SSRN Id4578803

ESG Rating Divergence, Investor Expectations, and Stock
ed
Returns
iew
Hongtao Chena,,1, Renxiang Hea, Yimei Huangb
a School of Economics and Management, Southeast University, 2 Southeast

University Road, Jiangning District, 211189 Nanjing, China
bAzman Hashim International Business School, Universiti Teknologi, Jalan Sultan
v
Yahya Petra, 54100 Kuala Lumpur, Malaysia
re
er
pe
ot
tn
rin
 Corresponding author at: School of Economics and Management, Southeast

University, 2 Southeast University Road, Jiangning District, Nanjing 211189, China.
ep
E-mail addresses: [email protected] (Hongtao Chen), [email protected]

(Renxiang He), [email protected] (Yimei Huang).
1 Please address correspondence to this author.
Acknowledgments
This work is supported by Social Science Foundation of Jiangsu Province (No.
Pr
22GLB001) and the Major Project of Philosophy and Social Science Research in
Colleges and Universities of Jiangsu Province (Education Department of Jiangsu
China, No.2020SJZDA059).
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4578803
ESG Rating Divergence, Investor Expectations, and Stock
ed
Returns
Abstract
iew
We investigate the relationship between ESG rating divergence and stock returns from
an investor’s perspective, to explore the impact of inconsistency among ESG rating
agencies on the capital market. We construct ESG rating divergence data using ratings
v
from three prominent ESG rating agencies in China. Our study is based on 54,679
re
company-quarter observations from 2018 to 2022, which covers 4,377 Chinese listed
companies. Our findings demonstrate a significant negative impact of ESG rating
divergence on stock returns, which we validate through a series of robustness tests and
er
endogenous analyses. Notably, we find that investors’ expectations mediate the
relationship between ESG rating divergence and stock returns. Further analyses show
pe
that only the divergence in social ratings have a significant inhibitory effect on stock
returns. In addition, ESG rating divergence significantly impedes subsequent average
ESG ratings. The adverse relationship between ESG rating divergence and stock returns
ot
is particularly pronounced in non-heavy pollution companies, non-state-owned

tn
companies, and companies with lower external attention.
Keywords: ESG rating divergence; Stock returns; Investor expectations; Information
asymmetry
rin
JEL classification: C12; C33; D81; D82; G14

ep
Pr
1 Introduction
ed
The concept of environmental, social, and governance (ESG) development has
gained recognition across many countries, and adhering to the principles of sustainable
iew
development has become a global trend in capital markets. However, while the ESG
concept is flourishing, there are also endless doubts about the rationality, authenticity,
and accuracy of ESG indicators. On June 26, 2023, the International Sustainability
v
Standards Board officially released the final drafts of the first two sets of international
re
financial reporting sustainability disclosure standards, which is a highly authoritative
standard for information disclosure and interpretation related to sustainable

er
development1. The ESG performance of companies has become a critical consideration
for investors in their decision-making process. Recent research indicates that investing
pe
in companies with strong ESG performance can yield additional returns while avoiding
companies with weak ESG performance can help mitigate risks (see, e.g., Bax et al.,
2023; Cornell, 2021; Gomez-Valencia et al., 2021; Shanaev and Ghimire, 2022; Yu et
ot
al., 2023). Therefore, investors have set high expectations for accurately assessing
companies’ ESG performance to ensure that their stock selections align with sustainable
tn
development principles, possess long-term value, and exhibit stable return growth.
However, scattered information, vague standards, and resource constraints hinder the
rin
accurate assessment of corporate ESG performance by investors. Consequently,
professional rating agencies, which collect and generate ESG rating data, have emerged
ep
as the primary reference for investors. Due to the diversity of ESG information and the
lack of a unified evaluation system, discrepancies in the selection criteria, metrics, and
Pr
1. See more details at

https://www.ifrs.org/news-and-events/news/2023/06/issb-issues-ifrs-s1-ifrs-s2/
weights of ESG information exist among these agencies (see, e.g., Brandon et al., 2021;
ed
Christensen et al., 2022; Kotsantonis and Serafeim, 2019; Liu, 2022; Serafeim and
Yoon, 2022). As a result, divergences have emerged in the ESG ratings of these
agencies, thereby highlighting uncertainties in their accurate measurement of
iew
companies’ ESG performance and confusing ESG investors.
With the market’s increasing emphasis on companies’ ESG performance,
controversies start to arise from ESG ratings. For instance, Tesla CEO Elon Musk has
v
publicly criticized the irrationality of ESG ratings on multiple occasions through
re
Twitter. In May 2022, Tesla, the electric vehicle company under Musk’s leadership,
was excluded from the S&P 500 ESG Index. When Tesla reentered the S&P 500 ESG
er
Index in April 2023, the company only received an ESG score of 37, which was far
lower than the score of 84 received by Philip Morris International, a cigarette

pe
manufacturer. Additionally, Tesla’s ESG ratings from MSCI and Sustainalytics showed
significant discrepancies, which fueled Musk’s skepticism and criticism. ESG
divergences have also garnered attention in emerging markets. For example, ESG
ot
ratings for Guizhou Moutai, a well-known Chinese alcohol company, exhibit
significant inconsistencies across different ESG rating agencies. Domestic rating

tn
agencies tend to assign high ratings, while foreign rating agencies, such as MSCI and
Bloomberg, generally provide lower ratings to the company due to so-called “sin”
rin
stocks (stocks associated with alcohol, tobacco, or gaming). These examples illustrate
the widespread existence of ESG rating divergences, which can lead to significant
ep
cognitive contradictions and profoundly impact a company’s reputation and market
expectations.
ESG rating divergences can produce a series of significant consequences. First,

Pr
different ratings imply difficulties for the market to accurately and fairly assess a
company’s ESG performance, which presents a considerable obstacle for investors and
ed
companies. Second, divergences among rating agencies may overshadow companies’
efforts in ESG and reduce their motivation for ESG improvement. Lastly, investors
relying on rating agency results may be misled, thereby impacting their investment
iew
behaviors and returns. Although some studies have explored the topic of ESG rating
divergences (see, e.g., Avramov et al., 2022; Berg et al., 2022; Brandon et al., 2021;
Christensen et al., 2022; Kotsantonis and Serafeim, 2019; Liu, 2022), they have mainly
v
concentrated on the causes of these divergences and ignored how they affect stock
re
returns. Further research is needed to explore how ESG divergences impact investor
returns.
er
For the empirical analysis, we use a sample of 4,377 Chinese listed companies
amounting to 54,679 quarterly observations from 2018 and 2022 to examine the impact
pe
of divergences among mainstream ESG rating agencies on stock returns. We provide
empirical evidence and supplement the research on the economic consequences of ESG
rating divergences.
ot
First, for our research sample, we select China’s financial market, which has
started exploring ESG aspects relatively late but has witnessed rapid advancements in
tn
this area (Wang, 2023). China’s financial market is very suitable for conducting our
research because of the significant increase in the number of its rating agencies, its need
rin
for unified rating systems and rules, and its relatively late entry into the ESG rating
market (Li et al., 2022). Second, we use ratings from three Chinese rating agencies to
ep
construct our divergence data. These three agencies are included in the largest stock
analysis software and financial databases used in China (WIND and iFinD) and update
their ratings every quarter. Compared with other rating agencies, these agencies offer
Pr
guaranteed visibility and professionalism with more frequent updates. Finally, we use
quarterly data to observe the immediate impact of ESG rating divergences. We consider
ed
the release of ESG ratings and the resulting divergences each quarter as short-term
events to analyze their effect on the market. Compared with annual data, quarterly data
better reflect the differences and trends among rating agencies; thus, the variables
iew
constructed from quarterly data can assess the degree of divergence more accurately.
Results show that stocks with higher levels of ESG rating divergences tend to have
lower stock returns, thereby suggesting a negative relation of ESG rating divergences
v
on investment returns. This conclusion remains significant after conducting robustness
re
checks and endogenous tests. Further analysis reveals that this relationship is only
significant in the aspects of social pillar, thereby suggesting that social issues are more
er
closely related to stock returns than the divergences in environmental and governance
matters. We also observe a significant negative relationship between ESG rating

pe
divergences and subsequent average ESG ratings, indicating that companies with higher
levels of ESG rating divergences tend to receive lower ratings in the future. Our
heterogeneity analyses highlight that this negative relationship is particularly

ot
pronounced among non-polluting sector companies, non-state-owned companies, and
companies with greater investor attention.

tn
This research offers contributions in three aspects. First, our study shifts the focus
from corporate ESG practices to ESG rating divergences and discusses the impact of
rin
external rating agencies. We find that the divergences in rating agencies’ interpretation
of ESG information can decrease stock returns through affecting investor expectations.
ep
Previous studies have focused on the market reactions triggered by corporate ESG
performance or ESG controversies, with only a few studies delving into the reasons
behind ESG rating discrepancies (see, e.g., Aouadi and Marsat, 2018; Broadstock et al.,
Pr
2021; Luo and Wu, 2022; Luo et al., 2015; Shanaev and Ghimire, 2022). Similarly, the
corresponding economic consequences and market reactions have attracted limited
ed
research attention. To fill these gaps and enrich the literature on ESG controversies, we
explore the market reactions resulting from rating agency divergences in an emerging
market context.
iew
Second, our study contributes to the literature on the relationship between
corporate ESG and information asymmetry. Unlike previous research that mostly
focuses on the impact of ESG-related information during its release and dissemination
v
(see, e.g., Cui et al., 2018; Meng and Zhang, 2022; Schiemann and Tietmeyer, 2022;
re
Zhang et al., 2022), we explore the market effects caused by differences in rating
agencies’ interpretation of ESG information. Despite the interpretation provided by

er
these agencies, the ESG information released by companies still creates information
asymmetry due to the differences in agencies’ ratings. This finding provides novel
pe
insights into the sources of information asymmetry, which may exist in various stages,
including “companies releasing ESG information → rating agencies interpreting such
information → the market receiving the rating results,” ultimately decreasing

ot
investment returns.
Lastly, our findings offer implications for various stakeholders, particularly

tn
investors. Specifically, we provide insights into how investors should perceive the ESG
ratings of target companies in major investment platforms or stock selection software.

rin
When constructing an investment portfolio, investors should observe the divergences
in ESG performance among various rating agencies and select those stocks with similar
ep
ratings. For companies, while focusing on ESG practices is their primary way to
improve their ESG performance, they also need to engage in dialogue with rating
agencies and reduce the divergences among them. For regulatory authorities, we
Pr
provide evidence highlighting the detrimental impact of divergences among ESG rating
agencies on the interests of investors and companies, thus underscoring the need for the
ed
further regulation and development of ESG rating rules.
The rest of the paper is organized as follows. Section 2 reviews the literature and
proposes the hypotheses. Section 3 presents the data and research design. Section 4
iew
discusses the results of the hypotheses tests and further analyses. Section 5 conducts
endogenous tests and robustness checks. Section 6 concludes the paper.
v
2 Literature review and hypotheses development
re
2.1 Literature review
er
Previous studies have analyzed ESG malfunction about greenwashing and
hypocrisy (see, e.g., Hafenbradl and Waeger, 2021; Lyon and Maxwell, 2011; She and
pe
Michelon, 2019; Wagner et al., 2009; Yu et al., 2020) but paid little attention to the role
of rating agencies. These researchers focus on the causes on the corporate side yet
ignore the impact of external agencies. Divergence in ESG ratings is a common

ot
phenomenon in mature (e.g., the US) and emerging financial markets (e.g., China), thus
tn
prompting investors to raise concerns regarding the reliability of these ratings. Investors
with a sense of social responsibility need to allocate their investment portfolios based
on the ESG performance of companies. At the same time, the ESG data provided by
rin
rating agencies significantly impact their investment behavior (Diaz et al., 2021).
Different impressions and profit expectations are formed for the same company when
ep
investors take various rating agencies’ information as reference. As the number of
rating agencies increases, the divergence in their evaluations of the ESG performance
Pr
of the same company becomes increasingly significant. Such divergence is driven by
the differences in these agencies’ rating criteria, underlying data, and valuation methods.
Berg et al. (2022) calculate the correlations among six mainstream ESG rating providers
ed
overseas and find that the correlation coefficients fluctuate between 0.38 and 0.71,
averaging only 0.54. In our correlation test of three mainstream ESG rating agencies in
China, we find that the highest correlation coefficient does not exceed 0.35, averaging
iew
only 0.31, thereby highlighting significant differences in the ratings provided by these
agencies.
v
2.1.1 ESG rating
re
Over the past decade, sustainable finance has become an emerging topic for
investors, thus leading to the entry of many ESG data providers that offer
er
comprehensive ratings of companies’ ESG performance. These ESG rating agencies
have become essential information intermediaries in the capital market that provide
pe
investors with ESG ratings as a reference for corporate ESG performance. Such
information helps these investors screen risks and opportunities in their investment
portfolios and adjust their investment strategies (Erhart, 2022). Rating agencies employ
ot
a team of professional ESG analysts to meticulously analyze fundamental information
and underlying data of companies, culminating in the formation of the ultimate rating.
tn
These teams use their own assessment methodologies to identify events related to ESG
issues and convert them into corresponding quantitative indicators. These hundreds or
rin
thousands of indicators are then transformed into E, S, and G pillars, and an overall
score or rating is assigned based on their respective weights (Christensen et al., 2022).
ep
Different types of investors can use ESG ratings. Avramov et al. (2022) explore three
types of investment institutions, namely, normally constrained institutions, hedge funds,
and other institutions, with the normally constrained institutions being more likely to
Pr
refer to a company’s ESG rating and complete socially responsible investments
accordingly. In addition, previous studies show that ESG ratings can affect capital
ed
allocation and a company’s cost of capital by changing investors’ return expectations.
Given that most users cannot independently evaluate a company’s ESG performance,
they rely heavily on the ratings assigned by ESG rating agencies. However, each rating
iew
agency provides varying references. According to KPMG, mainstream ESG data
suppliers worldwide typically use other ESG rating metrics when constructing their
ESG ratings. Thus, assigning different ESG ratings to the same company is expected.
v
For example, in MSCI’s ESG rating, Tesla is rated as medium, but for Sustainalytics,
re
Tesla is classified as high risk. Billio et al. (2021) attribute the divergences in ESG
ratings to the differences in underlying data, the selection of critical issues, and the
er
allocation of weighting. Dorfleitner et al. (2015) note that mainstream ESG rating
agencies show many differences in their measurement of the ESG performance of

pe
companies, and the results of different agencies have significant inconsistencies in
distribution and risk. Chatterji et al. (2016) observe a lack of consistency in the social
ratings given by six well-known rating agencies.

ot
2.1.2 ESG divergence

tn
Researchers have mostly focused on the causes of rating divergence, while
relatively few studies have explored its consequences, especially the response of the
rin
capital market. To the best of our knowledge, only Brandon et al. (2021) have explored
a topic similar to ours. The difference is that Brandon et al. (2021) focus on S&P 500
ep
constituent stocks and international ESG rating agencies, while we focus on Chinese
local ESG rating agencies. Early studies on ESG ratings reveal a significant lack of
convergence in ESG measurement standards (see, e.g., Chatterji et al., 2016; Dorfleitner
Pr
et al., 2015). These discrepancies can be categorized into measurement, scope, and
weight discrepancies based on their underlying causes. Among them, measurement
ed
discrepancy originates from the “rater effect,” which means that the evaluator’s overall
perception of the company influences the measurement of specific categories (Berg et
al., 2022). Yu and Luu (2021) use the Bloomberg ESG disclosure score as a measure of
iew
transparency and find that most of the changes in ESG disclosure are mainly determined
by company characteristics rather than country factors, such as corruption and political
rights. Previous studies generally converge on the increased uncertainty of corporate
v
future sustainability performance caused by ESG controversies, but divergent views
re
regarding the role of ESG disclosure are also present. Specifically, some researchers
believe that ESG disclosure can alleviate the uncertainty caused by divergence, and
er
companies with higher ESG rating can better withstand adverse market reactions after
experiencing ESG controversies, which help them accumulate “ethical” capital (see,
pe
e.g., Hummel et al., 2019; Schiemann and Tietmeyer, 2022). Meanwhile, other
researchers believe that disclosing additional information can lead to more cognitive
disagreements due to the lack of consensus in the rating system (Christensen et al.,
ot
2022). Companies with significant differences in their ESG ratings will experience a
substantial decline in their future ESG ratings, thereby highlighting a correlation

tn
between risk premium and ESG uncertainty (Liu, 2022). Empirical studies reveal that
the potential for reducing measurement diversity has not yet fully materialized due to
rin
the choice of ESG rating provider (see, e.g., Dumrose et al., 2022; Erhart, 2022), but a
specific taxonomy or best-in-class portfolio can be recommended.

ep
2.2 Hypotheses development

Pr
The validity and convergence of ESG rating data are highly controversial topics in
management studies. Brandon et al. (2021) suggest that the choice of ESG rating
10
providers may influence companies’ and investors’ sustainable financing and
ed
investment decisions. From the risk management perspective, high-quality ESG
practices can reduce internal risks related to operations and governance. For example,
releasing an ESG rating can lower the information and operational risks for a listed
iew
company, thereby establishing a foundation for sustainable development (Schiemann
and Tietmeyer, 2022). Good ESG performance enhances the ability of companies to
resist external risks, especially those related to environmental issues. The proactive
v
activism of ESG strategies can function as pre-reputation insurance that helps
re
companies avoid potential litigation costs and reputational damage, thus reducing the
downside risks (Gillan et al., 2021). Environmental information disclosure can reduce
er
the risk of future stock price crashes by alleviating the principal–agent problem,
signaling altruistic intentions to shareholders, and enhancing the authenticity of

pe
financial information through mechanisms (Zhang et al., 2022). Under this premise,
companies with strong ESG characteristics show greater resilience amid volatile market
conditions (Diaz et al., 2021). Meanwhile, ESG confuses the judgment of those
ot
investors who rely on a single rating result. Investors generally make ESG investments
through negative and positive screening. However, rating discrepancies may lead to the
tn
exclusion of certain low- and high-rated companies. These divergences also underscore
the need to reevaluate and reassess a company’s ESG performance, which signals a
rin
risk. For companies, significant differences in ESG ratings will lead to a substantial
decrease in their future ESG ratings, thereby highlighting the relationship between risk
ep
premium and ESG uncertainty (Liu, 2022). For investors, the perceived equity risk
increases along with ESG uncertainty, thus reducing the demand for equity. Rating
divergences also challenge investors to incorporate the ESG dimension into their
Pr
investment strategies, thereby resulting in poor capital market efficiency (Avramov et
11
al., 2022). This uncertainty also leads to an uncertainty premium to compensate for
ed
additional exposure, thus affecting the financial behavior of ESG-sensitive investors
(Liu, 2022). The reduced market liquidity further increases bid–ask spreads and
transaction costs, thereby increasing equity capital costs (Brandon et al., 2021).
iew
Companies with significant differences in their ESG ratings may also engage in
hypocritical or deceptive behavior. ESG controversies may place a company under
scrutiny of stakeholder sanctions, and the adverse reactions from stakeholders are not
v
always predictable (Schiemann and Tietmeyer, 2022). Therefore, due to higher risk
re
levels and weaker risk-bearing capacity, companies may be unable to access third-party
financial resources under favorable conditions. As the cost of debt financing increases,
er
investors will also reassess the possibility of investment, thus weakening their
confidence in the company’s growth prospects and leading to significant declines in

pe
stock returns (Erhart, 2022). On the basis of these arguments, we propose
H1: ESG rating divergence reduces stock returns.
From the information asymmetry perspective, a lower level of information

ot
asymmetry can reduce agency costs between principals and agents, regulatory costs
between investors and managers, and transaction costs among investors, thereby
tn
affecting corporate financial performance (Boone and White, 2015). Stakeholders may
be pleased to see companies disclosing more economic data on their own given that the
rin
widespread recognition of the meaning of financial information often reduces the
difference between stock and credit rating analysts, thereby leading to highly consistent
ep
conclusions (Cui et al., 2018). Previous studies have explored how corporate ESG
behaviors mitigate information asymmetry under different contexts. For instance,
Huang et al. (2022) find that companies reduce information asymmetry through various
Pr
means, such as management forecasts, press releases, analysts’ presentations,
12
standardized financial reporting, and voluntary information disclosure. However, the
ed
value and role of ESG in addressing ESG divergence resulting from the lack of
government regulations, market norms, and quantitative standards have not been
thoroughly studied. According to Avramov et al. (2022), if ESG scores are determinate
iew
and highly consistent, then these scores may lead to investors’ subjective perceptions
of ESG, thus exacerbating cognitive biases, further increasing information asymmetry,
and impacting market feedback. Christensen et al. (2022) focus on how the level of ESG
v
disclosure in companies drives consistency or inconsistency in ESG ratings and find
re
that a higher level of ESG disclosure corresponds to a more significant rating
discrepancy. This relationship holds for standardized and comparable ESG disclosure,
er
which show a negative correlation. Similarly, Cui et al. (2018) suggest that as the level
of corporate disclosure increases, providing more non-financial information and

pe
promoting greater transparency would alleviate the information asymmetry between a
company and its shareholders. Therefore, rating discrepancies represent an increase in
the amount of information available in the market and a reduction in information

ot
asymmetry.
Due to market information asymmetry, different investors hold varying

tn
perceptions of investment risks and expected returns. Generally, a more influential
market information corresponds to a more accurate and reliable decision, which would
rin
yield more benefits (Cui et al., 2018). According to Zhong et al. (2020), long-term
information asymmetry can lead to investors losing confidence in their investments due
ep
to economic losses, thus impeding public companies’ financing plans and causing
market stagnation. Therefore, investors’ expectations for the future are primarily based
on their judgment of the information they possess. Zhu and Niu (2016) document that
Pr
investor sentiment can change both expected earnings growth and returns and
13
consequently affect stock prices. Having enough information is highly conducive for
ed
investors to formulate correct market expectations and make rational investment
decisions. Therefore, we introduce trading volume into the ESG literature to measure
investor expectations. As an essential market characteristic, trading volume is
iew
significantly correlated with the predictability of returns (Phan et al., 2015). Peng et al.
(2020) emphasize that changes in stock price depend not only on the intrinsic value
represented by accounting information but also on investor behavior as measured by
v
stock trading volume. Groening and Kanuri (2018) argue that stock trading volume
re
conveys investor sentiment surrounding company activities and can serve as a proxy
for investor reactions to negative news. In sum, we derive the level of information
er
asymmetry in the market based on the changes in ESG rating divergences, which in
turn affects investors’ expectations and ultimately leads to fluctuations in stock returns
pe
for companies. On the basis of these arguments, we propose
H2: ESG rating divergence reduces stock returns by influencing investor
expectations.
ot
Insert Figure 1 about here

tn
3 Data and research design

rin
3.1 Sample selection and data sources
We utilize data from 2018 to 2022 and include all listed companies on the
ep
Shanghai and Shenzhen exchanges in China. We obtain our ESG rating data from the
Wind and iFinD databases, which are renowned financial data providers in China. Wind
Pr
takes a leading position in the mainland Chinese financial information services industry
and was named the Best ESG Solution in the China region by SRP
14
(StructuredRetailProducts.com) under Euromoney Institutional Investor PLC in 2023.
ed
As the industry’s first online financial information service provider listed on the
ChiNext board, iFinD serves institutional clients, including listed companies, fund
corporations, banks, governments, and universities. These two databases are widely
iew
recognized for their role in incorporating ESG ratings into company evaluations and for
serving as primary sources of ESG ratings for investors. We collect information on
stock returns, financial data, and stock market indicators from the CSMAR database,
v
which is one of earliest financial data providers in China. We carefully process all our
re
data to ensure sample representativeness. For instance, we exclude those samples with
zero values as the primary variable, samples from the financial industry, and samples
er
with abnormal operating conditions. We round all consecutive variables to the nearest
1% and 99% levels. Our final sample includes 4,377 companies with 54,679 quarterly
pe
observations.
3.2 Variables definition

ot
3.2.1 Dependent variable

tn
Our dependent variable is stock returns. Previous studies have discussed the
relationship between cash dividends and stock returns (see, e.g., DeAngelo et al., 1996;
rin
Jiang and Lee, 2007; Kane et al., 1984; Lettau and Ludvigson, 2005). Therefore, we
measure stock returns by subtracting the weighted value of the average market return
from the individual stock return while considering cash dividends reinvestment.
ep
Following Cheng et al. (2019), we download our stock return data from the CSMAR
database and construct our daily stock returns while assuming the reinvestment of cash
Pr
dividends.
15
ri,t  Pi,t (1 Fi,t  Si,t )Ci,t  Di,t , (1)
Pi,t-1Ci,tSi,tKi,t
ed
where i represents the individual stock, t represents the time, ri,t is the yield of
stock i on day t , Pi,t is the closing price of stock i on day t , Pi,t -1 is the closing
iew
price of stock i on day t - 1 , Di,t is the cash dividend per share of stock i on day t
as the ex-dividend date, Fi,t is the number of bonus shares per share of stock i on
v
day t as the ex-dividend date, Si,t is the number of allotments per share of stock i
re
on day t as the ex-dividend date, K i,t is the allotment price per share of stock i on
day t as the ex-dividend date, and Ci,t is the split number per share of stock i
when t is the ex-dividend date.

er
We then calculate the market average return rate as follows using the weighted
pe
average of the circulating market value:
 Vi, t - 1  Pi, t - 1  ri, t

Ri, t  i
, (2)
 Vi, t - 1  Pi, t - 1
i
ot
where Ri,t is the average market rate of return, Vi,t-1 is the number of outstanding
shares of stock i on day t - 1 , Pi,t-1 is the stock closing price on day t - 1 , and ri,t
tn
represents the return of stock i on day t while considering cash dividends
reinvestment.
rin
We subtract the market average return from the daily return of individual stocks
and then perform quarterly weighting to obtain the adjusted quarterly return of
ep
individual stocks as follows:

n
ri, t  Ri, t
StockReturn_ADJi, q  i 1
, (3)
Pr
n
where StockReturn_ADJi,q represents the adjusted individual return of stock i in
16
quarter q , ri,t represents the return of stock i on day t , Ri,t represents the
ed
average market return, and n is the number of trading days in the quarter.
3.2.2 Independent variable
iew
Our independent variable is ESG rating divergence. ESG ratings developed
relatively late in China, and significant differences can be observed among Chinese
rating agencies in terms of their establishment time, rating frequency, rating methods,
v
and influence. Therefore, how to choose representative rating agencies poses a
re
challenge for this study. We download our rating data from three major ESG rating
agencies, namely, Wind, Sino, and Quant, from the largest Chinese financial database
er
Wind and the mainstream stock analysis software iFinD. We choose these rating
agencies because they are included in the Wind and iFinD databases and have great
pe
influence and high visibility. In addition, these agencies have a quarterly rating
frequency and publish ratings on E, S, and G sub-items, thereby increasing the
transparency of rating process and the credibility of rating structure. Following

ot
Avramov et al. (2022), we uniformly assign the rating results of these three agencies to
tn
increase their comparability. Afterward, we compare the set values in pairs and use the
average of the comparison results to measure ESG rating divergence.
The pairwise comparison among different rating agencies is formulated as follows.

rin
For the rating results g1 and g2 , the rating divergences are
g1  g 2 2 g1  g 2 2
( g1  )  (g 2  )
ep
2 2 g1  g 2
Divergencei , q   . (4)
2 1 2
3.2.3 Mediating variables

Pr
We believe that investor expectations serve as the channel through which ESG
17
rating divergences affect individual stock returns. Changes in stock trading volume and
ed
amount are indicative of investor behavior. A decrease in either trading volume or
amount suggests that investors’ expectations are not being met (see, e.g., Durand et al.,
2019; Groening and Kanuri, 2018; Park and Lee, 2018). We measure investor
iew
expectations using the quarterly average number of stocks traded and the average
amount sold. We download our data on the daily trading volume and trading amount of
individual stocks from the CSMAR database and take the natural logarithm of the
v
quarterly average of these data to obtain the quarterly average trading volume
re
(Volumei,q) and quarterly average trading amount (Amounti,q).
3.2.4 Control variables er

We control for a series of variables related to stock returns. First, we control for
pe
variables related to the characteristics of a company, such as its listing age (Age), size
(Size), proprietary costs (Propcosts), and institutional investor shareholding ratio
(INST). Second, we control for a series of financial indicators, such as asset–liability

ot
ratio (Lev), return on assets (ROA), and cash flow ratio (Cashflow). We also control
some stock market indicators, including book-to-market ratio (BM) and price-to-
tn
earnings ratio (PE). Finally, we fix individual and quarterly variables and conduct
standard error clustering at the company level. We summarize our measurement

rin
methods and data sources for all our variables in Table 1. We winsorize all continuous
variables at the top and bottom 1% levels.

ep
Insert Table 1 about here
3.3 Descriptive statistics and correlations

Pr
Table 2 reports the descriptive statistics for all variables, including the dependent,
18
independent, mediator, and control variables, and the rating results of the three selected
ed
rating agencies. The mean of StockReturn_ ADJ is 0.094, which is much smaller than
its standard deviation of 2.887, thereby indicating a significant difference in investment
returns in China’s financial market. StockReturn_ RAW is 0.313, which is much higher
iew
than StockReturn_ ADJ (0.094), thereby indicating that investment returns are greatly
affected by the overall market. The mean of Divergence is 0.957, indicating a high
degree of divergence among rating agencies. A minimum value of 0 indicates that some
v
samples have obtained consistent results among the three rating agencies. By contrast,
re
a maximum weight of 2.357 suggests that some samples have serious divergence among
the three rating agencies, even spanning multiple levels. We also report the rating
er
results of the three rating agencies (ESG_Wind, ESG_Sino, and ESG_Quant) and the
average rating (ESG_AVG) and find that the mean of ESG_AVG is around 5, thereby
pe
indicating that Chinese listed companies have a moderate ESG level and warrant
improvement in this area.

ot
Table 3 reports the correlation among the variables. The correlation coefficients
tn
among the rating results of the three agencies are 0.317, 0.350, and 0.265, respectively,
with an average correlation coefficient of 0.311, which is lower than the average
rin
correlation coefficient of 0.54 obtained by Berg et al. (2022) from six mainstream
international ESG rating providers. Therefore, the low correlation among Chinese ESG
ep
rating agencies means a remarkable difference in their rating results, which leads to
serious rating disagreements. We also observe an unexpected result from our correlation
analysis. Specifically, 2 out of 3 rating agencies do not show a significant correlation

Pr
with the dependent variable (StockReturn_ADJ). The correlation coefficients among the
19
other variables are all less than 0.7, thereby indicating the absence of a multicollinearity
ed
problem. We further conduct variance inflation factor (VIF) tests on the variables in the
regression model. The final average VIF value is 1.45, which is much less than 10,
thereby further verifying the absence of multicollinearity problems.
iew
3.4 Model description
v
re
To test our hypotheses, we use a fixed effects model and aggregate the standard
error at the company level for the multiple regression. First, we explore the relationship
between ESG rating divergence and stock returns to test H1. We establish the following
model:
er
StockReturn_ADJi, q   0   1 Divergencei, q    kControlsi, q  FirmFE  QuarterFE   i, q. (5)
pe
If 1 is significantly negative, then H1 is supported, that is, a higher ESG rating
divergence is related to lower individual stock returns.

ot
We establish two equations to test the impact mechanism of ESG rating
divergences on stock returns. First, we construct the following model to regress

tn
companies’ quarterly trading volume and amount using ESG rating divergences;
afterward, we add mediating variables to the primary regression model for the
rin
regression:
Mediatori, q= 0  1Divergencei, q   kControlsi, q  FirmFE  QuaterFE   i, q, (6)
StockReturn_ADJi, q   0   1 Divergencei, q   2 Mediatori, q    kControlsi, q

ep
(7)
 FirmFE  QuarterFE   i, q,
where 1 represents the impact of ESG rating divergences on the stock trading
Pr
volume and amount. If 1 is significantly negative, then a higher ESG rating
divergence corresponds to a lower trading volume of the company, thereby indicating

20
that investors’ expectations for this company are highly pessimistic. 1 represents the
ed
impact of ESG rating divergence on stock returns while considering the mediating
effect, and  2 represents the direct impact of mediating variables on stock returns.
iew
4 Results and discussion
4.1 Hypotheses tests
v
re
The test results for H1 and H2 are shown in Table 4, with column (1) showing the
multiple regression results of the independent variable (Divergence) and the dependent
variable (StockReturn_ADJ) to test H1. We then explore the mediating effect of

er
individual stock quarterly trading volume (Volume) and individual stock quarterly
pe
trading amount (Amount) through columns (2)–(3) and (4)–(5), respectively, to test H2.
Column (2) presents the regression of the independent variable ESG divergence to the
intermediate variable quarterly trading volume of individual stocks, column (3) presents
ot
the regression of the independent variable Divergence to the dependent variable
StockReturn_ADJ after adding the mediating variable Volume, column (4) presents the
tn
regression of the independent variable Divergence to the mediating variable Amount,
and column (5) presents the regression of the independent variable Divergence to the
rin
dependent variable StockReturn_ADJ after adding the mediating variable Amount.
As shown in column (1) of Table 4, the regression coefficient of the independent
variable Divergence is -0.149 (p<0.01), thereby indicating that each unit increase in
ep
ESG rating divergence (Divergence) among different institutions will lead to a decrease
of 0.149 units in stock return (StockReturn_ADJ), thereby supporting H1. Columns (2)
Pr
and (4) show that the regression coefficients of the independent variable Divergence on
the mediating variables Volume and Amount are -0.011 (p<0.05) and -0.019 (p<0.01),
21
respectively, thereby indicating that for every unit increase in ESG rating divergence,
ed
both trading volume and amount decrease by 0.011 and 0.019 units, respectively. This
result is consistent with our inference that ESG rating divergence lowers investor
expectations. Columns (3) and (4) report the results of adding intermediate variables to
iew
the primary regression model. After adding Volume, the regression coefficient of the
independent variable decreases from -0.149 (p<0.01) to -0.142 (p<0.01), and after
adding Amount, the regression coefficient decreases from -0.149 (p<0.01) to -0.141
v
(p<0.01). The regression coefficients of Volume and Amount are significantly positive.
re
We can infer from these results that ESG rating divergence reduces stock returns by
reducing stock trading volume and trading amount, thus verifying H2.
er
The significant negative coefficient of ESG rating divergence indicates that severe
ESG rating divergence is related to a decrease in stock returns, and the mediating effect
pe
test supports investor expectations as a channel of action. The main regression results
are also in line with the conclusion of Serafeim and Yoon (2022), who find that ESG
rating divergence hinders the inclusion of value-related ESG news in prices. Our results
ot
also indicate that ESG rating divergence increases the degree of information asymmetry
and reduces the efficiency of market pricing mechanisms. In addition, the mediation
tn
effect test results indicate that in a highly ESG-uncertain environment, the investors’
demand for such stocks decreases, thereby forming more negative expectations and
rin
reducing trading volume and amount. The mediation effect results support the
conclusion of Avramov et al. (2022), who find that market premiums increase in an
ep
equilibrium state, while stock demand decreases under ESG uncertainty.

Pr
22
4.2 Further analyses
ed
4.2.1 E/S/G subdivision test
iew
Due to the short development history of Chinese ESG rating agencies, although
multiple ESG rating agencies are operating in the market, for data users, the underlying
data of most of these agencies are unknown, and their rating process presents a black
v
box. The three selected rating agencies are not only reliable in terms of rating update
re
frequency and popularity but also show advantages in data transparency. Compared
with other agencies that only provide annual comprehensive ratings, these three
agencies provide quarterly ratings and E/S/G sub-item rating data. ESG is a broad
er
concept encompassing three major aspects, and the impact of different rating
divergences on the market may vary. Therefore, we construct rating divergence

pe
indicators for E/S/G based on the same regression method and further observe the
relationship between rating divergence and stock returns.
StockReturn_ADJi, q   0   1 Divergence _ E / S / Gi, q    Controls  FirmFE  QuarterFE   i, q, (8)

ot
k i, q
where Divergence_E/S/G is computed through a pairwise comparison of the sub-item

tn
rating scores.
Table 5 reports the group regression results based on E/S/G segmented rating data.
The negative relationship between ESG rating divergence and stock returns is only
rin
significant at S (social), thereby suggesting that lower stock returns can only be
achieved when companies show differences in their social issues. These findings are
ep
consistent with those of the previous literature emphasizing the S issue (see, e.g., Diaz
et al., 2021; Gras and Krause, 2020; Lund-Thomsen and Lindgreen, 2014; Schiemann
Pr
and Tietmeyer, 2022). These results can also be ascribed to two reasons. First,
compared with the information at the E and G levels, S-level issues are more likely to
23
capture people’s attention due to their direct impact on shareholders’ rights and
ed
compliance with legal and regulatory requirements. Moreover, the related events at this
level have a relatively large frequency and scope, thus driving stock market sentiment
and ultimately affecting stock returns. Second, the S level involves many stakeholders,
iew
including but not limited to employees, suppliers, distributors and contractors,
customers and consumers, local communities, governments, the media, and NGOs
(Doh and Guay, 2006). Therefore, stakeholders may be more sensitive to a
v
company’s S rating and have the ability and motivation to make relevant
re
behaviors and reactions that will affect this company’s business activities.

er
4.2.2 ESG rating divergence and ESG average rating
pe
The divergence of ESG ratings represents the inconsistencies in the interpretation
of ESG information transmitted by companies to the market. In this case, will such
divergence affect the average ESG rating of a company? This issue can be examined
ot
from two aspects. On the one hand, if the main reason for the differences in ESG ratings
tn
is the lack and inconsistency of corporate ESG information disclosure, then if
companies attempt to alleviate these differences through ESG practices or a more
transparent information disclosure, these efforts should positively impact their final
rin
rating. On the other hand, if rating agencies’ rating methods are the main causes of these
divergences, then the efforts made by companies may still lead to more significant
ep
differences. As some researchers have found, ESG disclosure does not reduce but rather
increases the number of outstanding disagreements (see, e.g., Christensen et al., 2022;
Pr
Liu, 2022).
We use the independent variable Divergencei,q for the current period to regress the
24
quarterly average ratings ESG_AVG with 1–4 periods in advance and examine the
ed
impact of ESG rating divergence on the quarterly average ratings within a year. If the
regression coefficient of the independent variable is significantly positive, then those
companies with more considerable rating divergences will receive higher average
iew
ratings in the future. Companies may mitigate the risks brought about by rating
divergences by disclosing additional information or actively improving their ESG
performance. If the regression coefficient is significantly negative, then those
v
companies with severe cognitive differences among rating agencies will face more
re
challenges in improving their ESG rating, thus worsening their subsequent rating
performance. We construct our regression model as follows:
ESG _ AVGi, q + k，k = 1, 2, 3, 4

er
= 0   1 Divergencei, q    kControls  FirmFE  QuaterFE   i, q, (9)
where i represents the listed company, q denotes the quarter, and k denotes the number
pe
of quarters before the dependent variable. We measure ESG_ AVG by taking the
average of the ratings provided by the three selected ESG evaluation agencies.
Table 6 reports the regression results of ESG rating divergence on ESG average
ot
rating. The coefficient of ESG rating divergence is significantly negative and gradually
decreases over time, thereby suggesting that ESG rating divergence significantly
tn
suppresses the average ESG rating for each quarter of the year, and its inhibitory effect
gradually weakens over time. A more considerable ESG divergence does not improve
rin
the average rating given by an agency to the company in the future. On the contrary, a
more serious divergence corresponds to more challenges for the company to improve
ep
its average rating in the long term.

Pr
25
4.2.3 Heterogeneity analyses
ed
Given that the role of ESG rating divergences may vary among different
companies and market environments, to clarify those channels through which ESG
iew
rating divergences affect stock returns, we conduct heterogeneity tests based on the
industry characteristics, property rights, and external concerns of the sample companies.
The heterogeneity test results are shown in columns (1)–(6) of Table 7.
v
First, we group our sample companies based on their industry characteristics and
re
then regress the heavily polluted and non-heavily polluted industry samples to explore
the possible impacts of different industry characteristics. Given that companies in the
heavily polluting industry have always been under tremendous legal pressure, they pay
er
more attention to environmental information disclosure and practice and emphasize
their environmental contributions (see, e.g., Cho et al., 2012; Clarkson et al., 2008; Diaz
pe
et al., 2021). Meanwhile, companies in non-heavily polluting industries face fewer
environmental legality issues, thus allowing them to focus on social contributions,

ot
which are more relevant to social issues, such as labor relations, gender equality,
diversity, and supply chain sustainability (see, e.g., Bade et al., 2023; Baid and
tn
Jayaraman, 2022). These companies also have a highly diverse ESG disclosure, thereby
challenging both rating agencies and investors in forming accurate judgments and
rin
evaluations of their ESG performance. From the perspective of information
interpretation, companies in non-heavily polluting industries have a more
comprehensive range of information, have more diverse forms of expression, and are
ep
more prone to disagreements. Social issues have a more significant impact on stock
returns, so non heavily polluting industry enterprises that are more related to social
Pr
issues may present the same conclusion. Columns (1)–(2) of Table 7 report the
heterogeneity test estimations. The regression coefficients for ESG rating divergence
26
are significant, but the p-value is smaller in non-heavily polluting industry companies
ed
(p=0.013), close to 0.01. Meanwhile, for heavily polluting industry companies, the p-
value is only 0.081, indicating a more significant relationship between ESG rating
divergence and stock return. These regression results explain and support the E/S/G
iew
sub-item test results.
Second, state-owned and non-state-owned companies show many differences in
their governance outcomes, financing environment, legality, and social responsibility
v
requirements (Ang et al., 2022). Therefore, based on the property rights nature of state-
re
owned enterprises, we divide our sample into state-owned and non-state-owned
companies for the regression. The regression results are shown in columns (3)–(4) of
er
Table 7. The regression coefficient of the independent variable in the state-owned
companies sample is -0.170 (p<0.10), while that in the non-state-owned companies

pe
sample is -0.181 (p<0.01). These results indicate that ESG rating divergences have a
more significant impact on non-state-owned companies. Still, the effect on the stock
returns of state-owned companies is weaker, and these companies show a more stable
ot
performance. For non-state-owned companies facing more significant competitive
pressure and higher degree of marketization, on the one hand, a large number of these
tn
companies show substantial differences in their ESG information disclosure and ESG
practice, which often lead to rating divergences. On the other hand, ESG rating
rin
divergence has a greater impact on corporate reputation and triggers a stronger stock
market reaction, thus making non-state-owned companies enjoy higher ESG rating
ep
returns and bear the risks of higher ratings (Shahab et al., 2019). Meanwhile, state-
owned companies are required to shoulder more social responsibility, and their
operating environment is relatively stable given their close relationship with the
Pr
government. Therefore, state-owned companies generally have more stable
27
performance and profits, which weaken the impact of ESG rating divergences.
ed
Third, we consider the impact of external attention. Social attention refers to the
level of understanding of corporate information by the market and investors, and those
companies that receive more social attention have a higher level of corporate reputation
iew
and information transparency. When a company receives more attention, the
information gap between investors and rating agencies decreases, thus helping investors
overcome the negative effects of ESG rating divergences and helping rating agencies
v
provide consistent ratings. Following Zhao et al. (2023), we construct a listed company
re
online search index based on the internet search frequency of listed company names,
codes, and other related keywords as a measure of the external attention of a company.
er
For the regression analysis, we divide our sample into two groups based on the median
of external attention, with one group having a higher external attention and the other
pe
having a lower external attention. As shown in columns (5)–(6) of Table 7, the
regression coefficient of the independent variable in the high-attention sample is -0.110
(p<0.05), while that in the low-attention sample is -0.225 (p<0.01). From the
ot
perspective of information asymmetry, the ESG information of high-attention
companies is more fully absorbed by the market, thereby alleviating the degree of
tn
information asymmetry (see, e.g., Kolbel et al., 2017; Luo et al., 2022; Wong and Zhang,
2022). Meanwhile, for low-attention companies, a greater ESG information asymmetry

rin
can be observed between investors and rating agencies, hence posing additional
challenges to the interpretation and rating of these agencies and placing higher demands
ep
on investors to use ESG information when making investment-related decisions.
Therefore, although ESG rating divergence significantly suppresses stock returns in
companies with low and high external attention, external attention still effectively
Pr
alleviates the relationship between the two.
28
ed
5 Endogenous tests and robustness checks
iew
5.1 Endogenous tests
To address the potential endogeneity issues in the empirical analysis, we select the
following three instrumental variables to emphasize causal relationships: whether the
v
company has issued an independent social responsibility report before 2018 (Report),
re
whether the company has disclosed information following the GRI standard before
2018 (GRI), and whether the company has disclosed the construction and improvement
er
measures of the disclosure social responsibility system before 2018 (CSR). These
instrumental variables are all binary variables, and they respectively consider the carrier,
pe
standard, and content of ESG-related information disclosure, which play crucial roles
in how external markets interpret ESG information (Schiemann and Tietmeyer, 2022).
We also select data before 2018 (the sample starting year) to ensure that these
ot
instrumental variables do not impact stock returns within the sample interval. We use
tn
the 2SLS model and match the linear regression model with propensity score matching
(PSM) to alleviate potential reverse causality and sample selection bias. We test the
effectiveness of the instrumental variables. At the first stage of the 2SLS model, we
rin
estimate the regression coefficients of the instrumental variables to the independent
variable Divergence. All instrumental variables have significant regression coefficients,

ep
and the p-value of Kleibergen Paap rk LM statistic is less than 0.01, passing the non-
identifiability test. The F-value of the first-stage regression is 12.45, and the Cragg–
Pr
Donald Wald F statistical value (13.58) exceeds the critical importance of 10% (9.08),
thereby rejecting the original hypothesis of weak instrumental variables. The p-value of
29
the Hansen J test is 0.32, which rejects the null hypothesis that at least one instrumental
ed
variable is endogenous, proving the absence of over-identification issues, and
confirming the exogeneity of all instrumental variables.
iew
5.1.1 Two-stage least squares
Due to our use of data from the same period, our independent variable may be
influenced by the dependent variable, which may lead to an endogeneity problem of
v
reverse causality. To enhance the effectiveness of our causal inference, we add three
re
instrumental variables (Report, GRI, and CSR) to the 2SLS model for a two-stage
regression. Table 8 shows the regression results of 2SLS in two stages, with column (1)
er
showing the first-stage regression results. The instrumental variables Report, GRI, and
CSR all significantly impact the independent variable Divergence. Column (2) shows
pe
the second-stage regression results, and the regression coefficient of the independent
variable Divergence is -2.369 (p<0.05), indicating that after controlling for reverse
causality, ESG rating divergence still has a significant negative effect on stock returns,
ot
thereby supporting H1.

tn
5.1.2 Propensity score matching

rin
To address the issue of selection bias, we apply PSM to match the samples and to
form an experimental group and a control group with similar initial conditions. We use
ep
the 1:4 nearest neighbor matching method to match the leading financial indicators Lev,
ROA, and Size and the market indicators BM, PE, and INST of the sample companies.
Pr
We divide our sample into a treatment group and a control group. After matching, the
biases of the matching variables are all reduced to below 10%. The ATT value after
30
matching is -5.15, with an absolute value greater than the 1% significant critical value
ed
of 2.58, indicating an excellent matching effect. As shown in Table 9, the regression
coefficient of the independent variable Divergence in the matched treatment group is -
2.369 (p<0.05), indicating that after controlling for sample selection bias, ESG rating
iew
divergence still has a significant negative impact on stock returns, thereby supporting
H1.
v
re
5.2 Robustness checks
We use a series of robustness testing methods to verify the robustness of our

er
conclusions. Columns (1)–(4) of Table 10 report the results of a series of robustness
pe
tests, including replacing the dependent and independent variables, lagging the
independent variables by one period, and adding the fixed effects of industries and
provinces. The robustness test results all support our original conclusions, thereby
ot
proving the validity of our results.
5.2.1 Replacing the dependent variable measurement

tn
We use the quarterly average of individual stock returns (excluding market

rin
average returns) as our dependent variable and name it adjusted stock return
StockReturn_ADJ. We then replace the dependent variable and use the quarterly
average of individual stock returns without excluding market average returns as our
ep
new dependent variable StockReturn_RAW as a robustness test. As shown in column
(1) of Table 10, the regression coefficient of the independent variable Divergence is -
Pr
0.137 (p<0.01), indicating that ESG rating divergence has a characteristic corporate-
level impact and a significant negative impact on overall market returns.

31
5.2.2 Replacing the independent variable measurement
ed
We calculate the standard deviation of the data from the three rating agencies to
form our new independent variable Divergence_sd and test the robustness of our
iew
conclusions. Unlike the method proposed by Avramov et al. (2022), we calculate the
standard deviation of the data from the three rating agencies as follows to observe the
degree of dispersion among the data. For the rating data g1, g2, and g3 of these three
v
agencies, the standard deviation formula is
re
 g1  g 2    g 2  g 3    g1  g 3 
2 2 2
Divergence_sd  . (10)
3
As shown in column (2) of Table 10, the regression coefficient of the independent
er
variable Divergence_sd is -0.130 (p<0.01), which is consistent with the main regression
and supports our original conclusions.

pe
5.2.3 Lagging the independent variable by one period
We use the lagged one-period data of the independent variable to regress the
ot
current dependent variable stock returns given the clear time sequence relationship
between the lagged one-period independent variable and the dependent variable, which
tn
can significantly alleviate the endogeneity problem caused by time. Due to the late
development of ESG in China (Wang, 2023), Wind did not have an ESG rating database
rin
in 2017, thus reducing our sample size. As shown in column (3) of Table 10, the
regression coefficient of the independent variable Divergencei,q-1 is -0.183 (p<0.01),

ep
indicating that after considering possible reverse causality, ESG rating divergence still
significantly suppresses stock returns, thereby supporting our original conclusions.

Pr
32
5.2.4 Adding the fixed effect of industries and provinces
ed
In our original model, we control for individual and quarterly fixed effects.
However, industry and regional factors also play different roles. Due to the variance in
iew
business areas and policies, significant differences can be observed in the ESG
performance and ESG information of companies across various industries and regions,
which may affect our estimations (Diaz et al., 2021). Therefore, we continue to increase
v
the fixed effects of industries and provinces to minimize the impact of random
re
disturbances as much as possible. As shown in column (4) of Table 10, the regression
coefficient of the independent variable Divergencei,q-1 is -0.147 (p<0.01), which
supports our original conclusions.

er
pe
6 Conclusion
ot
We find that ESG rating divergence reduces stock returns, and the decrease in
trading volume and amount is the mechanism of action. First, our correlation test results
tn
on three mainstream rating agencies show that the correlation coefficients between their
ratings are all less than 0.4, with an average of 0.311, indicating a severe divergence in
rin
ESG ratings in China’s financial market. Second, we find that companies with higher
ESG rating divergences have lower investment returns, which contradicts the findings
of previous studies that only focus on the impact of rating levels on investment returns.
ep
We find that different ratings issued by multiple institutions can confuse investor
information judgments and affect investment returns. Therefore, investors not only
Pr
need to observe the ESG performance of the company when making investment
decisions but also need to check for any differences in the opinions of rating agencies.
33
Third, companies with higher ESG rating divergences have lower trading volume and
ed
amount in the market. ESG rating divergence affects investors’ views on these
companies. A higher rating divergence corresponds to the lower performance of a
company in the stock market, hence discouraging investors from investing in this
iew
company and ultimately decreasing its stock returns.
Our study can serve as a supplement to the literature on the economic
consequences of ESG malfunction. As indicated in the literature review in Section 2,
v
previous studies have mainly focused on corporate-level greenwashing and hypocrisy.
re
However, our paper offers a novel perspective by suggesting that the causes of ESG
malfunction may not solely lie within companies themselves but also in the errors of
er
external rating agencies, which lead to the lack of market awareness of companies’ ESG
efforts. We believe that the positive effects of ESG information disclosure can be
pe
influenced by rating agencies, and the inconsistencies among rating agencies can lead
to new information asymmetry that confuses market judgments. Further analysis
reveals that companies with more significant differences in their ESG ratings perform
ot
poorly in their subsequent average ratings, indicating that different standards from
rating agencies may overshadow these companies’ efforts in ESG.

tn
Our findings also offer a series of practical implications. First, for academic
research, conducting ESG-related tests using data from a single rating agency may
rin
result in serious practical biases given the significant differences among different rating
agencies. Therefore, subsequent ESG research should consider rating agencies’

ep
inconsistency and avoid its impact. Second, standardized and unified ESG disclosure
rules should be established in emerging markets as soon as possible to guide companies,
and the corresponding ESG rating rules should be developed to regulate the behavior
Pr
of rating agencies. The Chinese government is actively promoting the construction of
34
an ESG rating system and its standards, and on July 25, 2023, the government issued
ed
the “Notice on Forwarding the Research on the Preparation of ESG Special Reports for
State-Owned Listed Companies.” The ESG rating system of China is expected to be
further improved in the future. Third, for investors, the negative impact of ESG rating
iew
divergence on stock returns indicates the risk of rating divergence when making ESG
investments. ESG ratings issued by different agencies may also reduce investment
returns. Therefore, when making investment decisions, investors should fully consider
v
the ratings of multiple rating agencies and make judgments based on private
re
information.
Our empirical research also has some limitations, which provide directions for
er
future research. First, given that China’s market is still at its early stages of rapid ESG
development, there are limitations in the transparency and completeness of Chinese

pe
rating agencies’ data. Due to limited data availability, we only select data from three
rating agencies that have published quarterly and segmented data for our empirical
analysis. Future research can further expand the scope of rating agencies for
ot
investigation. Second, we assign values to data from different rating agencies and
construct the corresponding ESG rating divergence data. Given that rating agencies do
tn
not disclose their evaluation processes, our assignment results cannot reflect the
differences in the processes of these agencies and only represent the differences in their
rin
outcomes. Therefore, after the ESG rating system and standards are improved, future
research can construct more representative rating divergence data based on the rating
ep
process and products for testing. Third, we only explore one channel of investor
expectations, and future research can further investigate the mechanism behind the
effects of ESG rating divergence on stock returns.

Pr
35
Reference
ed
Ang, R., Shao, Z., Liu, C., Yang, C. H., & Zheng, Q. R. (2022). The relationship
between CSR and financial performance and the moderating effect of ownership
structure: Evidence from Chinese heavily polluting listed enterprises. Sustainable
Production and Consumption, 30, 117-129.
iew
http://dx.doi.org/10.1016/j.spc.2021.11.030
Aouadi, A., & Marsat, S. (2018). Do ESG Controversies Matter for Firm Value?
Evidence from International Data. Journal of Business Ethics, 151(4), 1027-1047.
http://dx.doi.org/10.1007/s10551-016-3213-8
v
Avramov, D., Cheng, S., Lioui, A., & Tarelli, A. (2022). Sustainable investing with
ESG rating uncertainty. Journal of Financial Economics, 145(2), 642-664.
re
http://dx.doi.org/10.1016/j.jfineco.2021.09.009
Bade, C., Olsacher, A., Boehme, P., Truebel, H., Burger, L., & Fehring, L. (2023).
Sustainability in the pharmaceutical industry-An assessment of sustainability
maturity and effects of sustainability measure implementation on supply chain
er
security. Corporate Social Responsibility and Environmental Management.
http://dx.doi.org/10.1002/csr.2564
Baid, V., & Jayaraman, V. (2022). Amplifying and promoting the "S" in ESG investing:
pe
the case for social responsibility in supply chain financing. Managerial Finance,
48(8), 1279-1297. http://dx.doi.org/10.1108/Mf-12-2021-0588
Bax, K., Sahin, O., Czado, C., & Paterlini, S. (2023). ESG, risk, and (tail) dependence.
International Review of Financial Analysis, 87.
http://dx.doi.org/10.1016/j.irfa.2023.102513
ot
Berg, F., Kolbel, J. F., & Rigobon, R. (2022). Aggregate Confusion: The Divergence
of ESG Ratings. Review of Finance, 26(6), 1315-1344.
tn
http://dx.doi.org/10.1093/rof/rfac033
Billio, M., Costola, M., Hristova, I., Latino, C., & Pelizzon, L. (2021). Inside the ESG
ratings: (Dis)agreement and performance. Corporate Social Responsibility and
Environmental Management, 28(5), 1426-1445. http://dx.doi.org/10.1002/csr.2177
rin
Boone, A. L., & White, J. T. (2015). The effect of institutional ownership on firm
transparency and information production. Journal of Financial Economics, 117(3),
508-533. http://dx.doi.org/10.1016/j.jfineco.2015.05.008
ep
Brandon, R. G., Krueger, P., & Schmidt, P. S. (2021). ESG Rating Disagreement and
Stock Returns. Financial Analysts Journal, 77(4), 104-127.
http://dx.doi.org/10.1080/0015198x.2021.1963186
Broadstock, D. C., Chan, K. L., Cheng, L. T. W., & Wang, X. W. (2021). The role of
Pr
ESG performance during times of financial crisis: Evidence from COVID-19 in

China. Finance Research Letters, 38. http://dx.doi.org/10.1016/j.frl.2020.101716
36
Chatterji, A. K., Durand, R., Levine, D. I., & Touboul, S. (2016). Do ratings of firms
converge? Implications for managers, investors and strategy researchers. Strategic
ed
Management Journal, 37(8), 1597-1614. http://dx.doi.org/10.1002/smj.2407
Cheng, Q., Du, F., Wang, B. Y. T., & Wang, X. (2019). Do Corporate Site Visits Impact
Stock Prices? Contemporary Accounting Research, 36(1), 359-388.
http://dx.doi.org/10.1111/1911-3846.12417
iew
Cho, C. H., Freedman, M., & Patten, D. M. (2012). Corporate disclosure of
environmental capital expenditures A test of alternative theories. Accounting
Auditing & Accountability Journal, 25(3), 486-507.
http://dx.doi.org/10.1108/09513571211209617
v
Christensen, D. M., Serafeim, G., & Sikochi, A. (2022). Why is Corporate Virtue in the
Eye of The Beholder? The Case of ESG Ratings. The Accounting Review, 97(1),
re
147-175. http://dx.doi.org/10.2308/tar-2019-0506
Clarkson, P. M., Li, Y., Richardson, G. D., & Vasvari, F. P. (2008). Revisiting the
relation between environmental performance and environmental disclosure: An
empirical analysis. Accounting Organizations and Society, 33(4-5), 303-327.
er
http://dx.doi.org/10.1016/j.aos.2007.05.003
Cornell, B. (2021). ESG preferences, risk and return. European Financial Management,
27(1), 12-19. http://dx.doi.org/10.1111/eufm.12295
pe
Cui, J., Jo, H., & Na, H. (2018). Does Corporate Social Responsibility Affect
Information Asymmetry? Journal of Business Ethics, 148(3), 549-572.
DeAngelo, H., DeAngelo, L., & Skinner, D. J. (1996). Reversal of fortune - Dividend
ot
signaling and the disappearance of sustained earnings growth. Journal of Financial

Economics, 40(3), 341-371. http://dx.doi.org/10.1016/0304-405x(95)00850-E
Diaz, V., Ibrushi, D., & Zhao, J. (2021). Reconsidering systematic factors during the
tn
Covid-19 pandemic - The rising importance of ESG. Finance Research Letters, 38,
101870. http://dx.doi.org/10.1016/j.frl.2020.101870
Doh, J. P., & Guay, T. R. (2006). Corporate social responsibility, public policy, and
NGO activism in europe and the United States: An institutional-stakeholder
rin
perspective. Journal of Management Studies, 43(1), 47-73.

http://dx.doi.org/10.1111/j.1467-6486.2006.00582.x
Dorfleitner, G., Halbritter, G., & Nguyen, M. (2015). Measuring the level and risk of
ep
corporate responsibility – An empirical comparison of different ESG rating

approaches. Journal of Asset Management, 16(7), 450-466.
http://dx.doi.org/10.1057/jam.2015.31
Dumrose, M., Rink, S., & Eckert, J. (2022). Disaggregating confusion? The EU
Pr
Taxonomy and its relation to ESG rating. Finance Research Letters, 48.
http://dx.doi.org/10.1016/j.frl.2022.102928
Durand, R., Paugam, L., & Stolowy, H. (2019). Do investors actually value
37
sustainability indices? Replication, development, and new evidence on CSR
visibility. Strategic Management Journal, 40(9), 1471-1490.
ed
http://dx.doi.org/10.1002/smj.3035
Erhart, S. (2022). Take it with a pinch of salt-ESG rating of stocks and stock indices.
International Review of Financial Analysis, 83.
iew
Gillan, S. L., Koch, A., & Starks, L. T. (2021). Firms and social responsibility: A review
of ESG and CSR research in corporate finance. Journal of Corporate Finance, 66.
http://dx.doi.org/10.1016/j.jcorpfin.2021.101889
Gomez-Valencia, M., Gonzalez-Perez, M. A., & Gomez-Trujillo, A. M. (2021). The
v
"Six Ws" of sustainable development risks. Business Strategy and the Environment,
30(7), 3131-3144. http://dx.doi.org/10.1002/bse.2794
re
Gras, D., & Krause, R. (2020). When does it pay to stand out as stand-up? Competitive
contingencies in the corporate social performance-corporate financial performance
relationship. Strategic Organization, 18(3), 448-471.
http://dx.doi.org/10.1177/1476127018805252
er
Groening, C., & Kanuri, V. K. (2018). Investor Reactions to Concurrent Positive and
Negative Stakeholder News. Journal of Business Ethics, 149(4), 833-856.
pe
Hafenbradl, S., & Waeger, D. (2021). The business case for CSR: A trump card against
hypocrisy? Journal of Business Research, 129, 838-848.
http://dx.doi.org/10.1016/j.jbusres.2019.08.043
Huang, L. X., Li, W., Wang, H., & Wu, L. S. (2022). Stock dividend and analyst
ot
optimistic bias in earnings forecast. International Review of Economics & Finance,

78, 643-659. http://dx.doi.org/10.1016/j.iref.2022.01.007
Hummel, K., Mittelbach-Hoermanseder, S., Rammerstorfer, M., & Weinmayer, K.

tn
(2019). Stock Market Reactions and CSR Disclosure in the Context of Negative
CSR Events. SSRN Electronic Journal. http://dx.doi.org/10.2139/ssrn.3467616
Jiang, X. Q., & Lee, B. S. (2007). Stock returns, dividend yield, and book-to-market
ratio. Journal of Banking & Finance, 31(2), 455-475.
rin
http://dx.doi.org/10.1016/j.jbankfin.2006.07.012
Kane, A., Lee, Y. K., & Marcus, A. (1984). Earnings and Dividend Announcements -
Is There a Corroboration Effect. Journal of Finance, 39(4), 1091-1099.
ep
http://dx.doi.org/Doi 10.2307/2327614
Kolbel, J. F., Busch, T., & Jancso, L. M. (2017). How Media Coverage of Corporate
Social Irresponsibility Increases Financial Risk. Strategic Management Journal,
38(11), 2266-2284. http://dx.doi.org/10.1002/smj.2647
Pr
Kotsantonis, S., & Serafeim, G. (2019). Four Things No One Will Tell You About ESG
Data. Journal of Applied Corporate Finance, 31(2), 50-58.
http://dx.doi.org/10.1111/jacf.12346
38
Lettau, M., & Ludvigson, S. C. (2005). Expected returns and expected dividend growth.
Journal of Financial Economics, 76(3), 583-626.
ed
http://dx.doi.org/10.1016/j.jfineco.2004.05.008
Li, S. C., Yin, P. F., & Liu, S. (2022). Evaluation of ESG Ratings for Chinese Listed
Companies From the Perspective of Stock Price Crash Risk. Frontiers in
Environmental Science, 10. http://dx.doi.org/10.3389/fenvs.2022.933639
iew
Liu, M. (2022). Quantitative ESG disclosure and divergence of ESG ratings. Frontiers
in Psychology, 13. http://dx.doi.org/10.3389/fpsyg.2022.936798
Lund-Thomsen, P., & Lindgreen, A. (2014). Corporate Social Responsibility in Global

Value Chains: Where Are We Now and Where Are We Going? Journal of Business
v
Ethics, 123(1), 11-22. http://dx.doi.org/10.1007/s10551-013-1796-x
Luo, K., & Wu, S. R. (2022). Corporate sustainability and analysts' earnings forecast
re
accuracy: Evidence from environmental, social and governance ratings. Corporate
Social Responsibility and Environmental Management, 29(5), 1465-1481.
http://dx.doi.org/10.1002/csr.2284
Luo, X. M., Wang, H. L., Raithel, S., & Zheng, Q. Q. (2015). Corporate Social
er
Performance, Analyst Stock Recommendations, and Firm Future Returns. Strategic
Management Journal, 36(1), 123-136. http://dx.doi.org/10.1002/smj.2219
pe
Luo, Y. D., Xiong, G. B., & Mardani, A. (2022). Environmental information disclosure
and corporate innovation: The "Inverted U-shaped" regulating effect of media
attention. Journal of Business Research, 146, 453-463.
http://dx.doi.org/10.1016/j.jbusres.2022.03.089
Lyon, T. P., & Maxwell, J. W. (2011). Greenwash: Corporate Environmental

ot
Disclosure under Threat of Audit. Journal of Economics & Management Strategy,

20(1), 3-41. http://dx.doi.org/10.1111/j.1530-9134.2010.00282.x
Meng, J., & Zhang, Z. X. (2022). Corporate environmental information disclosure and
tn
investor response: Evidence from China's capital market. Energy Economics, 108.
http://dx.doi.org/10.1016/j.eneco.2022.105886
Park, J. W., & Lee, C. W. (2018). Performance of stock price with changes in SRI
governance index. Corporate Social Responsibility and Environmental
rin
Management, 25(6), 1121-1129. http://dx.doi.org/10.1002/csr.1526
Peng, Z. H., Johnstone, D., & Christodoulou, D. (2020). Asymmetric impact of earnings
news on investor uncertainty. Journal of Business Finance & Accounting, 47(1-2),
ep
3-26. http://dx.doi.org/10.1111/jbfa.12428
Phan, D. H. B., Sharma, S. S., & Narayan, P. K. (2015). Stock return forecasting: Some
new evidence. International Review of Financial Analysis, 40, 38-51.
http://dx.doi.org/10.1016/j.irfa.2015.05.002
Pr
Schiemann, F., & Tietmeyer, R. (2022). ESG Controversies, ESG Disclosure and
Analyst Forecast Accuracy. International Review of Financial Analysis, 84.
39
Serafeim, G., & Yoon, A. (2022). Stock price reactions to ESG news: the role of ESG
ratings and disagreement. Review of Accounting Studies.
ed
Shahab, Y., Ntim, C. G., & Ullah, F. (2019). The brighter side of being socially
responsible: CSR ratings and financial distress among Chinese state and non-state
owned firms. Applied Economics Letters, 26(3), 180-186.
iew
http://dx.doi.org/10.1080/13504851.2018.1450480
Shanaev, S., & Ghimire, B. (2022). When ESG meets AAA: The effect of ESG rating
changes on stock returns. Finance Research Letters, 46.
http://dx.doi.org/10.1016/j.frl.2021.102302
v
She, C. Y., & Michelon, G. (2019). Managing stakeholder perceptions: Organized
hypocrisy in CSR disclosures on Facebook. Critical Perspectives on Accounting, 61,
re
54-76. http://dx.doi.org/10.1016/j.cpa.2018.09.004
Wagner, T., Lutz, R. J., & Weitz, B. A. (2009). Corporate Hypocrisy: Overcoming the
Threat of Inconsistent Corporate Social Responsibility Perceptions. Journal of
Marketing, 73(6), 77-91. http://dx.doi.org/10.1509/jmkg.73.6.77
er
Wang, L. L. (2023). Transmission Effects of ESG Disclosure Regulations Through
Bank Lending Networks. Journal of Accounting Research, 61(3), 935-978.
http://dx.doi.org/10.1111/1475-679x.12478
pe
Wong, J. B., & Zhang, Q. (2022). Stock market reactions to adverse ESG disclosure via
media channels. British Accounting Review, 54(1).
http://dx.doi.org/10.1016/j.bar.2021.101045
Yu, E. P. Y., & Luu, B. V. (2021). International variations in ESG disclosure - Do cross-
ot
listed companies care more? International Review of Financial Analysis, 75.

Yu, E. P. Y., Luu, B. V., & Chen, C. H. (2020). Greenwashing in environmental, social
tn
and governance disclosures. Research in International Business and Finance, 52.

http://dx.doi.org/10.1016/j.ribaf.2020.101192
Yu, H. X., Liang, C. Y., Liu, Z. H., & Wang, H. (2023). News-based ESG sentiment
and stock price crash risk. International Review of Financial Analysis, 88.
rin
Zhang, Z. Q., Su, Z., Wang, K., & Zhang, Y. J. (2022). Corporate environmental
information disclosure and stock price crash risk: Evidence from Chinese listed
ep
heavily polluting companies. Energy Economics, 112.

http://dx.doi.org/10.1016/j.eneco.2022.106116
Zhao, X. Z., Fang, L. B., & Zhang, K. (2023). Online search attention, firms' ESG and
operating performance. International Review of Economics & Finance, 88, 223-236.
Pr
http://dx.doi.org/10.1016/j.iref.2023.06.025
Zhong, H. Q., Yan, R. Y., Li, S., & Chen, M. (2020). The Psychological Expectation
of New Project Income Under the Influence of the Entrepreneur's Sentiment From
40
the Perspective of Information Asymmetry. Frontiers in Psychology, 11.
http://dx.doi.org/10.3389/fpsyg.2020.01416
ed
Zhu, B., & Niu, F. (2016). Investor sentiment, accounting information and stock price:
Evidence from China. Pacific-Basin Finance Journal, 38, 125-134.
http://dx.doi.org/10.1016/j.pacfin.2016.03.010
v iew
re
er
pe
ot
tn
rin
ep
Pr
41
e d
Tables
Type Name
Table 1 Variable Definition
Variable Measurement
i ew Data Source
Dependent variables Adjusted quarterly stock returns
Non-adjusted quarterly stock returns

StockReturn_ADJ
StockReturn_RAW
market returns.
e v
Quarterly weighted average of daily stock returns excluding average
Quarterly weighted average of daily stock returns without excluding

CSMAR
CSMAR
Independent variables
Mediator variables
Divergence among rating agencies
Quarterly stock trading volume

Divergence
Volume
average market returns.
r r
Avramov et al. (2022) define the divergence among rating agencies as the
average of the comparison results.
Quarterly average of daily stock trading volume.
WIND,
iFinD
CSMAR
e
Quarterly stock trading amount Amount Quarterly average of daily stock trading amount. CSMAR
Control variables Company age Age Natural logarithm of the difference in the number of quarters from the CSMAR
Company size
Proprietary costs
Size
Propcost
p e
company’s IPO date to the current quarter + 1.
Natural logarithm of total assets of the company at the end of the period.
Herfindahl–Hirschman Index, which is measured as the sum of the
squared market shares of all companies in each industry group.
CSMAR
CSMAR
Institutional investor shareholding

Return on assets
Leverage
Operating cash flow ratio
o t
INST
ROA
Lev
Cashflow
The proportion of total shares held by institutional investors.
Quarterly net profit/total asset balance.
Quarterly total liabilities/total assets.
Net cash flows from operating activities/total assets at the end of the
CSMAR
CSMAR
CSMAR
CSMAR
Book-to-market ratio
P/E ratio
t n BM
PE
period.
Total assets/market value.
The current value of the current closing price/(annual report value of net
profit of last year/closing value of current period of paid-in capital).
CSMAR
CSMAR
ir n
Notes: This table shows the measurement methods and data sources of the main variables in this paper. We construct ESG rating divergence data following the
method of Avramov et al. (2022). We also add the fixed effect of individuals and quarters in the model and cluster the standard error at the individual level.
e p
P r 42
Table 2 Summary Statistics
N Mean St.Dev Min Max
ed
StockReturn_ADJ 54,679 0.094 2.887 -13.654 33.511
StockReturn_RAW 54,679 0.313 3.225 -14.549 31.115
Divergence 54,679 0.957 0.498 0.000 2.357
ESG_Wind 54,679 5.504 0.859 3.000 9.000
ESG_Sino 54,679 4.202 1.068 1.000 8.000
ESG_Quant 54,679 5.567 1.347 1.000 9.000
iew
ESG_AVG 54,679 5.091 0.806 1.667 8.000
Volume 54,679 18.230 2.390 13.140 24.986
Amount 54,679 20.264 2.087 14.740 26.790
Age 54,679 3.527 0.907 1.099 4.787
Size 54,679 22.479 1.356 20.016 28.510
Propcosts 54,679 0.062 0.070 0.010 0.545
INST 54,679 42.386 25.317 0.057 98.563
v
ROA 54,679 0.029 0.049 -0.451 0.238
Lev 54,679 0.419 0.199 0.052 0.935
Cashflow 54,679 0.024 0.061 -0.174 0.270
re
BM 54,679 0.677 0.256 0.115 1.293
PE 54,679 34.595 22.473 3.241 97.157
ifhp 54,679 0.263 0.440 0.000 1.000
SOE 54,679 0.336 0.472 0.000 1.000
Notes: This table presents the descriptive statistics of the dependent variable, independent
er
variable, and the rating results of the three ESG agencies constituting the independent variable,
the average rating, the mediator variable, and the control variable after winsorizing the
continuous variable by the top and bottom 1%.
pe
ot
tn
rin
ep
Pr
43
Table 3 Correlations
e d
w
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)
(1)ESG_Wind 1.000
(2)ESG_Sino
(3)ESG_Quant
(4)StockReturn_ADJ
(5)Divergence
0.317***
0.350***
0.006
0.228***
1.000
0.265***
0.000
-0.518***
1.000
0.041***
0.367***
1.000
0.015*** 1.000
v i e
(6)Volume
(7)Amount
-0.102***
-0.084***
-0.070***
-0.058***
0.148***
0.147***
0.024***
0.032***
0.082***
0.078***
1.000
0.958*** 1.000
r e
r
(8)Age -0.098*** -0.087*** 0.256*** 0.034*** 0.128*** 0.549*** 0.523*** 1.000
(9)Size 0.107*** 0.191*** 0.479*** 0.002 0.129*** 0.352*** 0.328*** 0.480*** 1.000
e
(10)Propcosts -0.086*** -0.055*** 0.042*** -0.004 0.032*** 0.112*** 0.110*** 0.121*** 0.184*** 1.000
(11)INST 0.060*** 0.116*** 0.280*** 0.029*** 0.056*** 0.186*** 0.189*** 0.216*** 0.470*** 0.183*** 1.000
(12)ROA
(13)Lev
(14)Cashflow
0.114***
-0.120***
0.047***
0.159***
-0.096***
0.030***
0.091***
0.152***
0.091***
0.050***
0.015***
0.029***
-0.029***
0.102***
0.040***
-0.092***
0.218***
0.016***
p e
-0.065***
0.201***
0.024***
-0.149***
0.341***
0.007*
-0.002
0.501***
0.067***
-0.041***
0.100***
0.054***
0.079***
0.187***
0.100***
1.000
-0.305***
0.432***
1.000
-0.117*** 1.000
(15)BM
(16)PE
-0.101***
0.015***
0.038***
-0.107***
0.170***
-0.184***
-0.152***
0.105***
0.017***
0.002
o t
0.248***
-0.153***
0.207***
-0.141***
0.253***
-0.168***
0.512***
-0.357***
0.164***
-0.078***
0.153***
-0.161***
-0.220***
-0.128***
Notes: This table reports the Pearson correlations among the main variables and the ESG rating data from the three selected rating agencies. The coefficients
0.362***
-0.159***
-0.082***
-0.110***
1.000
-0.411*** 1.000
n
for all correlations are based on N = 54,679 (total sample). *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.
t
ir n
e p
P r 44
Table 4 Multiple Regression and Mediating Effect Test
(1) (2) (3) (4) (5)
ed
VARIABLES StockReturn_ADJi,q Volumei,q StockReturn_ADJi,q Amounti,q StockReturn_ADJi,q
Divergencei,q -0.149*** -0.011** -0.142*** -0.019*** -0.141***
(-4.11) (-2.44) (-4.24) (-3.82) (-4.19)
Volumei,q 0.604***
(13.60)
Amounti,q 0.424***
iew
(10.64)
Agei,q -0.156 -0.237*** -0.013 -0.312*** -0.024
(-1.49) (-22.69) (-0.14) (-26.03) (-0.26)
Sizei,q -0.141 -0.033** -0.121 0.064*** -0.168**
(-1.49) (-2.16) (-1.64) (3.89) (-2.27)
Propcostsi,q 0.919 0.439*** 0.654 0.440*** 0.732
(0.78) (3.82) (0.70) (3.33) (0.78)
v
INSTi,q 0.018*** -0.002*** 0.019*** -0.002*** 0.019***
(6.20) (-4.29) (7.51) (-4.03) (7.40)
ROAi,q 1.384*** 0.535*** 1.061*** 0.601*** 1.129***
re
(3.75) (4.89) (2.84) (5.26) (3.03)
Levi,q 1.568*** 0.015 1.559*** -0.190*** 1.648***
(6.06) (0.33) (6.98) (-3.79) (7.38)
Cashflowi,q 0.321 0.137*** 0.238 0.171*** 0.248
(1.01) (2.98) (0.78) (3.36) (0.82)
BMi,q -7.557*** -0.263*** -7.398*** -0.535*** -7.330***
PEi,q
(-46.75)
0.029***
er
(-13.10)
0.000
(-49.39)
0.029***
(-22.30)
0.000
(-48.57)
0.029***
(26.57) (1.07) (30.28) (1.03) (30.27)
Firm FE YES YES YES YES YES
pe
Quarter FE YES YES YES YES YES
Constant 6.523*** 20.004*** -5.560*** 20.389*** -2.123
(3.24) (62.51) (-3.13) (58.67) (-1.21)
Observations 54,679 54,679 54,679 54,679 54,679
Adj R2 0.143 0.981 0.148 0.969 0.146
Notes: This table shows the regression results of the bidirectional fixed effects model and the
mediation effect test model. Column (1) reports the estimation results for the main regression,
ot
which uses the following equation:

StockReturn_ADJi, q   0   1Divergencei, q   kControls  FirmFE  QuarterFE   i, q,
where the dependent variable is the quarterly return rate of individual stocks considering the
tn
reinvestment of cash dividends after deducting market average returns, and the independent
variable is based on divergent indicators constructed by the three primary ESG rating agencies.
We control for a series of control variables, including company characteristics, financial
indicators, and market performance, while fixing the individual and quarter based on individual
clustering standard errors.
rin
Columns (2) and (4) report the estimated impact of the independent variable Divergence on the
mediating variables Volume and Amount in the following models:
Volumei, q / Amounti, q = 0   1 Divergencei, q    kControls  FirmFE  QuaterFE   i, q,
The estimation results after adding the mediating variables in the regression model are shown
in columns (3) and (5):
ep
StockReturn_ADJi, q   0   1 Divergencei, q   2 Mediatori, q   Controls  FirmFE  QuarterFE   ,

k i, q
where 1 indicates the impact of Divergence on StockReturn_ADJ after adding mediator

variables, and 2 represents the influence of trading volume or amount on stock return. *, * *,
Pr
and * * * indicate significance at the 10%, 5%, and 1% levels, respectively.
45
Table 5 E/S/G Subdivision Test
(1) (2) (3)
ed
VARIABLES StockReturn_ADJi,q StockReturn_ADJi,q StockReturn_ADJi,q
Divergence_Ei,q -0.016
(-0.72)
Divergence_Si,q -0.053**
(-2.04)
Divergence_Gi,q 0.015
iew
(0.47)
Agei,q -0.135 -0.139 -0.139
(-1.29) (-1.32) (-1.33)
Sizei,q -0.147 -0.149 -0.149
(-1.55) (-1.57) (-1.58)
Propcostsi,q 0.906 0.882 0.924
(0.77) (0.75) (0.79)
v
INSTi,q 0.018*** 0.018*** 0.018***
(6.19) (6.18) (6.18)
ROAi,q 1.375*** 1.375*** 1.372***
re
(3.72) (3.72) (3.71)
Levi,q 1.543*** 1.541*** 1.538***
(5.97) (5.96) (5.95)
Cashflowi,q 0.308 0.309 0.310
(0.97) (0.98) (0.98)
BMi,q -7.540*** -7.545*** -7.537***
PEi,q
er
(-46.66)
0.029***
(-46.74)
0.029***
(-46.65)
0.029***
(26.45) (26.44) (26.44)
Firm FE YES YES YES
pe
Quarter FE YES YES YES
Constant 6.481*** 6.586*** 6.501***
(3.21) (3.27) (3.22)
Observations 54,679 54,679 54,679
Adj R2 0.142 0.143 0.142
Notes: This table reports the regression results of rating agencies’ divergent E/S/G ratings,
given by
ot
StockReturn_ADJi, q   0   1 Divergence _ E / S / Gi, q   Controls

k i, q  FirmFE  QuarterFE   i, q,
where the independent variable Divergence_E/S/G is computed from the sub-item rating scores
following the approach of Avramov et al. (2022). *, * *, and * * * indicate significance at the
tn
10%, 5%, and 1% levels, respectively.

rin
ep
Pr
46
Table 6 ESG Rating Divergence and ESG Average Rating
(1) (2) (3) (4)
ed
VARIABLES ESG_AVGq+1 ESG_AVGq+2 ESG_AVGq+3 ESG_AVGq+4
Divergencei,q -0.116*** -0.085*** -0.058*** -0.034***
(-21.16) (-14.83) (-9.65) (-5.38)
Agei,q 0.010 -0.005 -0.062*** -0.071***
(0.73) (-0.29) (-3.79) (-3.88)
Sizei,q 0.251*** 0.216*** 0.166*** 0.088***
iew
(20.45) (16.40) (11.90) (5.84)
Propcostsi,q -0.309* -0.166 0.027 0.295
(-1.89) (-0.98) (0.15) (1.51)
INSTi,q 0.001** 0.001*** 0.001** 0.001**
(2.33) (3.46) (2.05) (2.20)
ROAi,q -0.232*** 0.258*** 0.490*** 0.594***
(-3.74) (4.43) (8.23) (9.65)
v
Levi,q -0.566*** -0.547*** -0.447*** -0.291***
(-16.58) (-15.16) (-11.64) (-6.93)
Cashflowi,q -0.217*** -0.220*** -0.093* 0.009
re
(-4.74) (-4.66) (-1.90) (0.18)
BMi,q -0.157*** -0.164*** -0.206*** -0.263***
(-7.43) (-7.40) (-8.86) (-10.63)
PEi,q -0.000*** -0.000*** -0.000 0.000**
(-2.71) (-2.79) (-0.33) (2.34)
Constant -0.117 0.678** 1.979*** 3.693***
Firm FE
(-0.45)
YES
er (2.45)
YES
(6.72)
YES
(11.63)
YES
Quarter FE YES YES YES YES
Observations 48,961 44,531 40,434 36,668
pe
Adj R2 0.786 0.791 0.798 0.801
Notes: This table reports the regression results of the current ESG rating divergence on the
average ESG rating after 1–4 quarters, given by
ESG _ AVGi, q + k，k = 1, 2, 3, 4 = 0   1 Divergencei, q    kControls  FirmFE  QuaterFE   i, q,
where i represents the listed company, q denotes the quarter, and k denotes the number of
quarters before the dependent variable. ESG_AVG is measured by the average rating data from
ot
three ESG evaluation agencies. *, * *, and * * * indicate significance at the 10%, 5%, and 1%
levels, respectively.
tn
rin
ep
Pr
47
Table 7 Heterogeneity Test
e d
w
(1) (2) (3) (4) (5) (6)
VARIABLES Heavily polluting Non-heavy polluting State-owned Non-state-owned High attention Low attention
e
industries industries
Divergencei,q
Agei,q
-0.126*
(-1.75)
-0.034
(-0.17)
-0.104**
(-2.49)
-0.256**
(-2.08)
-0.107*
(-1.79)
-0.435
(-1.27)
-0.181***
(-3.81)
0.117
(0.87)
v i-0.110**
(-2.14)
-0.738***
(-2.76)
-0.225***
(-4.36)
0.087
(0.69)
Sizei,q
Propcostsi,q
-0.256
(-1.45)
-2.815
-0.065
(-0.59)
1.280
0.547***
(3.21)
-2.515
-0.384***
r
(-3.19)
2.081
e -0.028
(-0.22)
-1.170
-0.327**
(-2.37)
2.579*
INSTi,q
ROAi,q
(-1.06)
0.027***
(5.15)
-1.538*
(0.98)
0.013***
(3.88)
2.343***
er
(-1.52)
0.030***
(5.40)
-0.703
(1.21)
0.014***
(3.84)
1.879***
(-0.70)
0.031***
(7.27)
0.423
(1.68)
0.009**
(2.32)
2.307***
Levi,q
Cashflowi,q
(-1.82)
1.156**
(2.24)
-0.203
(5.83)
1.665***
(5.58)
0.602*
p e
(-0.85)
0.816*
(1.68)
-0.465
(4.48)
1.662***
(5.32)
0.480
(0.81)
1.469***
(4.03)
-0.049
(4.20)
1.792***
(4.84)
0.396
BMi,q
PEi,q
(-0.31)
-7.463***
(-25.56)
0.034***
(1.68)
-7.728***
(-39.51)
0.027***
o t (-0.94)
-8.885***
(-26.09)
0.025***
(1.15)
-7.624***
(-39.46)
0.032***
(-0.10)
-8.396***
(-34.39)
0.027***
(0.97)
-6.661***
(-30.31)
0.032***
Firm FE
Quarter FE
(14.98)
YES
YES
t n (21.92)
YES
YES
(12.36)
YES
YES
(23.47)
YES
YES
(17.27)
YES
YES
(21.22)
YES
YES
ir n
Constant 9.093** 5.240** -6.356 10.624*** 6.957** 8.899***
(2.38) (2.23) (-1.61) (4.22) (2.42) (3.12)
Observations 14,388 40,291 18,371 36,308 27,353 27,326
Adjusted R2 0.142 0.155 0.161 0.149 0.153 0.157
p
Notes: Columns (1)–(6) report the regression results based on industry characteristics, property rights, and external attention. Following Zhao et al. (2023), we
use the online search index to measure external attention and divide our sample into great-attention and low-attention groups for the regression based on median
e
size. *, * *, and * * * indicate significance at the 10%, 5%, and 1% levels, respectively.
P r 48
Table 8 2SLS Regression Test
(1) (2)
ed
First stage Second stage
VARIABLES Divergencei,q StockReturn_ADJi,q
Divergencei,q -2.369**
(-2.45)
Report 0.013*
(1.81)
iew
GRI 0.028***
(2.86)
CSR 0.017**
(2.29)
Agei,q 0.038*** 0.242***
(13.90) (5.60)
Sizei,q 0.038*** 0.239***
v
(14.79) (5.29)
Propcostsi,q 0.054* 0.445**
(1.73) (2.27)
re
INSTi,q -0.000*** 0.001*
(-2.60) (1.72)
ROAi,q -0.350*** 1.590***
(-6.64) (3.23)
Levi,q 0.117*** 1.045***
(8.70) (7.74)
Cashflowi,q
er 0.423***
(11.00)
1.243***
(2.62)
BMi,q -0.126*** -2.302***
(-11.92) (-16.15)
pe
PEi,q 0.001*** 0.012***
(7.58) (11.48)
Constant -0.026 -3.316***
(-0.52) (-10.17)
Observations 54,679 54,679
R2 -0.120
Notes: This table reports the regression results of the two stages of 2SLS. At the first stage, we
ot
use the instrumental and control variables to regress the independent variables and find that the
instrumental variables significantly affect the independent variables. We also test the
effectiveness of the instrumental variables. The Kleibergen Paap rk LM statistic in the
identification test is 37.30 (p<0.01), indicating no unrecognizable issues. In the weak
tn
identification test, the Cragg Donald Wald F statistic is 13.58, which is greater than the 10%
significant critical point of 9.08 and close to the 5% significant critical point of 13.91. In the
overidentification test, the Hansen J statistical value is 2.274 (p=0.321>0.10), indicating that
all instrumental variables are exogenous. *, * *, and * * * indicate significance at the 10%, 5%,
and 1% levels, respectively.
rin
ep
Pr
49
Table 9 Regression Test After 1:4 Nearest Neighbor Matching
(1)
ed
VARIABLES StockReturn_ADJi,q
Divergencei,q -0.175***
(-3.16)
Agei,q -0.783*
(-1.67)
Sizei,q -0.127
iew
(-0.76)
Propcostsi,q 1.385
(0.76)
INSTi,q 0.015***
(3.14)
ROAi,q 2.293**
(2.01)
v
Levi,q 1.845***
(3.76)
Cashflowi,q 0.912*
re
(1.93)
BMi,q -7.566***
(-26.42)
PEi,q 0.036***
(20.41)
Constant 7.995**
Observations
er (2.31)
29,358
Adj R2 0.213
Notes: This table reports the regression results of the experimental group after PSM. We use
pe
the 1:4 nearest neighbor matching method to match the sample companies’ main financial
indicators Lev, ROA, and Size and the market indicators BM, PE, and INST. After matching, the
biases of these indicators decrease to within 10%, and the ATT value is -5.15, with an absolute
value greater than the 1% significant critical value of 2.58, indicating a good matching effect.
*, * *, and * * * indicate significance at the 10%, 5%, and 1% levels, respectively.
ot
tn
rin
ep
Pr
50
Table 10 Robustness Test
(1) (2) (3) (4)
ed
VARIABLES StockReturn_RAWi,q StockReturn_ADJi,q StockReturn_ADJi,q StockReturn_ADJi,q
Divergencei,q -0.137*** -0.147***
(-3.77) (-4.06)
Divergence_sdi,q -0.130***
(-3.82)
Divergencei,q-1 -0.183***
iew
(-4.68)
Agei,q -0.180* -0.154 -0.686*** -0.140
(-1.70) (-1.47) (-5.38) (-1.34)
Sizei,q -0.140 -0.142 -0.163 -0.145
(-1.42) (-1.50) (-1.57) (-1.52)
Propcostsi,q 0.866 0.918 0.337 0.960
(0.73) (0.78) (0.27) (0.81)
v
INSTi,q 0.019*** 0.018*** 0.020*** 0.018***
(6.43) (6.19) (6.25) (6.22)
ROAi,q 1.371*** 1.385*** 1.794*** 1.352***
re
(3.73) (3.75) (4.59) (3.66)
Levi,q 1.599*** 1.567*** 1.748*** 1.554***
(6.06) (6.05) (6.39) (6.01)
Cashflowi,q 0.466 0.321 0.339 0.297
(1.48) (1.01) (1.03) (0.94)
BMi,q -7.754*** -7.555*** -8.045*** -7.581***
PEi,q
(-47.65)
0.030***
er (-46.75)
0.029***
(-45.61)
0.037***
(-46.71)
0.029***
(26.99) (26.56) (30.01) (26.55)
Firm FE YES YES YES YES
pe
Quarter FE YES YES YES YES
Province FE NO NO NO YES
Industry FE NO NO NO YES
Constant 6.862*** 6.538*** 9.053*** 6.570***
(3.28) (3.24) (4.10) (3.26)
Observations 54,679 54,679 48,961 54,679
Adj R2 0.321 0.143 0.162 0.142
ot
Notes: Columns (1)–(4) report the estimated results after replacing our measurement of the
dependent variable, replacing our measurement of the independent variable, lagging our
independent variable by one period, and adding fixed effects of industries and provinces. The
sample size in column (3) is reduced to 48,961 (independent variable lagged regression model)
tn
due to the late development of ESG in China. Wind also had no ESG rating database in 2017,
resulting in the lack of first-phase data. The regression coefficients of the independent variables
in all models are significantly negative, thereby supporting our original conclusions. *, * *, and
* * * indicate significance at the 10%, 5%, and 1% levels, respectively.
rin
ep
Pr
51
Figures
ed
Stock Trading
Volume
H2- H2-
iew
H1-
Rating Divergence Stock Returns
H2- H2-
Stock Trading
Amount
v
Figure 1 Mechanism of ESG rating divergence on stock returns.
Notes: This figure shows that ESG rating divergence affects stock returns through trading
re
volume and turnover. H1 proposes that ESG rating divergence reduces stock returns, while H2
proposes that investor expectations may be an influential channel of ESG rating divergence.
Given that the volume and amount of stock transactions can reflect investor sentiment and
indicate the possibility of future decisions (Durand et al., 2019; Groening and Kanuri, 2018;
Park and Lee, 2018), we use stock trading volume and amount as measures of investor
er
expectations to conduct our empirical tests separately.
pe
ot
tn
rin
ep
Pr
52

SSRN Id4578803

Uploaded by

Document Information

Original Description:

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 Id4578803

Uploaded by

Copyright:

Available Formats

ESG Rating Divergence, Investor Expectations, and Stock

a School of Economics and Management, Southeast University, 2 Southeast

 Corresponding author at: School of Economics and Management, Southeast

E-mail addresses: [email protected] (Hongtao Chen), [email protected]

an investor’s perspective, to explore the impact of inconsistency among ESG rating

companies. Our findings demonstrate a significant negative impact of ESG rating

returns. In addition, ESG rating divergence significantly impedes subsequent average

is particularly pronounced in non-heavy pollution companies, non-state-owned

companies, and companies with lower external attention.

Keywords: ESG rating divergence; Stock returns; Investor expectations; Information

JEL classification: C12; C33; D81; D82; G14

standard for information disclosure and interpretation related to sustainable

accurate assessment of corporate ESG performance by investors. Consequently,

1. See more details at

agencies, thereby highlighting uncertainties in their accurate measurement of

With the market’s increasing emphasis on companies’ ESG performance,

lower than the score of 84 received by Philip Morris International, a cigarette

significant discrepancies, which fueled Musk’s skepticism and criticism. ESG

ratings for Guizhou Moutai, a well-known Chinese alcohol company, exhibit

significant inconsistencies across different ESG rating agencies. Domestic rating

cognitive contradictions and profoundly impact a company’s reputation and market

ESG rating divergences can produce a series of significant consequences. First,

matters. We also observe a significant negative relationship between ESG rating

heterogeneity analyses highlight that this negative relationship is particularly

pronounced among non-polluting sector companies, non-state-owned companies, and

companies with greater investor attention.

agencies’ interpretation of ESG information. Despite the interpretation provided by

including “companies releasing ESG information → rating agencies interpreting such

information → the market receiving the rating results,” ultimately decreasing

Lastly, our findings offer implications for various stakeholders, particularly

ratings of target companies in major investment platforms or stock selection software.

When constructing an investment portfolio, investors should observe the divergences

endogenous tests and robustness checks. Section 6 concludes the paper.

ignore the impact of external agencies. Divergence in ESG ratings is a common

investors take various rating agencies’ information as reference. As the number of

of the same company becomes increasingly significant. Such divergence is driven by

a team of professional ESG analysts to meticulously analyze fundamental information

types of investment institutions, namely, normally constrained institutions, hedge funds,

refer to a company’s ESG rating and complete socially responsible investments

agencies show many differences in their measurement of the ESG performance of

ratings given by six well-known rating agencies.

2.1.2 ESG divergence

Researchers have mostly focused on the causes of rating divergence, while

perception of the company influences the measurement of specific categories (Berg et

rights. Previous studies generally converge on the increased uncertainty of corporate

substantial decline in their future ESG ratings, thereby highlighting a correlation

specific taxonomy or best-in-class portfolio can be recommended.

2.2 Hypotheses development

signaling altruistic intentions to shareholders, and enhancing the authenticity of

investment strategies, thereby resulting in poor capital market efficiency (Avramov et

hypocritical or deceptive behavior. ESG controversies may place a company under

confidence in the company’s growth prospects and leading to significant declines in

H1: ESG rating divergence reduces stock returns.

From the information asymmetry perspective, a lower level of information

widespread recognition of the meaning of financial information often reduces the

behaviors mitigate information asymmetry under different contexts. For instance,

means, such as management forecasts, press releases, analysts’ presentations,

of ESG, thus exacerbating cognitive biases, further increasing information asymmetry,

of corporate disclosure increases, providing more non-financial information and

company and its shareholders. Therefore, rating discrepancies represent an increase in

the amount of information available in the market and a reduction in information

Due to market information asymmetry, different investors hold varying