Mandatory CSR Disclosure and Information Asymmetry: Evidence from a Quasi-natural Experiment in China

Mingyi Hung* Jing Shi† Yongxiang Wang∞

February 2015

ABSTRACT

Using a quasi-natural experiment that mandates a subset of listed firms to issue corporate social responsibility (CSR) reports, this paper examines the effect of mandatory CSR disclosure on market information asymmetry in China, where we estimate information asymmetry using high-frequency trade and quote data. We find that contrary to the criticism that mandatory CSR disclosure lacks credibility and relevance in emerging markets, mandatory CSR reporting firms experience a decrease in information asymmetry subsequent to the mandate. In addition, the decrease in information asymmetry is more pronounced among firms with greater political/social risks, poorer information environments, and better CSR reporting quality. Additional analyses suggest that relative to mandatory CSR disclosure, voluntary CSR disclosure is part of a firm’s political/social strategy and has higher CSR reporting quality. However, the effect of voluntary CSR disclosure on information asymmetry is limited unless CSR reporting is widespread. * The Hong Kong University of Science and Technology, email: acmy@ust.hk † Australian National University, email: jing.shi@anu.edu.au ∞ University of Southern California, email: yongxiaw@marshall.usc.edu

Keywords: Mandatory CSR Disclosure, Information Asymmetry, Political/Social Risk, China. JEL classification: G15; M14; M48

Acknowledgments: We thank Harry DeAngelo, Raymond Fisman, Paul Gao, Ning Gong, Harrison Hong, Roger Loh, John Matsusaka, Dylan Minor, and T.J. Wong, and workshop participants at Nanjing University, Nanyang Technology University, Renmin University of China, University of Southern California, the Asian Finance Association 2013 Conference, the second Symposium on China’s Financial Market, and the 13th Annual Strategy and the Business Environment Conference at UT Austin for their insightful comments and constructive suggestions. All errors are our own.

1

Mandatory CSR Disclosure and Information Asymmetry: Evidence from a Quasi-

natural Experiment in China

1. Introduction

Corporate social responsibility (CSR) has become a key business practice, with its disclosure

now one of the most important reporting issues in global business environments (KPMG and

UNEP, 2010). Notably, a growing number of regulators worldwide begin to mandate CSR

reporting. 1 The CSR disclosure mandate, especially in emerging economies, has received

considerable attention because of the severity of their social, environmental, and governance

problems. The existing literature on CSR disclosure, however, mainly focuses on voluntary CSR

disclosure in developed markets. Thus, the effect of mandatory CSR disclosure in emerging

markets remains an unanswered question. Our study attempts to fulfill this void by examining

the impact of mandatory CSR disclosure on market information asymmetry in China.

Furthermore, by taking advantage of our setting, we aim to provide fresh insights into the

differences between voluntary and mandatory CSR disclosure.

Prior theoretical work suggests that voluntary disclosure and mandatory disclosure differ in

several important ways (Farrell and Saloner, 1985; Dye, 1990; Zhang, 2001; Fishman and

Hagerty, 2003). First, voluntary disclosure is an endogenous choice of the firm and mandatory

disclosure is an exogenous shock to the firm. Voluntary adopters self select into the disclosure

choice based on their own firm-specific factors, but mandatory adopters are forced to switch

1 Countries that mandate CSR-related disclosure include Australia, Belgium, China, Denmark, France, Malaysia,

Sweden, the U.K., and the U.S. The scope of the mandate has generally changed from the disclosure of a particular CSR issue (such as the 1986 Toxic Release Inventory in the U.S.) in earlier years to a broad-based CSR disclosure (such as the CSR mandate in Malaysia in 2007, and in China, Denmark, and Sweden in 2008) in recent periods. In 2013, the European Commission adopted a proposal for a directive enhancing the transparency of certain large companies on social and environmental matters.

2

under a ’one-size-fits-all’ regulation. Thus, the predicted relations can be different.2 Second,

voluntary disclosure allows firms to use the disclosure choice to communicate their type, but

mandatory disclosure suppresses this mechanism. Third, the value of voluntary disclosure (or the

propensity to provide voluntary disclosure) can be low when investors are not knowledgeable

enough to understand the information. In contrast, firms may benefit from the positive

information externality generated by mandatory disclosure due to the network effect. Consistent

with these arguments, a large body of empirical work examines both voluntary and mandatory

adoption of financial disclosure such as IFRS.3 This literature generally finds that the effect of

mandatory adoption tends to be weaker and depends on the underlying economics and

institutions, but mandatory adoption generates greater information externality such as

comparability (Daske et al., 2008; Li, 2010; DeFond et al., 2011; Wang, 2014). Thus, the effects

of voluntary and mandatory disclosure are distinct and it is important to examine both.

Proponents of mandatory CSR reporting argue that CSR disclosure is critical to achieve

stakeholder engagement and corporate accountability. Regulation is justified because

information is a public good and companies may not have sufficient incentives to voluntarily

disclose such information (Admati and Pfleiderer, 2000). Critics, in contrast, argue that

mandatory CSR disclosure lacks credibility and relevance, and is just another regulatory burden

on firms (Lin, 2010). Consequently, mandatory CSR reporting likely results in empty rhetoric

and biased information that may actually work against corporate accountability (Owen et al.,

2 For example, firms may provide more voluntary disclosure in response to increases in institutional ownership or

analyst following that in turn correlates with a decrease in information asymmetry, thereby leading to a negative relation between voluntary disclosure and information asymmetry. However, such an effect may not hold in a mandatory setting.

3 For example, see the review article by Hail et al. (2010, p. 363-366, sections “Evidence from voluntary IFRS adoption around the world” and “Evidence from mandatory IFRS adoption around the world”).

3

2001; Hess, 2007). Thus, it is an empirical question whether CSR mandate will increase or

decrease information asymmetry.

China offers a unique setting to test the effect of mandatory CSR disclosure on market

information asymmetry for several reasons. First, the Shanghai and Shenzhen stock exchanges

required a subset of listed Chinese firms to issue CSR reports as of the end of 2008, which

provides a quasi-natural experimental research setting that allows us to use a difference-in-

differences research design by comparing changes in the treatment group with changes in the

control group subsequent to mandatory CSR disclosure. Second, the CSR mandate in China

requires publication of a stand-alone CSR report on economic, environmental, social, and

governance performance and has implications on shareholders and various stakeholders. As the

world’s second largest economy, China is economically significant and provides rich market

depth for our empirical tests to explore cross-sectional variations in the effect of this disclosure

reform. Third, CSR-related activities are viewed by the Chinese government as a significant

force that can contribute to “building a harmonious society” because of the severity of the social

and environmental issues they address (Fisman and Wang, 2012). 4 Thus, in China CSR

performance is important for gaining support from the government, and its disclosure is

important for investors to assess political and social risks facing the firm in the future. However,

like the other emerging economies, China is commonly viewed as a country with weak legal

institutions and its disclosure regulation is subject to much skepticism due to concerns over

implementation credibility. By providing evidence from the Chinese capital markets, we

4 Examples of social and environmental issues include toxic products, labor abuse, chemical leaks, and

environmental disasters (Fisman and Wang, 2012). For example, the 2008 Great Sichuan Earthquake, which left more than 86,000 dead or missing and millions homeless, triggered a tidal wave of corporate donations and unprecedented demand for public disclosure on these activities (see ‘The Sichuan earthquake-salt in their wounds,’ Economist, May 14, 2009). Also in 2008, the melamine-laced milk scandal that sickened thousands of Chinese babies focused public attention on firms’ social responsibilities (see ‘Tainting of milk is open secret in China,’ Wall Street Journal, November 3, 2008).

4

contribute to this debate and improve our understanding of the importance of nonfinancial

disclosure in emerging markets.

We focus on market information asymmetry effects of mandatory CSR disclosure because

this is a fundamental question of disclosure. Market information asymmetry arises when some

investors possess private information about the firm’s value while other uninformed investors

only have access to public information. Disclosure essentially turns private information into

public information and may therefore reduce information asymmetry. Information asymmetry is

also an important topic on its own because a large body of literature suggests that the presence of

information asymmetry creates an adverse selection problem, deters efficient resource allocation,

reduces market liquidity, and increases a firm’s required rate of returns (Akerlof, 1970; Diamond

and Verrecchia, 1991; Healy and Palepu, 2001; Easley and O’Hara, 2004). Thus, an investigation

of the effect of mandatory CSR disclosure on information asymmetry is central to assess

potential effects of such disclosure on other economic consequences such as corporate

governance, liquidity, and the cost of capital.

We test the effect of mandatory CSR disclosure on information asymmetry using a sample of

A-share listed firms in China over the 2006 through 2010 period.5 We use high-frequency trade

and quote data to construct two proxies for information asymmetry based on microstructure

models of the bid-ask spread. Following the theoretical models of Kyle (1985) and Easley and

O’Hara (1987) and the empirical analysis of Glosten and Harris (1988), Sadka (2006), and Chan

et al. (2008), our first measure of information asymmetry captures the price impact of a trade and

our second measure captures the adverse selection component of the bid-ask spread. A higher

value indicates greater information asymmetry. The intuition behind these measures is that when

5 We exclude voluntary CSR reporting firms in our primary analysis testing the effect of mandatory CSR

disclosure on information asymmetry. We add back voluntary CSR reporting firms in additional analyses testing CSR reporting quality and the effect of voluntary disclosure on information asymmetry.

5

information asymmetry exists, informed traders submit order to profit from future price changes.

In response, market makers would widen the spread to recover their loss to the informed traders

from the uninformed investors.

Our research design combines a difference-in-differences (DID) method with a propensity-

score-matched (PSM) procedure, which we term “DID-PSM method.” The DID method

compares the changes in a firm’s information asymmetry measures for firms that are mandated

to release CSR reports (i.e., treatment firms) with the changes for firms that are not subject to the

mandate (i.e., benchmark firms). This comparison helps control for other concurrent macro-

economic shocks that may affect information asymmetry but are unrelated to the mandate. We

then further restrict the DID analysis to a PSM sample. Thus, our DID-PSM method addresses

the concern that our treatment firms are not randomly selected (i.e., they tend to be larger and

have better performance) by making non-CSR reporting firms more comparable with the

mandatory CSR reporting firms.

Our analysis shows that compared to benchmark firms not subject to the mandate, treatment

firms experience a decrease in information asymmetry subsequent to the CSR mandate. Thus,

contrary to the common criticism that mandatory CSR disclosure is simply window dressing or

greenwashing, our results suggest that mandatory CSR disclosure is informative and therefore

leads to a decrease in information asymmetry among investors in China.

To further understand our results, we perform various cross-sectional analyses to explore the

mechanisms through which mandatory CSR disclosure reduces information asymmetry. We

posit that CSR disclosure reduces information asymmetry by helping investors assess a firm’s

political or social risk.6 Since the Chinese government retains tight control over the corporate

6 Political risk generally involves execution of political power that threatens a company’s value. Social risk often

refers to challenges by stakeholders due to negative perceptions of business practices (Bekefi and Epstein, 2006).

6

sector with an eye toward maintaining social stability, firms with greater political or social risk

are more likely to incur large penalties and jeopardize their operation when they are involved in

social and environmental scandals (Calomiris et al., 2010; Chen et al., 2011). Thus, mandatory

CSR disclosure should be more valuable to the market and therefore lead to a larger reduction in

information asymmetry for firms with greater political or social risk. In addition, we predict that

firms in poorer information environments experience a greater decrease in information

asymmetry because these firms stand to benefit more from increased disclosure. Finally, all else

being equal, we predict that firms with better CSR reporting quality experience a greater

decrease in information asymmetry because the improvement in transparency, and therefore its

benefits, should be higher for these firms.

Consistent with our predictions, we find that the effect of mandatory CSR disclosure on

information asymmetry is more pronounced among treatment firms with greater political or

social risk (i.e., firms that have lower government ownership and are not politically connected),

poorer information environments (i.e., firms with fewer analyst following and smaller size), and

better CSR reporting quality (i.e., firms with more substantive disclosure content of CSR

reports). 7 Our analysis also finds that the effects of political/social risk and information

environments interact with CSR reporting quality, which in turn jointly drives the decrease of

information asymmetry.

We next expand our analysis to provide insights into the differences between voluntary and

mandatory CSR disclosure. Our findings suggest that voluntary CSR disclosure is part of firms’

We do not separate because the distinction between political and social risk is often blurred in China.

7 Our additional analysis (untabulated) finds no evidence that the decrease in information asymmetry differs across firms in provinces with high versus low legal development. This finding is consistent with the notion that the effect of mandatory CSR disclosure mainly comes from stakeholder monitoring, not legal enforcement. Consistent with this finding, our additional analysis finds that CSR reporting quality is positively associated with government ownership, but not associated with legal development.

7

political/social strategy and has higher CSR reporting quality. These findings provide support to

the notion that voluntary reporting firms self select into communicating their type and have

greater reporting incentives (Dye 1990; Zhang 2001). In addition, for firms not subject to the

CSR mandate, most firms voluntarily release CSR reports on or after the CSR mandate (i.e., late

voluntary reporting firms), and the effect of voluntary CSR disclosure on information asymmetry

is significant only for late voluntary CSR reporting firms. This finding is consistent with the

network effect and highlights the importance of externality in a disclosure setting (Farrell and

Saloner, 1985; Katz and Shapiro, 1985; Fishman and Hagerty, 2003).

Finally, our additional analysis of analyst coverage finds that treatment firms experience an

increase in analyst following subsequent to the CSR mandate. This finding suggests that

increased analyst coverage is a mechanism through which mandatory CSR disclosure reduces

information asymmetry.8 We also find that our conclusions continue to hold when we use a full

sample without PSM matching, run a placebo test, delete observations in 2008, control for

internal control reporting and time trend, and use bid-ask spread as an alternative information

asymmetry proxy.

Our findings make several contributions to the literature. First, our study adds to prior CSR

disclosure literature by being the first to document the effect of mandatory CSR disclosure on

information asymmetry. Using U.S. data, Dhaliwal et al. (2011) find that voluntary CSR

disclosure with superior social responsibility performance is associated with a reduction in firms’

cost of equity capital and improvement in analyst forecast accuracy. More related to our setting,

8 However, we find no evidence that mandatory CSR disclosure improves analysts’ earnings forecast accuracy.

One possible reason for this finding is that analyst forecasts in China, an emerging economy, are subject to greater noises or biases (Hope 2003; Ding et al., 2014). Another reason is that annual reports and CSR reports are simultaneously released in China so analysts are unable to incorporate information contained in CSR reports of the current year when making earnings forecasts of the current year. It is possible that analysts help reduce information asymmetry by incorporating the information in CSR disclosure through other communication channels such as brokerage presentations and client meetings.

8

Dhaliwal et al. (2012) use international data and find that CSR disclosure is positively associated

with analyst forecast accuracy and the relation is stronger in countries that are more stakeholder-

oriented. There are, however, important differences that distinguish our study from Dhaliwal et

al. (2012). The most important one is that we each investigate the effect of CSR reporting under

different reporting regimes. We focus on the mandatory reporting regime, whereas Dhaliwal et

al. (2012) focus on the voluntary regime. As discussed previously, the extant theoretical and

empirical work suggests that these settings are distinct and the predicted relations can be

different. Another difference is that our study examines the effect of CSR disclosure on

information asymmetry, whereas Dhaliwal et al. (2012) investigate the effect of CSR disclosure

on analyst forecast accuracy. While analyst forecasts are an important source of information for

market participants, their usefulness in emerging economies is an open question.9 Thus, evidence

based on analyst forecasts does not speak directly to the effect of disclosure on information

asymmetry. By using proxies based on high-frequency trade and quote data that are commonly

used in literature (Chan et al. 2008), we are able to more directly measure the impact of CSR

disclosure on information asymmetry between informed and uninformed investors. A third

difference is that we investigate how mandatory CSR disclosure differs from voluntary CSR

disclosure. We add to Dhaliwal et al. (2012) by showing that relatively to mandatory CSR

disclosure, voluntary CSR disclosure is associated with firms’ political/social strategy and higher

CSR reporting quality. However, the effect of voluntary CSR disclosure is limited when such

disclosure practices are not yet widely adopted.

9 Dhaliwal et al. (2012) find that the effect of voluntary CSR disclosure on forecast accuracy is generally

insignificant in their sample emerging economies. Specifically, Table 5 of Dhaliwal et al. (2012) indicates that, among the eight emerging economies examined in the study (Brazil, Chile, India, Malaysia, Mexico, Philippines, South Africa, and Thailand), the relation between CSR reporting and forecast accuracy is significant only in Brazil and Chile. While this finding is consistent with prior literature that shows that mandatory financial disclosure has little economic consequence in countries with weak institutional quality, it is also consistent with the notion that analyst forecasts are relatively noisy in such countries (Hope, 2003).

9

Second, by providing insights into the economic consequences of an important aspect of

CSR-related activities (namely, mandatory CSR disclosure), our study complements prior work

that examines the effects of CSR-related activities on stock returns, institutional investors’

portfolio choices, analyst forecast properties, and firm practices (Teoh et al., 1999; Hong and

Kacperczyk, 2009; Eccles et al., 2014). Compared to prior work, our quasi-natural experimental

setting also allows us to better identify the effect of mandatory CSR-related activities. Further,

despite the push for government-mandated CSR disclosure worldwide, there is limited evidence

on the effects of mandatory CSR disclosure.10 Our study improves our understanding of the

implications of mandatory CSR reporting by being the first to document its effect on a firm’s

information environments.

Finally, our study highlights the importance of political and social factors on the effect of

mandatory CSR disclosure. Our findings suggest that nonfinancial disclosure associated with the

CSR mandate in China reduces information asymmetry by helping investors assess political or

social risk. These results complement prior research that finds that political affiliations affect

CSR-related activities (Hong and Kostovetsky, 2012; DiGuili and Kostovetsky, 2012).

Section 2 discusses the institutional background, related literature, and our empirical

predictions. Section 3 presents our sample and data. Section 4 describes empirical results for the

average effect of mandatory CSR disclosure on information asymmetry. Section 5 reports cross-

sectional analyses and Section 6 reports analyses of voluntary CSR disclosure. Section 7 presents

additional analyses on analyst following and sensitivity tests. Section 8 concludes.

10 One exception is Ioannou and Serafeim (2014), who examine the effect of mandatory CSR reporting on

corporate disclosure practices in China, Denmark, Malaysia, and South Africa.

10

2. Institutional Background, Related Literature, and Empirical Predictions 2.1. Background on China’s CSR initiatives

While worsening working and environmental conditions in China were largely tolerated

during the rapid economic growth in the 1990s, they have taken a heavy toll in recent years and

have become an alarming issue. For example, according to a special report by China’s State

Environmental Protection Administration (SEPA) and the World Bank in 2007, the combined

health and non-health costs of outdoor air and water pollution in China amount to roughly

US$ 100 billion a year (5.8% of China’s GDP).11 In addition, due to weak legal institutions in

China, unresolved labor and environmental disputes often lead to protests and threaten social

instability. As a result of growing concerns about social and environmental problems in China

(Locke and Romis, 2007; Lin, 2010), anti-sweatshop and anti-pollution movements in the global

supply chain, and the view that CSR can contribute to “building a harmonious society” (the long-

term goal of the Chinese Communist Party), the Chinese government and stock exchange

regulators have launched an unprecedented number of CSR initiatives in recent years.

For example, the amended PRC Company Law, effective since January 2006, explicitly

refers to CSR.12 In addition, in December 2008 the Shanghai Stock Exchange (hereafter SHSE)

and the Shenzhen Stock Exchange (hereafter SZSE) issued “Notice Concerning Listed

Companies’ Preparation for 2008 Annual Reports” that mandated – for the first time – a subset

of listed firms to issue CSR reports along with annual reports starting from fiscal year 2008

(Noronha et al., 2012). The SHSE imposed the requirement on three types of its listed companies:

11 See ‘Cost of pollution in China’ by the World Bank. According to the article, the calculation of premature

deaths was removed from the final report for fear of provoking “social unrest” (see ‘750,000 a year killed by Chinese pollution,’ Financial Times, July 2, 2007).

12 Article 5 states that “In conducting business operations, a company shall comply with the laws and administrative regulations, social morality, and business morality. It shall act in good faith, accept the supervision of the government and general public, and bear social responsibilities”.

11

firms included in the SHSE “Corporate Governance Index,” financial firms, and firms with

overseas listed shares; the SZSE imposed the requirement on firms included in the SZSE 100

index. According to the relevant regulations (SZSE, Social Responsibility Instructions to Listed

Companies), the CSR report shall include, but not be limited to, the following: (1)

implementation of social responsibility activities related to employee protection, impact on the

environment, product quality, and community relationships; (2) assessment of implementation of

the Social Responsibility Instructions and reasons for the gap, if any; and (3) measures for

improvement and timetable. These requirements provide a unique research setting that allows us

to compare changes in information asymmetry for the treatment group (i.e., firms subject to the

CSR mandate) with changes in information asymmetry for the benchmark group (i.e., firms not

subject to the mandate) subsequent to 2008.

We note that our treatment firms are not randomly selected – they are constituents of key

stock exchange indexes and are typically large in size. Specifically, the SZSE 100 Index is

designed to represent the performance of top 100 A-share listed companies ranked by total

market capitalization, free-float market capitalization, and share turnover. The SHSE Corporate

Governance Index is designed to represent the listed companies with best governance practices

and is formed through voluntary application of listed companies, public opinion solicitation, and

expert assessment. Appendix A presents the membership criteria for these indexes. Because the

policy reform may not be viewed as a purely natural experiment for a standard DID research

design, we mitigate this concern by implementing a DID-PSM method that attempts to make our

treatment and benchmark groups more comparable based on the evaluation criteria of the index

membership. We caveat, however, the PSM procedure is based on observable variables. If our

treatment firms and control firms differ on unobservable dimensions, our DID-PSM method will

12

not be able to address such a concern. Thus, we provide further corroborating evidence to our

main argument using cross-sectional analyses and perform various robustness checks to mitigate

this concern.

The 2008 SHSE notice also requires companies to disclose the directors' self-assessment of

the internal control. However, this requirement mainly specifies the format guideline for internal

control reports because firms were already required to provide information on internal control by

SHSE and SZSE in 2006 and formally by the China Securities Regulation Commission in

2007.13 During our sample period, China also mandated the adoption of IFRS as of January 1,

2007. Since China mandates IFRS adoption for all listed Chinese firms in 2007, not just a subset

of listed companies, our DID design should mitigate the influence of IFRS adoption. In addition,

our use of PSM should also address the concern that mandatory IFRS adoption may affect our

treatment and control firms differently. Nonetheless, we perform a placebo test using a pseudo

adoption year of 2007 to further alleviate the concern that our treatment and control firms may

not be comparable. As discussed in Section 7, we find that our treatment and benchmark firms

exhibit a similar trend in information asymmetry prior to the CSR mandate. Thus, our results are

unlikely to be driven by other concurrent regulatory changes or time trend.

2.2. Related literature and empirical predictions

CSR reporting typically involves disclosure on a company’s economic, environmental, social,

and governance performance. CSR disclosures can help investors assess potential environmental

and social liabilities of the firm. In addition, since socially responsible firms appeal to socially

aware consumers and investors (Hong and Kacperczyk, 2009), CSR disclosure also serves as an

important way to communicate with stakeholders and investors. Consistent with the notion that

13 As reported in Section 7.2, we perform additional analysis controlling for internal control disclosure and find

our results continue to hold.

13

CSR disclosure is informative, prior studies find that CSR disclosure is associated with a

reduction in a firm’s cost of capital and analysts’ earnings forecast errors (Dhaliwal et al., 2011,

2012). However, due to data constraint, this literature generally focuses on voluntary disclosure,

where firms self select to provide information, in developed economies, where legal institutions

are generally strong. Since voluntary disclosure is an endogenous choice of the firm but

mandatory disclosure is an exogenous shock to the firm, prior literature suggests that their effects

can be different (Zhang, 2001). Further, the effect of mandatory disclosure critically depends on

the underlying economics and legal institutions (Hail et al., 2010). Thus, the informativeness of

mandatory CSR disclosure in emerging markets cannot be inferred from the prior CSR literature.

Mandatory CSR disclosure may not be informative and thus have a negligible effect on

market information asymmetry in China for at least two reasons. First, despite its impressive

economic growth, China is commonly viewed as having weak legal institutions and poor

protection of property rights, and its capital market is characterized by prevalent earnings

manipulation and high stock return synchronicity (Morck et al., 2000; Chen and Yuan, 2004).

Prior research therefore finds that the benefits of financial disclosure regulation tend to be

limited in China. For example, DeFond et al. (2012) and He et al. (2012) suggest that mandatory

IFRS adoption does not increase earnings’ usefulness in China. Second, CSR reports in China

typically are not audited and the quality of CSR disclosure tends to be low. According to a recent

report by CSR consulting firm SynTao, only 5% of the 535 CSR reports surveyed in 2009 were

independently audited and over 50% of the reports counted no more than ten pages (SynTao,

2009). Critics argue that a CSR report without a third-party inspection is no more than a

company brochure, and they urge companies to invite third-parties to supervise their CSR reports

(“China’s CSR reports need improvement,” Global Times, November 25, 2010). Not surprisingly,

14

there is much skepticism about the usefulness of mandatory CSR disclosure in China, with the

common concern being that these reports are simply window dressing (SRI, 2010).

One the other hand, there are reasons to believe that mandatory CSR disclosure may be

informative, and in turn reduce market information asymmetry, in China. First, increased

disclosure may help uninformed investors reduce their information disadvantage relative to more

informed investors (Diamond and Verrecchia, 1991). In our setting, when CSR reports are not

publically available and some investors (such as institutional investors) are better informed about

companies’ CSR activities due to lower information acquisition costs, there is high information

asymmetry in the market. Mandatory CSR disclosure makes it easier for less informed investors

to acquire information on companies’ CSR activities, thereby reducing their information

disadvantage and decreasing information asymmetry among different types of investors.14

Second, as a practical matter, CSR performance is important to gain support from

government and non-governmental organizations due to the severity of pollution and worker

safety problems in China (Fisman and Wang, 2012). Consequently, CSR disclosure is critical in

assessing a firm’s future social and political risks, and thus should help reduce information

asymmetry in the capital market. For example, coal mine accidents are common in China,

resulting in thousands of deaths each year.15 Nonfinancial disclosures such as the mortality rate

and spending on safety controls are therefore important in assessing the future prospects of

mining firms. Similarly, due to the heavy toll of worsening environmental conditions, disclosures

14 This empirical prediction is based on evidence from extant literature that increased disclosure reduces

information asymmetry. We note, however, that from a purely theoretical perspective, increased disclosure may exacerbate information asymmetry. Specifically, prior theoretical models suggest that to the extent that sophisticated investors have lower information processing costs and better ability to help interpret CSR disclosures, increased disclosure may actually increase the extent of information asymmetry between sophisticated and unsophisticated investors (Indjejkian, 1991; Kim and Verrecchia, 1994, 1997). Furthermore, a recent study by Kondor (2012) suggests that public disclosure might increase disagreement among agents in higher-order expectations when investors have different trading horizons and when they collect private information on different dimensions of the fundamentals.

15 See “Management blamed in China mine blast that kills 104,” CNN World, November 23, 2009.

15

on wastewater discharge and environmental protection projects are important in evaluating the

growth prospects of industrial firms in China. The 2010 toxic waste leakages of Zijin Mining

Group, China's leading gold and copper producer, illustrate this point. The two leakages from its

mine not only poisoned a river and devastated the surrounding area, but also resulted in a

significant decline in share price and a substantial loss of firm value. In addition, the leakages

prompted a government probe into the timing of the accident disclosures, and led to various

enforcement actions including imprisonment of several senior executives. 16 The company’s

expansion plan was also stalled due to deteriorating relations with the government. Not

surprisingly, in its 2010 sustainability report (29 pages long, in Chinese), Zijin Mining provides

extensive disclosure on both the incidents and corrective measures taken.

Last but not least, while mandatory CSR reports typically are not independently audited or

inspected, they are often subject to intense monitoring from stakeholders and can be verified

from various public sources (such as the inspection results from environmental protection

bureaus and media coverage of fraudulent business practices). Thus, stakeholder monitoring of

CSR disclosure may substitute for third-party verification and increase the credibility of the

reports.

3. Sample Selection and Descriptive Statistics

3.1. Data sources and sample selection

To select our sample, we begin with all Chinese firms listed on SHSE and SZSE from 2006

to 2010. We code 2006-2008 as the pre-adoption period and 2009-2010 as the post-adoption

period because mandatory CSR reports typically did not become publicly available until April

16 See “Zijin mining pollution draws Beijing’s Ire,” BusinessWeek, July 22, 2010.

16

2009. 17 We obtain stock trading data from GTA and financial information from its China

Security Market and Accounting Research (CSMAR) database. Our initial sample includes all

1,967 A-share (local shares) listed firms after excluding 23 financial firms and 106 firms with B-

shares (foreign shares) because they are subject to different regulatory oversight and market

trading mechanisms, respectively. We require firm-years to have necessary financial and trading

data for variables used in our analysis and exclude 107 firms that voluntarily issue CSR reports

during our sample period.18 These criteria result in a full sample of 1,516 firms (6,469 firm-

years), including 287 treatment firms (1,367 treatment firm-years) and 1,229 benchmark firms

(5,102 control firm-years).

To ensure that our control firms are reasonably comparable to our treatment firms, we

compile the benchmark sample used in our primary analysis by matching each treatment firm to

one control firm (with replacement) using a PSM approach. The PSM approach involves pairing

treatment and control firms based on similar observable characteristics (Dehejia and Wahba,

2002). We implement this procedure by first estimating a logit regression to model the

probability of being a treatment firm using the pre-shock period data. Following the guidelines of

the SZSE 100 index and SHSE Corporate Governance Index, as reported in Appendix A, the

logit model includes: (1) firm size (LogMV), share turnover (Turnover), stock returns (Return),

and return on assets (ROA), to capture market capitalization, liquidity, and performance, (2)

government ownership (Government ownership), controlling shareholders’ cash rights (Cash

17 All Chinese firms are required to have their fiscal years ended on December 31. Additional analysis finds that

our result is not sensitive to excluding the transition year, 2008. 18 We exclude voluntary CSR reporting firms in our primary sample because they confound the identification of

the treatment effect of mandatory CSR disclosure. Among the 107 voluntary reporting firms, 35 (72) firms voluntarily release CSR reports prior to the 2008 mandate (on and after the 2008 mandate). The reasons that firms may not voluntarily release CSR reports include concerns of direct costs and proprietary costs. The disclosure of CSR activities typically requires that various reporting mechanisms be put in place, which is a non-trivial task. In addition, CSR reports may attract pressure from various interested groups and allow competitors to infer operating efficiency and strategic plans of a firm.

17

flow rights), political connections (Political connection), and donation (Log(1+Donation)), to

capture governance characteristics and political/social strategies, and (3) analyst following

(Log(1+Analysts)) and information asymmetry (LogPI) prior to the mandate, to capture

information environment characteristics.19 We report variable definitions in Appendix B.

Appendix C reports the results of the logit model for the PSM procedure. Consistent with the

index membership selection criteria and implementation guidelines, we find that the likelihood

of being a treated firm is positively associated with market value of equity, share turnover, ROA,

government ownership, and analyst following. In addition, the results indicate that the extent of

information asymmetry prior to the mandate is not related to the likelihood of being a treated

firm. Next, we obtain the propensity score from the logit estimation and match each treatment

firm to the control firms using the nearest neighbor matching technique (with replacement, and

caliper set at 0.25*standard error of propensity score). This procedure results in our primary

sample, which we term “PSM sample,” of 712 firms (3,555 firm-years), including 230 treatment

firms (1,147 treatment firm years) and 482 control firms (2,408 control firm years).

3.2. Measures of information asymmetry

We use high-frequency trade and quote data to construct two proxies for information

asymmetry based on market microstructure models. These measures are more direct measures of

information asymmetry than bid-ask spreads because spread is a sum of at least three

components: information asymmetry, inventory costs, and order-processing costs that are not

necessarily positively correlated with each other (Lee et al., 1994; O’Hara, 1995; Brown and

19 Since the selection of SHSE corporate governance sector is mainly based on expert evaluation of governance

quality, and does not specify specific governance characteristics, we include ownership variables, such as cash flow rights and government ownership, that prior studies find important in explaining corporate governance outcomes in emerging economies (Claessens et al., 2002; Hung et al., 2012). In addition, we include variables such as analyst coverage and information asymmetry because the evaluation committee primarily consists of financial intermediaries (e.g., securities companies, fund management companies, and rating agencies).

18

Hillegeist, 2007). Our first proxy captures the price impact of a trade based on the theoretical

model in Kyle (1985)’s seminar paper. Specifically, Kyle (1985) proposes a linear relation

between order flows and price changes (i.e., price impact), with a higher price impact indicating

greater information asymmetry. Following Glosten and Harris (1988) and Chan et al. (2008), we

estimate the price impact measure using the following structural model:

(1)

Where P is transaction price, V is trade size, and Q is trade sign (i.e., +1 for a buyer-initiated

transaction and -1 for a seller-initiated transaction).20 is the price impact coefficient that

captures the degree of information asymmetry. Chan et al. (2008) document that for their sample

of 76 Chinese firms with A- and B- shares the mean estimate of is 9.66x10-7 Yuan per share

for the A-share market from January to December 2000. Given the small value of , we

calculate our price impact measure by taking a natural logarithm of multiplied by 10^7.

Specifically, we define our price impact measure, LogPI, as follows:

LogPI = Log (×107) (2)

Our second proxy for information asymmetry captures the adverse selection component of

the bid-ask spread, which is based on an extension of the price impact parameter as estimated in

equation (1). Following prior studies (Easley and O’Hara, 1987; Glosten and Harris, 1988; Lin et

al., 1995), we decompose the bid-ask spread to calculate its adverse selection component using

the following structural model:

(3)

20 Following prior studies (Lee and Ready, 1991; Chan et al., 2008), we use the following rule to classify trades as

buyer- or seller-initiated: (1) if a transaction occurs above the prevailing quote midpoint, we classify it as a purchase and vice versa, and (2) if a transaction occurs exactly at the quote midpoint, we sign it using the previous transaction price according to the tick test (i.e., a purchase if the sign of the last nonzero price change is positive and vice versa).

1( )t t t t t tP QV Q Q e

0 1 1 1 1 0 1( ) ( )t t t t t t t t t t tP c Q Q c QV Q V z Q z QV e

19

In this model, reflect the adverse selection component of the bid-ask spread,

where is the median order size. Sadka (2006) simplifies the calculation by using directly

as a measure of information asymmetry. We therefore follow Sadka to ease the calculation

before applying a logarithm transformation of multiplied by 10^8. Thus, we define our

adverse selection component of the bid-ask spread, LogAS, as follows:

LogAS=log(z1×108)

(4)

3.3. Descriptive statistics

Table 1 presents descriptive statistics of the variables used in our primary analysis. Panel A

of the table reports the summary statistics for the PSM sample. Panel B of the table reports the

mean values of the variables and their differences between our treatment group and benchmark

group. It shows that compared to the benchmark firms, the treatment firms on average have less

information asymmetry, are larger and less liquid, have better stock performance, and greater

state ownership during our sample period.21

Table 2 presents the correlations among variables used in our analysis. Consistent with prior

studies (Chan et al., 2008), the table shows that our two measures of information asymmetry are

highly correlated, with a correlation coefficient of 0.929. In addition, consistent with the view

that larger firms have less information asymmetry in the stock markets, we find that the

correlation is -0.167 between LogPI and LogMV and is -0.071 between LogAS and LogMV.

21 We match firms based on pre-adoption period so the difference between our treatment and benchmark firms

during the entire sample period can result from the effect of mandatory adoption. Our analysis also controls for differences in these firm characteristics by including them in the multivariate regression, thus mitigating the concern that the treatment and benchmark firms are not perfectly matched along every dimension. Our regression analysis in Table 3 indicates that the information asymmetry measure of our treatment firms is insignificantly different from that of the benchmark firms in the pre-adoption period.

*0 1 tz z QV

*V 1z

1z

20

4. Empirical Results

Our analysis regresses measures of information asymmetry on a dummy variable indicating

whether the period is post mandatory CSR reporting (Post), a dummy variable indicating

whether the firm is mandated to provide CSR reports (Treatment firms), their interaction term,

and control variables. Our baseline regression model is as follows:

LogPI/LogAS = β0+ β1(Post)+ β2(Treatment firms)+β3(Post×Treatment firms)+

βj(Controlsj) (5)

Of interest is the coefficient on the interaction term, β3, which captures the change in

information asymmetry for treatment firms relative to the change for benchmark firms

subsequent to the CSR mandate. A negative coefficient on β3 is consistent with a decrease in

information asymmetry and a positive coefficient on β3 is consistent with an increase in

information asymmetry. The coefficient on Post, β1, reflects the change in information

asymmetry for the benchmark group. The coefficient on Treatment firms, β2, represents the

difference in information asymmetry between the treatment group and the benchmark group in

the pre-period. We control for several firm characteristics that are likely correlated with

measures of information asymmetry: firm size, share turnover, stock returns, accounting

profitability, and government ownership. To mitigate the concern of correlated omitted variables,

we also estimate a full regression model with firm and year fixed effects (after suppressing the

coefficients on Post and Treatment firms).22 Appendix B reports variable definitions. We use

robust standard errors clustered by firm throughout our analyses.

Table 3 presents results of the regression analysis. Columns (1)-(2) show results using LogPI

as the dependent variable, with Column (1) reporting the results of our baseline regression model

22 We do not include Post and Treatment firms in this model because these variables are a linear combination of

year and firm fixed effects, respectively.

21

and Column (2) reporting the results of our full regression model. Columns (3)-(4) follow the

same pattern and show results using LogAS as the dependent variable. The table shows that the

coefficient on Post×Treatment firms is significantly negative at p ≤ 10% (two-tailed) throughout

various specifications, indicating that mandatory CSR reporting firms experience a decrease in

information asymmetry subsequent to the CSR mandate. 23 The decrease in information

asymmetry associated with mandatory CSR reporting is also economically significant.

Specifically, the coefficients in Columns (2) and (4), -0.136 and -0.217, indicate that relative to

the control firms, treatment firms experience a decrease of 12.7 percent and 19.5 percent in

information asymmetry, measured as LogPI and LogAS, respectively.24

As for the control variables, the coefficient on Treatment firms is insignificant at

conventional levels in Columns (1) and (3), indicating that information asymmetry of mandatory

CSR reporting firms does not differ significantly from that of the benchmark group in the pre-

period. This result suggests that the PSM matches are effective in narrowing the difference

between treatment and control firms in our variables of interest prior to the CSR mandate. We

also find that firm size (LogMV) and share turnover (Turnover) are negatively associated with

information asymmetry, consistent with our expectation that large firms and firms with higher

stock liquidity are associated with less information asymmetry. In addition, profitability (ROA) is

positively associated with information asymmetry, possibly because the incentives for private

information acquisition increase with profitability.25

23 Our additional analysis finds that 21 (one) treatment firms do not have 2008 (2009) CSR reports. Thus, we also

perform a robustness check after recoding the Post variable for treatment firm-years with missing CSR reports as zero. The analysis (untabulated) finds that our result remains qualitatively the same.

24 -12.7% = exp(-0.136)-1 and -19.5% = exp(-0.217)-1, because our dependent variable is a log-transformed variable.

25 Since loss firms are likely to have greater information asymmetry, we also perform additional analysis after further controlling for a dummy variable indicating reported loss. Consistent with this expectation, our analysis (untabulated) finds that the coefficient on the loss dummy is significantly positive. Importantly, the coefficient on Post×Treatment firms remains significantly negative with a similar magnitude of 0.13, as in Column (2) of Table 3.

22

In summary, the results of our analysis suggest that mandatory CSR reporting firms

experience a decrease in information asymmetry subsequent to the CSR mandate. Thus, contrary

to the common criticism that mandatory CSR disclosure is simply window dressing, our result

suggests that mandatory CSR disclosure in China is informative and helps improve a firm’s

information environments. This finding supports the view that social and environmental

disclosure, while lacking mature reporting framework, remains an important information source

to investors in an environment that is characterized by frequent social/environmental disasters

and large political/social penalties associated with these disasters.

5. Analyses Conditional on Political/social Risk, Information Environments, and CSR

Reporting Quality

5.1. The effect of mandatory CSR disclosure conditional on firms’ political/social risk

We propose that an important mechanism through which CSR disclosure reduces information

asymmetry in China is that investors find such disclosure useful in assessing a source of risk that

is critical for companies in emerging economies – political/social risk. Since the government

retains a tight control over the corporate sector and emphasizes social stability, firms with greater

political or social risk (e.g., those without strong political ties) not only are less likely to enjoy

favorable regulatory treatments and government support, but also are more likely to incur larger

penalties when violating labor and environmental standards (Calomiris et al., 2010; Chen et al.,

2011; Hung et al., 2012).26 Uninformed investors should therefore find CSR disclosure of firms

Finally, we note that the effects of firm characteristics are estimated imprecisely in our full regression models in Columns (2) and (4) of Table 3 because the firm fixed effects absorb most of the variation across firms. This concern is especially important when interpreting certain variables that are fairly stable over time, such as government ownership.

26 For example, Chen et al. (2011) find that compared to private companies, state-owned enterprises are punished less severely when violating government regulation. Calomiris et al. (2010, p.400) discuss similar cases.

23

with greater political or social risk to be more useful because such disclosure informs them about

what the firm has done to avoid negative CSR-related shocks. Consequently, we expect that

mandatory CSR disclosure of firms with greater political or social risk results in a greater

decrease in information asymmetry.

To test this prediction, we perform analysis conditional on a firm’s political/social risk. We

use two variables to capture political/social risk: government ownership and political

connections. The notion underlying these measures is that greater government ownership, and

stronger political connections can help firms mitigate unfavorable government intervention and

tension between the firm and its stakeholders (Ma and Parish, 2006). Following Calomiris et al.

(2010), we define a firm as politically connected if one of its top executives ever worked as a

(deputy) chief office in the city-level (and above) government, i.e., Chuji and above. We classify

a firm as having greater political/social risk if its government ownership is below the sample

firm-level median, or it is not politically connected. We then estimate our full model in Table 3

in each of these political/social risk partitions and test whether the coefficient on Post×Treatment

firms differs across the partitions. For parsimony, we only report results based on LogPI in this

and the remaining analyses. Results based on LogAS are qualitatively the same, which is not

surprising because these two measures are highly correlated as indicated in Table 2.

Table 4 summarizes the results of this analysis. Consistent with our expectation, the table

shows that the coefficient on Post×Treatment firms is more negative in the partition indicating

greater political/social risk (i.e., firms that have lower government ownership and are not

politically connected). Thus, the findings indicate that mandatory CSR reporting firms with

greater political or social risk experience a greater reduction in information asymmetry following

the CSR mandate.

24

5.2. The effect of mandatory CSR disclosure conditional on firms’ information environments

If mandatory CSR disclosure decreases information asymmetry, we expect the effects to be

more pronounced among firms in poorer information environments. This is because firms in

poorer information environments are likely to be more opaque prior to the disclosure mandate.

Thus, they are likely to experience greater improvements in transparency, which in turn results in

a larger decrease in information asymmetry after the CSR mandate.

To test this prediction, we perform analysis conditional on a firm’s information environment.

We use two variables to capture information environments: analyst following and firm size. We

classify a firm as having poorer information environment if its number of analyst following and

firm size is below the sample firm-level median. We then estimate our full model in Table 3 in

each of these information environment partitions and test whether the coefficient on

Post×Treatment firms differs across the partitions.

Table 5 summarizes the results of this analysis. Consistent with our expectation, we find that

the coefficient on Post×Treatment firms is more negative in the partition indicating poorer

information environments (i.e., firms that have a lower number of analyst following and smaller

size). Thus, the findings indicate that mandatory CSR reporting firms with poorer information

environments experience a greater reduction in information asymmetry following the CSR

mandate.

5.3. The effect of mandatory CSR disclosure conditional on CSR reporting quality

Since the effect of disclosure regulation depends on its implementation credibility, we expect

that firms with higher CSR reporting quality experience a greater reduction in information

asymmetry subsequent to the CSR mandate. To test this prediction, we first obtain CSR rating

information from RKS, a firm specializing in ranking CSR reports for listed Chinese companies.

25

RKS rates a CSR report by three categories: (1) content score, an evaluation score based on

specific CSR metrics for economic, environmental, and social performance, (2) macrocosm score,

an evaluation score based on CSR strategic effectiveness, stakeholder participation, and

information comparability, and (3) technique score, an evaluation score based on clarity,

consistency, and presentation formats. 27 While we report the results using all three CSR

disclosure scores for completeness, we are most interested in the content score because it reflects

the substance of a firm’s CSR activities and includes specific metrics such as pollution and toxic

waste indexes, product safety records, and labor protection measures.

We perform the analysis conditional on a firm’s CSR reporting quality by classifying a

treatment firm into a high CSR reporting quality subsample if its disclosure score is above the

sample median. Since the score is not applicable to control firms (because they do not have CSR

reports to be rated), we include all control firms in each partition. We then estimate our full

model in Table 3 in each of these partitions and test whether the coefficient on Post×Treatment

firms differs across the partitions.

Table 6 reports results of this analysis. Panel A of the table reports the descriptive statistics

on the content, macrocosm, and technique scores of CSR reports. Panel B reports the regression

results. We find that the coefficient on Post×Treatment firms is more negative in the subsample

of higher CSR content scores. This finding is consist with our expectation and suggests that

mandatory CSR reporting firms with more substantive CSR reporting experience a greater

reduction in information asymmetry following the CSR mandate. We do not find the coefficient

on Post×Treatment firms to differ between subsamples of firms with high and low macrocosm

27 See www.rksratings.com. The maximum content, macrocosm, and technique scores are 50, 30, and 20,

respectively. RKS provides ratings for CSR reports with fiscal years ended in 2008 or later.

26

and technique scores, suggesting that discussion of overall CSR strategies and presentation

format of CSR reports play a less important role in reducing information asymmetry.

5.4. Potential dominance of CSR reporting quality over firms’ political/social risk and

information environments

Since our analyses conditional on political/social risks and information environments do not

consider CSR reporting quality, it is possible that the effect of political/social risk or information

environment simply reflects the effect of reporting quality. To explore this issue, we examine the

coefficients on Post×Treatment firms across the subsamples that are first split by CSR content

score, then by government ownership or by analyst following.28

Table 7 reports the results of this analysis. For parsimony, we only report the coefficient on

Post×Treatment firms and its significance in a two-by-two analysis. Columns (1)-(3) report the

results partitioning the sample by CSR reporting quality and government ownership. We find

that government ownership is important in reducing information asymmetry only when CSR

reporting quality is high. In addition, CSR reporting quality is important in reducing information

asymmetry only when government ownership is low. These results suggest both low government

ownership and high reporting quality play a role in the decrease of information asymmetry

subsequent to the CSR mandate.

Columns (4)-(6) of Table 7 report the results partitioning the samples by CSR reporting

quality and analyst following. Similar to the results in Columns (1)-(3), we find that analyst

following is important in reducing information asymmetry only when CSR reporting quality is

high. In addition, CSR reporting quality is important in reducing information asymmetry only

28 We focus on government ownership and analyst following because we view these variables as our primary

measures of political/social risks and information environments. Using government ownership as a partitioning variable also yields a more balanced sample distribution across the partitions. We find our conclusions remain similar in additional analyses (untabulated) conditional on CSR reporting quality and political connections, and on CSR reporting quality and firm size.

27

when analyst following is low. These findings suggest that both low analyst coverage and high

reporting quality play a role in the decrease of information asymmetry subsequent to the CSR

mandate. Together, our results suggest that the effect of CSR reporting quality interacts with

political/social risk and information environments, which in turn jointly drives the decrease of

information asymmetry.

6. Analyses of Voluntary CSR Disclosure

6.1. Determinants and reporting quality of voluntary CSR disclosure

An important difference between a mandatory reporting setting and voluntary reporting

setting is that under a voluntary reporting regime, firms self select into the disclosure choice after

assessing their own costs and benefits (Dye, 1990; Zhang, 2001). To provide insights into this

difference, this section examines the determinants and reporting quality of voluntary CSR

disclosure.

To test the determinants of voluntary CSR disclosure, we first expand our sample by adding

back voluntary CSR reporting firms. We then perform a logit analysis regressing a dummy

variable indicating voluntary CSR reporting firms on the same set of variables used in our logit

model of Appendix C to facilitate the comparison with the selection of mandatory reporting

firms. In addition, we restrict this analysis to the pre-mandatory CSR reporting years, 2006-2008,

to avoid the confounding effect of mandatory CSR disclosure.

Panel A of Table 8 reports the results of this analysis. Column (1) shows that voluntary CSR

disclosure is positively associated with political connections, donation, analyst following,

information asymmetry, and negatively associated with government ownership. Column (2) uses

an alternative sample period, 2008, and finds that the results remain similar, except that the

28

coefficient on donation becomes insignificant. Overall, these results suggest that voluntary CSR

reporting is likely to be part of a firm’s strategies, combined with having politically connected

top executives and making larger donation, to increase its political/social power.29 In addition,

firms are motivated to voluntarily provide CSR reports in response to the demand from analysts

and to improve their information environments. In contrast, since government ownership is by

heritage, state-owned firms have less a need to voluntarily provide CSR disclosure to

communicate their political standing.

We assess the reporting quality of voluntary CSR disclosure by regressing the CSR

disclosure scores on a dummy variable indicating voluntary CSR reporting firms. In addition to

including the same firm-level control variables used in our main analysis in Table 3 and industry

fixed effects, we also include variables controlling for a firm’s provincial legal and market

development because prior studies suggest that the economic consequences of financial

disclosure depend on institutional quality. Since the RKS CSR ratings data begin with the 2008

CSR reports, we use all firm-years with 2008-2010 CSR reports in this analysis.

Panel B of Table 8 reports the results. Columns (1)-(3) present the results for the content,

macrocosm, and technique scores, respectively. We find that the coefficient on Voluntary

reporting firms is significantly positive in all columns, consistent with notion that voluntary CSR

reporting firms have greater reporting incentives than mandatory CSR reporting firms. Columns

(4)-(6) further examine whether the reporting quality of voluntary and mandatory CSR reporting

firms improves over time by adding a year variable (relative to the base year 2008) and its

interaction term with the dummy variable indicating voluntary reporting firms. We find that the

coefficient on Year is significantly positive in Column (4) but insignificant in Columns (5) and

29 In China, private non-profit organizations cannot raise money from public. Thus, donations mainly go to the

government controlled agents like the Red Cross Society of China, where the senior officials of the organization are appointed by the government.

29

(6), indicating that mandatory CSR reporting firms show an improvement in the CSR disclosure

content score over time, but no changes in the macrocosm and technique scores. These results

provide further support that the content score is more meaningful than the macrocosm and

technique scores. Furthermore, we find that the sum of the coefficients on Year and

Year×Voluntary reporting firms is insignificant in Column (4), suggesting that the CSR

reporting quality for voluntary reporting firms is fairly stable during 2008-2010.

In terms of control variables, Panel B of Table 8 finds that CSR disclosure score is positively

associated with firm size and government ownership, indicating that large firms and firms with

greater government ownership have more substantive CSR reporting. This finding is consistent

with the notion that CSR reporting quality is driven by stakeholder monitoring. In contrast, CSR

disclosure score is not associated with a firm’s provincial legal and market development. This

finding suggests that unlike mandatory financial disclosure whose effect depends on

implementation credibility via the legal system to ensure compliance with GAAP and constrain

earnings manipulation, the effect of mandatory CSR disclosure is less subject to legal

environments (likely because there is no specific measurement principles).

6.2. The effect of voluntary CSR disclosure on information asymmetry

Prior literature suggests that the value of voluntary disclosure can be low when investors are

not knowledgeable enough to understand the information (Fishman and Hagerty 2003). In

addition, firms may benefit from positive information externality generated by mandatory

disclosure due to the network effect (Farrell and Saloner, 1985; Katz and Shapiro, 1985; Wang,

2014). To shed light into this issue, this section tests the effect of voluntary CSR disclosure on

information asymmetry, both surrounding the voluntary switch and the CSR mandate.

30

We perform this test using a DID design by regressing our information asymmetry measure

(LogPI) on a variable indicating firm-years subsequent to voluntary CSR reporting

(Post_voluntary reporting). In addition, we separate voluntary reporting firms into (1) early

voluntary reporting firms (35 firms) that voluntarily release CSR reports prior to the 2008

mandate, and (2) late voluntary reporting firms (72 firms) that voluntarily release CSR reports on

or after the 2008 mandate (i.e., firms that release CSR reports for fiscal years ended in 2008 or

later). We then code two dummy variables, Post_early voluntary reporting and Post_late

voluntary reporting to indicate the periods subsequent to the voluntary switch for early voluntary

reporting firms and late voluntary reporting firms, respectively. Because this setting involves

multiple treatment groups and time periods (Wooldridge, 2007), our DID estimation includes a

full set of group effects (i.e., firm fixed effects) and a full set of time effects (i.e., year fixed

effects). In addition, we include firms not subject to the CSR mandate as our benchmark group.30

Table 9 reports the results of this analysis. Column (1) shows that the coefficient on

Post_voluntary reporting is significantly negative, indicating that voluntary CSR reporting firms

on average experience a decrease in information asymmetry subsequent to the voluntary switch.

Column (2) shows that the coefficient on Post_early voluntary reporting is insignificant, but the

coefficient on Post_late voluntary reporting is significantly negative. This finding suggests that

the effect of voluntary switch on information asymmetry becomes significant on and after the

CSR mandate.31 Overall, our results suggest that the impact of voluntary CSR disclosure on

market information asymmetry is limited unless such reporting is widespread.

30 The PSM procedure yields no matched firms, suggesting that voluntary CSR reporting firms are distinctly

different. 31 We note that the effect of early voluntary switch includes the potential spillover effect from the CSR mandate.

Additional analysis (untabulated) restricting the analysis to early voluntary reporting firms in the pre-period finds that the coefficient on Post_early voluntary reporting remains insignificant but becomes positive.

31

7. Other Additional Analyses and Sensitivity Tests

7.1. The effect of mandatory CSR disclosure on analyst following and forecast accuracy

A potential mechanism through which mandatory CSR disclosure reduces a firm’s market

information asymmetry is increased analyst coverage. Analysts are important information

intermediaries and prior literature suggests that firms with better disclosure quality attract greater

analyst following because increased disclosure reduces analysts’ information acquisition costs

(Lang and Lundholm, 1996). If our inference is correct that mandatory CSR disclosure provides

useful information to the market, we expect the releases of CSR reports to increase analyst

following, which in turn help reduce information asymmetry in the market.

We perform analysis examining the impact of mandatory CSR disclosure on analyst

following by replacing our dependent variable in Table 3 with analyst coverage, Log(1+

Analysts). Table 10 presents the results of this analysis. Panel A presents the descriptive statistics

for the additional variables and Panel B shows the regression results. Column (1) of Panel B

shows that the interaction term Post×Treatment firms is significantly positive at p ≤ 10% (two-

tailed), suggesting that treatment firms experience a greater increase in the number of analyst

following. Thus, the finding is consistent with our expectation that mandatory CSR reporting

reduces information asymmetry through increased analyst coverage.

Following Dhaliwal et al. (2012), we explore whether analysts use the information contained

in CSR disclosure to improve their forecast accuracy in our setting. Specifically, we examine

analyst earnings forecast accuracy for current-year earnings forecasts, one-year-ahead earnings

forecasts, and two-year-ahead earnings forecasts, measured as -1 multiplied by the absolute value

of analyst earnings forecast errors for year t, year t+1, and year t+2. We multiply the absolute

forecast error by -1 so a higher value indicates greater forecast accuracy. We calculate forecast

32

errors as actual earnings minus and consensus forecasted earnings, scaled by stock price at the

end of year t. For example, forecast errors for the first post-period, 2009, is calculated as actual

earnings for 2009 minus the consensus forecasted earnings, scaled by stock price at the end of

2009. We note, however, it is an open question whether the issuance of CSR reports may affect

analysts’ earnings forecast accuracy in China. One reason is that analyst forecasts in emerging

economies such as China may be subject to greater noises and biases (Hope, 2003; Ding et al.,

2014). Consistent with this reasoning, Dhaliwal et al. (2012) find that issuance of CSR reports is

largely unrelated to forecast accuracy in emerging economies in their sample. Another reason is

that CSR reports of year t cannot have any effects (and CSR reports of prior years have little

effects) on analyst current and long-term forecasts made in year t because annual reports and

CSR reports are simultaneously released in China.

Columns (2), (3), and (4) of Table 10, Panel B present the results of the impact of mandatory

CSR disclosure on forecast accuracy for current-year earnings forecasts, one-year-ahead earnings

forecasts, and two-year-ahead earnings forecasts, respectively. We find that the interaction term

Post×Treatment firms is insignificantly at the conventional levels across these three Columns,

suggesting that our treatment firms do not experience a change in analyst earnings forecast

accuracy in China. We note, however, it is possible that analysts help reduce information

asymmetry by incorporating the information in CSR disclosure through other communication

channels such as brokerage presentations and client meetings.

7.2. Robustness tests

We perform several sensitivity checks for alternative sample firms, alternative sample

periods, additional control variables, and alternative information asymmetry measures for our

33

primary results in Table 3. Table 11 reports the results. In the interest of parsimony, we do not

report the coefficients on firm-level controls in the sensitivity tests.

7.2.1. Alternative sample firms. Our primary analyses use the DID-PSM sample to mitigate the

concern that our treatment sample may not be comparable to the benchmark sample. To assess

the robustness of our results, we repeat our analysis after using the (unmatched) full sample. We

find that the coefficient on Post×Treatment firms is larger when using the full sample, perhaps

because the unmatched control firms is less comparable to the treatment firms. This finding also

suggests that the result based on the PSM sample in our primary analyses is more conservative

than the result based on the full sample.

7.2.2. Alternative sample periods. During our sample period, China also mandated the adoption

of IFRS as of January 1, 2007. Since the IFRS mandate applies to all listed firms, our

comparison of changes in information asymmetry for the treatment versus benchmark firms

mitigates the influence of possible confounding factors associated with mandatory IFRS

adoption. In addition, we do not expect IFRS to significantly affect Chinese firms’ information

asymmetry because prior studies suggest that mandatory IFRS adoption has little capital market

consequences in countries with weak legal enforcement (Daske et al., 2008; DeFond et al., 2011).

Nonetheless, to assess whether mandatory IFRS adoption may affect our treatment and

benchmark firms differentially, we perform a placebo test ending our sample period in 2008 (the

year before mandatory CSR reports are released). We then rerun our analysis by recoding the

post dummy variable from indicating the period 2009-2010 to indicating the period 2007-2008.

Unlike our result in Table 3, we find that the coefficient on Post×Treatment firms becomes

insignificant. Thus, our placebo test suggests that our treatment and benchmark firms exhibit a

34

similar trend in information asymmetry prior to the CSR mandate, which alleviates the concern

that our result may be influenced by mandatory IFRS adoption in China.

In addition, since 2008 is the peak of financial crisis and there is a lot of turbulence in global

financial markets in that year, we perform an analysis after deleting all observations in 2008. We

find that our results are qualitatively identical to those reported in Table 3.32

7.2.3. Additional controls. As noted earlier, the 2008 SHSE notice also requires companies to

disclose the directors' self-evaluation reports of the internal control, but this requirement mainly

specifies the format guideline for internal control reports because firms were already required to

provide information on internal control by SHSE and SZSE in 2006. To explore whether our

results are sensitive to controlling for internal control disclosure, we manually collect data on the

issuance of internal control reports for our sample firms. We then add an interaction term

between a dummy variable indicating firms that release internal control reports and a dummy

variable indicating years subsequent to the release of the internal control report.33 This analysis

shows that our results remain qualitatively unchanged after controlling for internal control

reporting.

Our DID estimation includes year dummies to control for year fixed effects. To explore

whether our results are sensitive to controlling for different time trends across the treatment and

benchmark groups, we further include another interaction term: Year×Treatment firms. We find

that our results are qualitatively identical to those reported in Table 3.

32 By ”qualitatively identical to those reported in Table 3,” we mean that the coefficient Post×Treatment firms is

negative and significant at p ≤ 10% (two-tailed) in Column (2) of Table 3, our full regression model using logPI as the dependent variable.

33 In additional analysis, we recode the dummy variable to indicate firms that release directors’ self evaluation of the internal control. This analysis (untabulated) finds our results remain qualitatively identical to those reported in Table 3.

35

7.2.4. Alternative information asymmetry proxy. Our primary analyses use price impact and the

adverse selection component of the bid-ask spread to capture information asymmetry because

bid-ask spread is an indirect proxy of information asymmetry and suffers from numerous

econometric and interpretation difficulties (Brown and Hillegeist, 2007). Nonetheless, we assess

the robustness of our results after repeating our analysis in Table 3 using bid-ask spread as an

alternative measure of information asymmetry. We find that our results are qualitatively identical

to those reported in Table 3.

8. Conclusion This paper examines the effect of mandatory CSR disclosure on firms’ market information

asymmetry in China. Following the market microstructure literature, we construct two measures

of information asymmetry: the price impact of a trade and the adverse selection component of

the bid-ask spread. In addition, we use a DID research design by taking advantages of the unique

regulatory setting that mandates a subset of listed Chinese firms to issue CSR reports from fiscal

year 2008. To mitigate the concern that our treatment firms are not random, we also combine the

DID design with a PSM procedure that attempts to make our treatment and benchmark groups

more comparable.

We find that contrary to common criticism that mandatory CSR disclosure lacks credibility

and is simply window dressing in emerging markets such as China, mandatory CSR reporting

firms experience a decrease in information asymmetry subsequent to the mandate, especially

among firms with greater political/social risk, poorer information environments, and better CSR

reporting quality. To shed light into the differences between mandatory and voluntary disclosure,

we also examine the determinants, reporting quality, and changes in information asymmetry for

36

voluntary CSR disclosure. Our analyses suggest that voluntary CSR disclosure is part of a firm’s

strategy to increase its political/social standing and has higher reporting quality. Further, the

effect of voluntary CSR disclosure on market information asymmetry is limited unless CSR

reporting is widespread. Overall, our study is the first to document the effect of mandatory

nonfinancial disclosure on information asymmetry. By taking advantage of our setting, we also

provide fresh insights into the determinants and economic consequences of voluntary versus

mandatory nonfinancial disclosure.

 

37

REFERENCES Admati, A., Pfleiderer, P., 2000. Forcing firms to talk: Financial disclosure regulation and

externalities. Review of Financial Studies 13, 479–515. Akerlof, G., 1970. The market for ‘lemons’: quality uncertainty and the market mechanism.

Quarterly Journal of Economics 90, 629–650. Bekefi, T., Epstein, M., 2006. Management Accounting Guideline: Integrating Social and

Political Risk into Management-decision Making. CMA-Canada and AICPA. Brown, S., Hillegeist, S., 2007. How disclosure quality affects the level of information

asymmetry. Review of Accounting Studies 12, 443–477. Calomiris, C., Fisman, R., Wang, Y., 2010. Profiting from government stakes in a command

economy: Evidence from Chinese asset sales. Journal of Financial Economics 96, 399–412. Chan, K., Menkveld, A., Yang, Z., 2008. Information asymmetry and asset prices: Evidence

from the China foreign share discount. The Journal of Finance 63, 159–196. Chaney, P.K., Faccio, M., Parsley, D., 2011. The quality of accounting information in politically

connected firms. Journal of Accounting and Economics 51, 58–76. Chen, D., Jiang, D., Liang, S., Wang, F., 2011. Selective enforcement of regulation. China

Journal of Accounting Research 4, 9–27. Chen, C.W.K., Yuan, H.Q., 2004. Earnings management and resource allocation: Evidence from

China’saccounting-based regulation of rights issues. The Accounting Review 79, 645–665. Claessens, S., Djankov, S., Fan, J., Lang, L., 2002. Disentangling the incentive and entrenchment

effects of large shareholdings. Journal of Finance 57, 2741–2771. Daske, H., Hail, L., Leuz, C., Verdi, R., 2008. Mandatory IFRS reporting around the world:

Early evidence on the economic consequences. Journal of Accounting Research 46, 1085–1142.

DeFond, M., Hu, X., Hung, M., Li, S., 2011. The impact of mandatory IFRS adoption on foreign

mutual fund ownership: the role of comparability. Journal of Accounting and Economics 51, 240–258.

DeFond, M., Gao, X., Li, O., Xia, L., 2012. Did China’s adoption of IFRS attract more foreign

institutional investment? Working paper, University of Southern California and National University of Singapore.

Dehejia, R. H., Wahba, S., 2002. Propensity score matching methods for non-experimental

causal studies. Review of Economics and Statistics 84, 151–161.

38

Dhaliwal, D., Li, O, Tsang, A., Yang, Y., 2011. Voluntary nonfinancial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting. The Accounting Review 86, 59–100.

Dhaliwal, D., Radhakrishnan, S., Tsang, A., Yang, Y., 2012. Nonfinancial disclosure and analyst

forecast accuracy: International evidence on corporate social responsibility disclosure. The Accounting Review 87, 723–759.

Diamond, D., Verrecchia, R., 1991. Disclosure, liquidity, and the cost of capital. Journal of

Finance 46, 1325–1360. DiGuili, A., Kostovetsky, L., 2012. Are Red or Blue companies more likely to go green? Politics

and corporate social responsibility. Working papers, University of Rochester.

Ding, F., Chen, M., Wu, Z., 2014. Do institutional investors use earnings forecasts from financial analysts? Evidence from China’s stock market. Emerging Markets Finance & Trade 50, 134–147.

Dye, R. A., 1990. Mandatory versus voluntary disclosures: The case of financial and real

externalities. The Accounting Review 65, 1–24. Du, X., 2012. Local protectionism and corruption lead to food safety issues. Legal Daily, July 6. Easley, D., O'Hara, M., 1987. Price, trade size, and information in securities markets. Journal of

Financial Economics 19, 69–90. Easley. D., O’Hara, M., 2004. Information and the cost of capital. The Journal of Finance 59,

1553–1583. Eccles, R. G., Ioannou, I., Serafeim, G., 2014. The impact of corporate sustainability on

organizational processes and performance. Management Science 60, 2835–2857. Fan, J., Rui, O., Zhao, M., 2008. Public governance and corporate finance: Evidence from

corruption cases. Journal of Comparative Economics 36, 343–364. Fan, J., Wong, T.J., Zhang, T., 2007. Politically connected CEOs, corporate governance, and

Post-IPO performance of China’s newly partially privatized firms. Journal of Financial Economics 84, 330–357.

Farrell, J., Saloner, G., 1985. Standardization, compatibility, and innovation. The Rand Journal

of Economics 16, 70–83. Fishman, M. J., Hagerty, K., 2003. Mandatory vs. voluntary disclosure in markets with informed

and uninformed customers. Journal of Law, Economics, & Organization 19, 45–63. Fisman R., Wang, Y.X., 2012. The mortality cost of political connections. Working paper,

Columbia University and University of Southern California.

39

Glosten, L. R., Harris, L.E., 1988. Estimating the components of the bid-ask spread. Journal of

Financial Economics 21, 123–142. Hail, L., Leuz, C., Wysocki, P., 2010. Global accounting convergence and the potential adoption

of IFRS by the U.S. (Part I): Conceptual underpinnings and economic analysis. Accounting Horizons 24, 355–394.

He, X., Wong, T.J., Young, D., 2012. Challenges for implementation of fair value accounting in

emerging markets: Evidence from IFRS adoption in China. Contemporary Accounting Research 29, 538–562.

Healy, P., Palepu, K., 2001. Information asymmetry, corporate disclosure, and the capital

markets: A review of the empirical disclosure literature. Journal of Accounting and Economics 31, 405–440.

Hess, D., 2007. Social reporting and new governance regulation: the prospects of achieving

corporate accountability through transparency. Business Ethics Quarterly 17, 453–476. Holmstrom, B., Tirole, J., 1993. Market liquidity and performance monitoring. Journal of

Political Economy 101, 678–709 Hong, H., Kacperczyk, M., 2009. The price of sin: The effects of social norms on markets.

Journal of Financial Economics 93, 15–36. Hong, H., Kostovetsky, L., 2012. Red and blue investing: Values and finance. Journal of

Financial Economics 103, 1–19. Hong, H., Kubik, J., Scheinkman, J., 2011. Financial constraints on corporate goodness.

Working paper, Princeton University and Syracuse University. Hope, O.-K. 2003. Disclosure practices, enforcement of accounting standards, and analysts'

forecast accuracy: An international study. Journal of Accounting Research 41, 235–272. Hung, M., Wong, T.J., Zhang, T., 2012. Political considerations in the decision of Chinese SOEs

to list in Hong Kong. Journal of Accounting and Economics 53, 435–449. Indjejikian, R.J., 1991. The impact of costly information interpretation on firm disclosure

decisions. Journal of Accounting Research 29, 277–301. Ioannou, I., Serafeim, G., 2014. The consequences of mandatory corporate sustainability

reporting: Evidence from four countries. Working paper. Harvard Business School. Katz, M., Shapiro, C., 1985. Network externalities, competition, and compatibility. The

American Economic Review 75, 424–440.

40

Kim, O., Verrecchia, R.E., 1994. Market liquidity and volume around earnings announcements. Journal of Accounting and Economics 17, 41–67.

Kim, O., Verrecchia, R.E., 1997. Pre-announcement and event-period private information.

Journal of Accounting and Economics 24, 395–419. Kondo, P., 2012. The more we know about the fundamental, the less we agree on the price.

Review of Economic Studies 79, 1175–1207. KPMG, 2011. KPMG International Survey of Corporate Responsibility Reporting. KPMG. KPMG, UNEP, 2010. Carrots and Sticks – Promoting Transparency and Sustainability. KPMG

and United Nations. Kyle, A. S., 1985. Continuous auctions and insider trading. Econometrica 53, 1315–1335. Lang, M., Lundholm, R., 1996. Corporate disclosure policy and analyst behavior. The

Accounting Review 71, 467–492. Lee, C. M. C., Mucklow, B., and Ready, M., 1994. Spreads, depths, and the impact of earnings information: An intraday analysis. Review of Financial Studies 6, 345–374. Lee, C., Ready, M.J., 1991. Inferring trade direction from intraday data. Journal of Finance 46,

733–746. Li, S., 2010. Does mandatory adoption of International Financial Reporting Standards in the

European Union reduce the cost of equity capital? The Accounting Review 85, 607–636. Lin, L., 2010. Corporate social responsibility in China: Window dressing or structural change?

Berkeley Journal of International Law 28, 64–100. Lin, J., Sanger, G.C., Booth, G., 1995. Trade size and components of the bid- ask spread. Review

of Financial Studies 8, 1153–1183. Lin, Y.F., Tan, G., 1999. Policy burdens, accountability, and the soft budget constraint.

American Economic Review P&P 89, 426–431. Locke, R., Romis, M., 2007. Improving work conditions in a global supply chain. MIT Sloan

Management Review 48, 54–63. Ma, D., Parish, W. L., 2006. Tocquevillian moments: Charitable contributions by Chinese

private entrepreneurs. Social Forces 85, 943–964. Margolis, J., Elfenbein, H., Walsh, J., 2007. Does it pay to be good? A meta-analysis and

redirection of the relationship between corporate social and financial performance. Working paper, Harvard Business School.

41

Morck, R., Yeung, B., Yu, W., 2000. The information content of stock markets: Why do

emerging markets have synchronous stock price movement? Journal of Financial Economics 58, 215–260.

Noronha, C., Tou, S., Cynthia, M.I., Guan, J., 2012. Corporate social responsibility reporting in

China: An overview and comparison with major trends. Corporate Social Responsibility and Environmental Management doi: 10.1002/csr.1276

O’Hara, M. 1995. Market Microstructure Theory. Blackwell Publishers. MA. Owen, D.L., Swift, T., Hunt, K., 2001. Questioning the role of stakeholder engagement in social

and ethical accounting, auditing, and reporting. Accounting Forum 25, 264–282. Sadka, R., 2006. Momentum and post-earnings-announcement drift anomalies: The role of

liquidity risk. Journal of Financial Economics 80, 309–349. SRI, 2001. Road to transparency: Status, trends, and challenges of CSR reporting in China.

Social Resources Institute, Beijing, China. SynTao, 2007. 2009. A Journey to Discover Values: A Study of Sustainability Reporting in

China. SynTao. Teoh, S. H., Welch, I., Wazzan, C. P., 1999. The effect of socially activist investment policies on

the financial markets: Evidence from the South African boycott. Journal of Business 72, 35–89.

Wang, C., 2014. Accounting standards harmonization and financial statement comparability:

Evidence from transnational information transfer. Journal of Accounting Research 54, 955–992.

Wooldridge, J., 2007. Difference-in-differences Estimation. NBER. Zhang, G. 2001. Private information production, public disclosure, and the cost of capital:

Theory and implications. Contemporary Accounting Research 18, 363–384

42

Appendix A Membership criteria for the SZSE 100 and SHSE corporate governance indexes

Excerpts from “Ground Rules for the SZSE 100 Index,” Shenzhen Securities Information Co., May 2014 1.2  SZSE 100 Index is designed to represent the performance of top 100 A-share listed companies in Shenzhen Stock Exchange ranked by total market capitalization, free-float market capitalization and turnovers.

Excerpts from “Notice issued SSE Corporate Governance Sector selection approach,” October 9, 2007 (original text in Mandarin and translated into English with highlight added). I. Basic Principles 1. Voluntary application. Voluntary application of listed companies to join the corporate governance section. 2. Fair and transparent. Based on company self-assessment, public comments, and expert assessment. II. Implementation The application materials are evaluated by securities companies, fund management companies, insurance companies, rating agencies and specialized research institutions. Annex I. SSE Corporate Governance Sector application requirements 1. The company’s controlling shareholders strictly follow state laws and regulations and

Shanghai Stock Exchange listing rules; 2. Company directors, supervisors, and senior management comply with state laws and

regulations and the Shanghai Stock Exchange Listing rules, are honest, trustworthy, and diligent;

3. The company has established a sound General Meeting, the Board of Directors, and Board of Supervisors;

4. The company has established a sound internal control system, has effectively implemented complete and timely disclosure of information, and its financial statements give a true and reliable view of business activities;

5. The company has established a sound high level management and employee incentive system; 6. The company has no major illegal activities in the most recent three years; the company and

its directors, supervisors and senior management personnel are not subject to administrative penalties imposed by China Securities Regulatory Commission and publicly condemned by the Exchange in the last thirty months.

43

Appendix B Variable definitions Variables Definitions Measures of information asymmetry LogPI A measure of information asymmetry that captures the price impact of

a trade, as estimated in Glosten and Harris (1988). The measure is defined as Log (×107), with being estimated from

where P is transaction price, V is trade

size, and Q is trade sign (i.e., +1 for a buyer-initiated transaction and -1 for a seller-initiated transaction).

LogAS A measure of information asymmetry that captures the adverse selection component of the spread decomposition, as estimated in Glosten and Harris (1988) and Lin et al. (1995). The measure is

defined as Log ( ×108), with being estimated from

Variables of interest Post A dummy variable equal to one for periods after 2008 (i.e., years 2009

and 2010). Treatment firms A dummy variable equal to one if the listed firm is mandated to issue

CSR reports starting from December 2008. Firm-level controls LogMV Natural logarithm of total market value of a firm’s equity. Turnover Annual share turnover, calculated as total number of shares traded

divided by the total number of shares outstanding for the period. Return Annual stock return.

ROA Return on assets. Government ownership Percentage of shares owned by the government.

Other variables Cash flow rights Cash flow rights of the ultimate controlling shareholders. Political connection A dummy variable equal to one if a firm's top executives ever worked

in the city-level (and above) government. Donation Amount of donations divided by total assets. Analysts The number of financial analysts following a firm. Content score An evaluation score, ranging from 0 to 50, focuses on the extent of

leadership and organizational systems in place for implementing CSR and on specific metrics for economic, environmental, and social performance.

Macrocosm score An evaluation score, ranging from 0 to 30, focuses on overall CSR strategy, comparability of CSR information, and innovativeness of CSR activities.

Technique score An evaluation score, ranging from 0 to 20, focuses on clarity, consistency, and availability of CSR information.

1( )t t t t t tP QV Q Q e

1z 1z

0 1 1 1 1 0 1( ) ( )t t t t t t t t t t tP c Q Q c QV Q V z Q z QV e

44

Appendix B, continued Legal development The legal environment index for a firm’s province, based on the

2008 National Economic Research Institute (NERI) Index of Marketization of China’s provinces.

Market development The market development index for a firm’s province, based on the 2008 National Economic Research Institute (NERI) Index of Marketization of China’s provinces.

Voluntary reporting firms A dummy variable equal to one if a firm voluntarily issues CSR reports.

Year Number of years relative to 2008.

Post_voluntary reporting A dummy variable equal to one for a firm-year subsequent to voluntary CSR reporting.

Post_early voluntary reporting

A dummy variable equal to one for a firm-year subsequent to voluntary CSR reporting for early voluntary reporting firms, defined as firms that voluntarily release CSR reports prior to the 2008 mandate.

Post_late voluntary reporting

A dummy variable equal to one for a firm-year subsequent to voluntary CSR reporting for late voluntary reporting firms, defined as firms that voluntarily release CSR reports on or after the 2008 mandate.

Forecast accuracyt, t+1, t+2 (-1) multiplied by the absolute value of analyst earnings forecast errors for year t, year t+1, and year t+2. Forecast errors for year t are calculated as actual earnings minus and consensus forecasted earnings for year t, scaled by stock price at the end of year t.

IC reporting firm A dummy variable equal to one if a firm issues an internal control report.

45

Appendix C The procedure to develop propensity-score-matching samples This table describes the propensity-score-matching approach. We implement this procedure by first estimating a logit regression to model the probability of being a treatment firm using the pre-shock period data. We then match each treatment firm to the control firms using the nearest neighbor matching technique (with replacement, and caliper set at 0.25*standard error of propensity score). The estimation result for the logit regression follows:

Variable Dep. var. = Mandatory CSR

reporting firms LogMV 1.243*** (0.168) Turnover 0.043** (0.019) Return -0.093 (0.062) ROA 2.576* (1.358) Government ownership 1.391*** (0.470) Cash flow rights -0.009 (0.007) Political connection 0.596 (0.390) Log(1+Donation) 0.010 (0.013) Log(1+Analysts) 1.558*** (0.307) LogPI 0.024 (0.116) Constant -38.911 (0.000) N 3,187 Pseudo R-squared 0.379

46

Table 1 Summary statistics, PSM sample This table provides descriptive statistics for the variables used in our analysis. See Appendix B for variable definitions. Panel A: Descriptive statistics (N=3,555) Variable Mean Std. dev. Median Min Max LogPI 0.460 1.056 0.431 -2.263 3.921 LogAS 1.503 1.244 1.547 -4.871 5.153 Post 0.411 0.492 0.000 0.000 1.000 Treatment firms 0.323 0.468 0.000 0.000 1.000 LogMV 14.900 1.145 14.846 11.415 18.972 Turnover 6.577 3.683 5.877 0.768 21.270 Return 0.800 1.184 0.654 -0.782 4.845 ROA 0.040 0.068 0.037 -0.339 0.226 Government ownership 0.222 0.231 0.166 0.000 0.971

Panel B: Mean value of treatment versus benchmark groups

Variable Treatment Group

(N=1,147) Benchmark Group

(N=2,408) Diff.

(Treatment-benchmark) LogPI 0.338 0.517 -0.179*** LogAS 1.385 1.560 -0.174*** LogMV 15.528 14.601 0.927*** Turnover 5.688 7.000 -1.312*** Return 0.830 0.785 0.045*** ROA 0.057 0.031 0.026 Government ownership 0.258 0.205 0.053***

47

Table 2 Correlation analysis, PSM sample This table reports Pearson correlation coefficients (N=3,555). See Appendix A for variable definitions. * indicates that the correlation is significant at least at the 10% level.

LogPI LogAS Post Treatment

firms LogMV Turnover Return ROA Government ownership

(1) (2) (3) (4) (5) (6) (7) (8) (9)(1) 1 (2) 0.929* 1 (3) -0.146* 0.019 1 (4) -0.079* -0.066* -0.009 1 (5) -0.167* -0.071* 0.429* 0.378* 1 (6) -0.064* -0.085* -0.157* -0.167* -0.370* 1 (7) 0.011 0.012 -0.017 0.018 0.206* -0.143* 1 (8) 0.284* 0.277* 0.006 0.180* 0.304* -0.225* 0.112* 1 (9) -0.007 -0.065* -0.353* 0.108* -0.144* 0.112* 0.054* 0.017 1

48

Table 3 The impact of mandatory CSR disclosure on information asymmetry, PSM sample This table presents the results of the impact of mandatory CSR disclosure on information asymmetry based on the PSM sample. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Dep. var. = LogPI Dep. var. = LogAS Variable (1) (2) (3) (4) Post -0.097** n.a. 0.279*** n.a. (0.045) (0.055) Treatment firms -0.060 n.a. -0.034 n.a. (0.074) (0.085) Post×Treatment firms -0.132** -0.136** -0.205*** -0.217***

(0.051) (0.053) (0.067) (0.069) LogMV -0.258*** 0.072 -0.260*** 0.241***

(0.036) (0.046) (0.042) (0.059) Turnover -0.026*** -0.044*** -0.026*** -0.032***

(0.006) (0.004) (0.007) (0.007) Return 0.012 -0.019 0.013 -0.061**

(0.014) (0.018) (0.017) (0.026) ROA 5.953*** 0.932*** 6.639*** 0.709**

(0.429) (0.269) (0.496) (0.349) Government ownership -0.284** 0.205** -0.346** 0.214

(0.121) (0.103) (0.144) (0.136) Fixed effects No Firm&Year No Firm&Year N 3,555 3,555 3,555 3,555 Adjusted R-squared 0.173 0.820 0.131 0.725

49

Table 4 The effect of mandatory CSR disclosure on information asymmetry, conditional on political/social risk This table presents the results of the impact of mandatory CSR disclosure on information asymmetry, conditional on a firm’s political/social risk. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively. Dep. var. = LogPI Government ownership Political connection

High government

ownership Low government

ownership Politically connected

Non-Politically connected

Post× Treatment firms (β1) -0.005 -0.239** 0.285 -0.159*** (0.086) (0.107) (0.198) (0.055) Test of difference in β1,

High-low

0.234*** 0.444*** LogMV 0.167** 0.002 0.059 0.072 (0.070) (0.072) (0.131) (0.047) Turnover -0.039*** -0.053*** -0.065*** -0.042*** (0.006) (0.007) (0.013) (0.005) Return -0.042 -0.025 -0.023 -0.020 (0.026) (0.031) (0.047) (0.019) ROA 0.435 0.948* 1.801 0.903*** (0.388) (0.485) (1.568) (0.273) Government ownership 1.200*** -1.204* 0.617* 0.200* (0.263) (0.651) (0.326) (0.108) Fixed effects Firm&Year N 1,769 1,786 196 3,359 Adjusted R-squared 0.853 0.835 0.874 0.816

50

Table 5 The effect of mandatory CSR disclosure on information asymmetry, conditional on information environments This table presents the results of the impact of mandatory CSR disclosure on information asymmetry, conditional on a firm’s information environment. See Appendix A for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Dep. var. = LogPI

Analyst following Firm size High analyst following Low analyst following Large firm Small firm

Post×Treatment firms (β1) 0.036 -0.267*** -0.047 -0.229*** (0.068) (0.100) (0.074) (0.085) Test of difference in β1,

High-low 0.303*** 0.182** LogMV -0.064 0.205*** 0.035 0.096

(0.051) (0.073) (0.071) (0.065) Turnover -0.054*** -0.028*** -0.041*** -0.042***

(0.006) (0.006) (0.007) (0.006) Return 0.009 -0.057** -0.005 -0.052*

(0.022) (0.029) (0.024) (0.027) ROA 2.694*** 0.227 1.709*** 0.608**

(0.524) (0.257) (0.538) (0.294) Government ownership 0.155 0.198 0.157 0.153

(0.143) (0.140) (0.170) (0.147) Fixed effects Firm&Year N 1,825 1,730 1,801 1,754 Adjusted R-squared 0.837 0.801 0.818 0.828

51

Table 6 The effect of mandatory CSR disclosure on information asymmetry, conditional on CSR reporting quality This table presents the results of the impact of mandatory CSR disclosure on information asymmetry, conditional on a firm’s CSR reporting quality. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Panel A: Descriptive statistics on additional variables Variable N Mean Std. dev. Median Min Max Content score 1,027 11.689 3.753 11.090 5.540 28.040 Macrocosm score 1,027 10.067 3.207 9.380 4.370 25.420 Technique score 1,022 6.880 1.637 6.530 4.440 16.110

Panel B: Regression analysis conditional on CSR reporting quality

Dep. var. = LogPI Content score Macrocosm score Technique score

High Low High Low High Low

Post×Treatment firms (β1) -0.209*** -0.084 -0.161** -0.112* -0.109 -0.193*** (0.079) (0.061) (0.072) (0.064) (0.067) (0.072) Test of difference in β1,

High-low -0.125** -0.049 0.084 LogMV 0.061 0.119** 0.069 0.111** 0.059 0.125**

(0.052) (0.047) (0.050) (0.048) (0.050) (0.049) Turnover -0.042*** -0.038*** -0.041*** -0.039*** -0.042*** -0.038***

(0.005) (0.004) (0.005) (0.004) (0.005) (0.004) Return -0.017 -0.034* -0.010 -0.041** -0.018 -0.032*

(0.021) (0.018) (0.020) (0.019) (0.020) (0.019) ROA 0.764*** 0.700*** 0.773*** 0.688*** 0.747*** 0.712***

(0.247) (0.252) (0.250) (0.248) (0.245) (0.254) Government ownership 0.216* 0.099 0.180 0.132 0.151 0.165

(0.115) (0.105) (0.114) (0.106) (0.113) (0.108) Fixed effects Firm&Year N 2,923 3,040 3,049 2,914 3,108 2,855 Adjusted R-squared 0.817 0.824 0.818 0.822 0.822 0.817

52

Table 7 Potential dominance of CSR reporting quality over firms’ political/social risk and information environments This table reports the analysis of mandatory CSR disclosure on information asymmetry, conditional on CSR reporting quality and government ownership, or on CSR reporting quality and analyst following. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Samples split by Coeff. on

POST×Treatment firms Government ownership Analyst following CSR reporting quality

High Low Diff. High Low Diff. (1) (2) (3) (4) (5) (6)

Content score

High -0.015

(N=1,360) -0.387***

(N=1,449) 0.372***0.019

(N=1,236) -0.680***

(N=1,573) 0.699***

Low -0.024

(N=1,547) -0.155

(N=1,607) 0.131 -0.011

(N=1,398) -0.152

(N=1,756) 0.141

Diff. 0.009 -0.232** 0.030 -0.528***

53

Table 8 Disclosure incentives and reporting quality of voluntary versus mandatory CSR reporting firms Panel A reports the analysis of the determinants of voluntary CSR disclosure. Panel B reports the analysis of CSR disclosure scores for voluntary and mandatory CSR reporting firms. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Panel A: Determinants of voluntary CSR reporting, voluntary CSR reporting firms+all other firms Dep. var. = Voluntary CSR reporting firms [2006-2008] [2008] Variable (1) (2) LogMV -0.079 0.262 (0.161) (0.263) Turnover 0.035 0.063* (0.027) (0.036) Return -0.000 -1.324 (0.079) (0.813) ROA 0.948 -0.799 (1.373) (2.193) Government ownership -1.378** -1.505** (0.542) (0.712) Cash flow rights -0.002 0.003 (0.008) (0.007) Political connection 0.949** 0.853* (0.407) (0.472) Log(1+Donation) 0.053** 0.024 (0.021) (0.025) Log(1+Analysts) 1.085*** 1.159** (0.333) (0.581) LogPI 0.281** 0.516*** (0.132) (0.166) Fixed effects Industry&Year Industry N 3,318 1,142 Pseudo R-squared 0.146 0.174

54

Table 8, continued Panel B: Regression analysis of CSR disclosure scores based on 2008-2010 CSR reports Dependent variable = Content Macrocosm Technique Content Macrocosm Technique

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

Voluntary reporting firms 1.205** 1.111*** 0.692*** 1.650*** 1.541*** 0.958***

(0.583) (0.405) (0.259) (0.619) (0.472) (0.282)

Year (β1) n.a. n.a. n.a. 0.867*** 0.167 -0.078

(0.203) (0.149) (0.077)

Year×Voluntary reporting firms (β2) n.a. n.a. n.a. -0.506 -0.436* -0.256*

(0.307) (0.234) (0.143)

β1+β2 0.361 -0.269 0.334**

LogMV 2.196*** 1.186*** 0.535*** 2.062*** 1.194*** 0.577***

(0.221) (0.161) (0.106) (0.239) (0.174) (0.114)

Turnover -0.008 -0.001 0.009 0.013 0.003 0.007

(0.054) (0.036) (0.021) (0.054) (0.036) (0.021)

Return 1.105*** -1.344*** -0.436*** 1.001*** -1.355*** -0.419***

(0.145) (0.103) (0.067) (0.140) (0.101) (0.064)

ROA -9.535** -2.216 -0.363 -8.876** -2.200 -0.524

(3.808) (2.566) (1.467) (3.784) (2.592) (1.488)

Government ownership 2.936*** 1.770** 0.905** 3.472*** 1.823** 0.812*

(1.092) (0.700) (0.436) (1.125) (0.721) (0.445)

Legal development 0.158 0.043 0.045 0.176 0.040 0.038

(0.171) (0.121) (0.077) (0.172) (0.121) (0.077)

Market development -0.040 0.060 -0.004 -0.068 0.066 0.008

(0.334) (0.227) (0.142) (0.334) (0.227) (0.142)

Fixed effects Industry

N 1,014 1,013 1,013 1,014 1,013 1,013

Adjusted R-squared 0.410 0.322 0.400 0.417 0.400 0.324

55

Table 9 The effect of voluntary CSR disclosure on information asymmetry The table reports the analysis on the effect of voluntary CSR disclosure on information asymmetry, using firms not subject to the CSR mandate as benchmark. The sample period is from 2005-2010. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively. Dep. var. = LogPI

(1) (2) Post_voluntary reporting -0.156** (0.072) Post_early voluntary reporting -0.118 (0.118) Post_late voluntary reporting -0.151* (0.086) LogMV 0.087 0.086

(0.060) (0.060) Turnover -0.006* -0.006*

(0.003) (0.003) Return 0.109*** 0.110***

(0.019) (0.019) ROA 0.639*** 0.640***

(0.148) (0.149) Government ownership -0.116 -0.115

(0.082) (0.082) Fixed effects Firm&Year N 5,279 5,279 Adjusted R-squared 0.803 0.804

56

Table 10 The effect of mandatory CSR disclosure on analyst following and forecast accuracy This table reports the analysis on the effect of mandatory CSR disclosure on analyst following and forecast accuracy. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively. Panel A: Descriptive statistics on additional variables Variable N Mean Std. dev. Median Min Max Log(1+Analysts) 3,555 1.941 1.236 1.946 0.000 4.682 Forecast accuracyt 2,732 -0.020 0.035 -0.008 -0.000 -0.228 Forecast accuracyt+1 2,697 -0.035 0.044 -0.020 -0.001 -0.262 Forecast accuracyt+2 2,406 -0.047 0.049 -0.031 -0.002 -0.270

Panel B: Regression analysis

Dependent variable = Log(1+Analysts) Forecast accuracyt Forecast accuracyt+1 Forecast accuracyt+2 (1) (2) (3) (4)

Post×Treatment firms 0.077* 0.003 -0.001 0.002 (0.046) (0.003) (0.003) (0.004) LogMV 0.505*** -0.003 0.001 0.007**

(0.038) (0.002) (0.003) (0.003) Turnover 0.018*** -0.002*** -0.002*** -0.001*

(0.004) (0.000) (0.000) (0.001) Return -0.142*** 0.004*** 0.005*** 0.009***

(0.015) (0.001) (0.001) (0.001) ROA 1.063*** 0.326*** 0.003 -0.086**

(0.262) (0.033) (0.030) (0.040) Government ownership 0.339*** 0.006 0.005 0.023***

(0.102) (0.005) (0.007) (0.008) Fixed effects Firm&Year N 3,555 2,732 2,695 2,406 Adjusted R-squared 0.869 0.467 0.419 0.479

57

Table 11 Robustness tests This table presents results of sensitivity tests. The dependent variable is LogPI. See Appendix B for variable definitions. Robust standard errors clustered by firm are reported in parentheses. *,**,*** indicate significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Coefficient on

Post×Treatment firms Obs. Adj. R2

Alternative sample firms 1. Using full sample -0.202*** 6,469 0.822 (0.047) Alternative sample periods 2. A placebo test using sample period 2006-2008 0.180 2,095 0.866 (0.131) 3. Deleting 2008 -0.129** 2,843 0.809 (0.065) Additional controls 4. Further controlling for Post_IC×IC reporting firm -0.150*** 3,555 0.820 (0.053) 5. Further controlling for Year×Treatment firms -0.180*** 3,555 0.82 (0.059) Alternative information asymmetry proxies 6. Using bid-ask spread as a proxy for information asymmetry -0.002** 3,555 0.677 (0.001) Firm-level controls Included Fixed effects Firm&Year