The Consequences of Mandatory Corporate Sustainability Reporting: Evidence from Four Countries

Transcription

1 The Consequences of Mandatory Corporate Sustainability Reporting: Evidence from Four Countries Ioannis Ioannou George Serafeim Working Paper August 20, 2014 Copyright 0211, 2012, 2014 by Ioannis Ioannou and George Serafeim Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.

2 The Consequences of Mandatory Corporate Sustainability Reporting: Evidence from Four Countries Ioannis Ioannou and George Serafeim Abstract We examine the effect of mandatory sustainability reporting on corporate disclosure practices. Specifically, we examine regulations mandating the disclosure of environmental, social, and governance information in China, Denmark, Malaysia, and South Africa using differences-in-differences estimation with propensity score matched samples. We find significant heterogeneity in corporate disclosure responses across those four countries. Relative to propensity score matched control firms, treated firms in China and South Africa increased disclosure significantly. We also find increased propensity to receive assurance to increase disclosure credibility in the case of South Africa, and increased propensity to adopt reporting guidelines to increase disclosure comparability in both China and South Africa. In contrast, treated firms in Denmark and Malaysia did not increase disclosure. Danish firms responded by embedding environmental and social factors in their supply chain management, and by signing on the United Nations Global Compact while Malaysian firms adopted reporting guidelines. We do not find any evidence that the disclosure regulations adversely affected shareholders. Instrumental variables regressions suggest that increases in disclosure driven by the regulation are associated with increases in firm value. Our results highlight the role of local context and institutional differences in how firms in different countries respond to reporting regulations. Keywords: sustainability reporting, mandatory reporting, corporate sustainability, corporate social responsibility Ioannou is an Assistant Professor of Strategic and International Management at London Business School. Serafeim is an Associate Professor of Business Administration at Harvard Business School. We are grateful to Bob Eccles, Paul Healy, Mike Krzus, Costas Markides, and seminar participants in the Business and Environment Initiative at the Harvard Business School, Boston University, Academy of Management, and at London Business School for helpful comments. All remaining errors are our own. 1

3 I. Introduction During the last two decades, reporting of sustainability information has become widespread. While less than 100 firms reported such information twenty years ago, by 2013 more than 6,000 companies around the world were issuing sustainability reports. Governments and stock exchanges have promoted sustainability reporting by adopting regulations or listing requirements that often mandate this form of disclosure. In this paper, we focus on understanding the effect of such mandatory disclosure regulations on firms nonfinancial disclosure practices and organizational processes. 1 An increasing number of regulations are emerging across countries, mandating the disclosure of environmental, social, and governance data. We collect data for four countries, China, Denmark, Malaysia, and South Africa that mandated such disclosures before 2011; as a result, we are able to collect data both before as well as after the enforcement of the regulation. In contrast to previous studies that have examined the effect of disclosure regulations on stock market variables, which are typically assumed to be reflective of changes in the underlying disclosure (Daske et al. 2008), we model actual disclosure changes. Therefore, our setting enables us to test the first order and direct effect of the disclosure regulation. As with all regulatory shocks that intend to affect certain organizational processes for firms that are covered by the regulation, the main challenge is to separate the effect of the regulation from other confounding effects. To do so, we implement propensity score matching procedures to construct control groups for the treated firms. Specifically, in the year immediately prior to the enforcement of the regulation, we match on firm size, profitability, Tobin s Q, pre-existing environmental, social, and governance (ESG) disclosure, financial leverage, and industry membership, to construct a more comparable group of control firms. This is especially important for our setting because ESG disclosure has significantly and voluntarily increased over the past ten years. Therefore, to observe increased disclosure by treated firms over time does not constitute evidence that the disclosure regulation had any impact. Moreover, because firms in different 1 The terms sustainability, environmental, social and governance (ESG), non-financial or corporate social responsibility (CSR) reporting are all been used interchangeably, to describe reports with different degrees of focus on environmental, social or corporate governance issues. 2

4 countries might have been affected either by voluntary disclosure initiatives, or major environmental or social crises that put pressure on firms to disclose more, or even disclosure regulations specifically around climate change, we present results using control firms from a set of multiple countries. Specifically, we utilize firms from the US, Japan, India, Australia, and the UK as controls and we separately report our empirical results for each group. We find that firms in China and South Africa increase ESG disclosure significantly after the regulation relative to control firms. Moreover, they significantly increase the probability of getting assurance on their ESG disclosures (result is significant only for South Africa) and they are significantly more likely to adopt the set of reporting guidelines of the Global Reporting Initiative (GRI), which represents the most widely adopted set of reporting guidelines for nonfinancial information. Therefore, we conclude that firms in these countries not only increase disclosure following the regulatory change but also, they seek to improve the reliability and comparability of these disclosures. In contrast, we do not find any increase in disclosure by firms in Denmark or Malaysia relative to the control groups. Instead, we present evidence that firms in these countries responded differently. Danish firms signed on the United Nations Global Compact (UNGC) given that complying with the Communication of Progress in the UNGC was explicitly stated in the regulation and was regarded as a sufficient action that would satisfy its disclosure requirements. Moreover, we find that Danish firms emphasize the incorporation of social and environmental criteria in their supply chain management. Because in Malaysia corporate social responsibility (CSR) is still understood as a set of philanthropic activities (e.g. community giving), we examine whether Malaysian firms increased disclosure around community spending. We find directionally consistent but insignificant results. Malaysian firms do however exhibit an increased propensity to adopt GRI guidelines after the regulation. Both Danish and South African firms exhibited relatively high ESG disclosure even before the regulation. We speculate that the subsequent differential response could be attributed to South Africa mandating integrated rather than sustainability reporting and that for firms with already high ESG disclosure, integration with financial data might be necessary for increasing disclosure even further. On the 3

5 other hand, both Chinese and Malaysian firms have almost nonexistent ESG disclosure prior to the regulation. However, while disclosure of ESG information was understood and framed in economic terms in China, it was understood in the context of philanthropy in Malaysia, therefore generating very different responses by firms across these two countries. We also estimate instrumental-variables regressions for firms in China and South Africa to understand the financial implications from the increase in disclosure. In the first stage, we estimate the effect of the disclosure regulation on ESG disclosure and, in the second stage, we examine the effect of the predicted ESG disclosure on Tobin s Q. We find that Tobin s Q exhibits a positive and significant relation with the predicted component of the ESG disclosure. The results are highly significant for China and moderately significant for South Africa. These results suggest that the effect of disclosure regulation has been valueenhancing rather than value-destroying for the firms in our sample. The results of this paper contribute to the academic literature on the effects of disclosure regulation (Armstrong et al. 2008; Daske et al. 2008; Joos and Leung 2013). However, in contrast to previous studies that examine the effect of financial disclosure regulation, here we study a fundamentally different reporting framework and a different type of information disclosure. In the sustainability setting, the target audience for the mandated disclosures is not strictly confined to a firm s investor base, as is typically the case with financial reporting. Rather, by issuing a sustainability report, firms target a wide and diverse range of nonshareholding stakeholders in addition to shareholders; as a result, we suggest that a different set of motivations is driving the decision to proceed with such disclosures compared to the set of motivations driving traditional financial reporting. In addition, this paper contributes to an emerging literature that seeks to understand the characteristics and consequences of nonfinancial disclosure (Simnett, Vanstraelen, and Chua 2009; Dhaliwal et al. 2011; Cheng, Ioannou and Serafeim 2014). Our results also contribute to a literature that examines firm noncompliance with required disclosures relating to executive compensation, internal control weaknesses, legal liabilities and environmental liabilities (Rice and Weber 2012; Barth et al. 1997; Robinson et al. 2011). Importantly, our findings have implications for regulators and policy makers who have already mandated or are considering mandating sustainability or integrated reporting. Our 4

6 findings suggest that the way and extent to which firms might respond to disclosure regulation critically depends on the local context as well as the in-depth understanding of the motivations for ESG disclosure. The remainder of the paper is organized as follows. Section II briefly reviews the history of sustainability reporting to date. Section III presents the data sources and sample. Section IV discusses the research design. Section V presents our findings and section VI concludes. II. Historical Background and Disclosure Regulations Both in the US and in Europe during the 1960s and 1970s, sustainability reporting was driven by a renewed awareness of responsibilities that remained unfulfilled by governmental institutions, and some that were directly attributable to business organizations. Early attempts with social reporting, primarily in the Netherlands and France, paved the way for the introduction of environmental reports in countries such as Germany, Austria, and Switzerland. During the 1980s, ethical investment funds in the UK and the US implemented an investment approach broadly known as negative screening excluding firms from their investment universe based on the firms social and ethical performance. Originally based on religious principles, investment decisions by such funds excluded firms operating in sin industries like alcohol and tobacco. Towards the end of the 1980s and following the 1989 Exxon Valdez disaster, the US-based Coalition for Environmentally Responsible Economies (CERES) developed The CERES/Valdez Principles on behalf of the Social Investment Forum (SIF), introducing a set of environmental reporting guidelines. In 1997, CERES and the United Nations Environment Program (UNEP) launched the Global Reporting Initiative (GRI) to develop reporting guidelines for the triple bottom line : accounting for economic, as well as environmental and social performance. The goal was to establish sustainability reporting at par with financial reporting in terms of rigor. Not surprisingly, the 1990s witnessed an explosive growth in voluntary corporate sustainability reports. In recent years, growing social (e.g., poverty, deteriorating social equality, and corruption) and environmental (e.g., climate change, water usage, and waste) global challenges have generated pressures on companies to adopt a more systematic treatment of sustainability reporting by disclosing how they are 5

7 utilizing, developing (or depleting) and, more generally, affecting human capital and natural resources. Moreover, as a result of several high-profile corporate scandals and the recent global financial crisis, there has been a general feeling of distrust regarding companies ability to self-regulate (Edelman Trust Barometer, 2009) and a belief that current company disclosures provide information mostly about past performance rather than the future prospects of the company (e.g. Kaplan and Norton, 1992; Simnett, Vanstraelen and Chua, 2009). Meanwhile, investors and information intermediaries in capital markets, such as sell-side analysts, increasingly integrate ESG data in their valuation models, creating demand for sustainability reporting (Ioannou and Serafeim, 2014). As a result, an increasing number of countries around the world are mandating the disclosure of ESG information. In the next section, we present the cases of four countries that adopted such regulations between 2007 and We proceed with our discussion by grouping together countries with high versus low levels of ESG disclosure before the regulation. The Case of Denmark and South Africa Both Denmark and South Africa are countries in which ESG reporting has been widespread before the regulation, at least among the large firms in the economy. For example, in Denmark the Center for Quality in Business Regulation estimated that approximately 1,030 out of the 1,100 companies affected by the regulation were already working on social responsibility issues. Both countries are characterized by relatively high quality corporate reporting (Leuz, Nanda and Wycosci 2003). Corporate governance and protection of outside investor rights are considered world-class in South Africa, largely due to the influence of British common law that was transferred through colonization. Corporate governance is also regarded as reasonably strong in Denmark, with an exceptionally high quality legal enforcement regime. However, while Denmark was not facing any significant social or environmental problems, South Africa had to tackle high levels of corruption, poverty, social inequality, and widespread HIV infection among its population. In Denmark, the Minister for Economic and Business Affairs introduced an Act that amended the Danish Financial Statements Act, in October of Large companies, meaning businesses that satisfy two out of the three criteria of a) total assets more than DKK 143 million, b) net revenues of DKK 286 6

8 million, and c) an average number of full-time employees of 250, were required to supplement their management s review with a report on social responsibility. Corporate social responsibility was defined in the report as voluntarily include considerations for human rights, societal, environmental and climate conditions as well as combatting corruption in their business strategy and corporate activities. It was not therefore mandated that companies adopt or implement such policies; however, if companies did not have such policies they were required to disclose this fact in their management s review. Because many of these policies were predominantly relevant to Danish companies conducting business abroad due to the absence of significant environmental and social challenges domestically emphasis was given on supply chain relations. Moreover, the amendment put emphasis on the United Nations Global Compact (for nonfinancial companies) and the United Nations Principles for Responsible Investment (for financial companies). It stated that exemptions from disclosure requirements would be granted to companies that have acceded to the principles of these two organizations and published a report on their progress as part of their accession. The amendment entered into force and applied for the financial years commencing on 1 st of January 2009 or later. In South Africa, the Johannesburg Stock Exchange (JSE) mandated the disclosure of sustainability information starting in the 2010 financial year and subsequently mandated the disclosure of integrated reporting starting in 2011, through the issuance of the King III Report on Corporate Governance in King III was preceded by the issuance of the King I Report on Corporate Governance in 1994, and by the King II Report on Corporate Governance in While the two earlier King reports did not specify mandatory requirements for companies, their guidelines were selectively or holistically adopted by the JSE as listing requirements (Eccles, Serafeim, and Armbrester 2012). Importantly, although both King I and King II discussed sustainability issues as part of corporate management, King III emphasized to a great extent the ideas of leadership, sustainability, and corporate citizenship. In contrast to Denmark s Act, that required disclosure of ESG issues in a supplementary and nonintegrated way King III stated that reporting on sustainability issues was to be interwoven with financial reporting (Eccles, Serafeim, and Armbrester 2012). Therefore, the integrated report would describe the 7

9 value creation process inside the organization, critically putting its economic performance into context. In doing so, companies would have to discuss the environment in which they operated as well as their impact on stakeholders, and the strategies for mitigating any potentially negative impacts on society. In addition, King III stated that the sustainability elements in an integrated report should be independently assured, although it recognized that such information is more complex and more difficult to audit. Finally, King III emphasized that integrated reporting was not only an end-of-year disclosure practice but it was more about encouraging or even pressuring companies to integrate ESG factors into their operations throughout the year. Consequently, in South Africa, integrated reporting required a larger investment in terms of funds, human capital, and organizational processes commitment. The JSE made integrated reporting mandatory for all listed companies on an apply or explain basis allowing companies that did not issue an integrated report to explain why this was the case. Similarly to a company in Denmark therefore, a company in South Africa could simply explain why they would not make any ESG disclosures. While in both countries companies were mandated to disclose the policies that they have in relation to a series of ESG issues as well as to report on the actions that they take to achieve the objectives of their policies, no specific guidelines were provided or standards were set to require disclosure along a specific set of metrics. The Case of China and Malaysia Companies in China as well as companies in Malaysia had very low levels of ESG reporting before the regulation. While both countries represent emerging economies and both have experienced impressive economic growth in the past few decades, they also have significant differences. Malaysia is a more open economy with less government intervention compared to China. Both the quality of the corporate governance structures and corporate reporting are considered higher compared to China. On the other hand, both social as well as environmental challenges are more severe in China due to increasing levels of social inequality, concerns about human rights violations, and acute environmental pollution. In China, the Shanghai Stock Exchange (SHSE) and the Shenzhen Stock Exchange (SZSE) mandated certain listed firms to disclose ESG information starting for financial year end in December

10 SHSE mandated reporting of ESG information for firms included in the SHSE Corporate Governance Index, firms with overseas listed shares, and financial firms. SZSE mandated ESG reporting for firms included in the Shenzhen 100 Index. In fact, in 2006 the Chinese government revised Article 5 of the Company Law requiring companies to undertake social responsibility in the course of business. In January of 2008, the State-Owned Assets Supervision and Administration Commission of the State Council released the Guide Opinion on the Social Responsibility Implementation for the State-Owned Enterprises controlled by the government. Both the reporting regulations and the previous government actions emphasized the economic benefits of Corporate Social Responsibility (CSR), how CSR could be a driver for harmonious growth, and how it could help organizational creativity, reputation, and employee engagement. Rather than emphasizing the moral obligations of companies, government officials focused on economic incentives and the positive effects of CSR initiatives. In Malaysia, the stock exchange Bursa Malaysia made ESG disclosure a listing requirement for all listed firms starting in 31 st of December This followed the Malaysian Prime Minister s speech announcing the requirement for listed companies to report on their CSR initiatives. Specifically, there is an obligation for firms to disclose a description of their CSR activities or, if they have none, to issue a statement publicly acknowledging the absence of such activities. It is important to note that CSR activities are even currently perceived and implemented by the majority of Malaysian companies as philanthropic initiatives and are rarely, if ever, a core component of a company s business model or understood and integrated in a strategic manner. Similar to the regulation adopted for Danish and South African companies, no specific guidelines were provided to require disclosure on specific metrics in China or Malaysia. A very important difference however, is that the Chinese and Malaysian disclosure regulations did not discuss the role of assurance and the responsibilities of the auditors in the sustainability context. III. Data and Sample 9

11 We collect data on reporting practices and other organizational processes around ESG issues from Bloomberg. Bloomberg is by a large margin the most widely used data provider for stock market, financial, and other corporate data. ESG data originate from company-sourced fillings including sustainability or CSR reports, annual reports, company websites, and a proprietary Bloomberg survey that requests corporate data directly from the companies. In contrast to other data providers, Bloomberg does not estimate or derive from mathematical models any of the ESG data. Rather, all data points are transparent and can be traced back to their source in a company document. Importantly for our study, Bloomberg has the widest coverage of all ESG datasets, assessing the ESG disclosure of more than 10,000 companies around the world. By comparison, MSCI covers less than 6,000 companies and Thomson Reuters fewer than 4,000. Due to this rather limited overall coverage, these other datasets have a much smaller number of companies covered in the countries that we are interested in for our purposes. We begin our data collection process by downloading the members of the Bloomberg product index, which includes 10,472 companies from around the world. This is a set of companies for which Bloomberg analysts have collected ESG data since 2005; therefore our data span the period For all these companies we also collect accounting and stock market data from Bloomberg. All data are measured in US dollars. Bloomberg calculates an ESG Disclosure score and its three sub-scores (E, S and G) to quantify a company s transparency in reporting ESG information. Environmental data relate to emissions, water, waste, energy and operational policies around environmental impact. Social data relate primarily to employees, products and impact on communities. Governance data relate to board structure and function, firm s political involvement, and executive compensation. This score is based on 100 out of 219 raw data points that Bloomberg collects, and is weighted to emphasize the most commonly disclosed fields. The weighted score is normalized to range from zero, for companies that do not disclose any ESG data, to 100 for those who disclose every data point collected. Bloomberg accounts for industry-specific disclosures by normalizing the final score based only on a selected set of fields applicable to the industry type; for example, 10

12 Total Power Generated is counted into the disclosure score of utility companies only. We use these disclosure scores as our main dependent variables of interest. We identify treated firms in each country through the criteria that each regulation stated and required for a company to be covered. Within the group of firms with available ESG data in Bloomberg, we identify 144 Chinese, 29 Danish, 43 Malaysian, and 101 South African firms. By construction, these firms are the largest firms in each economy and collectively cover most of the market capitalization in the stock exchange that are listed. Indicatively, as of the end of 2012, the aggregate market capitalization of the treated Chinese firms was $1,880 billion (50% of total market capitalization of all Chinese stock markets and 90% of total market capitalization of all treated Chinese firms). The respective numbers for treated Danish, Malaysian and South African firms was $220 (90% of total market capitalization of Danish stock exchange), $250 (60% of total market capitalization of Bursa Malaysia), and $460 billion (70% of total market capitalization of JSE). IV. Research Design To identify the effect of the disclosure regulation on firms reporting practices, we use a differences-indifferences approach whereby we track a reporting practice before and after the regulation both for the treated and the control firms. Moreover, we use propensity score matching to ensure that the treated and the control group are as comparable as possible on a number of observable characteristics. Specifically, we match in the year before the regulation on firm size (natural logarithm of total sales), profitability (ROA), leverage (total liabilities over total assets), market expectations about growth opportunities (Tobin s Q), the level of the ESG disclosure, and industry membership (financial vs. non-financial sectors). We select for each firm the closest neighbor based on the following model: Treatment i = α + β 1 Size i + β 2 ROA i + β 3 Leverage i+ β 4 Tobin s Q i+ β 5 ESG i+ µ j (1) Treatment takes the value of 1 if firm i is covered by the regulation or otherwise it takes the value of zero. µ j is an indicator variable that takes the value of one for industry j. Each logit regression is estimated separately for pairs of treated and control countries. Because firms in different countries might have been 11

13 affected by a) various voluntary disclosure initiatives, b) major environmental or social crises that put pressure on firms to disclosure more, or c) disclosure regulations specifically about climate change issues, we present results using control firms from a set of multiple countries. In particular, we utilize firms from the US, Japan, India, Australia, and the UK as controls and we separately report the empirical findings for each control group. There are 2,663, 1,863, 683, 428, and 394 potential control firms from the US, Japan, India, Australia, and the UK, respectively. Within this set of countries probably the cleanest control group is US firms, which for the years included in our sample have not been subject to mandated disclosure of any ESG metrics. In contrast, disclosures of firms in the UK might have been affected by the 2006 Companies Act broad requirements to identify and disclose environmental and social factors that could affect the business in the director s report; firms in Japan have increased disclosure of carbon emissions and other environmental information due to the 2005 Mandatory GHG accounting system; the largest firms in India were mandated in 2011 to start producing sustainability reports the following year. Overall, we choose firms from these five countries because they have the largest number of firms covered by the Bloomberg ESG database enabling us to match all of the treated firms in the countries of interest. Moreover, we present results using local firms as a control group. For China we have a large pool of Chinese control firms. For Denmark, we define as a local control group firms from the other Scandinavian countries (i.e. Norway, Finland and Sweden). For Malaysia, we use companies from Singapore, Thailand and Indonesia as controls. Finally, for South Africa we use UK companies as local controls because of the tight business connections between the two countries and the similarities in corporate governance regimes. However, we note that the local control group has some drawbacks especially in the case of China where control firms are domiciled in the same country as the treated firms. Past research has shown that corporate social responsibility practices, such as reporting on those practices, are implemented because of institutional pressures and peer effects, where competitors mimic what other companies are doing (Misani 2010). Therefore, even if certain Chinese companies are not covered by the regulation, they might still increase disclosure if the Chinese firms that are covered by the regulation (i.e. their competitors or companies in the same peer group) increase their disclosure. This is because customers, 12

14 employees, investors, and local communities now form expectations from companies to be transparent about its environmental and social impact. Table 1 presents summary statistics pertaining to the matching algorithm. Specifically, we present the average level of each covariate of interest for the treated firms in the year of matching (i.e. the year prior to the first year of the regulation) and moreover, we tabulate the average difference in means between treated and a) unmatched (Δ No Match) and b) matched control (Δ Match) group as well as the corresponding t-statistic. We match treated with control firms in the year prior to regulation as follows: 2009 for South Africa, 2008 for Denmark, 2007 for China and 2006 for Malaysia. Because all Malaysian firms have zero ESG disclosure prior to the regulation, we do not use this variable in the logit model for this country. Instead, we restrict all candidate control firms to be drawn from a pool of firms that also have zero ESG disclosure in Chinese firms have low ESG disclosure in the year of matching while both South African and Danish firms have considerably higher levels of disclosure. Indicatively, we note that among all firms in the Bloomberg sample, the disclosure levels of the treated South African and Danish firms rank at the top quartile in their respective year of matching, given that the 75 percentile of all companies has an ESG score of approximately 21. The statistics of Table 1 suggest that the matching procedure works reasonably well with mean differences for many covariates being significant across the treated and unmatched control group but insignificant between the treated and the matched control group. In cases where matched and treated group are still significantly different, the differences are smaller suggesting that the matching process increases the similarity on the observable characteristics between the two groups actually used in the empirical analysis. To identify the effect of the disclosure regulation on ESG reporting we estimate the following model through ordinary least squares: Reporting it = α i + µ t + γ jt + β 1 Treatment i x Mandate t + β 2 Size it + β 3 Leverage i (2) Reporting it is the disclosure or other reporting practice for firm i in year t. In particular, we explore the impact of regulation on a) a composite index of ESG disclosure as well as individual indices for b) environmental, c) social and d) governance disclosure. Mandate t is an indicator variable capturing whether 13

15 in year t the regulation mandates disclosure of ESG information and zero otherwise. From equation (1), Treatment i takes the value of 1 if firm i is covered by the regulation and zero otherwise. To mitigate the concern of correlated omitted variables, we include firm fixed effects (α i), year fixed effects (µ t) and yearindustry paired fixed effects (γ jt, where we categorize all industries in our sample into two groups: financial and non-financial industries). The average treatment effect is the estimated β 1 on the interaction term Treatment i*mandate t, which captures the change in disclosure for treated firms after the regulation relative to the change for control firms. A positive coefficient on β 1 is therefore consistent with an increase in the disclosure variable following the regulation. We control for key time-varying firm characteristics that are likely to be correlated with disclosure levels: firm size and leverage. 2 Finally, we use robust standard errors clustered at the firm and year level throughout our analysis to mitigate serial correlation within a firm across years and across firms within a year. V. Results Disclosure Changes Table 2 presents the estimated coefficient β 1 and its statistical significance based on the regression analysis. Specifically, each column of table 2 estimates equation (2) using a different set of control firms each time (i.e. local firms and firms from the US, Japan, India, Australia, and the UK). For each of the four countries of interest, we report results for four distinct measures of disclosure as the dependent variables: the composite ESG disclosure score and its three components E, S, and G. The findings show that the regulation has a highly significant positive effect on treated firms in South Africa suggesting that these firms increased their level of ESG disclosure following the regulation compared to the control group. In China, we also find a highly significant increase in disclosure by treated firms compared to control firms from any of the other countries except China itself. We speculate that this 2 We also included other time-varying firm characteristics such as profitability, ownership concentration, and stock return volatility but none of the variables was significant. All results reported here remain unchanged. 14

16 lack of significance when the control group includes other Chinese firms may potentially be attributed to spillover effects of regulation. Past literature has shown that sustainability practices and reporting diffuse because of institutional pressures and mimetic behavior (Misani 2010). In this setting, it appears as if when a competitor issues a sustainability report, other companies follow, irrespective of whether they are directly affected by the regulation. The empirical findings are consistent with such spillover effects. In fact, the Chinese control firms in our sample increased ESG disclosure significantly after the regulation: we propensity score match these Chinese control firms with US, Japanese, Indian, Australian, and UK firms and we find that the Chinese control firms increase disclosure significantly relative to the control groups after the regulation. We also note that while the effect is highly significant for all disclosure measures in South Africa, for China the most consistently significant effect is on social disclosures (although environmental and governance disclosure obtain significance in several of the specifications as well). The results of table 2 also indicate that regulation in Denmark and Malaysia did not significantly affect disclosures by treated firms relative to any control group: across almost all specifications the estimated coefficient β 1 does not obtain significance. However, these findings do not necessarily suggest that the regulation had no impact at all. In fact, treated firms might have responded in different ways; an issue that we empirically explore in subsequent analysis. Interestingly, we note that firms in both South Africa as well as Denmark exhibit relatively high ESG disclosure even before the regulation. Accordingly, we speculate that their subsequent differential response to regulation (i.e. significant increase in South Africa and no effect in Denmark) could be attributed to the fact that the South African regulation actually mandated integrated rather than sustainability reporting. For countries with already high ESG disclosure, integration with financial data (rather than separate disclosure) might be necessary for subsequently increasing disclosure even further. We explore this explanation in two ways. First, while sustainability reporting was a listing requirement on an apply or explain basis in 2010, integrated reporting was for financial year ending on or after March of Therefore, the disclosure regulation in 2010 for South Africa is similar with the disclosure regulation 15

17 in Denmark. We find that disclosure increased significantly in 2011 and 2012 but not in 2010 consistent with integrated reporting being a more effective mechanism to increase disclosure in an environment with already high levels of ESG disclosure. Second, we identify 29 firms that had voluntarily issued integrated reports before These firms should be affected less by the mandate to issue integrated reports since they are already practicing integrated reporting. Indeed, we find that these firms increased disclosure after the regulation significantly less than the rest of the South African firms, consistent with integrated reporting as opposed to sustainability reporting being the catalyst for the increased disclosure in South Africa. Table 3 is similar to table 2 in that it presents the estimated coefficient β 1 based on estimations of equation (2) while the key difference is that we distinguish between treated firms with high versus low disclosure scores (i.e. above and below the median level) in the year of matching. Accordingly, the models in the first panel for each country are estimated based on a subsample of high disclosure treated firms and their corresponding control group, whereas the models in the second panel are estimated on a subsample of low disclosure firms and their control firms. Since all Malaysian firms have zero ESG disclosure prior to the regulation, we cannot perform this analysis for Malaysia. Our findings reveal significant differences between high and low disclosure firms with regards to responding to the regulation for firms in South Africa and China. In South Africa, firms that had higher levels of environmental or social disclosure increased their environmental or social disclosure by more compared to firms with low levels of disclosure. In contrast, firms that had higher levels of governance disclosure increased their governance disclosure by less. This asymmetry is consistent with the disclosure regulation generating differential responses across different types of information, even within the nonfinancial information category. Firms that were laggards in terms of governance disclosure increased their disclosure significantly more and, after the regulation, reached levels of governance disclosure similar to the leaders. In contrast, the leaders in environmental and social disclosure further widened the gap after the disclosure regulation. This is consistent with governance disclosures being less costly to obtain and make available compared to environmental or social information. For example, while board or compensation level information is more readily available, information around environmental impact, or 16

18 employee metrics are more difficult to obtain, aggregate and release. A similar asymmetry is observed in the case of China where firms with low governance disclosure increased governance disclosure relatively more, while firms with high environmental or social disclosure increased environmental or social disclosure more. Reliability and Comparability In table 4, we explore whether in addition to having an impact on disclosure practices (or perhaps as a substitute for), the regulation also has an impact on firms propensity to seek to improve the reliability and comparability of their ESG disclosures. Specifically, in the models of table 4 we replace the dependent variables on disclosure with two new variables capturing a) whether firms receive assurance on their ESG data and b) whether they adopt the reporting guidelines of the GRI, which represents the most widely adopted set of reporting guidelines for nonfinancial information. Accounting information that is assured is considered to be more reliable because of lower noise and/or lower bias. Assurance can decrease both the probability of unintentional errors in calculations (by double-checking the numbers and underlying assumptions) and the likelihood of intentional misreporting because of managerial incentives. We estimate linear probability models because of the large number of fixed effects in our models. We find that treated firms in South Africa are significantly more likely to receive assurance or adopt the GRI guidelines compared to control groups of firms from any other country. Moreover, we find that treated Chinese firms are significantly more likely to adopt the reporting guidelines of the GRI following the regulation. These results therefore, seem to suggest that in South Africa and China, firms not only increased disclosure but also sought to increase its reliability and comparability. On the other hand, and perhaps consistent with the finding of no changes in disclosure, we find no evidence that firms in Denmark were more likely to seek assurance or adopt the GRI following the regulation. We do find however, some evidence that Malaysian firms are more probable to adopt GRI guidelines after the regulation. However, adoption of GRI guidelines was already relatively higher in Denmark with 15% adopting them in the year before the regulation and nonexistent in Malaysia before the regulation. 17

19 What happened in Denmark and Malaysia? As we already mentioned the fact that Danish and Malaysian firms did not increase significantly their levels of disclosure or supplemented disclosure with assurance or by following the GRI guidelines (in the case of Denmark), does not necessarily imply that regulation was entirely ineffective. Instead, it is conceivable that due to idiosyncrasies of the regulation, or the local context itself, firms in these countries responded differently. For example, in Denmark, the regulation explicitly stated that compliance with the Communication of Progress in the UNGC was considered as sufficient action for satisfying the regulation s disclosure requirements. Moreover, and as discussed earlier, many of the sustainability policies explicitly referred to in the regulation were predominantly relevant to Danish companies conducting business abroad due to the absence of significant environmental and social challenges domestically. Accordingly, Danish firms may have put more emphasis on supply chain relations and/or becoming signatories of the UN Global Compact, rather than increasing disclosure levels or assurance. We directly explore whether these alternative responses indeed materialized in Denmark. Specifically, in table 5, we report the estimated coefficient β 1 based on estimations of equation (2) with three alternative dependent variables that capture a) whether a firm is a signatory of the UN Global Compact, b) whether a firm is using environmental criteria in supply chain management and c) whether a firm is using social criteria in supply chain management. All three data items are sourced from Bloomberg. We find a highly significant and positive effect of the regulation on the propensity of Danish firms to become signatories of the UN Global Compact, as expected. Also, we find that Danish firms were significantly more likely to adopt social criteria in supply chain management compared to most of the control groups and moderately more likely to adopt environmental criteria. Similarly, we suggest that regulation may have generated different responses by firms in Malaysia; a country in which CSR activities are perceived and implemented by the majority of Malaysian companies as philanthropic initiatives and rarely constitute a core component of a company s business model or are 18

20 understood and integrated in a strategic manner (UNICEF 2009). 3 In such a context, it is likely that affected Malaysian firms responded to the regulation by increasing their philanthropic donations. In table 6, we present estimation results for models in which the dependent variable is an indicator variable that takes the value of one if the firm discloses the amount of funds spent in community projects. However, we are unable to confirm our prediction: we find no significant results across all of the specifications. This might be because Malaysian firms might disclose nonmonetary measures of their community involvement (housing services, education, hospitals etc.), which we are unable to measure systematically. Financial Effects Given all of the above findings, and in order to gain a more in-depth understanding of how mandatory corporate sustainability reporting is affecting companies, we also estimate instrumental variable (IV) specifications to trace the financial implications from increased disclosure. The results of table 1 for ESG disclosure in South Africa and China essentially constitute the first stage of our IV approach and confirm that Treatment i*mandate t is a valid instrument (i.e. highly significant). Accordingly, in the second stage, we instrument for ESG disclosure using this interaction term and report results for the impact of increased ESG disclosure on Tobin s Q, the dependent variable of interest. Similarly for Denmark, based on the results of table 5, we construct a composite index as the sum of UN Global Compact Signatory, Environmental criteria in supply chain management and Social criteria in supply chain management. This index ranges from zero to three. We instrument for this index using the interaction term Treatment i*mandate t and in the second stage, we estimate the impact of this instrumented index on Tobin s Q. Because we are unable to find any effect of the regulation on Malaysian firms, we do not conduct the analysis since the instrument is not relevant and as a result not valid. We suggest that if the disclosure regulation forced firms to disclose proprietary information then we would expect to find a negative effect on firm valuation. Moreover, a negative effect on valuation is 3 For more information about CSR in Malaysia see 19

21 expected if these disclosures were costly for the firm in terms of revealing bad news or direct costs in the form of information systems, personnel, and management time. In general, if the disclosure regulation forced firms to adopt organizational processes that are costly and with little benefit for the company then the disclosure regulation would negatively affect the valuation of a firm (Eccles, Ioannou and Serafeim 2014). In contrast, we expect a positive effect on valuation if increased disclosure brought benefits either in terms of improving a firm s access to finance (Cheng, Ioannou and Serafeim 2014), or by forcing the firm to a) discover new risks and opportunities thereby improving process efficiency that leads to cost savings, b) manage better reputation risk thereby improving brand value, or c) identify market opportunities and as a result growing sales (Eccles, Ioannou and Serafeim 2014). Consistent with this argument, we note that as companies change their disclosure level they might also be changing other internal management practices. For example, the disclosure regulation and/or other ESG regulations that become effective around the same time might be forcing them to implement initiatives to reduce their carbon emissions, increase employee engagement and reduce turnover, change suppliers, invest in product quality and safety procedures etc. Therefore, our tests are capturing financial effects both from the increased level of disclosure and any internal changes that the management is making. 4 Table 7 presents the results of this analysis across the different control groups. We find significant results for China and South Africa indicating an increase in Tobin s Q for firms that increased disclosure relative to the control group following the regulation. The only non-significant results are generated when the control group includes Indian firms. We note that Indian firms are potentially the most problematic control group, given that in 2011 it was announced that starting from 2012 the largest Indian firms would have to issue sustainability reports. In other words, our findings suggest that by increasing disclosure for treated firms and potentially affecting ESG management practices, the disclosure regulation generated longrun benefits for companies across these countries. We find no such effect for firms in Denmark using the 4 We also note that if some companies implement internal changes in response to the regulation but do not increase disclosure then our results might be biased. However, it seems unlikely that a disclosure regulation would force firms to change management practices and at the same time leave disclosure practices unaffected. 20

22 composite index of environmental and social supply chain management and signing on the UN Global Compact. Table 7 panel B shows results from IV regressions where the first stage is each sub-component of the overall ESG disclosure index. We estimate these models to understand which of these subcomponents may be driving the overall effect on Tobin s Q. We run separate instrumental variable models for each subcomponent. For South Africa we find that increased environmental, social and governance disclosures due to the regulation are associated with increases in Tobin s Q (except when the control group includes Indian firms). The results are similar for China for all the specifications where the instrument in the first stage is relevant. VI. Conclusion In this paper we investigate the effect of disclosure regulations that mandate sustainability reporting on firms disclosure practices and other organizational processes. We collect data for firms in China, Denmark, Malaysia, and South Africa. We find that the disclosure regulations had different effects across the countries in our sample. Firms in China and South Africa countries characterized by severe social and environmental challenges increased disclosure significantly compared to control firms. Moreover, they supplemented the increased disclosure with efforts to increase the comparability and credibility of the disclosed information. In contrast, firms in Denmark and Malaysia countries characterized by relatively less severe social and environmental challenges exhibited no increase in disclosure relative to control firms. Danish firms signed on the United Nations Global Compact and concentrated on supply chain management while Malaysian firms adopted reporting guidelines. The economic effect of the disclosure regulations appear to be positive. Instrumenting the increases in disclosure because of the regulation reveals a positive correlation between instrumented disclosure and Tobin s Q. The results are robust across several specifications both in China and South Africa. While the regulation might have imposed costs on some firms, our estimates suggest that on average, the effect of the regulation on companies has been value-enhancing rather than value-destroying. 21

23 Several opportunities exist for future research. First, an increasing number of countries are adopting disclosure regulations similar to the ones considered here. Because firms respond differently across countries, it is important to evaluate how companies respond differently across distinct contexts. Second, more research is needed in order to understand how companies change resource allocations and investment decisions as a response to changes in disclosure regulations. Finally, a fruitful avenue for future research is investigating changes in the demand for non-financial information. While the disclosure regulations in some cases increase the supply of such information, we still know little about how they affect the demand across different stakeholders. 22

24 References Barth, M. E., M.F. McNichols and G. P. Wilson Factors influencing firms' disclosures about environmental liabilities. Review of Accounting Studies 2: Cheng, Beiting, Ioannis Ioannou, and George Serafeim. "Corporate Social Responsibility and Access to Finance." Strategic Management Journal 35, no. 1 (2014): Eccles, Robert G., Ioannis Ioannou, and George Serafeim. "The Impact of Corporate Sustainability on Organizational Processes and Performance." Management Science (forthcoming). Eccles, Robert G., George Serafeim, and Pippa Armbrester. "Integrated Reporting in South Africa." Harvard Business School Case , September Ioannou, Ioannis, and George Serafeim. "The Impact of Corporate Social Responsibility on Investment Recommendations: Analysts' Perceptions and Shifting Institutional Logics." Strategic Management Journal (forthcoming). Philip P. M. Joos and Edith Leung Investor Perceptions of Potential IFRS Adoption in the United States. The Accounting Review: March 2013, Vol. 88, No. 2, pp Misani, Nicola The Convergence of Corporate Social Responsibility Practices (February 6, 2009). Management Research Review, Vol. 33, No. 7, pp Rice, S. C. and Weber, D. P How Effective Is Internal Control Reporting under SOX 404? Determinants of the (Non-)Disclosure of Existing Material Weaknesses. Journal of Accounting Research, 50: Simnett, R., A. Vanstraelen, and W.F. Chua Assurance on Sustainability Reports: An International Comparison, Accounting Review, 84(3): UNICEF Corporate Social Responsibility Practices in Malaysia. 23

25 Table 1: Mean of covariates used in the propensity-score matching and differences between treated and a) unmatched and b) matched control group China Denmark Malaysia Treat Local US Japan India Australia UK Δ No Δ Δ No Δ Δ No Δ Δ No Δ Δ No Δ Δ No Δ Match Match Match Match Match Match Match Match Match Match Match Match t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat t-stat ESG Disclosure Sales Leverage ROA Tobins Q ESG Sales Leverage ROA Tobins Q ESG Sales Leverage

26 South Africa ROA Tobins Q ESG Sales Leverage ROA Tobins Q This table tabulates the average level of each variable for the treated group that is included in a logit model as described by equation (1) in the text. The logit is estimated separately for each treated-control country pair in the year before the first year of the regulation. Δ No Match is the difference in the means between the treated group and the potential control group while Δ Match is the difference in the means between the treated group and the propensity-score matched control group based on propensities estimated from the logit model. ESG is the ESG disclosure score from Bloomberg. Sales is the natural logarithm of firm sales. Leverage is total liabilities over total assets. ROA is net income plus net interest expense over total assets. Tobin s Q is the market value of a firm s assets over total book value of assets. *,**,*** indicate significance at 10%, 5%, and 1% levels (two-tailed), respectively. 25

27 Table 2: The impact of mandatory corporate sustainability reporting on ESG, Environmental, Social and Governance disclosure Local US Japan India Australia UK South Africa (Local: UK) ESG Disclosure 11.55*** 10.50*** 8.831*** 9.644*** 11.40*** 11.55*** (1.187) (1.338) (1.276) (1.435) (1.264) (1.187) Environmental Disclosure 8.221*** 8.361*** 5.292*** 6.810*** 8.577*** 8.221*** (1.314) (1.602) (1.566) (1.528) (1.448) (1.314) Social Disclosure 13.07*** 14.42*** 15.27*** 13.76*** 11.20*** 13.07*** (1.794) (1.941) (1.798) (2.043) (1.925) (1.794) Governance Disclosure 17.04*** 11.02*** 9.823*** 11.62*** 17.68*** 17.04*** (1.986) (2.079) (1.502) (1.919) (1.631) (1.986) China (Local: China) ESG Disclosure *** 3.072** 3.987*** 5.214*** 3.903*** (0.799) (0.918) (1.186) (1.037) (0.776) (1.123) Environmental Disclosure *** * *** (0.678) (0.891) (1.416) (1.108) (0.830) (1.038) Social Disclosure *** 11.04*** 12.89*** 9.241*** 6.430*** (1.231) (1.211) (1.392) (1.422) (1.202) (1.478) Governance Disclosure *** *** 10.90*** (1.256) (1.662) (1.397) (1.381) (1.350) (1.848) Denmark (Local: Sweden/Norway/Finland) ESG Disclosure (2.626) (2.278) (2.654) (2.874) (2.526) (2.542) Environmental Disclosure (3.145) (2.757) (3.234) (2.979) (2.945) (2.637) Social Disclosure *** *** * (3.623) (3.056) (3.399) (3.482) (3.352) (3.257) Governance Disclosure *** (3.107) (3.276) (2.724) (3.748) (2.879) (4.036) 26

28 Malaysia (Local: Singapore/Indonesia/Thailand) ESG Disclosure *** (2.191) (2.333) (2.720) (2.476) (2.004) (2.355) Environmental Disclosure ** * *** (1.787) (1.856) (2.742) (2.216) (1.654) (2.149) Social Disclosure *** *** (2.762) (3.065) (3.344) (3.188) (2.772) (3.043) Governance Disclosure *** *** (3.637) (3.902) (3.403) (3.401) (3.054) (3.498) This table tabulates the estimated coefficient on the interaction term between an indicator variable that takes the value of one for treated firms and an indicator variable that takes the value of one for years after the regulation. The dependent variable is the ESG disclosure score or its subcomponents. Equation (2) in the text describes the full model which includes firm fixed effects, year fixed effects, time-varying sector indicators, a firm size control, and a firm leverage control. *,**,*** indicate significance at 10%, 5%, and 1% levels (twotailed), respectively. 27

29 Table 3: The impact of mandatory corporate sustainability reporting on ESG, Environmental, Social and Governance disclosure, conditional on levels of disclosure in the year prior to the regulation US Japan India Australia UK South Africa ESG Disclosure High 11.62*** 12.66*** 8.431*** 12.00*** 13.75*** (2.010) (1.617) (2.067) (1.773) (1.596) Environmental Disclosure High 8.872*** 11.30*** 8.100*** 10.98*** 11.26*** (2.363) (1.808) (2.201) (1.956) (1.759) Social Disclosure High 17.10*** 21.22*** 13.75*** 12.73*** 17.87*** (2.387) (2.133) (2.746) (2.473) (2.037) Governance Disclosure High 5.948* 5.428*** *** 11.42*** (3.019) (1.987) (2.210) (1.927) (2.560) ESG Disclosure Low 9.302*** 4.983** 10.51*** 11.40*** 8.852*** (1.688) (1.902) (1.876) (1.828) (1.664) Environmental Disclosure Low 7.668*** ** 6.885*** 4.506*** (1.767) (2.469) (2.016) (1.981) (1.669) Social Disclosure Low 11.19*** 8.625*** 13.82*** 9.477*** 6.941** (2.861) (2.772) (2.738) (2.976) (2.810) Governance Disclosure Low 15.84*** 14.01*** 20.97*** 23.09*** 22.16*** (2.730) (1.916) (2.701) (2.406) (2.936) China ESG Disclosure High 3.551** ** 2.178* (1.366) (2.058) (1.297) (1.302) (1.973) Environmental Disclosure High 8.232*** ** 7.091** (2.273) (6.012) (2.614) (3.160) (4.938) Social Disclosure High 14.21*** 11.44*** 15.25*** 10.41*** (3.537) (4.058) (3.573) (3.454) (4.836) Governance Disclosure High * ** (2.256) (2.021) (2.096) (2.180) (2.863) 28

30 ESG Disclosure Low 7.475*** 5.615*** 4.331*** 6.331*** 6.981*** (1.098) (1.397) (1.411) (0.880) (1.186) Environmental Disclosure Low 4.591*** * *** (0.926) (1.480) (1.193) (0.852) (0.921) Social Disclosure Low 13.34*** 10.81*** 12.46*** 8.949*** 6.314*** (1.294) (1.500) (1.556) (1.295) (1.517) Governance Disclosure Low 5.457*** 12.19*** *** 16.40*** (2.088) (1.556) (1.630) (1.523) (2.236) Denmark ESG Disclosure High 5.887* * 6.510* 8.887** (3.357) (3.620) (3.777) (3.798) (3.488) Environmental Disclosure High ** 9.787*** (4.216) (4.145) (4.200) (3.430) (3.206) Social Disclosure High 7.684* *** (3.804) (4.598) (4.027) (3.991) (3.783) Governance Disclosure High ** (4.838) (4.193) (5.759) (3.391) (5.413) ESG Disclosure Low * (1.861) (2.602) (2.770) (2.027) (2.891) Environmental Disclosure Low ** (2.315) (3.744) (2.753) (3.653) (3.412) Social Disclosure Low (5.592) (5.695) (5.786) (5.864) (6.039) Governance Disclosure Low ** (3.956) (3.567) (4.397) (4.723) (5.209) This table tabulates the estimated coefficient on the interaction term between an indicator variable that takes the value of one for treated firms and an indicator variable that takes the value of one for years after the regulation. The dependent variable is the ESG disclosure score or its subcomponents. Equation (2) in the text describes the full model which includes firm fixed effects, year fixed effects, time-varying sector indicators, a firm size control, and a firm leverage control. A firm is identified as High Disclosure if it discloses more than the median firm in the year before the first year of the regulation. A firm is identified as Low Disclosure if it discloses equal to or less the median firm in the year before the first year of the regulation.*,**,*** indicate significance at 10%, 5%, and 1% levels (two-tailed), respectively. 29

31 Table 4: The impact of mandatory corporate sustainability reporting on assurance and propensity to adopt the GRI reporting guidelines US Japan India Australia UK South Africa Assurance 0.114*** 0.145*** ** ** 0.108*** (0.0364) (0.0321) (0.0380) (0.0394) (0.0344) GRI 0.273*** 0.301*** 0.285*** 0.313*** 0.367*** (0.0483) (0.0455) (0.0465) (0.0429) (0.0410) China Assurance ** (0.0120) (0.0214) (0.0235) (0.0175) (0.0188) GRI 0.125*** ** 0.106*** 0.121*** (0.0337) (0.0406) (0.0384) (0.0345) (0.0368) Denmark Assurance ** (0.0700) (0.0605) (0.0734) (0.0821) (0.0732) GRI ** ** (0.0877) (0.0898) (0.0873) (0.0966) (0.0853) Malaysia Assurance ** * (0.0134) (0.0117) (0.0363) (0.0204) (0.0294) GRI 0.150** ** 0.134** (0.0596) (0.0750) (0.0707) (0.0599) (0.0582) This table tabulates the estimated coefficient on the interaction term between an indicator variable that takes the value of one for treated firms and an indicator variable that takes the value of one for years after the regulation. The dependent variable is Assurance, an indicator variables that takes the value of one if the firm has received assurance on its ESG disclosures, or GRI, which is an indicator variable that takes the value of one if the firm follows the Global Reporting Initiative guidelines. Equation (2) in the text describes the full model which includes firm fixed effects, year fixed effects, time-varying sector indicators, a firm size control, and a firm leverage control. *,**,*** indicate significance at 10%, 5%, and 1% levels (two-tailed), respectively. 30

32 Table 5: The impact of mandatory corporate sustainability reporting on propensity to become a UNGC signatory and to adopt environmental and social criteria in supply chain management for Danish firms Denmark US Japan India Australia UK UN Global Compact Signatory 0.431*** 0.369*** 0.380*** 0.468*** 0.462*** (0.0977) (0.0998) (0.106) (0.0921) (0.0887) Environmental criteria in supply chain management 0.184* 0.161* 0.194** 0.258** 0.193* (0.0938) (0.0937) (0.0856) (0.104) (0.100) Social criteria in supply chain management 0.218* 0.264** 0.346*** 0.337*** 0.241** (0.122) (0.118) (0.107) (0.104) (0.113) This table tabulates the estimated coefficient on the interaction term between an indicator variable that takes the value of one for treated firms and an indicator variable that takes the value of one for years after the regulation. The dependent variable is a) UN Global Compact Signatory, an indicator variable that takes the value of one if the firm is a signatory of the UN Global Compact, b) Environmental criteria in supply chain management, an indicator variable that takes the value of one if the firm has applied environmental criteria when choosing its supply chain partners and c) Social criteria in supply chain management, an indicator variable that takes the value of one if the firm has applied social criteria when choosing its supply chain partners. Equation (2) in the text describes the full model which includes firm fixed effects, year fixed effects, time-varying sector indicators, a firm size control, and a firm leverage control. *,**,*** indicate significance at 10%, 5%, and 1% levels (two-tailed), respectively. 31

33 Table 6: The impact of mandatory corporate sustainability reporting on community spending by Malaysian firms Malaysia US Japan India Australia UK Philanthropy (Community Spending) *** (0.0560) (0.0722) (0.0663) (0.0625) (0.0704) This table tabulates the estimated coefficient on the interaction term between an indicator variable that takes the value of one for treated firms and an indicator variable that takes the value of one for years after the regulation. The dependent variable is Philanthropy (Community Spending), an indicator variable that takes the value of one if the firm discloses the level of community spending. Equation (2) in the text describes the full model which includes firm fixed effects, year fixed effects, time-varying sector indicators, a firm size control, and a firm leverage control. *,**,*** indicate significance at 10%, 5%, and 1% levels (two-tailed), respectively. 32

34 Table 7: Instrumental variables analysis the impact of disclosure on Tobin s Q Panel A: Instrumented ESG Disclosure US Japan India Australia UK South Africa * *** *** ( ) ( ) ( ) ( ) ( ) China *** *** *** *** ( ) (0.0233) ( ) ( ) (0.0134) Denmark (0.199) (0.0742) (0.123) (0.0652) (0.0758) 33

35 Panel B: Instrumented sub-components of ESG Disclosure US Japan India Australia UK South Africa E * *** *** ( ) (0.0109) ( ) ( ) ( ) S * *** *** ( ) ( ) ( ) ( ) ( ) G * *** *** * ( ) ( ) ( ) ( ) ( ) China E *** N/A N/A *** N/A (0.0102) (0.0159) S *** *** *** *** ( ) ( ) ( ) ( ) ( ) G N/A *** N/A *** *** ( ) (0.0133) ( ) This table tabulates the estimated coefficient on the instrumented ESG disclosure score (Panel A) or its subcomponents (Panel B) in the second stage of an instrumental variables regression. The dependent variable is Tobin s Q calculated as the sum of the market value of equity and the book value of the liabilities over the total book value of assets. Equation (1) in the text describes the first stage. N/A means that the instrument in the first stage is not relevant. *,**,*** indicate significance at 10%, 5%, and 1% levels (two-tailed), respectively. 34