the
literature is in line with the view that ESG performance goes hand in hand with firm value.
Based on those studies reporting a positive relation between ESG performance and
v
firm value, this relation can be achieved by increasing cash flows or by decreasing the cost of
re
capital. Customers may want to buy goods from firms that have good reputations in ESG, and
employees are more productive when they work for such firms. Many investors may also
want to provide funds to such firms, resulting in having access to cheaper capital. On the
er
other hand, when a company has many ongoing controversies, it is reasonable to forecast a
decrease in its financial performance in response to the reactions of its stakeholders. In the
case of listed companies on the exchange, the capital market can also overreact to such events
pe
and make the impact even greater. A decrease in reputation due to ESG controversies leads to
a decrease in trust. Hence, stakeholders may undertake actions against the company. Based
on stakeholder theory and legitimacy framework, ESG controversy can affect the firm value
negatively (Aouadi & Marsat, 2018). In addition, negative ESG incidents result in long-term
ot
financial loss (Capelle-Blancard & Petit, 2017). A substantial drop in shareholders’ value is
caused by eco-harmful behavior (Flammer, 2013).
tn
Moreover, investors tend to focus on the negative effect that negative incidents can make
a bigger impact on shareholders’ value rather than positive ones. Furthermore, it is even more
clear that ESG controversies harm firm value. Negative events have greater, more consistent,
rin
and more intense consequences than positive events (Baumeister et al., 2001; Peeters &
Czapinski, 1990).
Besides, many behavioral finance studies have suggested that negative events trigger faster
ep
and cause stronger responses from audiences than positive events (Bird et al., 2007; Flammer,
2013; Klassen & McLaughlin, 1996; Kothari et al., 2009; Krüger, 2015b; Reyes, 2019;
Sabbaghi, 2020). Our research proves that investors will take stronger reactions in terms of
Pr
measuring the degree of controversy of companies through media channels and allowing direct
access to negative information (Bird et al., 2007; Flammer, 2013; Klassen & McLaughlin, 1996;
Kothari et al., 2009; Krüger, 2015b; Reyes, 2019; Sabbaghi, 2020). Empirically, bad news has
7
a greater impact on the stock price volatility of firms with ESG engagement than good news
((Sabbaghi, 2020). Moreover, investors will be more sensitive to negative information in Korea
ed
where herding behavior is strong. As a result, the relationship between ESG controversies and
firm value establishes the following assumption.
iew
H1: ESG controversies are negatively associated with firm value.
One of the reasons why companies are involved in ESG engagement is that they mitigate
the risk of reputation damage following a negative event. This is because the reputation of
v
companies actively carrying out ESG activities will improve, and negative effects can be
re
mitigated when an unfavorable event occurs. Hence, ex-ante ESG can protect the firm from the
ex-post repercussions of a negative event due to the prior accumulation of moral capital with
stakeholders. However, we understand that there are no prior studies on the direct moderating
effect of such ESG activities yet.
er
Meanwhile, moderating effects exist for CSR, which is recognized as a concept like ESG.
Many studies find various aspects of risk reduction by engaging in CSR activities (Feldman et
pe
al., 1997; Husted, 2005; McGuire et al., 1988; Orlitzky & Benjamin, 2001). CSR activities can
be transmitted as a sign of protecting various stakeholders, establishing moral reputation capital,
and enabling stable management in the wake of negative events during day-to-day business
operations (Donaldson and Preston(1995) and Jones(1995). Ex-ante CSR activities can protect
ot
against the negative effects of ex-post controversies due to the prior accumulation of moral
capital with stakeholders (Godfrey, 2005). Moreover, the insurance-like effect is crucial with
tn
the consideration of heterogeneity in the risk exposure of different firms and the extent to which
the firm can gain pragmatic and moral legitimacy with stakeholders (Koh et al., 2014). In
addition, CSR performance is more likely to play a protective role rather than to increase the
rin
financial value of a company (Ortiz‐de‐Mandojana & Bansal, 2016). In particular, the stock
and bond prices of firms engaging in CSR can benefit from insurance-like effects and the CSR
engagement on a continuous and long-term basis provides insurance-like effects on both the
ep
stock and bond prices of firms in the face of negative events (Shiu & Yang, 2017). However,
stakeholders are likely to perceive ESG and CSR performance as similar concepts without
clearly distinguishing them.
Pr
Hence, ESG performance, like CSR performance, will have insurance-like protection.
Investors among stakeholders can recognize that investment execution is based on ESG, not
CSR, so the insurance-like effect of ESG performance and CSR performance may be somewhat
8
different.
However, it is difficult for stakeholders to grasp whether companies are sincere or
ed
disguised in ESG activities. In addition, CSR has a spill-over or 'halo' effect on consumer
judgments, which mediates the negative impact on consumers' brand evaluation. Moreover,
CSR engagement enhances the firm value of even controversial industry firms (Cai et al., 2012).
iew
Therefore, insurance-like protection can be expected not only before the occurrence of the ESG
controversy but also after it. In addition, it seems clear that there is a continuity of insurance-
like effect for ESG performance. Hence, we establish the following hypothesis.
v
H2. ESG performance positively moderates the impact of ESG controversies on firm value.
re
If the insurance-like effect exists in ESG engagement, what characteristics of companies
will have more significant effects? Perhaps highly visible or recognized companies by the
er
public can be an example. Consistent with this point of view, the potential influence of CSR
on firm value is proportional to the degree of visibility in corporate social behavior (Barnett,
2014; Cho et al., 2013; Di Giuli & Kostovetsky, 2014; Servaes & Tamayo, 2013). A company's
pe
social behavior is evaluated as a rating of corporate social performance (CSP) and it can be
used as a tool that affects stakeholders, therefore, higher corporate morality is likely to
influence consumers, employees, and investors. Such visibility will be determined by unique
characteristics such as firm size and industry, but may also vary depending on the ownership
ot
structure, awareness, reputation, etc. Media can serve as a major channel for determining a
company's visibility. Because stakeholders will not be aware of what is happening inside the
tn
company and especially in the case of negative events, they will not be able to have access to
the truth. Therefore, only when external sources such as the media disclose both ESG
performance and ESG controversies, stakeholders can evaluate the reputation and trust of
rin
companies. In this regard, they acknowledge how CSR information is linked to firm value
through media attention (Jiao, 2010).
Regarding the unique firm characteristics, large companies will have higher visibility than
ep
small and medium-sized companies (SMEs). Because large companies have an absolute
majority of stakeholders, public interest will focus more on large companies. In addition, it
seems clear that the service (non-manufacturing) industry, which has more actual contact with
Pr
consumers, will draw more public attention than the manufacturing industry. By the same token,
business-to-consumer (B2C) companies will draw more attention from the public than B2B
(business-to-business) companies11. In addition, a large business group by FTC commonly
called ‘chaebol’ with fair trade issues and tunneling issues between affiliates due to the
ed
excessive expansion may be more easily noticed than a non-family business group. According
to the Financial Supervisory Service of Korea, chaebol is defined as large conglomerates
controlled by a family of the founder(s) or a controlling shareholder(s), with more than two
iew
trillion won in total assets. A company with high foreign ownership may draw more investors’
attention as many individual investors keep track of foreign investors' herding behavior. High
search volume or wider analyst coverage may be associated with firm visibility by the same
reasoning.
v
The intersection of corporate social performance (CSP) and controversial issues has a
re
remarkable impact on the firm value of highly visible companies (Aouadi & Marsat, 2018).
With high Google search volume, high analyst coverage, and CSR awards, those companies
that have a visibility effect will affect the interaction between controversial issues and social
er
performance, which has a strong positive (+) effect on firm value. Based on the above
discussion, we propose the following hypothesis:
pe
H3: The moderating effect of ESG performance will be more clearly demonstrated in
companies with high visibility.
3. Data and Research Model

ot
3.1 Data
The sample period of this study is from 2013 to 2018. Due to the COVID-19 pandemic,
tn
most Korean firms suffered unexpected difficulties in their business activities in early 2020
and their financial performance deteriorated significantly. As we focus on the role of ESG
controversy, we avoid FY2019 which can be affected by the outbreak starting in December
rin
2019. We also use 353 listed companies with ESG ratings to evaluate the ESG performance
and ESG moderating effect. We exclude the corporations that belong to the financial sector
such as banks, brokerage firms, and insurance companies. We also exclude companies with
ep
capital erosion and merger and acquisition engaged in the past from the sample.
Pr
11 Large business-to-consumer (B2C) firms receive the lion’s share of attention for their activities (Haddock‐Fraser, 2012),
and this attention is considered a key motivator to improve corporate, social, and environmental performance (Goettsche et
al., 2016).
10
We use the annual ESG ratings announced by the Korea Corporate Governance
Service (KCGS).12 The KCGS began evaluating the governance structure of companies in 2003.
ed
It has developed an ESG evaluation model containing environmental and social factors,
integrated ESG of listed companies, and disclosed ratings by sector every year since 2011. The
main categories for evaluating the environmental factor consist of environmental strategies and
iew
organizations, environmental management (management), environmental management
(activity), environmental performance, and stakeholder response. The main categories for the
social factor are related to workers, suppliers and competitors, consumers, and communities.
The governance factor covers the protection of shareholder rights including ownership
v
structure and internal transactions with related parties, managerial compensation structure, the
re
composition and operation of the board of directors, corporate disclosure, auditing quality, and
distribution policy.
KCSG gives the ESG rating in seven levels (S, A+, A, B+, B, C, D). Furthermore,
er
KCGS also announces each rating of E, S, and G in the same manner.
Table 1 shows the distribution of ESG composite ratings by year. The average ESG
composite rating of companies is B. As the number of companies subject to ESG evaluation
pe
increases recently, the number of companies that are given a D rating is also increasing. In this
study, 6 points to 1 point from A+ to D are assigned to each level of E, S, and G, and the sum
of the scores in these three categories is used as the ESG score13.
ot
[Insert Table 1 here]

tn
We use DataGuide (https://dataguide.fnguide.com) to obtain the firm-level financial

and accounting data including basic financial data such as Tobin's Q, market-to-book value
ratio, advertising expense to sales ratio, R&D expense to sales ratio, sales growth rate, debt-to-
rin
total asset ratio, and cash dividend ratio.

In addition, we use KISVALUE (https://www.kisvalue.com) for foreign ownership and
the largest shareholder's ownership including family members and related parties of the
ep
controlling shareholder(s). We also find large conglomerates-related information from Fair
12Domestic ESG ratings are provided by Sustinvest and Eco-Partners in addition to the Korea Corporate Governance
Service(KCGS). However, the data from the KCGS is the most available for use in terms of the number of companies, time
Pr
series period, and association.

13 Until recently, the Korea Corporate Governance Service officially announced E and S as "B or lower ratings," which
combine all B, C, and D. Accordingly, even though B or lower ratings exceeded half of the total, some researchers used this
rating, causing the lack of association due to limited data availability.
11
Trade Commission (https://www.egroup.go.kr). The number of analysts covering specific
companies is obtained from FNGuide (https://www.fnguide.com). Google search engine’s
ed
keyword popularity is available to the public via “Google Trends” (https://trends.google.com).
Google Trends indicates how often a particular word is searched relative to the total volume of
searches in specific periods and locations. The search frequency is a direct and unambiguous
iew
measure of attention (Da et al., 2011). Recent empirical finance literature shows that the search
frequency is significantly correlated with stock returns (Gwilym et al., 2016; Nguyen et al.,
2019).
3.2 ESG controversies news sampling
v
DeepSearch (https://deepsearch.com) systematically collects and analyzes news
re
articles, press releases, I.R. books, analyst reports, patent information, and credit information
on more than 1.4 million Korean public and private firms. We collect original news stories via
DeepSearch API (https://deepsearch.com/product/api) which includes all media coverage of
er
Korean-listed companies on the Korea Exchange (KRX). For each news article, we obtain a
unique identification (UID), title, publisher, date and time of release, a summary of the article,
name of the related firm, stock code of the firm, and polarity (polarity label*polarity score).
pe
One of our co-authors currently works for DeepSearch, Inc. and is responsible for data
collection and data analysis for this research. We hand-collect 23,837 news stories with ESG
controversial issues for 294 companies that have ESG composite and individual ratings by the
Korea Corporate Governance Service (KCGS) for the period 2013-2018. This corresponds to
ot
more than 330 ESG controversy news items per month.

We initially collect 228,580 newspaper articles that contain our main keyword
tn
‘controversy’ (‘논란’ in Korean) with a single company name from 2013 to 2018. We exclude
articles mentioning multiple company names to get rid of noisy stock market news reports. We
also remove news items that appear in other sections than Economy and Society as we assume
rin
ESG controversy-related issues should be assigned to either the Economy section or Society
section in the newspaper in any case. We select 72,866 potential ESG controversy articles
specifying a single company name after removing companies without ESG ratings. However,
ep
not all the collected news reports may not be ESG-related. Moreover, not all news reports may
not contain ESG controversial issues. However, it is still an extremely difficult and time-
consuming task to collect news stories with ESG controversy from that large set of documents.
Pr
DeepSearch API collects and classifies news based on an algorithm called the convolutional
12
neural network (CNN) model to capture the polarity of a piece of news following a standardized
process.
ed
In recent years, deep-learning methods for sentiment classification have been widely
adopted due to their excellent performance. (Kim, 2014) claimed that a convolutional neural
network (CNN) model14 with one layer of convolution performed remarkably well for
iew
document classification. He showed experimental results that a simple CNN model with static
vectors (CNN-static) achieved higher accuracies than the traditional machine-learning-based
models and other deep neural network structures. There are many other experimental pieces of
research showing that CNN models reach state-of-the-art results in polarity classification (Dos
v
Santos & Gatti, 2014; Kalchbrenner et al., 2014; Mohammad et al., 2013; Severyn & Moschitti,
re
2015; Tang et al., 2014; Wehrmann et al., 2017). DeepSearch API adopts the convolutional
neural network model (CNN-static) with 'morpheme vector' as an input rather than 'word
vector' which is mainly used in English. The morpheme vector refers to a vector representation
er
for the morpheme and can be derived by applying an existent word vector derivation
mechanism to the sentences divided into constituent morphemes. DeepSearch API applies
CNN as a feature extraction and text classification algorithm to achieve greater predictive
pe
accuracy of polarity.
Polarity ranges from [-1, +1], where -1 indicates negative news and +1 does positive
news with the strongest confidence. Polarity is calculated as the product of the polarity label
and polarity score. The polarity label indicates the tone of the article. For example, a label is
ot
assigned -1 if the learning algorithm predicts firm-specific news with a negative tone, 0 if a
neutral tone, and +1 if a positive tone. The Polarity score ranges between 0 and 1 and indicates
tn
the confidence of the polarity. A value close to 1 represents higher confidence. In this study,
we use news articles with a negative tone (polarity label = -1) and the confidence of the polarity
higher than 0.5 to focus on negativity and to increase accuracy in sampling. The sample
rin
selection process is explained in Appendix B.
[Insert Appendix B here]

ep
<Table 2> below reports the distribution of our sample across two dimensions. Panel
A displays the distribution of ESG controversial news articles per year and Panel B provides
Pr
14The CNN model was invented for image processing but has subsequently been demonstrated to be effective for detecting
context sentiment. CNN uses convolutional layers to filter inputs for useful information. See “Deep Learning for Sentiment
Analysis: A Survey” (2018). Available online: https://arxiv.org/abs/1801.07883/.
13
an industry breakdown of the sample by each category (E, S, G). Environment articles are only
103 (0.4%), while social articles are 7,868 (33.0%), and governance articles are 15,866 (66.6%).
ed
iew
We presume that governance-related scandals by the owner of Korean large
conglomerates attract substantial media attention, leading to a higher number of governance
news articles such as corruption, bribery, collusion, and accounting fraud.
v
3.3 Research Model
re
In this study, a fixed-effect panel regression model is used15. First, the effect of the
occurrence of controversial issues on the firm value is derived from the following equation.
er
TQi,t = α + βCIi,t―1 + ∑ γkCNTi,t―1 + ∑ δmindustryFE + ∑ θnyearFE + ei,t―1 (1)
TQ represents Tobin’s Q and CI does the score of the controversial issue. CNT refers
pe
to control variables. industryFE is a fixed effect by industry whereas yearFE is a fixed effect
by year. Tobin’s Q is a natural logarithm of the ratio of the market value of total assets to the
book value of total assets (the sum of the book value of total liabilities and the market value of
common and preferred stocks). In addition to Tobin’s Q, the ratio of market value to book
ot
value (MB) is also used as a dependent variable. CI is a natural logarithm of the proportion of
ESG-related controversial news articles in the total number of articles on individual companies
tn
from 130 media sources. As for CNT, advertising-to-sales ratio (AD), R&D-to-sales ratio (RD),
sales growth ratio (SG), debt-to-total asset ratio (DA), a proportion of foreign ownership in the
total number of shares outstanding (FOR), a proportion of largest shareholder’s stake in the
rin
total number of shares outstanding (IOWN), a natural logarithm of total assets (SIZE), cash
dividend ratio (DIV), a natural logarithm of the number of analysts covering a particular
company, and beta ratio (BE) are used. industryFE is a dummy variable for industries
ep
representing B2B manufacturing, B2B non-manufacturing, B2C manufacturing, and B2C non-
Pr
15 As a result of the Lagrange Multiplier Test and the Hausman Test, the null hypothesis is rejected, and a fixed-effect model
is used. In the empirical analysis, we test the robustness of the fixed-effect model results compared with the results of the
random effect model and the 2SLS (two-stage least squares) regression analysis.
14
manufacturing. yearFE is also a dummy variable representing the year16. Analysis of the
moderating effect of ESG performance is conducted through Equation (2) below.
ed
TQi,t+1 = α + β1CIi,t + β2ESGi,t + β3(CIi,t × ESGi,t) + ∑ γkCNTi,t + ∑ δmindustryFE +
∑ θnyearFE + ei,t (2)
iew
The ESG score (ESG) represents the degree of ESG performance of a company by
converting E, S, and G individual ratings into a score and summing up all these individual
scores. CIi,t × ESGi,t refers to the interaction term of the score of controversy and the score of
v
ESG that is interpreted as a moderating effect. It also refers to the reduced proportion of firm
re
value by controversial issues. ESGt is the ESG score in year t meaning that "t-1" refers to a
time lag of 1 year in the past, while "t+1" refers to a time lag of 1 year in the future. In a
chronological sequence of time periods, "t" represents the current time when the controversy
er
occurs. The control variables and industry and year fixed effects are the same as in Equation
(1). The descriptive statistics of variables are summarized in <Table 3>.
pe
The average number of controversies by year by firm is 14.8, the maximum value is
3,264. The average score of the controversy (CI) is a natural logarithm of it by dividing it by
ot
the total number of articles is -5.86. The average of the ESG score (ESG) which is the sum of
individual ESG individual scores is 8.93, indicating the average B rating for each category. The
tn
average of Tobin's Q is 1.30, the median is 1.02, the logarithmic value of it 0.78, and its median
value is 0.70. In addition, the average MB ratio is 1.512. Additionally, 25% of the sample firms
belong to a large business group, with an average foreign ownership ratio of 11.1%, the largest
rin
shareholder's stake including related parties of the largest shareholder(s) is 41.5%, and the
analyst coverage is 0.58.
ep
4. Empirical Study
4.1 Controversial issue and firm value
ESG controversies harm firm value as expected. <Table 4> shows that both Tobin’s Q
Pr
and the ratio of market value to book value of companies decrease next year (t+1) with the
16 For a more detailed description of the variables, see Appendix A.
15
occurrence of ESG controversies in the current year (t). This is a result of supporting
Hypothesis 1 and is consistent with the results of (Capelle-Blancard & Petit, 2017; Krüger,
ed
2015a), in which the information is quickly delivered to the capital market, making the
stakeholders recognize and react negatively.
iew
In Equation (1), the dependent variable referring to firm value is Tobin’s Q (TQ) or
the ratio of market value to book value (MB). The score of the company i′s t-year ESG
v
controversy is measured as CIt. The relationship between the current year's controversy score
re
(CIt) and the firm value of the next year (TQt+1 or MBt+1) is shown in the first and third columns,
and the relationship between the previous year's controversy score (CIt―1) and the firm value
of the current year is shown in the second and fourth columns respectively. As a result, the
er
ESG controversy has a negative effect on firm value in the current year but does not have a
significant effect on the firm value in the next year. Even in the case of (TQ), the previous
year's controversy score coefficient is not statistically significant but is zero. It means that
pe
investors' negative perception of the ESG controversy is only short-lived due to the role of the
media. In other words, when bad ESG news is released, market participants are quickly aware
of it due to rapid and intensive media reports, but the persistence of the information is not long
ot
enough. Looking at the economic significance of the coefficient, when the score of the
controversial issue indicated by the natural logarithm value increases by 1%, Tobin Q indicated
by the natural logarithm value decreases by 1.8%, and the ratio of market value to book value
tn
decreases by 2.8%.
Turning to the control variables related to ownership structure, the foreign ownership
ratio shows a positive relationship with firm value, but the large business group by FTC and
rin
controlling shareholder ownership are irrelevant to firm value. Among the control variables
related to financial ratios, debt ratio, cash dividend ratio, and beta have a positive relationship
with firm value while firm size shows a negative relationship with firm value. In addition,
ep
analyst coverage and firm value are positively related as well.
4.2 The moderating effect of ESG performance

Pr
We prove that the moderating effect of ESG performance of companies with

controversies exists. The result of the moderating effect and its persistence of ESG performance
16
using Equation (2) are summarized in <Table 5>. The dependent variable is Tobin’s Q (TQ) or
the ratio of market value to book value (MB), and the joint effect of the score of ESG
ed
controversy and ESG rating is shown as the CI × ESG term. The first dependent variable,
Tobin’s Q, is shown in column 1, 3, and 5, and the second dependent variable, the ratio of
market value to book value, is in column 2, 4, and 6. The ESG performance in the previous
iew
year (ESG―1) is displayed in column 1 and 2, the current year (ESG0) in column 3 and 4, and
the next year(ESG+1) in column 5 and 6.
v
re
As a result, the coefficient of the interaction term (CI × ESG―1) of the previous year's
ESG rating and the current controversy score (CI) is statistically significant as 0.004 for
Tobin’s Q and 0.006 for the market-to-book value. In other words, the ESG controversy(CI)
er
attenuates firm value by showing a coefficient of -0.048 for Tobin’s Q and -0.076 for the ratio
of market-to-book value. However, when the controversy is combined with the ESG rating of
the previous year, it turns out that the diminishing portion of firm value is alleviated by
pe
approximately 10% (-0.048 vs 0.004). High-performing ESG in the past helps reduce the firm
value which is consistent with the moderating effect of CSR (Godfrey, 2005; Shiu & Yang,
2017).
ot
We figure out that the moderating effect of ESG performance continues to be affected
by not only the ESG rating of the previous year but also the ESG rating of the current and next
year. The first interaction term in columns 3 and 4 (CI × ESG0) uses the ESG rating of the
tn
current year when the controversy occurs in the same year and the second interaction term in
columns 5 and 6 (CI × ESG+1) uses the ESG rating of the next year are also smaller than those
in column 1 and 2 (CI × ESG―1), but still positive. On the other hand, the firm value shows a
rin
statistically significant positive relationship with the ESG performance in the previous year
(ESG―1) with a coefficient of 0.021 (TQ) and 0.029 (MB), while positive weak significance or
no significance in the current year (ESG0) and the next (ESG+1). In other words, for firm value,
ep
the ESG performance itself has a positive effect only in the previous year, but it seems clear
that ESG performance combined with controversies play a crucial role in restoring firm value
damaged by the occurrence of a controversial issue(s) not only in the previous year but also in
Pr
the current and next year. However, the joint effect is greater in the previous year than the
current or next year which means continuing the effect to some extent after a certain moderating
17
effect occurs. The significance of other control variables is the same as the result from the
analysis of the relationship between the score of ESG controversy and firm value examined in
ed
Section 1.
4.3 The moderating effect of each E/S/G performance
iew
We demonstrate the moderating effect of such ESG performance in each category of
E, S, and G. <Table 6> is the result of analyzing the effect of the interactive term between
distinct environmental, social, and governance scores and controversy in each category of E,
S, and G on firm value. The dependent variable is Tobin’s Q and for each variable of E, S, and
v
G among the independent variables, column 1 to 3 are the result of using the previous year,
re
column 4 to 6 are the current year, and column 7 to 9 are the next year. Therefore, the interactive
terms of each E, S, and G score and the ESG controversy(CI × E, CI × S, CI × G) respectively
all show statistically significant and positive except for the interactive term of the controversy
er
and environmental rating (CI × E―1) in the previous year and governance rating in the current
year (CI × G0). In other words, not only ESG's overall performance but also E, S, and G's
individual performance has the moderating effect that alleviates the damage of firm value
pe
caused by the event of controversy and the effect is persistent. However, like the ESG overall
score, the coefficient of the interactive term between the E, S, and G scores in the previous year
with ESG controversies is the largest, indicating that the moderating effect of each E, S, and G
performance in the previous year is greater than that of the current or next year.
ot

tn
In particular, the moderating effect on the social category (CI × S) is greater than that
of the environment (CI × E) or governance category (CI × G). The coefficient of the interactive
rin
term (CI × S―1) between social score and controversy in the previous year is 0.007, which is
statistically significant and larger than that of environmental or governance score with ESG
controversies. In addition, both the decreasing firm value (-0.038) due to the controversial issue
ep
and the increasing firm value (0.047) by social performance (S) are greater than those of the
environment (E) or governance (G). This is the supporting result that stakeholders are most
sensitive to social issues among ESG activities of companies. The effects of E, S, and G
Pr
individual performance on firm value, like overall ESG performance, show statistical
18
significance in the previous year but most of them do not show the same results in the current
or next year.
ed
On the other hand, it is also evident that the diminishing firm value due to the
controversial issue is offset by both the overall ESG performance and the individual
performance of E/S/G. However, the moderating effect by each distinct performance of E, S,
iew
and G is greater than that by the overall ESG performance. In other words, in the event of a
controversy, the decline in firm value, which is attenuated by the overall ESG performance in
the previous year, is 8.3%, but the decline in firm value, which is reduced by each E, S, and G
performance in the previous year, reaches 13.8%, 18.4%, and 18.2% respectively. By looking
v
at columns 1 to 3, the coefficients of a controversial issue (CI) for firm value were -0.029, -
re
0.038, and -0.033 respectively, concerning E, S, and G activities in the previous year, but the
interactive term coefficients for combining these activities and controversy are 0.004 (CI × E―1
) and 0.007 (CI × S―1), and 0.006 (CI × G―1) respectively.
er
4.4 Firm characteristics with the moderating effect of ESG performance
This section analyzes the characteristics of companies whose decline in firm value is
pe
due to ESG performance when controversy arises. Corporate characteristics are divided into
two groups based on (1) ESG score, (2) ownership structure, and (3) firm visibility to examine
the differences in ESG moderating effects for each group. The empirical results are recorded
in <Table 7> which replicates the analysis in <Table 5> on subgroup of the data. As expected,
ot
it seems clear that ESG performance has a moderating effect in groups with high ESG scores,
large conglomerates. high Google search volume, and high analyst coverage. Recent studies
tn
have used a variety of proxies for firm visibility such as firm size, search frequency, media
coverage, institutional ownership, and the firm’s social reputation. Hence, the result of our
subgroup analysis demonstrates the corporate characteristics with the moderating effect of ESG
rin
performance is linked to firm visibility.

ep
First, regarding the ESG score, as expected, the high ESG score group has a moderating
effect thanks to the high volume of ESG activities, but the low ESG score group has no
Pr
moderating effect. In addition, in the high ESG score group, controversies negatively and
significantly impact firm value. Accordingly, stakeholders react more sensitively when adverse
19
events occur in companies with excellent ESG performance, but positively evaluate the ESG
performance of these companies.
ed
With regards to the ownership structure, the ESG moderating effect exists only for
large conglomerates while there is no difference in foreign ownership. In other words, the
coefficient of the interactive term between the occurrence of controversies and ESG
iew
performance of companies belonging to the large business group is 0.003 while the coefficients
of the interactive term of a non-member of the large business group and both high and low
foreign ownership group are no significance statistically. When negative events occur with
either the large business group so-called 'chaebol’ or ‘non-chaebol’, the firm value is equally
v
damaged, while the moderating effect of ESG performance exists only in companies belonging
re
to the large business groups. In the case of ‘chaebol’, even if some affiliates engage in ESG
activities, stakeholders tend to regard them as activities of the entire large business group. On
the other hand, regarding foreign ownership, for companies with a high foreign ownership ratio,
er
the firm value is damaged amid controversies, but the moderating effect of these companies'
ESG performance does not exist. Foreign shareholders may have doubts about the authenticity
of ESG activities although investors respond sensitively to negative events against companies
pe
with high foreign ownership.
With regards to recognition, it is evident that the higher the frequency of Google search
and the higher the analyst coverage, the more moderating effect on ESG performance exists.
As a result of analyzing the search frequency of each company provided by Google Trend, the
ot
firm value falls amid negative events in the high-frequency group, and there is an effect of
recovering diminishing value owing to ESG performance. In addition, with the number of
tn
analysts providing investment recommendations about a particular company, its firm value also
falls in the wake of negative events in both the high and low analyst coverage group, but the
moderating effect from ESG performance exists only in the high group. This is directly
rin
proportional to the moderating effect of firm visibility and ESG performance, and it is in line
with (Aouadi & Marsat, 2018) that stakeholders focus more on the ESG performance of highly
visible firms.
ep
Moreover, when it comes to the size of the company, there is no moderating effect of
ESG performance in both ‘big’ and ‘small’ groups. In the small group, the coefficient of the
interactive term between the controversy and ESG performance is negative, which is not
Pr
statistically significant though, but ESG performance produces an adverse effect. Stakeholders
may doubt the authenticity of the ESG performance of companies that are not large enough and
perceive it as 'window dressing'. With regards to profitability, the result demonstrates when
20
controversies arise from financially less-performing companies (‘low ROA’), the firm value
decreases and the moderating effect driven by ESG performance also exists. It shows that
ed
investors are even concerned about the risk of bankruptcy when negative events occur in
companies with low financial performance. Regarding the industry facing controversies, the
firm value falls within both manufacturing (‘Mfg’) and non-manufacturing (‘Non-mfg’) firms
iew
but the value diminishes less caused by ESG performance in manufacturing firms. These results
are contrary to the expectation that the moderating effect will be stronger in the non-
manufacturing industry which is closer to consumers. As major affiliates of large
conglomerates belonging to the manufacturing industry such as Samsung Electronics and
v
Hyundai Motor Company in the Korean market, they are more visible than ‘non-chaebol’ in
re
the non-manufacturing industry. According to Korean Standard Industrial Classification
(KSIC), the manufacturing industry includes semiconductors, chemicals, textiles, and
consumer goods. These manufacturing firms often lead to environmental controversies such as
er
water pollution, air pollution, and waste mistreatment and social issues such as 'gapjil’17 and
push-out sales. Hence, stakeholders will pay more attention to those companies which may
cause corporate wrongdoings. In addition, B2B companies faces the reduction of firm value
pe
due to the occurrence of controversies, but there is no offsetting effect because of statistically
insignificant ESG performance. B2C also exhibit no significant relationship between the
occurrence of controversies and the firm value. This result is contrary to the expectation that
B2C that is directly connected to consumers will have a stronger moderating effect, because
ot
investors are more interested in B2B transactions which are much larger than B2C in terms of
size and volume of transactions in Korea.
tn
5. Robustness Test
5.1 The moderating effect of CSR performance
rin
In this section, we investigate whether the moderating effect of ESG performance

differs from that of CSR. We use Equation (2) to verify the moderating effect of ESG
performance with the dependent variables of Tobin's Q (TQ) and the market-to-book value
ep
ratio (MB). The independent variables remain the same as Equation (1) including the score of
controversy (CI). The only difference is the CSR score versus the ESG score. There are not
Pr
17 Gapjil is a combination of the word “Gap” and the suffix “-jil,” which refers to an immoral or unethical action. The term
Gap derives from the relationship between “Gap” and “Eul,” which are formal expressions used by the standard labor
contract. In general, Gap is typically in a position as an employer who secures superior status, while Eul is an employee who
is of lower status. Gapjil refers to an abuse of power based on an imbalanced power relationship.
21
many institutions providing CSR ratings in Korea and the most widely used is the KEJI index
of the Korea Economic Justice Institute. Korea Economic Justice Institute (KEJI) discloses
ed
CSR scores for those companies based on its own selection criteria among listed companies in
Korea Stock Price Index (KOSPI) market. More specifically, KEJI evaluates Korean
manufacturing firms in seven categories: financial soundness (25 points), fairness (20 points),
iew
social contribution (15 points), consumer protection and satisfaction (15 points), environmental
commitment (10 points), and employee satisfaction (15 points). The institute adds up each
score of individual categories, converts the total score into 100 points scale based on its rating
formula, calculates a final score weighted by the indicator of each category, and selects the top
v
200 companies based on the total score. In this study, we use 227 companies with at least two
re
KEJI-CSR scores between 2013 and 2018. The average CSR score of these companies is 63.9,
with a median of 63.6, a maximum of 72.3, and a minimum of 59.2.
The empirical result using CSR scores is summarized in <Table 8> and it clearly shows
er
that the moderating effect does not exist with CSR performance, unlike other previous literature.
In both Tobin’s Q (TQ) and market-to-book value (MB), the interactive coefficients of CSR
performance and controversy in the previous year (CI × CSR―1) are 0.005 and 0.003
pe
respectively with statistical significance. However, there is no moderating effect in the
interactive term of CSR performance and controversy in the current year (CI × CSR0) and that
of CSR performance and controversy in the next year (CI × CSR+1). In other words, unlike the
ot
moderating effect of ESG performance, when a controversy arises, there is a moderating effect
of CSR performance in the previous year, but the persistence of the effect does not exist. As a
result, we observe that investors are more sensitive to ESG performance than CSR performance
tn
considering the rise in the firm value caused by ESG or CSR, the decline in the firm value due
to controversies, and their persistence of moderating effect.
rin
5.2 Robustness checks: Random effect and 2-stage least square model
ep
We employ a fixed-effect panel data model considering ESG performance by year and
industry. However, other than the year effect and industry effect, there may be an unrecognized
endogenous issue between ESG variables and the firm value. For example, ESG performance
Pr
may not increase the firm value by offsetting negative events but there could be an inverse
causality between them. In other words, companies with good ESG performance may not
22
increase their firm value due to the interaction with ESG controversies, but companies with
high firm value tend to be very enthusiastic about conducting ESG activities, which may lead
ed
to high ESG performance. In addition, unique financial characteristics among companies may
stimulate ESG activities.
Accordingly, in this section, we conduct robustness checks to avoid the endogenous
iew
issue using (1) the random effect model and (2) the 2SLS (two-stage least square) model with
the industry average ESG score by year as an instrument variable. As shown in <Table 9>, the
moderating effect and persistence of ESG performance exist even if the endogenous issue is
controlled. In the case of the random effect model, as in column 1 to 3, not only ESG in the
v
previous year but also the interactive term coefficients of the current year and next year with
re
controversies are statistically significant. Of course, as in the previous results, the interactive
term coefficient (CI × 𝐸𝑆𝐺―1) is 0.004 in table 9, which is larger than that of other periods
(0.002 and 0.002) and the range of compensation for the decline in the firm value (0.004 vs -
er
0.044) is also greater. As shown in Table 10, we find the moderating effect of ESG performance
in the 2SLS model. We estimate instrumental variable regression to mitigate the endogeneity
concern. Following the previous research (Aouadi & Marsat, 2018; Ghoul et al., 2017), we
pe
instrument the ESG score by the mean industry-year ESG score while excluding the focal firm.
In the first stage, we estimate the first-stage OLS regressions to predict the value of ESG
performance. We regress the ESG score on the instrument (mean industry-year ESG score),
other control variables, ESG controversies, and fixed effects. We then replace the ESG score
ot
with its fitted value as estimated in the first-stage regression. In the second stage, this predicted
value is used as an independent variable. As for the moderating effect, the interactive term
tn
between the controversy and the predicted value of ESG is statistically significant not only in
the previous but also in the current and next years.
rin
5.3 Robustness checks: 3-year average ESG performance

ep
Finally, the moderating effect is checked by using the three-year average ESG score,
not just a single-year ESG score. In the case of ESG rating, there is only a change when a firm-
specific influential event occurs, but the variability of rating is not that high. In addition, as
Pr
seen in the previous studies, we find that the persistence of the moderating effect exists, so
there may be interest in the possibility of expanding the persistence. (Chatterji et al., 2009) also
23
observed the moderating effect of CSR at the 5% significance level in the previous three years'
average score as well as the CSR score in the previous year. As a result, as shown in <Table
ed
11> below, we observe that the moderating effect of ESG performance exists not only in the
fixed-effect (FE) model but also in the random-effect (RE) model and the 2SLS. However,
compared to the single-year ESG score, the moderating effect of the three-year average ESG
iew
performance is slightly smaller. In other words, in the case of the fixed-effect model, the
moderating effect from the three-year average ESG performance is 5.6% (0.003 vs -0.053),
which is marginally higher than the 5.2% (0.004 vs -0.048) decline in the firm value offset by
ESG performance in a single year. We observe the same result in the 2SLS model as well as
v
the random effect (RE) model.
re
6. Conclusion
er
Understanding and valuing the importance of ESG performance can help corporate
managers to ensure the survival and legitimacy of firms by meeting the needs of stakeholders.
pe
By addressing ESG performance proactively, corporates can increase firm value and gain
stakeholders’ trust. If corporates only value financial performance, prioritize short-term gains
, and disregard ESG activities, and prioritize short-term gains, this can negatively affect the
perceptions of stakeholders. Hence, firms involved in controversial issues may confront
ot
legitimacy threats and may need to engage in ESG activities to restore their reputation and
legitimize their actions.
tn
Our research examines the impact of environmental, social, and governance (ESG)
controversies on firm value. We also analyze the moderating effect of ESG performance on
this relationship. More precisely, we examine whether ex-ante ESG activitiesrlde alleviates the
rin
negative effect on firm value even with controversies. Besides, we investigate whether the
moderating effect exists due to ESG performance in the previous year, the current year, and
the next year. Moreover, we find what firm characteristics own the ESG moderating effect. We
ep
use deep learning method called convolutional neural network (CNN) to determine polarity of
each news article and to select negative ESG controversial news more accurately from over 22
million news items.
Pr
Consequently, the firm value in the current year declines due to the occurrence of
controversies. The diminishing firm value is partially recovered due to pre-ESG activities
before the outbreak of the controversy. In addition, ESG performance in the current and next
24
year also helps contribute to the recovery of the firm value. In other words, companies with
controversies have a moderating effect owing to ESG performance and the persistence of the
ed
effect also exists. However, we discover that the moderating effect of ESG performance in the
previous year is much stronger than that of the current and next year. Such moderating effects
exist in the individual performance of each category E, S, and G. Like overall ESG performance,
iew
there is a greater moderating effect in the previous year over ESG performance than in the
current and next year. The moderating effect of social performance (S) is stronger than that of
environmental performance (E) and governance performance (G). We observe the moderating
effect of ESG performance in companies with (1) high ESG performance, (2) large business
v
group, (3) high Google search volume, (4) high analyst coverage, (5) low profitability, and (6)
re
the manufacturing industry. These characteristics may relate to firm visibility.
On the other hand, unlike the ESG moderating effect, the CSR moderating effect exists
only before the occurrence of controversies and the degree of moderating effect is much weaker
er
than that of ESG. In addition, the moderating effect of ESG is robust when we employ various
methodologies such as the random effect model and 2SLS in addition to the fixed-effect model,
and we also find the same result with a 3-year average ESG score over a single-year ESG
pe
score.
To sum up, we focus more on the moderating effect of ESG performance rather than the
CSR moderating effect and contribute to an improved understanding of the effects of ESG
controversies on the firm value. We also test the moderating effects of each category of E/S/G
ot
and discover that the moderating effect of the social category is the strongest among them. In
addition, we believe that it is a novel approach to build a database of ESG controversial news
tn
items by adopting the latest data science method. Finally, our research presents several
limitations as follows. First, when we conduct robustness checks using fixed-effects and the
instrumental variable approach to mitigate endogeneity concerns, potential unobserved factors
rin
may influence the level of controversies which in turn may also affect the relationship between
ESG performance and the firm value. Second, this study is based on a small sample focusing
on a single country with relatively a short period of time. Third, we only use the volume of
ep
negative news items (media frequency) as a proxy for ESG controversies and do not consider
the media severity of each news item.
We suggest that future research overcome these limitations by extending sample size ,
Pr
sample period, and number of countries and adding the severity proxy of media coverage.
25
ed
v iew
re
er
pe
ot
tn
rin
ep
Pr
26
REFERENCES
Albuquerque, R., Koskinen, Y., & Zhang, C. (2019). Corporate social responsibility and firm
ed
risk: Theory and empirical evidence. Management Science, 65(10), 4451–4469.
Aouadi, A., & Marsat, S. (2018). Do ESG Controversies Matter for Firm Value? Evidence
from International Data. Journal of Business Ethics, 151(4), 1027–1047.
https://doi.org/10.1007/s10551-016-3213-8
iew
Barnett, M. L. (2014). Why Stakeholders Ignore Firm Misconduct: A Cognitive View.
Journal of Management, 40(3), 676–702. https://doi.org/10.1177/0149206311433854
Baron, D. P. (2007). Corporate social responsibility and social entrepreneurship. Journal of
Economics & Management Strategy, 16(3), 683–717.
Baron, D. P. (2008). Managerial contracting and corporate social responsibility. Journal of
Public Economics, 92(1–2), 268–288.
v
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger
than good. Review of General Psychology, 5(4), 323–370.
re
Benabou, R., & Tirole, J. (2010). Individual and corporate social responsibility. Economica,
77(305), 1–19. https://doi.org/10.1111/j.1468-0335.2009.00843.x
Bird, R., Hall, A. D., Momentè, F., & Reggiani, F. (2007). What corporate social
responsibility activities are valued by the market? Journal of Business Ethics, 76(2),
189–206. er
Borghesi, R., Houston, J. F., & Naranjo, A. (2014). Corporate socially responsible
investments: CEO altruism, reputation, and shareholder interests. Journal of Corporate
Finance, 26, 164–181.
Buchanan, B., Cao, C. X., & Chen, C. (2018). Corporate social responsibility, firm value, and
pe
influential institutional ownership. Journal of Corporate Finance, 52, 73–95.
Cai, Y., Jo, H., & Pan, C. (2012). Doing Well While Doing Bad? CSR in Controversial
Industry Sectors. Journal of Business Ethics, 108(4), 467–480.
https://doi.org/10.1007/s10551-011-1103-7
Capelle-Blancard, G., & Petit, A. (2017). Every Little Helps? ESG News and Stock Market
Reaction. Journal of Business Ethics, 1–23. https://doi.org/10.1007/s10551-017-3667-3
ot
Chatterji, A. K., Levine, D. I., & Toffel, M. W. (2009). How well do social ratings actually
measure corporate social responsibility? Journal of Economics & Management Strategy,
18(1), 125–169.
tn
Cho, S. Y., Lee, C., & Pfeiffer, R. J. (2013). Corporate social responsibility performance and
information asymmetry. Journal of Accounting and Public Policy, 32(1), 71–83.
https://doi.org/10.1016/j.jaccpubpol.2012.10.005
Da, Z., Engelberg, J., & Gao, P. (2011). In search of attention. The Journal of Finance, 66(5),
1461–1499.
rin
Di Giuli, A., & Kostovetsky, L. (2014). Are red or blue companies more likely to go green?
Politics and corporate social responsibility. Journal of Financial Economics, 111(1),
158–180. https://doi.org/https://doi.org/10.1016/j.jfineco.2013.10.002
Donaldson, T., & Preston, L. E. (1995). The stakeholder theory of the corporation: Concepts,
ep
evidence, and implications. Academy of Management Review, 20(1), 65–91.

Dos Santos, C., & Gatti, M. (2014). Deep convolutional neural networks for sentiment
analysis of short texts. Proceedings of COLING 2014, the 25th International Conference
on Computational Linguistics: Technical Papers, 69–78.
Fatemi, A., Fooladi, I., & Tehranian, H. (2015). Valuation effects of corporate social
Pr
responsibility. Journal of Banking & Finance, 59, 182–192.

Ferrell, A., Liang, H., & Renneboog, L. (2016). Socially responsible firms. Journal of
Financial Economics. https://doi.org/10.1016/j.jfineco.2015.12.003
27
Flammer, C. (2013). Corporate social responsibility and shareholder reaction: The
environmental awareness of investors. Academy of Management Journal, 56(3), 758–
ed
781. https://doi.org/10.5465/amj.2011.0744
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.
Gao, L., & Zhang, J. H. (2015). Firms’ earnings smoothing, corporate social responsibility,
iew
and valuation. Journal of Corporate Finance, 32, 108–127.
Ghoul, S. El, Guedhami, O., & Kim, Y. (2017). Country-level institutions, firm value, and the
role of corporate social responsibility initiatives. Journal of International Business
Studies, 48(3), 360–385.
Gillan, S., Hartzell, J. C., Koch, A., & Starks, L. T. (2010). Firms’ environmental, social and
governance (ESG) choices, performance and managerial motivation. Unpublished
v
Working Paper, 10.
Godfrey, P. C. (2005). The Relationship Between Corporate Philanthropy And Shareholder
re
Wealth: A Risk Management Perspective. Academy of Management Review, 30(4), 777–
798. https://doi.org/10.5465/amr.2005.18378878
Goettsche, M., Steindl, T., & Gietl, S. (2016). Do customers affect the value relevance of
sustainability reporting? Empirical evidence on stakeholder interdependence. Business
Strategy and the Environment, 25(3), 149–164.
er
Groening, C., & Kanuri, V. K. (2013). Investor reaction to positive and negative corporate
social events. Journal of Business Research, 66(10), 1852–1860.
Gwilym, O. A., Hasan, I., Wang, Q., & Xie, R. (2016). In Search of Concepts: The Eﬀects of
Speculative Demand on Stock Returns. European Financial Management, 22(3), 427–
pe
449.
Haddock‐Fraser, J. (2012). The role of the news media in influencing corporate
environmental sustainable development: an alternative methodology to assess
stakeholder engagement. Corporate Social Responsibility and Environmental
Management, 19(6), 327–342.
ot
Hong, H., Kubik, J. D., & Scheinkman, J. A. (2012). Financial constraints on corporate
goodness. National Bureau of Economic Research.
Hsu, P., Liang, H., & Matos, P. (2018). Leviatan Inc. and sustainability, Working Paper,
Singapore Management University.
tn
Humphrey, J. E., Lee, D. D., & Shen, Y. (2012). Does it cost to be sustainable? Journal of
Corporate Finance, 18(3), 626–639.
Iliev, P., & Roth, L. (2021). Do Directors Drive Corporate Sustainability? Available at SSRN
3575501.
rin
Jiao, Y. (2010). Stakeholder welfare and firm value. Journal of Banking and Finance, 34(10),
2549–2561. https://doi.org/10.1016/j.jbankfin.2010.04.013
Jo, H., & Na, H. (2012). Does CSR Reduce Firm Risk? Evidence from Controversial Industry
Sectors. Journal of Business Ethics, 110(4), 441–456. https://doi.org/10.1007/s10551-
012-1492-2
ep
Jones, T. M. (1995). Instrumental stakeholder theory: A synthesis of ethics and economics.

Academy of Management Review, 20(2), 404–437.
Kalchbrenner, N., Grefenstette, E., & Blunsom, P. (2014). A convolutional neural network
for modelling sentences. ArXiv Preprint ArXiv:1404.2188.
Kim, Y. (2014). Convolutional neural networks for sentence classification. ArXiv Preprint
Pr
ArXiv:1408.5882.
Klassen, R. D., & McLaughlin, C. P. (1996). The Impact of Environmental Management on
Firm Performance. Management Science, 42(8), 1199–1214.
28
https://doi.org/10.1287/mnsc.42.8.1199
Koh, P., Qian, C., & Wang, H. (2014). Firm litigation risk and the insurance value of
ed
corporate social performance. Strategic Management Journal, 35(10), 1464–1482.
Kothari, S. P., Shu, S., & Wysocki, P. D. (2009). Do managers withhold bad news? Journal
of Accounting Research, 47(1), 241–276. https://doi.org/10.1111/j.1475-
679X.2008.00318.x
Krüger, P. (2015a). Corporate goodness and shareholder wealth. Journal of Financial
iew
Economics, 115(2), 304–329. https://doi.org/10.1016/j.jfineco.2014.09.008
Krüger, P. (2015b). Corporate goodness and shareholder wealth. Journal of Financial
Economics, 115(2), 304–329.
Masulis, R. W., & Reza, S. W. (2015). Agency problems of corporate philanthropy. The
Review of Financial Studies, 28(2), 592–636.
Minor, D., & Morgan, J. (2011). CSR as reputation insurance: Primum non nocere.
v
California Management Review, 53(3), 40–59.
Mohammad, S. M., Kiritchenko, S., & Zhu, X. (2013). NRC-Canada: Building the state-of-
re
the-art in sentiment analysis of tweets. ArXiv Preprint ArXiv:1308.6242.
Nguyen, C. P., Schinckus, C., & Nguyen, T. V. H. (2019). Google search and stock returns in
emerging markets. Borsa Istanbul Review, 19(4), 288–296.
Ortiz‐de‐Mandojana, N., & Bansal, P. (2016). The long‐term benefits of organizational
resilience through sustainable business practices. Strategic Management Journal, 37(8),
er
1615–1631.
Peeters, G., & Czapinski, J. (1990). Positive-negative asymmetry in evaluations: The
distinction between affective and informational negativity effects. European Review of
Social Psychology, 1(1), 33–60.
pe
Reyes, T. (2019). Negativity bias in attention allocation: Retail investors’ reaction to stock
returns. International Review of Finance, 19(1), 155–189.
Sabbaghi, O. (2020). The impact of news on the volatility of ESG firms. Global Finance
Journal, 100570.
Servaes, H., & Tamayo, A. (2013). The impact of corporate social responsibility on firm
ot
value: The role of customer awareness. Management Science, 59(5), 1045–1061.

https://doi.org/10.1287/mnsc.1120.1630
Severyn, A., & Moschitti, A. (2015). Unitn: Training deep convolutional neural network for
twitter sentiment classification. Proceedings of the 9th International Workshop on
tn
Semantic Evaluation (SemEval 2015), 464–469.

Shiu, Y., & Yang, S. (2017). Does engagement in corporate social responsibility provide
strategic insurance‐like effects? Strategic Management Journal, 38(2), 455–470.
Tang, D., Wei, F., Qin, B., Liu, T., & Zhou, M. (2014). Coooolll: A deep learning system for
rin
twitter sentiment classification. Proceedings of the 8th International Workshop on

Semantic Evaluation (SemEval 2014), 208–212.
Wehrmann, J., Becker, W., Cagnini, H. E. L., & Barros, R. C. (2017). A character-based
convolutional neural network for language-agnostic Twitter sentiment analysis. 2017
International Joint Conference on Neural Networks (IJCNN), 2384–2391.
ep
Pr
29
Appendix A. Variable Definitions
Variable Definition
ed
TQ Tobin’s Q. Ln(market value of equity + book value of assets – book value of equity – balance
sheet deferred taxes)/book value of assets)
MB The ratio of Market value to Book value. Ln(market value of equity/book value of equity)
CI Controversy Issues. The logarithm of the number of negative news related to ESG controversies
is divided by the number of the total news published in 130 newspapers.
iew
ESG Sum of converted scores from each E, S, and G grade of the company announced by KCGI. The
converted score from each grade is as follows. A+:6, A:5, B+:4, B:3, C:2 D:1
AD Advertising-to-sales ratio. Advertisement cost divided by sales.
RD R&D-to-sales ratio. R&D cost divided by sales.
SG Sales growth ratio Growth rate of sales.
DA Debt-to-total asset ratio. Long-term debt is divided by total assets.
v
FOR Foreign investor ownership. A proportion of foreign ownership in the total number of shares
outstanding.
re
CHAEBOL Dummy variable equal to 1 for firms included in conglomerates by the FTC (Fair Trade
Committee, and 0 otherwise.
IOWN A proportion of the largest shareholder’s stake in the total number of shares outstanding.
Ownership ratio for the largest stockholders or special relatives stockholders.
SIZE A natural logarithm of total assets.
er
DIV Cash dividend ratio of common stock.
ANALYST A natural logarithm of the number of analysts covering a particular company.
BETA Beta coefficient
ID Industry dummy variables (B2B manufacturer, B2B service, B2C manufacturer, B2Cservice).
Year dummy variables (2013, 2014, 2015, 2016, 2017, 2018)
pe
YD
ot
tn
rin
ep
Pr
30
Appendix B. ESG controversies sample selection process
Number of Samples (news
ed
Sample Selection Criteria
items)
Collect news items containing the main keyword "controversy" in the
1 228,580
Economy and Society section
iew
Delete news items containing multiple company names or no company
2 (120,025)
name to avoid noise
3 Delete news on companies without ESG ratings provided by KCGS (35,689)
4 Remove news items with polarity >-0.5 (49,029)
5 Final sample 23,837
v
re
er
pe
ot
tn
rin
ep
Pr
31
Table 1. ESG composite ratings
ed
This table presents the distribution of observations and ESG composite ratings provided by the Korea Corporate Governance
Service (KCGS) per year.
Year A+ A B+ B C D Total
2013 6 34 91 344 218 0 693
iew
2014 2 44 88 351 209 0 694
2015 1 29 82 400 308 9 829
2016 8 33 108 323 229 12 713
2017 5 39 118 291 258 22 733
v
2018 8 41 101 256 286 35 727
re
Total 30 220 588 1,965 1,508 78 4,389
er
pe
ot
tn
rin
ep
Pr
32
Table 2. Sample distribution
ed
This table reports the distribution of our sample across two dimensions. Panel A displays the distribution of ESG controversial
news articles per year and Panel B provides an industry breakdown of the sample by each category (E, S, G).
Panel A: Distribution by year

Year E S G ESG
iew
2013 15 604 2,115 2,734(11.5%)
2014 11 427 1,674 2,112(8.8%)
2015 18 1,283 1,580 2,881(12.1%)
2016 10 1,004 3,528 4,542(19.1%)
2017 13 865 3,675 4,553(19.1%)
2018 36 3,685 3,294 7,015(29.4%)
Total 103(0.4%) 7,868(33.0%) 15,866(66.6%) 23,837(100.0%)
Panel B: Distribution by industry
v
Industry E S G ESG
Industrials 5 24 584 613 (2.6%)
Consumer Discretionary 36 4,561 4,653 9,248 (38.8%)
re
Software & Service 6 565 2,459 3,030 (12.7%)
Materials & Energy 40 1,454 3,526 5,020 (21.1%)
Technology & Hardware 6 767 2,749 3,522 (14.8%)
Consumer Staples & Utilities 2 361 501 864 (3.5%)
Health Care 8 136 1,394 1,538 (6.5%)
Total 103(0.4%)
er 7,868(33.0%) 15,866(66.6%) 23,837 (100%)
pe
ot
tn
rin
ep
Pr
33
Table 3. Descriptive statistics of variables
ed
This table reports the descriptive statistics of variables for 1,976 firm-year of 353 companies. It shows the mean and median
value of each variable, its standard deviation as well as the minimum and maximum of the distribution.
Variables Mean Median SD MIN MAX

TQ .779 .779 .289 .249 2.608
iew
MB 1.512 1.043 1.627 0 16.181
CI -5.861 -5.861 2.161 -13.61 0
ESG 8.925 8.925 2.506 3 18
AD .01 .01 .024 0 .294
RD .01 .01 .023 0 .178
SG .048 .048 .253 -.886 8.184
v
DA .405 .405 .201 .001 .982
FOR .111 .111 .133 0 .897
CHAEBOL .247 .247 .432 0 1
re
IOWN .415 .415 .18 0 1
SIZE 20.241 20.241 1.571 16.64 26.112
DIV .009 .009 .015 0 .322
ANALYST .577 .577 .822 0 3.135
.778 .778 .424 -.692 2.324
BETA
er
pe
ot
tn
rin
ep
Pr
34
Table 4. ESG Controversy and Firm Value
ed
This table reports the results of a linear regression model with firm value (Tobin’s Q, and the ratio of market value to book
value) as the dependent variable. ESG controversy is a natural logarithm of the proportion of ESG-related controversial issue
articles in the total number of articles on individual companies.
TQt+1 MBt+1
iew
CIt -.018***(.005) -.028***(.007)
CIt―1 0(.005) -.004(.007)
AD -2.258*(1.174) -1.591(1.126) -2.193(1.36) -1.287(1.304)
RD 2.24(1.418) 2.313(1.597) 2.157(1.554) 1.979(1.918)
.007(.014) .001(.013) .015(.022) .008(.018)
v
SG
DA .323***(.101) .229*(.132) .823***(.14) .711***(.175)
re
FOR .586***(.146) .608***(.179) .822***(.199) .778***(.232)
CHAEBOL .008(.02) .021(.024) .008(.037) .018(.039)
IOWN -.041(.055) -.065(.064) -.036(.072) -.098(.081)
SIZE
DIV
-.124***(.043)
2.175***(.66)
er
-.15***(.054)
2.006(.692)
-.254***(.064)
2.797***(.92)
-.276***(.077)
2.613***(.937)
ANALYST .097***(.016) .081***(.017) .137***(.023) .109***(.024)

pe
BETA .039***(.015) .03*(.016) .067***(.023) .054**(.024)
Constant 2.858***(.841) 3.57***(1.065) 5.158***(1.262) 5.868***(1.525)
Observation 1,976 1,976
Adj. R2 .204 .227

ot
Industry dummy Yes Yes Yes Yes
Year dummy Yes Yes Yes Yes

*** p<0.01, ** p<0.05, * p<0.1
tn
rin
ep
Pr
35
Table 5. ESG controversies (CI), ESG performance (ESG), and firm value
ed
value) as the dependent variable. CI is a natural logarithm of the proportion of ESG-related controversial issue articles in the
total number of articles on individual companies. ESG is the firm's ESG score which is the degree of ESG performance of a
company by converting E, S, and G individual ratings into a score and summing up all these individual scores. The variable
of interest is the interaction term CI×ESG.
iew
ESG―1 ESG0 ESG+1
Dependent TQt+1 MBt+1 TQt+1 MBt+1 TQt+1 MBt+1
CIt -.048*** -.076*** -.04*** -.063*** -.042*** -.063***

ESGt .021** .029** .009 .014 .015* .018
CIt X ESGt .004*** .006*** .002** .004** .003** .004***
v
AD -1.623 -1.309 -2.292* -2.221 -1.482 -1.352
2.404 2.078 2.243 2.146 .953 .685
re
RD
SG .003 .009 .007 .014 .006 -.006
DA .246* .739*** .33*** .84*** .253** .783***
FOR .659*** .848***

er .583*** .807*** .47*** .73***
CHAEBOL .017 .011 .008 .008 .006 -.013

IOWN -.058 -.088 -.034 -.034 .005 .03
SIZE -.143*** -.267*** -.12*** -.258*** -.108** -.249***
pe
DIV 2.037*** 2.667*** 2.164*** 2.774*** 2.158*** 2.707***
ANALYST .077*** .104*** .093*** .128*** .075*** .103***
BETA .029* .051** .041*** .067*** .04** .066**
Constant 3.132*** 5.298*** 2.704*** 5.128*** 2.454** 4.944***

ot
Observation 1,620 1,620 1,974 1,974 1,621 1,621

Adj. R-square .191 .208 .207 .231 .164 .2
Industry dummy Yes Yes Yes Yes Yes Yes

tn
Year dummy Yes Yes Yes Yes Yes Yes

rin
ep
Pr
36
Table 6. ESG controversies (CI), Individual E/S/G performance (E/S/G), and firm value
ed
This table reports the results of a linear regression model with firm value (Tobin's Q,) as the dependent variable. CI is a natural
logarithm of the proportion of ESG-related controversial issue articles in the total number of articles on individual companies.
E/S/G is the firm's score which is the degree of each E/S/G performance of a company by converting E, S, and G individual
ratings into a score and summing up all these individual scores. The variable of interest is the interaction term CI×E, CI×S,
and CI×G.
iew
t―1 t0 t+1
E-1 S-1 G-1 E0 S0 G0 E+1 S+1 G+1
Dependent TQt+1 TQt+1 TQt+1
CI -.029*** -.038*** -.033*** -.03*** -.034*** -.025*** -.029*** -.029*** -.031***
v
E .019 .007 .014
S .047*** .028* .021
re
G .036* .012 .034**
CI × E .004 .004* .004*
CI × S .007*** .005** .004*
CI × G
Control
Yes Yes
.006*
Yes
er Yes Yes
.002
Yes Yes Yes

.005**
Yes
variables
Observation 1,620 1,620 1,620 1,974 1,974 1,974 1,621 1,621 1,621
pe
Adj. R2 .188 .19 .188 .208 .206 .203 .163 .162 .163
Industry
Yes Yes Yes Yes Yes Yes Yes Yes Yes
dummy
Year
Yes Yes Yes Yes Yes Yes Yes Yes Yes
dummy
ot
tn
rin
ep
Pr
37
Table 7. The impact of the interaction of ESG controversies and ESG score on firm value by firm characteristics
e d
This table reports the results of fixed-effects time series regression for firm value as measured by Tobin's Q on ESG controversies, ESG score, an interaction between ESG controversies and ESG
levels, respectively denoting the statistical significance of the difference between the coefficients of the two subsamples.
i ew
score, and other selected control variables. All regressors are one-year lagged to account for a possibly endogenous interdependence and winsorized at the 1 and 99 % level to mitigate the effect
of outliers. The sample is split concerning several measures of firm-specific investors' attention as measured by ESG score, conglomerate, foreign ownership, google trends, analyst following,
size of the firm, profitability, and industry type. Robust standard errors adjusted for clustering by the firm are in parentheses. ***, **, and * denote statistical significance at the 1, 5, and 10 %
high
ESG Score
low yes
Conglomerate
no
Foreign Ownership
high low
e
high
v
Googling
low
Analyst Coverage
high low
r
CI -.056*** -.028 -.037** -.035** -.029* -.035 -.043** -.028 -.031* -.033*
ESG .017 .001 .011 .006 .011 .007 .014 -.009 .009 -.006
CI × ESG
Control variables
Observation
.004**
Yes
978
.001
Yes
985
.003*
Yes
484
.002
Yes
1,479
er .003
Yes
982
.001
Yes
981
.003*
Yes
983
0
Yes
980
.002*
Yes
760
.001
Yes
1,203
Adj. R2
Industry Dummy
.288
Yes
.273
Yes
.339
Yes
.227
Yes
p e .375
Yes
.202
Yes
.227
Yes
.213
Yes
.266
Yes
.204
Yes
Year Dummy Yes
big
Firm Size
Yes
small
Yes
high
ROA
o t Yes
low
Yes
Mfg
Industry #1
Yes
Non-Mfg
Yes
B2C
Industry #2
Yes
B2B
Yes Yes
CI
ESG
-.019
.008
.001
-.018
.02
-0.002
t n -.039
.005
.003
-.044***
.015
.003*
-.057***
.018*
.004***
-.031*
.007
.002
-.062
.034
.005
-.033***
.003
.002
ir n
CI × ESG
Control variables Yes Yes Yes Yes Yes Yes Yes Yes
Observation 978 958 981 982 1,279 684 336 1,597
Adj. R2
p
.308 .24 .292 .191 .189 .359 .32 .198
e
Industry Dummy Yes Yes Yes Yes Yes Yes Yes Yes
P r
Year Dummy Yes Yes Yes Yes
38
Yes Yes Yes Yes
Table 8. ESG controversies (CI), CSR performance (CSR), and firm value
ed
value) as the dependent variable. CI is a natural logarithm of the proportion of ESG-related controversial issue articles in the
total number of articles on individual companies. CSR is the firm’s CSR score which is the degree of CSR performance of an
individual company. The variable of interest is the interaction term CI×CSR.
CSR―1 CSR0 CSR+1
iew
Dependent TQt+1 MBt+1 TQt+1 MBt+1 TQt+1 MBt+1
CIt -.29** -.185** -.057 .027 .115 .178

CSRt .015 .014 .008 -.002 -.007 -.015
CIt X CSRt .005** .003** .001 -.001 -.002 -.003
v
Control variables Yes Yes Yes Yes Yes Yes
Observation 230 230 528 528 461 461
re
Adj. R-square .326 .301 .207 .203 .191 .224
Industry dummy Yes Yes Yes Yes Yes Yes
Year dummy Yes Yes Yes Yes Yes Yes

er
pe
ot
tn
rin
ep
Pr
39
Table 9. Robustness checks: Random effect model
ed
RE
ESG-1 ESG0 ESG1
CI -.044***(.012) -.035***(.011) -.034***(.01)
iew
ESG .023***(.008) .01(.007) .013*(.007)
CI × ESG .004**(.001) .002**(.001) .002**(.001)
Control variables Yes Yes Yes
Observation 1,620 1,974 1,621
v
Adj. R2 .174 .190 .147
re
Industry dummy Yes Yes Yes
Year dummy Yes Yes Yes
er
pe
ot
tn
rin
ep
Pr
40
Table 10. Robustness checks: Two-stage least square model
ed
This table explores the relationship between Tobin’s Q, both ESG controversies and the ESG score and the interaction between
ESG controversies and ESG score while addressing endogeneity issues. The first model investigates the inclusion of firm-
fixed effects. The second and third models are estimated in two stages. The first-stage regression involves regressing the ESG
score on all exogenous independent variables, ESG controversies, fixed effects, and the instrument (mean industry-year ESG
score excluding the focal firm). The second-stage regression results use the predicted values of the ESG score from the first-
stage regressions. Only Model 1 and Model 3 include firm-fixed effects. All variables are defined in Appendix A. Robust
standard errors adjusted for clustering by the firm are in parentheses. ***, **, and * denote statistical significance at the 1, 5,
iew
and 10 % levels, respectively.
RE 2SLS
ESG-1 ESG-1
ESG CI × ESG TQ
v
CI -.044***(.012) CI 0.311(.190) -1.001 (1.455) -.042 (.040)
re
ESG 0230***(.008) ESG_IDAV 0.736*** (.171) 0.907 (1.214)
CI × ESG_IDAV -0.029(.021) 1.073*** (.164)
CI × ESG .004***(.001)
er
ESG_hat .075** (.038)
CI × ESG_hat .002* (.004)

pe
Control variables Yes Control Variables Yes Yes Yes
Observation 1,620 Observation 1,606 1,606 1,606
Adj. R2 .174 Adj. R2 .101 .704 .193
Industry dummy Yes Industry dummy Yes Yes Yes

ot
Year dummy Yes Year dummy Yes Yes Yes

tn
rin
ep
Pr
41
Table 11. Robustness Checks: Three-year average ESG score
ed
TQ
FE RE 2SLS
CI -.053**(.02) -.044**(.017) -.088***(.033)
.018(.015) .027**(.012) .099***(.029)
iew
ESG
CI × ESG .003*(.002) .004**(.002) .008**(.004)
Control variables Yes Yes Yes
Observation 1,974 1,974 1,964
Adj. R2 .207 .307 .212

Industry dummy Yes Yes Yes
v
Year dummy Yes Yes Yes
re
er
pe
ot
tn
rin
ep
Pr
42

SSRN Id4500037

Uploaded by

Copyright:

Available Formats

SSRN Id4500037

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SSRN Id4500037

Uploaded by

Copyright:

Available Formats

The Effect of ESG Controversies on Firm Value: Evidence from Korea

applying a method of discounting existing ESG scores amid controversy.

6 Bird et al(2007), Flammer(2013), Klassen and McLaughlin(1996), Krüger(2015), Reyes(2019), Sabbaghi(2020)

2. Theoretical Background and Hypotheses

3. Data and Research Model

[Insert Table 1 here]

We use DataGuide (https://dataguide.fnguide.com) to obtain the firm-level financial

total asset ratio, and cash dividend ratio.

controlling shareholder(s). We also find large conglomerates-related information from Fair

series period, and association.

more than 330 ESG controversy news items per month.

selection process is explained in Appendix B.

[Insert Appendix B here]

16 For a more detailed description of the variables, see Appendix A.

analyst coverage and firm value are positively related as well.

4.2 The moderating effect of ESG performance

We prove that the moderating effect of ESG performance of companies with

[Insert Table 5 here]

4.3 The moderating effect of each E/S/G performance

[Insert Table 6 here]

performance is linked to firm visibility.

[Insert Table 7 here]

In this section, we investigate whether the moderating effect of ESG performance

[Insert Table 8 here]

[Insert Table 10 here]

5.3 Robustness checks: 3-year average ESG performance

evidence, and implications. Academy of Management Review, 20(1), 65–91.

responsibility. Journal of Banking & Finance, 59, 182–192.

Jones, T. M. (1995). Instrumental stakeholder theory: A synthesis of ethics and economics.

value: The role of customer awareness. Management Science, 59(5), 1045–1061.

Semantic Evaluation (SemEval 2015), 464–469.

twitter sentiment classification. Proceedings of the 8th International Workshop on

2013 6 34 91 344 218 0 693

2015 1 29 82 400 308 9 829

2016 8 33 108 323 229 12 713

2017 5 39 118 291 258 22 733

Panel A: Distribution by year

Variables Mean Median SD MIN MAX

CIt―1 0(.005) -.004(.007)

AD -2.258*(1.174) -1.591(1.126) -2.193(1.36) -1.287(1.304)

RD 2.24(1.418) 2.313(1.597) 2.157(1.554) 1.979(1.918)

.007(.014) .001(.013) .015(.022) .008(.018)

DA .323***(.101) .229*(.132) .823***(.14) .711***(.175)

CHAEBOL .008(.02) .021(.024) .008(.037) .018(.039)

IOWN -.041(.055) -.065(.064) -.036(.072) -.098(.081)

ANALYST .097***(.016) .081***(.017) .137***(.023) .109***(.024)

Constant 2.858***(.841) 3.57***(1.065) 5.158***(1.262) 5.868***(1.525)

Observation 1,976 1,976

Adj. R2 .204 .227

Industry dummy Yes Yes Yes Yes

Year dummy Yes Yes Yes Yes

Dependent TQt+1 MBt+1 TQt+1 MBt+1 TQt+1 MBt+1

CIt -.048*** -.076*** -.04*** -.063*** -.042*** -.063***

SG .003 .009 .007 .014 .006 -.006

DA .246* .739*** .33*** .84*** .253** .783***

FOR .659*** .848***

CHAEBOL .017 .011 .008 .008 .006 -.013

ANALYST .077*** .104*** .093*** .128*** .075*** .103***

BETA .029* .051** .041*** .067*** .04** .066**

DA .323***(.101) .229(.132) .823(.14) .711*(.175)

ANALYST .097(.016) .081(.017) .137(.023) .109(.024)

Constant 2.858(.841) 3.57(1.065) 5.158(1.262) 5.868(1.525)

CIt -.048* -.076* -.04* -.063* -.042* -.063*

DA .246* .739* .33* .84* .253 .783***

FOR .659* .848*

ANALYST .077* .104* .093* .128* .075* .103*

BETA .029* .051 .041* .067* .04 .066**

Constant 3.132* 5.298* 2.704* 5.128* 2.454 4.944*

CI -.029* -.038* -.033* -.03* -.034* -.025* -.029* -.029* -.031***

CI × S .007* .005 .004*

CIt -.29 -.185 -.057 .027 .115 .178

CI -.044(.012) -.035(.011) -.034***(.01)

CI × ESG .004(.001) .002(.001) .002**(.001)