Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India

Shivaram Rajgopal; Prasanna L. Tantri

doi:10.2139/ssrn.3909219

Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India

Rajgopal, Shivaram;Tantri, Prasanna L. 2021-08-22 00:00:00 Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India Shivaram Rajgopal Prasanna Tantri August 22, 2021 Shivaram Rajgopal can be reached at sr3269@gsb.columbia.edu. Prasanna Tantri can be reached at prasanna tantri@isb.edu. We thank the Editor, Christian Leuz, an anonymous Associate Editor, and an anonymous referee for helpful suggestions. We also thank Sakshi Gabba, Sharad Hotha, Gautham Kan- thasamy, Saiharsha Katuri, Aditya Murlidharan, Shradhey Prasad, and Nishka Sharma for excellent research assistance. We acknowledge helpful comments from Ray Ball and workshop participants at Booth School of Business, University of Chicago. We also acknowledge helpful comments from David Yermack and Confer- ence participants at NYU-IIM C India Conference. We are grateful to the Center for Analytical Finance, Indian School of Business for providing the data and the necessary nancial assistance for this project. Rajgopal thanks the Columbia Business School for nancial support. Any remaining errors are ours. Electronic copy available at: https://ssrn.com/abstract=3909219 Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility ? Evidence from India Abstract We investigate the implementation of a government of India mandate that requires rms to spend at least 2% of their pro ts on corporate social responsibility (CSR). We nd that mandated rms that voluntarily engaged in CSR before the mandate reduce their CSR spending signi cantly after the mandate. The erstwhile voluntary CSR spenders increase advertising expenditure plausibly to oset the lost signaling value of voluntary CSR. The 2% mandate negatively impacts valuations and operating performance. Our results show that regulatory intervention in CSR diminishes its signaling value and leads to a reduction in voluntary CSR spending. Electronic copy available at: https://ssrn.com/abstract=3909219 1 Introduction \I don't think you generate CSR by putting statutory requirements. I think there is enough social consciousness among the larger Companies to drive it on the basis of what they consider their responsibility."{ Azim Premji, Chairman of Azim Premji Foundation. Governments in several countries have begun playing an active role in corporate social responsibility (CSR). Some have nudged corporations to spend funds on CSR, while others have moved to mandate disclosures. A few countries, such as India (Manchiraju and Raj- gopal (2017), Dharmapala and Khanna (2018)) and Indonesia (Waagstein (2011)) have gone a step further and have promulgated laws that make not only disclosure but also spending on speci ed CSR activities mandatory. We study the Indian CSR mandate in this paper. A signi cant part of the current academic debate on CSR is centered around the rms' motives underlying voluntary CSR and the value consequences thereof (see Kitzmueller and Shimshack (2012), Orlitzky, Schmidt, and Rynes (2003), McWilliams, Siegel, and Wright (2006), Margolis, Elfenbein, and Walsh (2009), and Perrini, Russo, Tencati, and Vurro (2011)). Christensen, Hail, and Leuz (2019), however, point out that most extant stud- ies based on voluntary CSR are subject to selection issues. Furthermore, most CSR-related regulations worldwide mandate only disclosure. Therefore, an analysis of the impact of forced CSR spending in the Indian context is likely to shed light on the motives behind CSR spending. In particular, the government of India forced \eligible" rms to spend at least 2% of Source: http://economictimes.indiatimes.com/news/company/corporate-trends/azim-premji-against- law-on-mandatory-csr-spending-by-corporates/articleshow/7782555.cms. Wipro is one of the largest Infor- mation technology companies in India. Recently, the European Union member states have agreed to pass legislation requiring cor- porations to report their CSR activities in detail. Similar laws have been passed or are be- ing contemplated in countries such as China (Chen, Hung, and Wang (2018)) and Canada. Source: https://www.theguardian.com/sustainable-business/eu-reform-listed-companies-report- environmental-social-impact, http://corporatejustice.org/; http://corporatejustice.org/news/1174- getting-non- nancial-reporting-right-eu-commission-guidelines-clarify-expectations-towards- business; https://www.globalreporting.org/information/policy/Pages/EUpolicy.aspx; https://mastereia.wordpress.com/2014/04/10/mandatory-environmental-corporate-social-responsibility- can-canada-become-a-leader/; https://www.greenbiz.com/news/2009/01/07/mandatory-csr-reporting- denmarks-largest-companies Electronic copy available at: https://ssrn.com/abstract=3909219 3 average three years' pro ts on CSR. The eligibility threshold is either INR 50 million in pro t, INR 5 billion in net worth, or INR 10 billion in sales. Firms that exceed one or more of the above thresholds are subject to the 2% CSR spending rule. We focus on rms that voluntarily spent more than the minimum 2% limit before the law was passed. We can track CSR spending before the rule was enacted because Indian rms are required to disclose such spending as per the applicable accounting standards. The CSR law imposed by India is quite stringent in intent and enforcement. The mandate requires that the rm's board justify any failure to comply with the CSR requirement (\the comply or explain" model). However, the law does not specify what can be considered a reasonable explanation, creating room for ambiguous interpretation of such explanation by ocials of the federal ministry of corporate aairs (MCA). More importantly, the law elevates the responsibility of CSR spending from the management to the board. The violation of the CSR requirement is, therefore, considered a non-ful llment of directors' duties. If the law has the requisite bite, we expect rms that spent less than 2% before the mandate (labeled \low CSR rms") to spend close to 2% of pro ts after the mandate. We nd such a result. However, the expected reaction of rms that spent more than 2% (labeled \high CSR rms") in the pre-mandate period is not clear ex-ante. We expect high CSR rms to cut spending in the post mandate period if (i) if the mandate dilutes the underlying motive for voluntary CSR spending; or (ii) if alternative ways of satisfying such a motive become relatively more attractive; or (iii) if 2% is perceived as the society's new norm for CSR spending. Otherwise, the voluntary spending of high CSR rms should not change. Data on voluntary CSR spending are drawn from the Prowess database maintained by the Center for Monitoring Indian Economy (CMIE). In addition, we obtain data on mandated CSR spending from the federal Ministry of Corporate Aairs (MCA). In the post-mandate period, we nd that high CSR rms signi cantly reduce their CSR spending to around 2% of pro ts in univariate tests. On the other hand, as expected, low CSR rms increase their CSR spending to the 2% level. INR stands for Indian Rupee. Electronic copy available at: https://ssrn.com/abstract=3909219 The ideal identi cation strategy is unfortunately not available to us. Both high and low CSR rms are \treated" in an experimental sense as the 2% law applies to both these types of rms. Furthermore, comparing the entire mandated group as treated with the non- mandated control group does not work well because the post-mandate incentives of high and low CSR rms within the mandated group are not the same as explained next. As a compromise, in our baseline dierence-in-dierence (di-in-di, henceforth) analysis, the treated (control) rms are those that are (not) mandated to invest in CSR, within the sample of high CSR rms (see Figure 1). By construction, both the treatment and control samples in this experiment do not have to mechanically increase CSR spending to comply with the 2% law. Crucially, this sample includes voluntary high spenders who are not covered by the 2% mandate. The dependent variable in our research design is the magnitude of CSR spending at a rm-year level. Our focus is on the interaction between the post-mandate indicator variable and the indicator variable representing rms required to spend on CSR. We test for and rule out the existence of pre-trends. We acknowledge that the non-mandated group will react to the actions of the mandated group. Hence, we do not claim to measure the complete causal eect of the mandate. Instead, we document a dierence in the relative impact on CSR spending between the two groups We nd a statistically signi cant and economically meaningful 32.57% decline in CSR spending in the post-mandate period. Already compliant rms covered by the 2% rule cut CSR spending signi cantly. We expect low CSR rms covered by the mandate to spend more and nd such a result. These inferences hold when we use a triple interaction framework with all three margins (timing of the mandate, mandate eligibility, and pre-mandate CSR spending) considered together. In the second part of the paper, we try to understand why high CSR rms cut spending. We conjecture that the mandate dilutes strategic value of voluntary CSR which, in turn, leads to a negative impact on valuation and operating performance (Deng, Kang, and Low (2013)). Event study and other tests con rm that the imposition of mandatory CSR leads to a negative stock price reaction and lower operating performance for mandated rms within Electronic copy available at: https://ssrn.com/abstract=3909219 the universe of high CSR rms. This opens the question of whether a strategic motive suered from the 2% mandate and led high CSR rms to cut spending. We conjecture that voluntary CSR spending is a unique method of signaling both virtue and higher quality of their products simultaneously (Kausar, Shro, and White (2016), Kotler and Lee (2008)). The 2% mandate diluted such a special signal. We present a mosaic of evidence consistent with this conjecture. First, we examine the CSR sections of the annual reports and social media postings of rms. We nd a signi cant (i) decline in the number of such communications, in general; (ii) fall in the number of words that convey product or service quality; (iii) decline in words that signal virtue; and (iv) increase in standardization of CSR communication. Second, mandated high-CSR rms that reduce CSR expenditure, on account of its diluted signaling power, signi cantly increase advertising expenditure after the 2% rule. Moreover, the Rupee increase in advertising expenditure is very close to the Rupee decline in CSR expenditure for the mandated high-CSR rms. Despite such a one-to-one replacement, the mandated high-CSR rms suer a loss of value and operating performance. This fact pattern is consistent with the hypothesis that voluntary CSR, as opposed to advertising, is more eective at communicating product quality and virtue simultaneously. An important open question remains after presenting such evidence: why don't mandated high CSR rms continue to enjoy signaling bene ts by spending more than the mandated 2% level on CSR even after the mandate? We believe that compliance costs with the new law are a signi cant deterrent. CSR spending requires the board's approval post mandate. Additional approval lters and the requirement to spend at least 2% of pro ts every year also curtail exibility. Most importantly, in practice, CSR projects are not divisible into mandated and non-mandated parts. It may not be possible to run the mandated and non- mandated CSR projects in separate silos within an organization. Thus, compliance costs apply not only to the mandated part but increase in proportion to the entire CSR expendi- ture. Furthermore, the excess spending over 2% is likely viewed as a buer against a strict Electronic copy available at: https://ssrn.com/abstract=3909219 interpretation of CSR de nition by the government ocers, and hence, become less eective for signaling. Therefore, alternative ways of signaling, such as advertising, may become more attractive after the mandate. However, we acknowledge that we cannot cleanly disentangle the direct loss of signaling power of CSR from higher compliance costs. In follow-up work, we considered alternative strategic channels for CSR spending such as customers' willingness to overpay, workers' willingness to work for lower salaries, and the ability of CSR to deter value loss due to political interventions. None of the above channels seems to statistically explain the cut in CSR spending by high spenders, its replacement by advertisement, and change in CSR related communication. We also cannot rule out the possibility that voluntary CSR is partly motivated by (i) stakeholder altruism (Reinhardt, Stavins, and Vietor (2008), B enabou and Tirole (2010)); (ii) managerial moral hazard (Cheng, Hong, and Shue (2013), Masulis and Reza (2014)); or (iii) revelation of society's expectation related to CSR spending through the mandate. We in- terpret our four empirical ndings (a decline in voluntary CSR spending, negative impact on valuations and operating performance, change in CSR communication, and the replacement of CSR by advertisements) as tentative evidence that voluntary CSR is not driven solely by altruism, moral hazard or newly set social norms for CSR spending. Firms consciously used CSR for signaling quality and virtue, potentially in addition to other motives. Our paper contributes to the literature on the strategic value of CSR (Deng, Kang, and Low (2013), Cheng, Ioannou, and Serafeim (2014), Lins, Servaes, and Tamayo (2017), Dimson, Karaka s, and Li (2015), Blacconiere and Northcut (1997), Christensen, Floyd, Liu, and Maett (2017), Dhaliwal, Li, Tsang, and Yang (2011), Elliott, Grant, and Rennekamp (2017), Flammer (2015)). In their review of the CSR reporting literature, Christensen, Hail, and Leuz (2019) point out that most CSR spending is voluntary, which leads to selection- related issues. We identify a unique setting where the government mandates CSR spending. Manchiraju and Rajgopal (2017) and Dharmapala and Khanna (2018) also study the short-term stock price impact on the passage of the mandatory CSR rule in India. We focus on the impact of the mandate on actual CSR spending and show that erstwhile high spenders Electronic copy available at: https://ssrn.com/abstract=3909219 cut CSR spending post-mandate. More importantly, we examine plausible motives behind voluntary CSR spending and suggest that voluntary CSR, unlike advertising, uniquely signals both virtue and product quality. As in the case of literature dealing with the voluntary adoption of audits (Kausar, Shro, and White (2016), Dedman and Kausar (2012), Lennox and Pittman (2011)) and IFRS (Daske, Hail, Leuz, and Verdi (2013), Kim, Tsui, and Cheong (2011), Christensen, Hail, and Leuz (2013), Hung and Subramanyam (2007), Florou and Pope (2012)), we compare the implications of voluntary and mandated CSR spending. We propose that CSR spending due to a government mandate makes it a less valuable signaling tool. Despite a substitution of CSR by advertisement, the overall nancial performance of treated rms deteriorates, suggesting a unique role for voluntary CSR in signaling virtue and product quality. 2 Institutional Background and the Event Section 135 of the newly introduced Indian Companies Act of 2013 has mandated that rms above a threshold (de ned in terms of net worth, sales, and pro t) have to spend 2% of their average past three years' pro t on CSR activities. The eligibility threshold was de ned as either INR 50 million (USD 0.78 million) in pro t, INR 5 billion (USD 0.078 billion) in net worth, or INR 10 billion (USD 0.156 billion) in sales. Before the decree, rms were required only to disclose spending as per existing accounting standards. Every rm covered by the mandate is required to set up a CSR policy. Although the new Companies Act came into force on August 29, 2013, the CSR mandate was made eective from 2014-2015 (i.e., the year beginning April 1, 2014). We are not aware of any economically meaningful reason for xing the limit at 2%. It appears that the government picked a round number for simplicity. The law requires non-compliant rms to explain in their annual reports the reasons behind their non-compliance. However, the law does not specify guidelines to determine whether an explanation is valid, leaving room for regulatory discretion in interpretation. The Act de nes We assume an exchange rate of INR 63 to USD 1, which was prevalent when the Act was passed. Electronic copy available at: https://ssrn.com/abstract=3909219 CSR broadly but leaves the details to the boards of the individual rms (see Manchiraju and Rajgopal (2017) for information about the speci c provisions). Under the new law, the responsibility to manage CSR spending rests with the board and not the management. The ministry of corporate aairs (MCA) has issued show-cause notices to more than 1,000 rms, charging them with violations of the CSR law. Notices have been issued even when rms have preferred to explain rather than to comply on the grounds that the stated explanations are not satisfactory. Section 135, which imposes mandatory CSR, does not im- pose penal provisions if the spending mandate is violated. However, the MCA has charged the alleged non-compliant rms under a dierent section 134, which speci es directors' re- sponsibilities for nancial statements, and contains strict penal provisions. Although, in theory, mandatory CSR works on a \comply or explain" model, in practice, it is safer for the companies to comply than to explain. 3 Data, Variable De nition And Sample Construction We use the Prowess database maintained by the Center For Monitoring Indian Economy (CMIE) for the pre-mandate period. For the post-mandate period, we use the ministry of corporate aairs (MCA) data for rms covered by the ministry and the CMIE Prowess database in other cases. Both these databases source information from the annual reports of rms. Expectedly, both the data sets report the same Rupee numbers for CSR outlays th th for most cases. As shown in Table A1 of the online appendix, both the 5 and the 95 percentile thresholds of the dierence between reported CSR amounts in the two databases for a given rm are zeroes. We manually check the reasons behind the dierences between the two datasets in some extreme cases. The MCA data set is more accurate in all cases. However, we found data entry errors and incorrect round-os of decimals in the Prowess database in rare situations. 5https://www. nancialexpress.com/industry/government-issued-notices-to-1018- rms-for-csr-non- com- pliance/589099/ In one case, the Prowess database has left out one part of CSR expenditure reported in the annual report. In another case, the Prowess database used the budgeted CSR numbers presented in the annual Electronic copy available at: https://ssrn.com/abstract=3909219 Given these ndings, we use the MCA database for the post-mandate period for the rms covered by the ministry. Because the ministry does not maintain pre-mandate data and data for non-mandated rms, we have to rely on the Prowess dataset for (i) the pre-mandate period CSR spending of all rms; and (ii) pre and post-mandate period CSR spending of all non- mandated rms. The ministry data is available for years spanning 2014-2015 to 2018-2019. Compilation of the data thus gathered gives us CSR spending at a rm-year level. The variable thus created is labeled the \CSR amount." Because of data integrity issues at the extreme ends of the CSR spending distribution, we winsorize the variable at 1% and 99% in our primary analysis. We provide variable de nitions in Table 1. 3.1 Sample Construction We report the details underlying the construction of the sample in Table 2. The Prowess database contains information on 43,051 rms. Many of these are shell companies formed for money laundering. The ministry dataset covers 12,097 companies. We found 10,154 rms of these in the Prowess database using the unique corporate identity number (CIN) as the matching variable. Our sample starts from the nancial year 2009-2010. The years between 2009-2010 and 2013-2014 are labeled as the pre-regulation years. Years 2014-2015 and 2018- 2019 comprise the post-mandate period. Thus, we cover a sum total of 10 years. The table shows that the merged data set contains 39,309 rms and 236,044 rm-year observations with usable data in both pre and post-mandate periods. We employ two additional lters. We leave out observations where CSR information is missing instead of treating them as zeros because we are unsure why these data points are absent. We have data on CSR spending for 44,769 of the 236,044 rm-year observations. Of these, 16,251 (28,518) observations belong to rm-years where the average pre-period ratio of the amount spent on CSR and average past three year pro ts is more than 2% (less than report rather than the actual amount spent. We provide ve examples of the largest deviations between these two databases in Table A2 of the online appendix. http://www. rstpost.com/business/over-1-62-lakh-shell-companies-deregistered-over-half-from- mumbai-delhi-hyderabad-3907583.html Electronic copy available at: https://ssrn.com/abstract=3909219 or equal to 2%). The table also provides information related to the number of observations for which CSR to pro ts ratio is equal to or greater than the 5%, 7.5%, and 10% thresholds. 4 Empirical Strategy and Results 4.1 Univariate Tests We begin by calculating the CSR Ratio, de ned as the ratio of the CSR amount and average pro t before tax for the past three years, for each rm-year. We average the CSR ratio over the pre-regulation period. Firms with an average CSR Ratio of greater than 2% form the \high CSR" group. Firms below this threshold form the \low CSR" group. As a further robustness check, we use three other threshold cutos based on 5%, 7.5%, and 10% of the average three years' pro ts. In Table 3, we nd that the high CSR rms cut back their spending on CSR signi cantly post-mandate. However, the low CSR rms increase CSR spending up to 2% in the post- mandate period. For example, column 2 shows that rms that spent between 0% and 1% of pro ts on CSR before the mandate increase spending by 1.7 percentage points, on average, after the mandate. In column 4, we report that rms that used to spend between 2% and 3% before the mandate maintain almost the same level of spending even after the mandate. Interestingly, from column 5, we nd a signi cant decline in CSR spending after the mandate. In column 5, rms that used to spend between 3% and 4% in the pre-mandate period reduce spending to 3% in the post-mandate period. In column 7, rms that used to spend between 5% and 6% also reduce spending by 2.5 percentage points in the post-mandate period. A similar trend is seen in other columns where we consider rms that used to spend a higher proportion of pro ts on CSR in the pre-mandate period. Electronic copy available at: https://ssrn.com/abstract=3909219 4.2 Diculty With Comparing High And Low CSR Firms A potential empirical strategy is a di-in-di test within the set of rms subject to the 2% mandate where the high CSR rms (low CSR rms) are designated as the treated (control) group. One problem with this strategy is that both the treated and control groups are directly impacted by the mandate, although in opposite directions. The univariate results, presented in Table 3, clearly show that high CSR rms cut spending on CSR whereas low CSR rms increase spending. Therefore, it is dicult to tell whether our results are attributable to (i) an increase in spending by the control group; or (ii) a decline in spending by the treated group. Hence, we do not use this identi cation strategy. A second strategy of comparing the entire mandated group as the treated class and the entire non-mandated group as the control class does not work either because high and low CSR rms behave dierently within the mandated group. 4.3 Mandated Vs Non-Mandated- Within High CSR Firms To test the implication of the mandate on high CSR rms, we adopt an identi cation strategy that compares mandated and non-mandated rms within the high CSR group. The mandate applies only to those rms which satisfy at least one of the qualifying criteria described in Section 2. We consider such mandated rms as the treated group. We acknowledge that non- mandated rms will have to consider the reaction of the mandated rms while formulating their plans. Hence, our tests can only capture the incremental eect on the treated rms relative to the control sample and not the complete causal eect of the mandate. We discuss this issue in Section 6.3. We conduct a di-in-di comparison within high CSR rms between the mandated and non-mandated groups (depicted in Figure 1). As shown in the gure, we conduct a similar but separate di-in-di test within the sample of low CSR rms. The treated group within the set of low CSR rms will increase CSR spending mechanically to comply with the law, but the control group does not have a similar obligation. Electronic copy available at: https://ssrn.com/abstract=3909219 We estimate the following standard di-in-di regression equation by rst limiting the sample to the high CSR rms. (1) Y = + Post Treatment + X + + + it 1 t i 2 it 3 i 4 t it The dependent variable is the amount (in INR) of CSR expenditure made by a rm i in the year t. The variable Post is a dummy variable, which takes the value of one for post-regulation years and zero otherwise. Treatment is a dummy variable that takes the value of one if a rm satis es at least one of the three CSR mandate conditions based on average values during the pre-mandate years and zero otherwise. Firms that do not satisfy any of the 2% rule's qualifying conditions form the control group. The interaction between the above Post and Treatment variables is the explanatory variable of interest. Note that t i represents the rm xed eects, represents the year xed eects, and X represents i t it rm-level time-varying variables such as pro t and total net worth. Furthermore, we use total assets for scaling the dependent variable in subsequent tests. We use net worth as a proxy for size because we can nd more non-missing observations for net worth relative to that for sales. Finally, the standard errors are clustered at the industry level and adjusted for heteroskedasticity. The results are reported in Panels A and B of Table 4. The CSR amount is the dependent variable in Panel A. In columns 1 and 5, we use the 2% threshold to de ne high CSR rms. As expected, we nd a sharp decrease in the dierence between the treatment and control rms in the post-regulation period compared to the dierence in the pre-regulation period. The magnitude of the decline is INR. 7.87 million, which is 32.57% of the average CSR amount in the pre-mandate period. Therefore, the fall is economically meaningful. We present the results with higher thresholds of 5% (columns 2 and 6), 7.5% (columns 3 and 7), and 10% (columns 4 and 8) of past pro ts. The results strengthen with an increase in the threshold spending levels, suggesting that voluntary high spenders cut CSR spending signi cantly in response to the CSR mandate. In Panel B, we use the ratio between CSR amount and Electronic copy available at: https://ssrn.com/abstract=3909219 total assets as the dependent variable. Column 1 shows that the ratio decreased by 14 basis points. The decline represents an economically meaningful 51.9% of the pre-mandate levels. To test whether the treated rms drive the impact, we compare the pre and post-mandate periods within the sample of mandated high CSR rms. We estimate the following regression equation: (2) Y = + Post + X + + it 1 t 2 it 3 i it All the terms have the same de nitions as in equation 1. We cannot include year xed eects because the tests compare between pre and post period expenditure. The results are reported in Panels A and B of Table A3 of the online appendix. In column 1, where we limit the sample to rms spending at least 2% of their pro ts on CSR, the CSR spending declines by about INR 7.63 million after the mandate. The decline represents 31.56% of the average CSR amount in the pre-mandate period. We nd similar results even when we use the ratio between CSR amount and total assets as the dependent variable in Panel B. These results con rm that almost all of the decline in CSR spending documented in Table 4 is driven by the treated group. We perform the same di-in-di analysis as in equation 1 among low CSR rms. The comparison is between mandated and non-mandated rms, as before. We expect the man- dated rms to invest more in a di-in-di sense mechanically to comply with the legal requirement to hike spending. We nd such a result (reported in Table A4 of the online appendix). Next, we incorporate all three variations in one speci cation whereby the terms \Post-Pre," \Mandated-Non- mandated," and \High CSR-Low CSR," are considered in a triple interaction framework as shown at the bottom of Figure 1. We expect the triple inter- action term, the Post* Mandated* High CSR indicator variable, to be negative, and we nd that result (in Table A5 of the online appendix). Thus, our results stem from spending cuts initiated by the high CSR rms among the mandated group after the law was promulgated. The results taken together suggest that (i) the CSR mandate leads to a reduction in Electronic copy available at: https://ssrn.com/abstract=3909219 voluntary spending; (ii) the reduction stems from a cut in CSR spending post mandate by the erstwhile high spenders and (iii) the pre-mandate low spenders increase spending as required to by the law. 4.3.1 Dierence-In-Dierences Pre-Requisites And Robustness a. The Test of Pre-Trends: To mitigate the possibility that a mechanical continuation of a pre-existing trend drives our results, we plot the ratio of CSR amount and total assets for the treated and the control groups in Figure 2. We consider ve years before and ve years after the mandate. As shown in the gure, there does not appear to be any clear pre-trend. Further, we notice a sharp decline in CSR spending of the treated group in the post-mandate period. Second, in our baseline regression setup, we introduce indicators for individual years. Then, we interact each of the pre-year indicator variables with the treatment variable. We nd all the interaction terms, except the interaction between post-mandate and treatment indicator variables, to be statistically indistinguishable from zero. This result, reported in Table A6 of the online appendix, also helps us rule out the existence of pre-trends. b. Comparing Treated And Control Groups: We acknowledge that the treated rms are systematically larger than the control rms because the mandate applies to larger rms. While the non-existence of pre-trends helps ameliorate concerns in this regard, we conduct an additional test. We compare several operating and nancial ratios for the treated and control rms in Table 5. These ratios include operating margin, leverage, stock turnover, return on equity, advertisement to sales, and pro t per employee. None of the above ratios is signi cantly dierent between the treated and the control rms. This analysis shows that the two sets of rms are similar in terms of operating eciency, although they dier in size. c. Additional Robustness Tests We perform several additional robustness tests. First, to account for possible anticipation Electronic copy available at: https://ssrn.com/abstract=3909219 of the law on account of the discussion in the media before its implementation, we omit the years 2012-2013 and 2013-2014 from the sample, which is when the new Companies Act was discussed and introduced. The CSR provision came into force eective 2014-2015. We estimate equation 1 using this subsample and nd that the results are consistent with our hypotheses (reported in Table A7 of the online appendix). Note that we exclude the years 2012-2013 and 2013-2014 while arriving at the treated and control groups for the above test. Second, we conduct placebo tests with false treatment years within the pre-mandate period. We report the results using 2011-2012 as a false treatment year in Table A8 of the online appendix. We estimate regression equation 1. The coecient of the interaction term between the post-mandate indicator variable and the treatment indicator variable is statistically indistinguishable from zero. Furthermore, we get similar results when we use other false treatment years. Finally, control rms that lie very close to the cuto during the pre-mandate period may cross over during the post-mandate period and fall within the purview of the 2% rule. Thus, we potentially misclassify a few treated rms as control rms. To account for such a possibility, we leave out control rms very close to the cuto and re-estimate regression equation 1. We present the results in Table A9 of the online appendix. The results remain unchanged. 5 Why do high spenders reduce spending on CSR post the mandate? In the second part of the paper, we examine why the erstwhile high spenders cut CSR spending after the mandate. We hypothesize that voluntary CSR has a strategic value that gets diluted after the mandate (\the strategic CSR hypothesis"), which, in turn, negatively impacts rm value and operating performance. A decline in both operating performance and shareholder value would be consistent with a strategic CSR hypothesis. However, these Electronic copy available at: https://ssrn.com/abstract=3909219 results are insucient to establish a strategic motive at work and also cannot rule out non- strategic motives. In follow-up work, we explore speci c operational changes that plausibly explain the exact strategic motive underlying voluntary CSR. 5.1 Impact On Operating Performance We use return on assets (ROA) and return on equity (ROE) as measures of operating per- formance. Column 1 of Table 6 shows that ROE declines by 2.29 percentage points for the treated group in a di-in-di sense. Because the average ROE is 19.9% in the pre-mandate period, the decline represents an economically meaningful 11.51%. Similarly, in column 2, we nd a 0.93 percentage points decline in ROA or an economically meaningful 8.08% of the average ROA during the pre-mandate period. In columns 4 and 5, we include one-year lags of pro ts and net worth as control variables. Finally, in columns 5 and 6, we use the natural logarithm of sales as the dependent variable. We do not nd a signi cant change in sales. 5.2 Impact on Stock Valuations The dilution of strategic motivation of high CSR rms to cut spending should have clear shareholder value implications. Manchiraju and Rajgopal (2017) examine eight events lead- ing up to the CSR law and nd that the stock price reaction of the entire set of mandated rms when compared to non-mandated rms, is negative. Note that stock prices of man- dated rms could react negatively for two reasons: (i) they are forced to spend 2% of their average pro ts on CSR, which is akin to an increase of 2% in corporate taxes; and/or (ii) mandatory CSR makes a strategic motive behind voluntary CSR less valuable. Manchiraju and Rajgopal (2017) cannot distinguish between these explanations in their setting. We attempt to do so by comparing the stock market reaction of mandated and non- mandated rms' stocks but we limit the sample to high CSR rms. In this setting, any adverse price reaction of the mandated rms' stocks is likely attributable to the loss of strategic value and other compliance costs, rather than the increased CSR spending burden Electronic copy available at: https://ssrn.com/abstract=3909219 as high CSR sample rms already spend more than 2% of their pro ts on CSR. As noted by Manchiraju and Rajgopal (2017), the introduction of mandatory CSR was a part of new company legislation passed by the Parliament. Naturally, all the provisions of the bill were debated in detail. Hence, it is hard to use one date to conduct an event study. Instead, we rely on the sudden announcement made by the nance minister on July 17, 2019, that, henceforth, non-compliance with CSR provisions will constitute a criminal oense. Till then, non-compliance with CSR was considered a civil oense. The nance minister reversed her position on August 23, 2019. This event provides an excellent testing ground for the shareholder value implications of voluntary CSR for the following reasons. First, after the rst announcement by the minister, business executives faced the prospect of a jail term for non-compliance with the CSR provision. Therefore, the value of voluntary CSR should have signi cantly diminished after her rst announcement. Second, the relatively arbitrary reversal of her position mitigates other endogenous factors that might confound our event study results. It is hard to argue that some other unobservable factor moved precisely in line with the minister's announcements on both occasions. We present our results in Table 7. In Panel A, we consider the rst announcement which made non-compliance with CSR regulations a criminal oense. We compare all mandated with all non-mandated rms. We estimate a regression where the dependent variable is the excess return three days around the event of a stock compared to the market benchmark. In columns 1 and 2, the explanatory variable (\treatment") identi es mandated rms. We include industry- xed eects and cluster standard errors at an industry level. We nd that the mandated rms, in general, react negatively to the event (-1.2%), consistent with Manchiraju and Rajgopal (2017). In columns 3 and 4, we restrict the sample to high CSR rms and nd that the stock prices of mandated high CSR rms decline by about 0.8% relative to that of non-mandated high CSR rms. By construction, high CSR rms are generous spenders, and hence, none of the treated or the control rms needs to increase CSR spending to comply with the new law. We use Nifty 50, Indias most widely tracked market index, as the market benchmark. Electronic copy available at: https://ssrn.com/abstract=3909219 Therefore, it seems reasonable to attribute the decline in share prices to the loss of strategic value of CSR and additional compliance costs. In Panel B, we study the reaction to the reversal of position by the nance minister on August 23, 2019. The remaining details remain unchanged. Investors appear to reverse almost the entire under-performance of mandated high CSR stocks when the nance minister changed her mind. This pattern provides further support to our hypothesis that mandatory CSR negatively impacts a strategic channel that made voluntary CSR eective before the new law came into force. We explore what that strategic channel might be in the following section. 5.3 Possible Strategic Reasons In a quest for a potential strategic motivation behind voluntary CSR before the mandate, we begin with (i) a close examination of various forms of public communications related to CSR; and (ii) the types of expenditure that replace voluntary CSR spending after the mandate. 5.3.1 Loss Of Signaling Value? The literature has argued that the voluntary adoption of audits (Kausar, Shro, and White (2016), Dedman and Kausar (2012), Lennox and Pittman (2011)) has signaling value. Sim- ilarly, Kotler and Lee (2008, 2005) suggest that voluntary CSR positively impacts potential customers' views about the rm, its' strengths and product oerings. Therefore, we investi- gate the possibility that pre-mandate CSR was perceived as a credible signal about a rm's overall product quality. We extend this thesis to suggest that voluntary CSR provides a unique opportunity to signal both high quality and virtue. This bundle cannot be perfectly substituted away by conventional advertising. However, regulation and enforcement that circumscribes the exact nature of CSR potentially dilutes the power of such a signal. As noted in section 2, the ministry routinely issues show-cause notices to rms arguing that the CSR expenditure Electronic copy available at: https://ssrn.com/abstract=3909219 claimed by rms is not socially desirable but merely a commercial expenditure incurred in the ordinary course of business. Furthermore, even within the rm, CSR spending requires the approval of the board and has to comply with the CSR policy laid down by the board. Thus, both internal and external constraints could hamper the creative ability of the organization to use CSR for signaling purposes when compared to other means such as advertisements. CSR Communications- Social Media: We begin with a study of the rm's communication related to CSR. If the 2% rule dilutes CSR's signaling ability, we expect to observe (i) a reduction in CSR related communication; (ii) a reduction in number of words that convey positive CSR related product or service attributes; and (iii) an increase in the standardization of CSR communication. On the other hand, if the mandate makes no dierence to the signaling value or if there is no signaling value to CSR in the rst place, we expect no change in the communication relating to CSR. We identify the ocial Twitter handles of rms and scrape the tweets tweeted during the sample period. We label a tweet as CSR-related if any words related to CSR are mentioned in the tweet. In section 9 of the online appendix, we provide a detailed account of the process used to identify CSR related words. We ask whether the proportion of CSR related tweets comes down in a di-in-di sense. We estimate regression equation 1 with the proportion of CSR-related tweets among all tweets as the dependent variable and present the results in column 1 of Table 8. We nd a 2 percentage points decline in the proportion of CSR related tweets for the mandated rms. The general reduction in CSR-related tweets suggests that CSR-related communication becomes less valuable after the mandate. Next, we investigate the proportion of words within CSR tweets that signal the high quality of a rm's products or services. We use standard marketing dictionaries to identify such words, as explained in section 9.2 of the online appendix. As shown in column 2 of Table 8, we nd a 2 percentage points decline in the proportion of words that signal product or service quality within CSR related tweets in a di-in-di sense. Finally, we check whether virtue-signaling also reduces post the mandate. We identify words that signal charity or good intentions of the rms within CSR-related tweets (see Section 9.2 of the online appendix). Electronic copy available at: https://ssrn.com/abstract=3909219 As shown in column 3 of Table 8, we nd a one percentage point decline in the proportion of words that signal virtue. In the subsequent three columns, where we include rm-level control variables, we nd similar results. These results suggest that voluntary CSR was used to convey higher product quality and virtue, and the eectiveness of the signal reduces after the mandate. CSR Communications- Annual Reports: We then examine the CSR sections of the annual reports. We expect to nd higher standardization of CSR reporting in the annual reports in the post-mandate period. To test the above hypothesis, we calculate the cosine similarity of the text of the CSR section in the annual reports between rms every year. For each mandated (non-mandated) rm, we calculate the cosine similarity of the CSR section of its annual report with that of all other mandated (non-mandated) rms in a year and calculate the average of the cosine similarity score. The comparison is limited to rms whose annual reports are available. The average thus calculated is the dependent variable for each rm-year. Higher standardization is likely to make the reporting more similar. We nd an increase in the standardization in CSR reporting in a di-in-di sense for the mandated group in the post-mandate period. We report the above results in column 1 of Table 9. Increased standardization is consistent with reduced use of CSR for signaling purposes. In columns 2 and 3, we ask whether mandated rms reduce signaling their product quality and virtue. We follow the same methodology as in the case of social media posts. We nd a close to 8 percentage points (6 percentage points) reduction in the signaling of product quality (virtue). Increase in Advertising: A likely consequence of a reduction in signaling ability of CSR after the mandate and the consequent decrease in CSR spending is an increase in expenditure on advertisements. Therefore, we test whether mandated rms increase spending on advertising by estimating regression equation 1. The sample, as before, is restricted to high CSR rms. In Table 10, the level of advertisement expenditure (the ratio between advertisement spending and total We obtain annual reports from www.moneycontrol.com. Electronic copy available at: https://ssrn.com/abstract=3909219 assets) is the dependent variable in Panel A (B). The organization of the table exactly mimics Table 4. We nd a signi cant increase in spending on advertisements in all speci cations. The increase translates to 24.85% of the average advertisement expenditure in the pre- mandate period, and is therefore, materially signi cant. Next, we attempt to understand the relative magnitude of advertisement expenditure in- curred by rms to replace CSR. To this end, we sum the CSR expenditure and advertisement expenditure and use the newly created variable as the dependent variable in the regression equation 1. An increase (decrease) in a di-in-di sense indicates that the increase in adver- tisement expenditure is higher (lower) than the reduction in CSR spending. No change would mean close to a one to one replacement. We present the results in Table A10 of the online appendix. We nd that that the di-in-di coecient is statistically indistinguishable from zero. In other words, it appears that the replacement of CSR expenditure by advertisement expenditure is almost one-to-one. Note that in section 5.2, we nd that mandatory CSR leads to a loss of shareholder value. It seems reasonable to infer that voluntary CSR spending is more valuable than advertisements as a signaling device. Hence, even a close to one-to-one replacement of CSR by advertising does not prevent the loss of shareholder value. The question of why rms do not increase advertisement beyond a one-for-one replace- ment remains. We conjecture that nancial constraints could be a part of the explanation. Perhaps some customers respond only to voluntary CSR as opposed to advertising. The mandate might have also shut the rms' ability to signal virtue. Hence, even a higher level of advertisement spending cannot replace the value loss due to reduced CSR. Finally, rms could have potentially increased spending on other types of signaling such as certi cations and quality control. Such extra spending could plausibly explain the decline in operating performance. Unfortunately, we lack detailed, granular data to verify these conjectures. Electronic copy available at: https://ssrn.com/abstract=3909219 5.3.2 Other Strategic Reasons- Customer Overpayment? Kitzmueller and Shimshack (2012) suggest that some customers are willing to pay more for the products of rms engaged in CSR-related activities. We ask whether the reduction in CSR expenditure by erstwhile high spenders re ects consumers' unwillingness to pay more for products sold by rms whose CSR activities are mandated by the government. The proxy we use is the ex-CSR pro t margin (Panel A of A11). Note that the change in the CSR spending post-mandate does not mechanically in uence this variable. If consumers are unwilling to pay more for a product with mandatory CSR, the ex-CSR margin (pro t margin without considering CSR) should decline. We conduct a di-in-di test of the form of equation 1 with ex-CSR margin as the dependent variable and report the results in Panel A of A11 of the online appendix. We nd no change in a di-in-di sense. The result implies no change in consumers' willingness to pay for products in response to mandated CSR. 5.3.3 Other Strategic Reasons- Labor Donations? We next consider the \labor donation" argument, which posits that some employees are willing to accept lower wages to work for socially responsible rms (Greening and Turban (2000)). We ask whether the dilution of the labor donations channel in response to the mandate leads to reduced spending on CSR by the mandated rms within the high CSR rms. If the labor donation channel works, we would expect wages for the mandated rms within the high CSR group to increase in the post-mandate period. However, we do not nd such an increase as reported in Table A12 of the online appendix. 5.3.4 Other Strategic Reasons- Politics The politics view of CSR posits that CSR deters interventions by activists (Davidson III, Worrell, and El-Jelly (1995)) and governments (Khanna and Anton (2002)). Such interven- In Panel B, we use the margin before CSR and salaries as the dependent variable and nd similar results. The variable accounts for any change in salaries due to the mandate. We elaborate more on this point in Section 5.3.3. Electronic copy available at: https://ssrn.com/abstract=3909219 tions can make a rm's product or service unpopular. However, it is challenging to estimate in advance the level of CSR spending activists and governments expect from the rm. An explicit 2% rule clearly signals what governments and activists expect in terms of CSR. Therefore, rms that invested in CSR with a political motive likely converge their spending to the 2% limit. Those who invested more (less) reduce (increase) CSR expenditure. We rely on highly polluting rms as a sub-sample of companies that likely invest in CSR to deter governments and activists from unwanted intervention. We obtain data relating to highly polluting rms from the database maintained by the Ministry of Environment. Our focus is on the triple interaction between Treatment, Post, and PollutingFirms. In the results reported in Table A13 of the online Appendix, the triple interaction term is statistically indistinguishable from zero. Therefore, reduction in CSR spending is likely not driven by politically sensitive rms. 5.3.5 Revelation of Social Norms We now consider the more general possibility of learning about society's expectations from the mandate. Before the 2% mandate, we observe heterogeneity in CSR spending plausibly because rms have diculty in assessing the social norms for CSR spending. Thus, we would expect to observe a reduction (increase) in spending by high (low) CSR rms after the mandate. In an analogous situation, Rose and Wolfram (2002) nd that an attempt to restrict CEO pay to USD one million using tax policy had the perverse eect of treating USD 1 million as a new expected norm whereby CEO pay increased in cases where it was less than USD 1 million before. If realization of a new social norms is the dominant explanation, both mandated and non- mandated voluntary high spenders should have moved symmetrically towards the 2% limit as the norms apply to all rms. Our results show that, even among high spenders, mandated rms cut CSR spending more. While 43% of non-mandated high spenders cut spending after the mandate, the proportion is more than 80% for the mandated high spenders. Mandated Source: http://pib.nic.in/newsite/PrintRelease.aspx?relid=137373 Electronic copy available at: https://ssrn.com/abstract=3909219 high spenders also cut more of the actual Rupees spent. In sum, we believe the dilution of signaling value of CSR likely explains fall in CSR spending over and above the revelation of societal expectations. Our belief is based on a combined assessment of the following empirical observations (i) excess reduction of CSR spending by mandated high spenders; (ii) the replacement of CSR by advertising; (iii) the loss of shareholder value and operating performance linked to the 2% rule; and (iv) the higher rates of standardized CSR communications. 6 Discussion 6.1 Signaling Impact on Spend Above 2% Thus far, we have interpreted the cut in voluntary CSR spending by the erstwhile high spenders as a dilution in the ability of CSR to signal both product quality and virtue. A skeptic can ask what stops high CSR rms from deriving signaling bene ts on voluntary CSR spending above the mandated 2% level? We believe that regulatory intervention has increased compliance costs and reduced the signaling bene ts of CSR across the board, for both the 2% mandated part and any voluntary spending above that 2% threshold for four reasons. First, CSR spending is not easily divisible into strict 2% and above 2% compartments. As an illustration, consider a mandated rm that intends to spend 10% of pro ts in a year on a CSR mission with signaling bene ts. Assume that the future spending on this project is contingent on the perceived strategic bene ts. Before the mandate, the marketing department could have worked out and executed a CSR project. Post mandate, it is hard to design a CSR budget such that 2% falls within compliance with the law, and the remaining budget is intended to be spent to signal virtue and product quality. Second, before the mandate the rm did not have to worry about any externally imposed de nition of CSR that is also subject to ex-post regulatory interpretation. As mentioned Electronic copy available at: https://ssrn.com/abstract=3909219 in Section 5.3.1, the ministry of corporate aairs now has discretionary powers to question what constitutes CSR for a particular rm. Hence, additional spending over and above 2% level potentially constitutes slack that the rm can use to protect itself against regulatory questioning. Such slack-based spending is unlikely to have the same signaling value as unfettered CSR outlays. Third, the board of directors did not have to get involved in the activity earlier, and their approval was not explicitly required. Relatedly, the rm could exibly continue or stop CSR projects in future years. Now, the 2% will have to be spent every year. Hence, the marketing department may have to commit spending for a certain number of years to obtain board approval. It might be dicult for the board to nd a worthy CSR opportunity at the last minute if a prior project were discontinued. Fourth, the issue of tax-deductibility of the entire CSR expenditure is sub-judice. The revenue department claims that the entire CSR expenditure is not tax-deductible. Because of such additional costs, rms may look for a exible alternative way of signaling. Accordingly, mandated high CSR rms increasingly use advertisements as a replacement for CSR spending. These rms are likely to view CSR as a tax and hence reduce product quality and virtue signaling in CSR communication. As noted in Section 5.3.1, we do nd signi cant changes in CSR communication that show de-emphasize on signaling product quality and virtue. 6.2 Compliance And Signaling The discussion in section 6.1 shows that mandated high CSR rms could be impacted both due to loss of direct signaling value and also higher cost of compliance, making CSR a less preferred way of signaling. We do not have a de nitive way of disentangling the two costs partly because these costs reinforce one another. With that caveat in mind, we test whether larger rms behave any dierently when compared to smaller rms. Larger rms should be able to absorb the xed costs of compliance better. If compliance costs were the prime Electronic copy available at: https://ssrn.com/abstract=3909219 driver behind reduced CSR spending documented in Table 4, larger rms should be less impacted by the mandate. We estimate a triple interaction speci cation (Treated X Post X Large Firms) and report the results in Table A14 of the online appendix. In our main speci cation, where we use the CSR Amount as the dependent variable, larger rms cut CSR spending more than the smaller rms. When we consider the ratio of CSR Amount and assets, we nd that the large rms cut CSR spending as much as the small rms. These ndings are inconsistent with a simple compliance cost explanation. 6.3 Are Control Group Firms Also Aected? A change in the marketing and CSR strategy of one set of rms in an industry impacts all rms in the ecosystem. Even non-mandated rms are likely to change their CSR, commu- nications, and marketing strategy. Therefore, our identi cation strategy of comparing the mandated and non-mandated rms can only capture the relative dierence in the new rules impact and not the complete causal eect. First, as discussed above in Section 6.1, compliance costs are higher for mandated rms. Several restrictions such as the need to obtain the board of directors' approval, scrutiny by the ocials of the ministry of corporate aairs, and others apply only to mandated rms. As noted in Section 2, the ministry has issued over 1,000 notices rejecting claimed expenses as CSR. Non-mandated rms do not face any of these hurdles. As long as they follow accounting standards, no regulatory questions can be asked about the nature of such expenditure. Second, as noted in section 2, mandated rms lose exibility concerning CSR spending as they must spend at least 2% of pro ts every year and strictly adhere to the law. For instance, a strategy of spending a large amount on CSR once in many years and nothing in between becomes impossible for the mandated rms. The non-mandated rms can continue with such a strategy. Third, we nd a higher impact of the law on stock price reaction and operating perfor- mance of mandated set among the high CSR rms. Fourth, mandated rms increase ad- Electronic copy available at: https://ssrn.com/abstract=3909219 vertising expenditure in proportion to the decline in CSR expenditure in the post-mandate period. Finally, textual analysis suggests that the mandated set among the high CSR rms reduces the usage of words that convey product quality and virtue through CSR compared to the non-mandated group among high CSR rms. 6.4 Non Strategic Motives The existence of strategic outcomes alone is not sucient to rule out the existence of non-strategic motives such as stakeholder altruism (Reinhardt, Stavins, and Vietor (2008), B enabou and Tirole (2010)) and manager moral hazard (Cheng, Hong, and Shue (2013), Masulis and Reza (2014)). For instance, a rm may derive signaling bene ts even when altruism motivates its CSR spending. Therefore, we cannot design sharp tests to cleanly separate strategic and non-strategic motives. Our results can only reject non-strategic motives as the sole explanation for voluntary CSR. However, we do not believe that the data support the claim that non-strategic motives primarily drove voluntary CSR spenders, and the signaling bene t was only an unintended consequence. If that were indeed the case, impacted rms would not have increased adver- tisement spending and changed communication strategy in response to the mandate. 7 Conclusion In this study, we examine the impact of a regulatory edict related to minimum CSR spending on the actual CSR spending of rms. We rely on the recent law passed in India that requires all rms above a certain threshold to spend at least 2% of their average three years' pro ts on CSR. We examine the impact of this law on rms that were voluntarily engaged in CSR before the regulation was passed relative to those that were not. Within voluntary high spenders in the pre-mandate period, we compare rms mandated to spend on CSR and those that are not. Electronic copy available at: https://ssrn.com/abstract=3909219 We nd that voluntary spenders reduce their CSR spending signi cantly after the man- date to the legally prescribed limit of 2% of the average three years' pro ts. On the other hand, rms that spent less on CSR during the pre-regulation period increase their spending slightly to meet the new requirement. Our ndings suggest that the imposition of mandatory CSR crowds out voluntary spending on CSR. In the second part of the paper, we attempt to understand the channel at work behind such cuts. The negative share price reaction of the mandated rms that used to spend more than 2% of their pro ts in the pre-mandate period indicates that some strategic motive be- hind CSR becomes less eective after the mandate. We nd that aected rms increase their advertising expenditure. They also change the tone of the CSR communication signi cantly away from signaling product quality and virtue to one of compliance. These results indicate that voluntary CSR signals product quality and virtue to stakeholders. Regulators might want to consider the possible impact of a proposed intervention on the CSR spending of rms that voluntarily engage in pro-social behavior. In the short run, a mandate may plausibly lead to increased total CSR spending because the edict brings a larger number of rms into the mandatory CSR net. In fact, a recent KPMG report shows that the total CSR spending increased following the regulation. However, if the compulsion to spend on CSR crowds out voluntary spending, then such a mandate may lead to a reduction in CSR spending in the long run. Firms that would have voluntarily spent on CSR, with some persuasion by NGOs, may not do so when regulation is imposed. In such a case, the magnitude of CSR spending in the pre-regulation period may not serve as the appropriate counterfactual. Many Indian rms would have potentially initiated CSR investments voluntarily in the absence of a mandate in response to growing prosperity, given that India is among the fastest-growing large economies in the world. We acknowledge that we cannot make de nitive welfare claims. The overall welfare depends on several factors such as (i) the utility of increased spending when the country Source-KPMG Report can be found here: https://assets.kpmg.com/content/dam/kpmg/in/pdf/2018/02/CSR- Survey-Report.pdf Electronic copy available at: https://ssrn.com/abstract=3909219 is underdeveloped relative to spending at a higher stage of development; (ii) the relative eciency of private spending compared with that of government spending on social issues; (iii) the impact of CSR on tax compliance; and (iv) other factors. We do not have credible evidence on these fronts. Further, as acknowledged before, our identi cation strategy allows us to estimate only the dierential impact between the mandated and non-mandated rms as opposed to the complete causal impact of the mandate. Finally, we do not have high-quality survey or eld evidence to test what kind of stakeholders were impacted by dierent types of signaling. We rely instead on carefully constructed archival proxies for drawing inferences on signaling. Our study opens up interesting follow-up questions. First, how does the government of India intend to assess the social impact of mandatory CSR spending? Second, does the imposition of a mandate reduce the responsiveness of rms to local disasters? Top management may feel that they have discharged their social responsibility by complying with the regulation and, hence, do not need to do more. Answers to these questions may reveal additional costs of mandating CSR over and above those identi ed in this study. Electronic copy available at: https://ssrn.com/abstract=3909219 References Benabou, R., and J. Tirole (2010): \Individual and corporate social responsibility," Economica, 77(305), 1{19. Blacconiere, W. G., and W. D. Northcut (1997): \Environmental information and market reactions to environmental legislation," Journal of Accounting, Auditing & Fi- nance, 12(2), 149{178. Chen, Y.-C., M. Hung, and Y. Wang (2018): \The eect of mandatory CSR disclosure on rm pro tability and social externalities: Evidence from China," Journal of Accounting and Economics, 65(1), 169{190. Cheng, B., I. Ioannou, and G. Serafeim (2014): \Corporate social responsibility and access to nance," Strategic Management Journal, 35(1), 1{23. Cheng, H., H. Hong, and K. Shue (2013): \Do managers do good with other people's money?," Discussion paper, National Bureau of Economic Research. Christensen, H. B., E. Floyd, L. Y. Liu, and M. Maffett (2017): \The real eects of mandated information on social responsibility in nancial reports: Evidence from mine- safety records," Journal of Accounting and Economics, 64(2-3), 284{304. Christensen, H. B., L. Hail, and C. Leuz (2013): \Mandatory IFRS reporting and changes in enforcement," Journal of Accounting and Economics, 56(2-3), 147{177. (2019): \Adoption of CSR and Sustainability Reporting Standards: Economic Analysis and Review," Discussion paper, National Bureau of Economic Research. Daske, H., L. Hail, C. Leuz, and R. Verdi (2013): \Adopting a label: Heterogeneity in the economic consequences around IAS/IFRS adoptions," Journal of Accounting Research, 51(3), 495{547. Davidson III, W. N., D. L. Worrell, and A. El-Jelly (1995): \In uencing managers to change unpopular corporate behavior through boycotts and divestitures: A stock market test," Business & Society, 34(2), 171{196. Dedman, E., and A. Kausar (2012): \The impact of voluntary audit on credit ratings: evidence from UK private rms," Accounting and Business Research, 42(4), 397{418. Deng, X., J.-k. Kang, and B. S. Low (2013): \Corporate social responsibility and stake- holder value maximization: Evidence from mergers," Journal of Financial Economics, 110(1), 87{109. Dhaliwal, D. S., O. Z. Li, A. Tsang, and Y. G. Yang (2011): \Voluntary non nancial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting," The Accounting Review, 86(1), 59{100. Dharmapala, D., and V. Khanna (2018): \The impact of mandated corporate social responsibility: evidence from Indias Companies Act of 2013," International Review of Law and Economics, 56, 92{104. Electronic copy available at: https://ssrn.com/abstract=3909219 Dimson, E., O. Karakas , and X. Li (2015): \Active ownership," The Review of Financial Studies, 28(12), 3225{3268. Elliott, W. B., S. M. Grant, and K. M. Rennekamp (2017): \How disclosure fea- tures of corporate social responsibility reports interact with investor numeracy to in uence investor judgments," Contemporary Accounting Research, 34(3), 1596{1621. Flammer, C. (2015): \Does corporate social responsibility lead to superior nancial perfor- mance? A regression discontinuity approach," Management Science, 61(11), 2549{2568. Florou, A., and P. F. Pope (2012): \Mandatory IFRS adoption and institutional in- vestment decisions," The Accounting Review, 87(6), 1993{2025. Greening, D. W., and D. B. Turban (2000): \Corporate social performance as a com- petitive advantage in attracting a quality workforce," Business & society, 39(3), 254{280. Hung, M., and K. Subramanyam (2007): \Financial statement eects of adopting in- ternational accounting standards: the case of Germany," Review of Accounting Studies, 12(4), 623{657. Kausar, A., N. Shroff, and H. White (2016): \Real eects of the audit choice," Journal of Accounting and Economics, 62(1), 157{181. Khanna, M., and W. R. Q. Anton (2002): \What is driving corporate environmentalism: Opportunity or threat?," Corporate Environmental Strategy, 9(4), 409{417. Kim, J.-B., J. S. Tsui, and H. Y. Cheong (2011): \The voluntary adoption of Interna- tional Financial Reporting Standards and loan contracting around the world," Review of Accounting Studies, 16(4), 779{811. Kitzmueller, M., and J. Shimshack (2012): \Economic perspectives on corporate social responsibility," Journal of Economic Literature, 50(1), 51{84. Kotler, P., and N. Lee (2005): \Best of breed: When it comes to gaining a market edge while supporting a social cause,corporate social marketing leads the pack," Social Marketing Quarterly, 11(3-4), 91{103. (2008): Corporate social responsibility: Doing the most good for your company and your cause. John Wiley & Sons. Lennox, C. S., and J. A. Pittman (2011): \Voluntary audits versus mandatory audits," The Accounting Review, 86(5), 1655{1678. Lins, K. V., H. Servaes, and A. Tamayo (2017): \Social capital, trust, and rm per- formance: The value of corporate social responsibility during the nancial crisis," The Journal of Finance. Manchiraju, H., and S. Rajgopal (2017): \Does Corporate Social Responsibility (CSR) Create Shareholder Value? Evidence from the Indian Companies Act 2013," Journal of Accounting Research. Electronic copy available at: https://ssrn.com/abstract=3909219 Margolis, J. D., H. A. Elfenbein, and J. P. Walsh (2009): \Does it pay to be good... and does it matter? A meta-analysis of the relationship between corporate social and nancial performance," Working Paper. Masulis, R. W., and S. W. Reza (2014): \Agency problems of corporate philanthropy," The Review of Financial Studies, 28(2), 592{636. McWilliams, A., D. S. Siegel, and P. M. Wright (2006): \Corporate social respon- sibility: Strategic implications," Journal of Management Studies, 43(1), 1{18. Orlitzky, M., F. L. Schmidt, and S. L. Rynes (2003): \Corporate social and nancial performance: A meta-analysis," Organization Studies, 24(3), 403{441. Perrini, F., A. Russo, A. Tencati, and C. Vurro (2011): \Deconstructing the rela- tionship between corporate social and nancial performance," Journal of Business Ethics, 102(1), 59{76. Reinhardt, F. L., R. N. Stavins, and R. H. Vietor (2008): \Corporate Social Re- sponsibility Through an Economic Lens," Review of Environmental Economics and Policy, 2(2), 219{239. Rose, N. L., and C. Wolfram (2002): \Regulating executive pay: Using the tax code to in uence chief executive ocer compensation," Journal of Labor Economics, 20(S2), S138{S175. Waagstein, P. R. (2011): \The mandatory corporate social responsibility in Indonesia: Problems and implications," Journal of Business Ethics, 98(3), 455{466. Electronic copy available at: https://ssrn.com/abstract=3909219 Figure 1: Comparison Between Mandated And Non-Mandated Firms Within Sub Samples This gure depicts our identi cation strategy. High (low) CSR spenders are those that spend more (less) than 2% of their average three-year pro ts on CSR in the pre-mandate period. The rms that are required by law to spend on CSR form the mandated group and those that are not so required form the non-mandated group. Electronic copy available at: https://ssrn.com/abstract=3909219 Figure 2: Pre And Post Trend- Comparison between Mandated and Non- Mandated Firms Within High CSR Category This gure depicts the pre and post trend between the treated and the control groups in terms of CSR spending. The sample is restricted to high CSR spenders in the pre-mandate period. In other words, all rms in the sample spend more than 2% of their average three- year pro ts on CSR in the pre-mandate period. The rms that are required by law to spend on CSR form the treated group and those that are not so required form the control group. The vertical axis plots the ratio of the CSR amount and the total assets and the horizontal axis denotes the years. The blue line represents the treated group and the orange line the control group. 2010 2012 2014 2016 2018 2020 Year Treatment Control Electronic copy available at: https://ssrn.com/abstract=3909219 CSR_Asset .002 .003 .004 .005 .006 34 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 1: Variable Definition In this table, we de ne the key variables. Variable De nition High CSR Firms Firms that spent more than a threshold, usually 2%, in terms of CSR-to-pro t ratio in the pre-regulation period. We also use other thresholds. Pro t here refers to the average pro ts of the previous three years. Low CSR Firms Firms that spent less than a threshold, usually 2%, in terms of CSR-to-pro t ratio in the pre-regulation period. Post-regulation period Financial years 2014-2015 and after. Mandated Firms Firms that breach in any one or more of the criteria speci ed by Section 135 of the Companies Act. These are INR 50 million in pro ts; INR 5 billion in net worth; INR 10 billion in sales. The values are arrived at based on annual averages in the pre-mandate period. CSR- Amount INR spending on CSR. Pre-regulation period data comes from Prowess; Post-regulation period data comes from the ministry for rms covered by the ministry and from Prowess database for other rms. CSR Ratio The ratio between CSR and average pro ts after tax in the previous three years. Pro ts Pro t After Tax at a rm-year level. Net worth Book value of equity at a rm-year level. TABLE 2: Sample Construction In this table, we report details about the sample used. Variable Value Firms in Prowess Database 43,051 Firms in Ministry Database 12,097 Firms in MCA that could be merged with Prowess 10,154 Firms in both the pre and post-mandate Period 39,309 Number of sample years 10 Total observations in the merged data set 236,044 Total observations with non-missing average CSR numbers in the pre-mandate period 44,769 Observations with CSR more than 2% of average three year pro ts 16,251 Observations with CSR less than 2% of average three year pro ts 28,518 Observations with CSR more than 5% of average three year pro ts 9,108 Observations with CSR more than 7.5% of average three year pro ts 6,873 Observations with CSR more than 10% of average three year pro ts 5,523 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 3: Univariate Comparison Between High And Low CSR Firms In this table, we present univariate comparisons between the high CSR and the low CSR rms. CSR Ratio, as de ned in Table 1, is the dependent variable. The sample is restricted to rms that earn positive pro ts across all the sample years. Firms are grouped based on their pre-regulation spending on CSR. In column 1, we consider rms that spent nothing on CSR in the pre-mandate period. In column 2, we consider rms that spent between 0 to 1 percent of previous three years' average pro ts in the pre-regulation period. Similarly, in each column, we consider a progressively higher range. We compare the dierence between pre-regulation and post-regulation expenditure and also report the t-statistics. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. Group-Based Pre-Period CSR Ratio 0 0-1 1-2 2-3 3-4 4-5 5-6 6-7 7-10 More than 10 Pre 0.000 0.004 0.015 0.025 0.036 0.046 0.056 0.066 0.087 0.276 Post 0.024 0.021 0.024 0.026 0.030 0.036 0.031 0.038 0.050 0.103 Di(post-pre) -0.024 -0.017 -0.009 -0.001 0.006 0.010 0.025 0.029 0.037 0.173 T-Stat -11.124 -28.556 -8.074 -0.697 2.491 2.672 6.807 5.098 8.004 24.777 No. of observations 2878 15040 5718 3079 1975 1501 1031 676 1230 4857 Electronic copy available at: https://ssrn.com/abstract=3909219 37 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 4: Comparison Between Mandated and Non-Mandated Firms Within High CSR Firms In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In panel A, the CSR Amount, as de ned in Table 1, is the dependent variable. In Panel B, the ratio between CSR amount and total assets is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the last three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is a mandated rm and zero otherwise. The rm-year level controls included in columns 5 to 8 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. Dependent Variable Panel A: CSR Amount Treatment X Post-Mandate -7.87*** -18.58*** -24.63*** -27.80*** -7.49*** -18.98*** -25.17*** -29.00*** [-5.84] [-7.60] [-7.51] [-7.87] [-5.33] [-8.13] [-7.98] [-7.93] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,678 8,708 6,544 5,251 15,673 8,704 6,541 5,248 R-squared 0.71 0.66 0.67 0.68 0.72 0.68 0.68 0.68 Dependent Variable Panel B: CSR/Total Assets Treatment X Post-Mandate -0.0014*** -0.0040*** -0.0056*** -0.0067*** -0.0014*** -0.0041*** -0.0057*** -0.0067*** [-3.1038] [-5.4022] [-5.6622] [-5.5534] [-3.2415] [-5.6697] [-6.1635] [-6.3933] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,073 8,288 6,216 4,971 15,070 8,285 6,213 4,968 R-squared 0.56 0.57 0.58 0.59 0.56 0.57 0.58 0.59 TABLE 5: Summary Comparison In this table, we compare the treated and the control groups in terms of various accounting ratios. Mandated (Non-mandated) rms form the treated (control) groups. The sample is restricted to rms that spend more than 2% of their average three-year pro ts on CSR in the pre-mandate period. Variable Control Treatment Dierence T- Stat Margin 0.12 0.13 -0.01 -0.95 Leverage 1.13 0.63 0.50 1.24 Turnover 1.40 0.91 0.48 0.81 Return On Equity 0.41 0.35 0.06 0.25 Advertisement To Sales Ratio 0.02 0.01 0.00 0.82 Pro t Per Employee 0.44 0.28 0.16 0.10 Electronic copy available at: https://ssrn.com/abstract=3909219 39 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 6: Comparison Between Mandated and Non-Mandated Firms Within High CSR Firms In Terms of Oper- ating Performance In this table, we compare the operating performance before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The return on equity (ROE) (the return on assets (ROA))(natural logarithm of sales) is the dependent variable in columns 1 and 4 (2 and 5) (3 and 6). The sample is restricted to rms investing above 2% of last three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls such as one year lagged pro ts and net worth in columns 4 and 5. In column 6, we use the current year pro ts and net worth as control variables. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) Dependent Variable ROE ROA log Sales ROE ROA log Sales Treatment x Post-mandate -0.0229* -0.0093*** 0.0720 -0.0263* -0.0099*** 0.0605 [-1.7267] [-2.8758] [1.4981] [-1.7999] [-2.8863] [1.4045] Lag Pro t After Tax 0.0000 0.0000 [0.8915] [1.6313] Net Worth -0.0000 0.0000 0.0000 [-1.2108] [1.6392] [1.0428] Pro t After Tax 0.0000 [1.5784] Firm Fixed Eects Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Observations 23,629 23,629 23,521 22,389 22,387 23,520 R-squared 0.3768 0.4606 0.8850 0.3927 0.4774 0.8854 TABLE 7: The Value Impact Of Mandatory CSR In this table, we present the results relating to stock price reaction to mandatory CSR. We use the July 17, 2019 announcement that made the violation of mandatory CSR a criminal oense as the event in Panel A and the announcement of the withdrawal of the above provision on August 23, 2019, as the event in Panel B. The three-day excess return around the event is the dependent variable. Treatment is a dummy variable that takes the value of one if the rm under consideration is mandated to invest in CSR and zero otherwise. In columns 1 and 2, we consider all rms. In columns 3 and 4, the data are restricted to rms that used to spend more than 2% of their average three-year pro ts on CSR in the pre-mandate period. To account for liquidity, we include turnover as a control variable in columns 2 and 4. We employ industry xed eects and cluster the errors at an industry level. Robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. Panel A: CSR violation to be a criminal oence (1) (2) (3) (4) Dependent Variable Excess Return Mandatory -0.005*** -0.012*** -0.005* -0.008* (0.002) (0.003) (0.003) (0.004) Turnover 0.002 0.008*** 0.008*** 0.013*** (0.002) (0.003) (0.002) (0.003) Industry F.E Yes Yes Yes Yes Sample All Firms High CSR Firms Observations 1,994 1,994 824 824 R-squared 0.094 0.125 0.047 0.061 Panel B: CSR violation not to be a criminal oence Dependent Variable Excess Return Mandatory 0.012*** 0.014*** 0.013*** 0.017*** (0.003) (0.004) (0.004) (0.006) Turnover -0.000 0.001 -0.003 0.000 (0.003) (0.003) (0.003) (0.003) Industry F.E Yes Yes Yes Yes Sample All Firms High CSR Firms Observations 1,994 1,994 824 824 R-squared 0.091 0.093 0.068 0.059 Electronic copy available at: https://ssrn.com/abstract=3909219 41 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 8: Test of Signaling Hypothesis- Social Media Communication In this table, we examine the social media communications of rms through their ocial Twitter handles. Each observation represents a rm-year. In columns 1 and 4, the proportion of CSR related tweets over all tweets is the dependent variable. In columns 2 and 5 (3 and 6), the proportion of words signaling product or service quality (virtue) within CSR related tweets is the dependent variable. The procedure used to identify CSR related tweets, words signaling quality of products and services, and words signaling virtue is detailed in Section 9 of the online appendix. The sample is restricted to rms investing above 2% of the proportion of the previous three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. The rm-year level controls included in columns 4 to 6 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) Dependent Variable prop csr prop product prop charity prop csr prop product prop charity Treatment x Post -0.02* -0.02* -0.01** -0.03* -0.02* -0.01** [-1.87] [-1.99] [-2.24] [-1.99] [-1.89] [-2.20] Firm Level Controls No No No Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Observations 704 500 409 704 500 409 R-squared 0.58 0.39 0.61 0.58 0.39 0.61 42 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 9: Test of Signaling Hypothesis- Annual Reports In this table, we examine the CSR sections of annual reports of rms. Each observation represents a rm-year. In columns 1 and 4, average cosine similarity at a rm-year level is the dependent variable. For each mandated (non-mandated) rm, we calculate the cosine similarity of the CSR section of the annual report with every other mandated (non-mandated) rm in a year and calculate the average of the cosine similarity score. The average so calculated is the dependent variable. In columns 2 and 5 (3 and 6), the proportion of words signaling product or service quality (virtue) within CSR section of the annual report is the dependent variable. The procedure used to identify words signaling quality of products and services, and words signaling virtue is detailed in Section 9 of the online appendix. The sample is restricted to rms investing above 2% of the proportion of the previous three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. The rm-year level controls included in columns 4 to 6 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) Dependent Variable Similarity Product Charity Similarity Product Charity Treatment x Post-mandate 0.21*** -0.19*** -0.08*** 0.22*** -0.19*** -0.09*** [3.78] [-6.73] [-6.54] [3.86] [-6.21] [-6.04] Firm Level Controls No No No Yes Yes Yes Firm Fixed Eect Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Observations 242 242 242 242 242 242 R-squared 0.70 0.66 0.78 0.70 0.66 0.78 43 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 10: Test of Signaling Hypothesis- Advertisement In this table, we compare advertising expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In Panel A (B), the INR amount spent on advertising (the ratio between advertisement expenditure and total assets) is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. The rm-year level controls included in columns 5 to 8 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Panel A: Advertising Expenditure Treatment X Post-mandate 31.97*** 32.64*** 36.79*** 38.66** 31.56*** 33.55*** 37.03*** 38.00*** [4.97] [2.80] [2.76] [2.60] [5.25] [3.18] [2.92] [2.80] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 16,300 8,964 6,711 5,397 16,292 8,961 6,708 5,394 R-squared 0.87 0.89 0.89 0.88 0.87 0.90 0.89 0.88 Dependent Variable Panel B: Advertisement/Total Assets Treatment X Post-mandate 0.0052** 0.0071* 0.0050 0.0081* 0.0051** 0.0070* 0.0049 0.0079* [2.1987] [1.8668] [1.0296] [1.7337] [2.1703] [1.8227] [0.9889] [1.6888] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,657 8,554 6,398 5,135 15,652 8,552 6,396 5,133 R-squared 0.59 0.59 0.57 0.62 0.59 0.59 0.57 0.62 Online Appendix Does Mandated Corporate Social Responsibility Reduce Intrinsic Motivation? Evidence from India Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A1: Comparison of Difference Between CSR Amount As Per Two Databases In this table, we present the distribution of the dierence between the CSR amount as per the prowess database and the ministry database for the same rm. The dierence is normalized using pro ts. Statistic Value Mean 0 Median 0 Q1 0 Q3 0 5th Percentile 0 95th Percentile 0 1st Percentile -10.7 99th Percentile 0.15 Electronic copy available at: https://ssrn.com/abstract=3909219 46 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A2: Comparison of Difference Between CSR Amount As Per Two Databases-Examples In this table, we provide ve examples of large deviations between the Prowess and MCA database. Company Name Year Actual CSR (INR mill) Ministry CSR (INR mill) Prowess CSR (INR mill) Reliance Industries Limited 2015 7,610 7,610 15,660 Reliance Industries Limited 2016 6,520 6,520 13,280 NTPC Limited 2016 4,918 4,918 8,272 South Eastern Coal elds Limited 2015 404 404 1,313.3 Power Finance Corporation 2015 516.8 516.8 1,178.3 47 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A3: Pre And Post Comparison Within Mandated High CSR Firms In this table, we compare CSR expenditure before and after the government mandate for mandated high CSR rms. Each observation represents a rm-year. In Panel A, CSR Amount, as de ned in Table 1, is the dependent variable. In Panel B, the ratio between CSR amount and total assets is the dependent variable. The sample is restricted to mandated rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The rms that are required to spend on CSR as per law are considered mandated. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value of one for years after the regulation change, and zero otherwise. The rm-year level controls included in columns 5 to 8 are pro t after tax and net worth. We also include rm xed eects in all 8 columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Panel A: CSR Amount post-mandate -7.63*** -18.58*** -24.84*** -28.06*** -7.48*** -18.93*** -25.14*** -28.97*** [-5.60] [-7.61] [-7.52] [-7.73] [-5.31] [-8.09] [-7.95] [-7.90] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 5,869 2,486 1,633 1,215 5,867 2,485 1,633 1,215 R-squared 0.68 0.62 0.62 0.63 0.70 0.64 0.64 0.64 Dependent Variable Panel B: CSR To Assets Ratio post-mandate -0.0039*** -0.0077*** -0.0099*** -0.0114*** -0.0039*** -0.0077*** -0.0099*** -0.0114*** [-9.7106] [-10.5799] [-9.5238] [-9.0795] [-10.9308] [-11.5832] [-11.3180] [-11.2556] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 5,755 2,426 1,588 1,181 5,755 2,426 1,588 1,181 R-squared 0.46 0.51 0.53 0.56 0.46 0.51 0.54 0.56 48 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A4: Comparison Between Mandated and Non-Mandated Firms Within Low CSR Firms In this table, we compare CSR expenditure before and after the government mandate for rms that used to spend less than 2% of average three years' pro ts before the mandate. CSR Amount, as de ned in Table 1, is the dependent variable in columns 1, 2, 3 and 4. The ratio between CSR amount and the total assets is the dependent variable in columns 5, 6, 7, and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. In columns 1, 2, 5, and 6, we compare the post and pre-mandate spending by rms that are required to spend on CSR by law but were spending less than the stipulated level in the pre-mandate period. Each observation represents a rm-year. In columns 3, 4, 7, and 8 we conduct di-in-di tests. Treatment and Control are as de ned in Table 4. We include rm-year level controls in columns 2, 4, 6, and 8. We include rm xed eects in all columns and year xed eects in columns 3, 4, 7, and 8. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR/Total Assets Post-Mandate 9.23*** 8.81*** 0.0004*** 0.0004*** [9.84] [10.04] [4.2066] [4.2267] Post-Mandate X Treatment 12.98*** 12.49*** 0.0006*** 0.0006*** [9.97] [10.16] [5.2417] [5.2892] Firm-Level Controls No Yes No Yes No Yes No Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects No No Yes Yes No No Yes Yes Observations 27,932 27,913 27,932 27,913 27,527 27,509 27,527 27,509 R-squared 0.68 0.69 0.69 0.70 0.5533 0.5547 0.5568 0.5582 49 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A5: Impact of the CSR Mandate- Triple Interaction In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms in a triple interaction framework. Each observation represents a rm-year. CSR Amount, as de ned in Table 1, is the dependent variable in columns 1 and 2. The ratio between CSR amount and total assets is the dependent variable in columns 3 and 4. High CSR is a dummy variable that takes the value of one for rms investing above 2% of the previous three years' average pro ts in the pre-mandate period and zero otherwise. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in columns 2 and 4. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) Dependent Variable CSR Amount CSR/Total Assets Treatment x Post x High CSR -20.91*** -20.51*** -0.0020*** -0.0020*** [-12.09] [-12.41] [-4.6661] [-4.6519] Mandate x Post 13.00*** 12.54*** 0.0006*** 0.0006*** [9.96] [10.26] [5.3128] [5.3554] High CSR x Post -0.79*** -0.79*** -0.0025*** -0.0025*** [-3.31] [-3.29] [-10.3162] [-10.3050] Firm-Level Controls No Yes No Yes Firm Fixed Eects Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Observations 43,610 43,586 42,600 42,579 R-squared 0.70 0.70 0.5841 0.5846 50 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A6: Comparison Between Mandated and Non-Mandated Firms Within High CSR Firms- Test of Pre- mandate Trend In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The ratio between CSR Amount, as de ned in Table 1, and total assets is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in column 1, 5% in column 2, 7.5% in column 3, and 10% in column 4. Post- mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Pre Year 1 is a dummy variable taking the value one for the year just before the mandate and zero otherwise. Pre Year 2 is a dummy variable taking the value one for the year two years before the mandate and zero otherwise. Pre Year 3 is a dummy variable taking the value one for the year three years before the mandate and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is a mandated rm and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) Dependent Variable CSR/Total Assets post-mandate X Treatment -0.0014*** -0.0043*** -0.0058*** -0.0066*** [-2.6571] [-4.6457] [-4.4503] [-4.2774] Pre Year 3 X Treatment 0.0001 -0.0006 -0.0003 0.0000 [0.1501] [-0.6293] [-0.2653] [0.0248] Pre Year 2 X Treatment 0.0000 -0.0002 0.0001 0.0005 [0.0710] [-0.1855] [0.0988] [0.2709] Pre Year 1 X Treatment -0.0001 -0.0002 0.0002 0.0005 [-0.2374] [-0.2557] [0.1583] [0.2980] Firm Level Controls Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Observations 15,070 8,285 6,213 4,968 R-squared 0.56 0.57 0.58 0.59 51 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A7: Comparison Between Mandated and Non-Mandated Firms- High CSR- Dropping Years In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In columns 1 to 4 (5 to 8) CSR amount (the ratio between CSR amount and total assets) is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. We leave out years 2012-2013 and 2013-2014. We do not consider the above years while identifying the treated and control groups in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR/Total Assets post-mandate X Treatment -6.63*** -19.29*** -24.43*** -29.09*** -0.0011** -0.0037*** -0.0052*** -0.0065*** [-3.46] [-5.55] [-5.68] [-5.95] [-2.0871] [-3.8028] [-4.7241] [-4.6288] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 10,074 5,576 4,131 3,322 9,657 5,276 3,901 3,124 R-squared 0.72 0.68 0.68 0.69 0.5662 0.5760 0.5820 0.5936 52 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A8: Comparison Between Mandated and Non-Mandated Firms- High CSR-Placebo Test In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. We use a false treatment year of 2011-2012. We restrict the sample to pre-mandate period. Each observation represents a rm-year. CSR amount (the ratio between CSR amount and total assets), as de ned in Table 1, is the dependent variable in columns 1 to 4 (5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-placebo treatment period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the placebo treatment year, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR /Total Assets Treatment X Post-mandate -2.02 -3.65 -3.77 -3.76 -0.0002 -0.0008 -0.0012 -0.0013 [-1.52] [-1.44] [-1.04] [-1.19] [-0.4258] [-0.8285] [-0.8319] [-0.8894] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 5,855 3,508 2,664 2,177 5,583 3,310 2,509 2,046 R-squared 0.87 0.88 0.87 0.87 0.6824 0.6699 0.6676 0.6604 53 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A9: Comparison Between Mandated and Non-Mandated Firms- High CSR-Excluding Firms Close to the Threshold In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. CSR amount (the ratio between CSR amount and total assets), as de ned in Table 1, is the dependent variable in columns 1 to 4 (5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. We exclude control group rms close to the threshold. We rst calculate the distance of a rm from the threshold following Manchiraju and Rajgopal (2017) and exclude those that are less than one standard deviation of such distances for all rms from the threshold. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR /Total Assets Treatment X Post-mandate -5.67*** -12.54*** -16.47*** -21.20*** -0.0016*** -0.0040*** -0.0055*** -0.0068*** [-6.41] [-7.41] [-6.71] [-6.90] [-3.2012] [-4.9494] [-5.1719] [-5.4145] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 14,346 8,242 6,232 5,021 13,759 7,829 5,909 4,743 R-squared 0.56 0.58 0.58 0.59 0.5738 0.5844 0.5921 0.6011 54 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A10: Test of Signaling Hypothesis- CSR And Advertising Expenditure In this table, we compare the sum of advertising and CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The sum of the amount spent on advertising and CSR (the ratio between the sum of the amount spent on advertising and CSR and total assets) is the dependent variable in columns 1 to 4 (5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Advertisement and CSR Amount Advertisement and CSR to Assets Ratio Treatment x Post-mandate 1,053.01 869.55 1,302.59 1,751.16 -0.09 869.55 1,302.59 1,751.16 [1.38] [1.00] [0.99] [1.00] [-0.32] [1.00] [0.99] [1.00] Firm Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 20,001 11,326 8,573 6,898 19,217 11,326 8,573 6,898 R-squared 0.10 0.11 0.11 0.11 0.10 0.11 0.11 0.11 55 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A11: Comparison Between Mandated and Non-Mandated Firms Within High CSR- The Consumer Channel In this table, we compare the ex-CSR margin before and after the government mandate between treatment and control rms. In Panel A (B), margin before CSR (margin before CSR and salary) is the dependent variable. Each observation represents a rm-year. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in columns ve to eight. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Panel A: Ex CSR Margin Treatment X Post-mandate 0.19 -0.84 -1.62 -2.39 -0.64 -1.37 -2.04 -2.72 [0.55] [-0.86] [-0.95] [-1.05] [-1.14] [-1.12] [-1.08] [-1.07] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Observations 23,519 13,672 10,440 8,418 23,518 13,671 10,439 8,417 R-squared 0.13 0.14 0.13 0.13 0.13 0.14 0.13 0.13 Dependent Variable Panel B: Margin including CSR Exp and Wages Treatment X Post-mandate -0.88 -1.81 -2.54 -3.33 -0.88 -1.81 -2.54 -3.33 [-1.28] [-1.17] [-1.09] [-1.09] [-1.28] [-1.17] [-1.09] [-1.09] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 23,254 13,571 10,387 8,369 23,254 13,571 10,387 8,369 R-squared 0.14 0.14 0.13 0.13 0.14 0.14 0.13 0.13 56 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A12: Comparison Between Mandated and Non-Mandated Firms Within High CSR- Labor Donations Chan- nel In this table, we compare the wage expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in columns 5 to 8. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Wages Treatment X Post-mandate 0.0006 0.0011 0.0015 0.0018 -0.0000 0.0001 0.0002 0.0004 [1.1855] [1.1761] [1.1001] [1.0992] [-0.1142] [0.2620] [0.2686] [0.3335] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Observations 22,006 12,752 9,724 7,836 22,005 12,751 9,723 7,835 R-squared 0.2175 0.2154 0.1964 0.1881 0.2172 0.2150 0.1958 0.1873 TABLE A13: Preempting Public Or Private Politics In this table, we compare the CSR expenditure before and after the government mandate between treatment and control rms in a triple interaction framework. CSR Amount, as de ned in Table 1 (the ratio between CSR amount and total assets), is the dependent variable in columns 1 to 4 (columns 5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in column 1, 5% in column 2, 7.5% in column 3, and 10% in column 4. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. Red is a dummy variable that takes the value of one for rms belonging to polluting industries. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR/Total Assets Treatment X Post-mandate -7.77*** -17.91*** -22.47*** -26.04*** -0.0011** -0.0037*** -0.0051*** -0.0055*** [-5.03] [-6.42] [-6.33] [-6.78] [-2.3399] [-5.0419] [-5.6535] [-5.3400] Red X Post 0.38 0.39 0.46 0.58 0.0008* 0.0012* 0.0014 0.0020** [1.46] [0.81] [0.69] [0.73] [1.6826] [1.7018] [1.6311] [2.1020] Post X Red X Treatment 0.28 -2.76 -6.72 -7.15 -0.0009 -0.0011 -0.0018 -0.0034 [0.10] [-0.59] [-1.12] [-1.01] [-1.2113] [-0.8661] [-1.0080] [-1.6414] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,519 8,598 6,452 5,166 14,953 8,205 6,147 4,906 R-squared 0.72 0.68 0.68 0.68 0.5645 0.5786 0.5872 0.5981 Electronic copy available at: https://ssrn.com/abstract=3909219 58 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A14: Comparison Between Mandated and Non-Mandated Firms- Compliance And Signaling In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In columns 1 and 2 (3 and 4) CSR amount (the ratio between CSR amount and total assets) is the dependent variable. The sample is restricted to rms investing above a 2% in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. Large rms is a dummy variable that takes the value of one for rms above the medium in terms asset values in the pre-mandate period and zero otherwise. We include rm-year level controls in columns 2 and 4. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) Dependent Variable CSR Amount CSR Amount CSR Asset CSR Asset Post x Mandate -3.26*** -3.19*** -0.0025** -0.0025** [-3.06] [-3.00] [-2.1755] [-2.1596] Post x Large Firms -0.49 -0.50 0.0001 0.0001 [-1.15] [-1.18] [0.2097] [0.2180] Large Firms x Mandate x Post -4.85*** -4.68*** 0.0012 0.0012 [-2.89] [-2.85] [1.0260] [1.0201] Controls No Yes No Yes Year Fixed Eect Yes Yes Yes Yes Firm Fixed Eect Yes Yes Yes Yes Observations 15,678 15,673 15,073 15,070 R-squared 0.71 0.72 0.5638 0.5638 8 Industry Level CSR Spending Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A15: Industry Level CSR Spending: In this table, we calculate CSR expenditure by rms before the mandate at an industry level. Here Industry classi cation follows the three digit National Industrial Classi cation (NIC- 2008) by Ministry of Corporate Aairs (MCA). Column 1 presents the total CSR Amount, that is aggregate value of CSR expenditure by an Industry. Column 2 presents average CSR Amount per rm, that is aggregate CSR expenditure (Column 1) by total rm years in an industry. Column 3 presents average CSR ratio, that is aggregate CSR ratio, de ned as per table1, by total rm years in an industry. NIC Code Industry Total CSR Average CSR Average CSR Amount(INR Amount(INR Proportion Million) Million) 11 Growing of non-perennial crops 561.60 4.07 0.0026 12 Growing of perennial crops 31.70 0.79 0.0020 13 Plant propagation 0.10 0.05 0.0029 14 Animal production 92.10 1.00 0.0024 16 Support activities to agriculture and post-harvest crop activities 23.40 0.90 0.0021 23 Gathering of non-wood forest products 155.00 6.46 0.0016 24 Support services to forestry 24.10 2.01 0.0106 32 Aquaculture 123.50 3.63 0.0073 51 Mining of hard coal 7848.20 115.41 0.0095 52 Mining of lignite 339.60 84.90 0.0031 61 Extraction of crude petroleum 716.50 65.14 0.0008 71 Mining of iron ores 2598.60 66.63 0.0040 72 Mining of non-ferrous metal ores 1335.90 46.07 0.0071 81 Quarrying of stone, sand and clay 81.70 1.00 0.0021 89 Mining and quarrying n.e.c. 1267.60 37.28 0.0061 101 Processing and preserving of meat 2142.80 52.26 0.0156 102 Processing and preserving of sh, crustaceans and molluscs 139.00 3.31 0.0073 103 Processing and preserving of fruit and vegetables 1.60 0.15 0.0034 104 Manufacture of vegetable and animal oils and fats 398.30 2.36 0.0012 105 Manufacture of dairy products 58.10 0.95 0.0027 106 Manufacture of grain mill products, starches and starch products 83.50 0.98 0.0017 107 Manufacture of other food products 1687.80 2.96 0.0028 108 Manufacture of prepared animal feeds 30.20 0.86 0.0026 110 Manufacture of beverages 964.80 4.14 0.0026 120 Manufacture of tobacco products 1192.60 31.38 0.0035 131 Spinning, weaving and nishing of textiles 1753.90 2.40 0.0017 139 Manufacture of other textiles 80.80 0.64 0.0013 141 Manufacture of wearing apparel, except fur apparel 266.70 1.91 0.0029 143 Manufacture of knitted and crocheted apparel 36.40 0.81 0.0006 151 Tanning and dressing of leather; manufacture of luggage, handbags, sad- 111.80 3.73 0.0126 dlery and harness; dressing and dyeing of fur 152 Manufacture of footwear 134.10 2.16 0.0041 161 Sawmilling and planing of wood 0.40 0.20 0.0011 162 Manufacture of products of wood, cork, straw and plaiting materials 119.10 1.89 0.0016 170 Manufacture of paper and paper products 750.20 2.57 0.0023 181 Printing and service activities related to printing 287.40 13.06 0.0015 191 Manufacture of coke oven products 0.80 0.11 0.0025 192 Manufacture of re ned petroleum products 4191.60 38.11 0.0016 201 Manufacture of basic chemicals, fertilizer and nitrogen compounds, plastics 4929.90 7.90 0.0062 and synthetic rubber in primary forms 202 Manufacture of other chemical products 6597.70 12.15 0.0042 203 Manufacture of man-made bres 1004.80 23.92 0.0018 210 Manufacture of pharmaceuticals, medicinal chemical and botanical prod- 7637.30 10.31 0.0033 ucts 221 Manufacture of rubber products 215.30 1.84 0.0015 222 Manufacture of plastics products 637.60 1.54 0.0014 231 Manufacture of glass and glass products 155.20 3.23 0.0011 239 Manufacture of non-metallic mineral products n.e.c. 3867.30 10.62 0.0019 241 Manufacture of basic iron and steel 2926.90 3.35 0.0010 242 Manufacture of basic precious and other non-ferrous metals 4496.20 27.58 0.0032 243 Casting of metals 415.10 1.63 0.0014 251 Manufacture of structural metal products, tanks, reservoirs and steam 1335.80 12.97 0.0012 generators Electronic copy available at: https://ssrn.com/abstract=3909219 NIC Code Industry Total CSR Average CSR Average CSR Amount(INR Amount(INR Proportion Million) Million) 252 Manufacture of weapons and ammunition 0.10 0.10 0.0029 259 Manufacture of other fabricated metal products; metalworking service ac- 259.70 1.06 0.0016 tivities 261 Manufacture of electronic components 46.90 0.72 0.0029 262 Manufacture of computers and peripheral equipment 20.30 1.56 0.0008 263 Manufacture of communication equipment 269.40 5.99 0.0024 264 Manufacture of consumer electronics 562.90 43.30 0.0005 265 Manufacture of measuring, testing, navigating and control equipment; 448.70 4.67 0.0019 watches and clocks 266 Manufacture of irradiation, electromedical and electrotherapeutic equip- 32.00 1.19 0.0027 ment 271 Manufacture of electric motors, generators, transformers and electricity 355.80 2.34 0.0012 distribution and control apparatus 272 Manufacture of batteries and accumulators 583.60 18.24 0.0021 273 Manufacture of wiring and wiring devices 292.20 2.40 0.0016 274 Manufacture of electric lighting equipment 46.10 2.00 0.0007 275 Manufacture of domestic appliances 59.70 1.57 0.0024 279 Manufacture of other electrical equipment 171.90 1.87 0.0019 281 Manufacture of general purpose machinery 1749.70 5.07 0.0021 282 Manufacture of special-purpose machinery 841.00 3.05 0.0020 291 Manufacture of motor vehicles 1438.70 49.61 0.0005 292 Manufacture of bodies (coachwork) for motor vehicles; manufacture of 442.20 2.70 0.0010 trailers and semi-trailers 293 Manufacture of parts and accessories for motor vehicles 552.70 1.42 0.0015 301 Building of ships and boats 521.40 18.62 0.0005 302 Manufacture of railway locomotives and rolling stock 18.70 0.89 0.0017 303 Manufacture of air and spacecraft and related machinery 0.00 0.00 0.0000 309 Manufacture of transport equipment n.e.c. 989.00 20.60 0.0014 310 Manufacture of furniture 22.90 2.86 0.0055 321 Manufacture of jewellery, bijouterie and related articles 941.90 4.40 0.0012 323 Manufacture of sports goods 5.30 0.53 0.0017 325 Manufacture of medical and dental instruments and supplies 20.70 1.15 0.0017 329 Other manufacturing n.e.c. 3.00 0.43 0.0009 351 Electric power generation, transmission and distribution 14008.10 25.66 0.0023 360 Water collection, treatment and supply 42.10 14.03 0.0002 370 Sewerage 0.20 0.10 0.0053 382 Waste treatment and disposal 1.10 1.10 0.0202 410 Construction of buildings 4400.20 4.99 0.0022 421 Construction of roads and railways 2018.90 13.20 0.0018 422 Construction of utility projects 1521.00 6.98 0.0015 429 Construction of other civil engineering projects 2571.90 6.26 0.0025 431 Demolition and site preparation 139.10 4.64 0.0017 432 Electrical, plumbing and other construction installation activities 0.50 0.50 451 Sale of motor vehicles 177.80 1.57 0.0018 453 Sale of motor vehicle parts and accessories 37.40 1.04 0.0008 454 Sale, maintenance and repair of motorcycles and related parts and acces- 0.50 0.10 0.0012 sories 461 Wholesale on a fee or contract basis 801.60 2.17 0.0048 462 Wholesale of agricultural raw materials and live animals 294.50 2.45 0.0045 463 Wholesale of food, beverages and tobacco 176.90 1.47 0.0045 464 Wholesale of household goods 585.80 1.41 0.0028 465 Wholesale of machinery, equipment and supplies 705.40 2.20 0.0035 466 Other specialized wholesale 2229.50 3.60 0.0023 469 Non-specialized wholesale trade 1217.00 2.28 0.0047 471 Retail sale in non-specialized stores 9.20 0.58 0.0027 472 Retail sale of food, beverages and tobacco in specialized stores 65.10 10.85 0.0174 Electronic copy available at: https://ssrn.com/abstract=3909219 NIC Code Industry Total CSR Average CSR Average CSR Amount(INR Amount(INR Proportion Million) Million) 474 Retail sale of information and communications equipment in specialized 38.50 1.75 0.0009 stores 475 Retail sale of other household equipment in specialized stores 21.10 0.50 0.0013 476 Retail sale of cultural and recreation goods in specialized stores 4.30 1.08 0.0008 477 Retail sale of other goods in specialized stores 199.90 2.30 0.0024 479 Retail trade not in stores, stalls or markets 1.30 0.16 0.0057 491 Transport via railways 777.10 33.79 0.0030 492 Other land transport 160.50 1.56 0.0024 493 Transport via pipeline 179.60 8.98 0.0008 501 Sea and coastal water transport 49.40 1.34 0.0023 511 Passenger air transport 64.60 5.38 0.0009 521 Warehousing and storage 1885.00 15.45 0.0016 522 Support activities for transportation 2383.60 7.24 0.0028 532 Courier activities 24.90 1.92 0.0050 551 Short term accommodation activities 610.00 1.89 0.0023 563 Beverage serving activities 1.70 0.34 0.0037 581 Publishing of books, periodicals and other publishing activities 625.00 7.18 0.0023 591 Motion picture, video and television programme activities 87.20 1.25 0.0013 601 Radio broadcasting 0.00 0.00 0.0000 602 Television programming and broadcasting activities 186.70 7.47 0.0005 611 Wired telecommunications activities 31.20 1.42 0.0029 612 Wireless telecommunications activities 1267.60 57.62 0.0003 619 Other telecommunications activities 9.80 0.58 0.0015 620 Computer programming, consultancy and related activities 2613.80 7.97 0.0030 631 Data processing, hosting and related activities; web portals 5.60 0.40 0.0030 639 Other information service activities 372.70 4.10 0.0019 641 Monetary intermediation 3678.20 3.56 0.0028 643 Trusts, funds and other nancial vehicles 3714.40 5.79 0.0041 649 Other nancial service activities, except insurance and pension funding 4705.60 5.63 0.0025 activities 651 Insurance 5.50 5.50 0.0378 661 Activities auxiliary to nancial service activities, except insurance and 1166.60 3.61 0.0019 pension funding 663 Fund management activities 29.80 0.56 0.0007 682 Real estate activities on a fee or contract basis 0.30 0.15 0.0004 691 Legal activities 0.30 0.30 0.0010 702 Management consultancy activities 903.10 6.02 0.0058 711 Architectural and engineering activities and related technical consultancy 540.00 3.83 0.0039 731 Advertising 67.90 1.89 0.0047 732 Market research and public opinion polling 0.80 0.10 0.0069 749 Other professional, scienti c and technical activities n.e.c. 0.00 0.00 0.0000 781 Activities of employment placement agencies 38.90 1.62 0.0098 791 Travel agency and tour operator activities 34.70 0.72 0.0013 822 Activities of call centres 0.20 0.10 0.0010 823 Organization of conventions and trade shows 32.90 6.58 0.0064 829 Business support service activities n.e.c. 66.40 1.90 0.0064 841 Administration of the State and the economic and social policy of the 190.80 4.65 0.0031 community 851 Primary education 5.20 0.74 0.0006 853 Higher education 205.40 4.28 0.0091 854 Other education 0.60 0.20 0.0007 861 Hospital activities 469.50 2.50 0.0041 862 Medical and dental practice activities 2.50 0.31 0.0006 869 Other human health activities 7.50 0.75 0.0029 900 Creative, arts and entertainment activities 3.90 1.95 0.0013 932 Other amusement and recreation activities 117.90 3.47 0.0051 941 Activities of business, employers and professional membership organiza- 13.40 2.68 0.0136 tions 949 Activities of other membership organizations 392.50 39.25 0.0195 Electronic copy available at: https://ssrn.com/abstract=3909219 9 A Note On Dictionaries Used 9.1 Twitter Scraping We use the prowess database to get a list of all rm names. Then, the names of the com- panies are processed to remove unnecessary spaces within the names, which were present in the original list. These processed names are used as input to search the Twitter handles associated with those names. To download these tweets, we use Tweepy, a python library that is used to interface with the Twitter API. To lter out the Twitter handles with more precision, we use a string similarity metric. It generates similarity scores between 0 and 1, 1 being the same and 0 being no similarity present. We use a threshold of 0.8 to match prowess names with names mentioned in the Twitter handles. After ltering based on this threshold, we obtain approximately 8000 handles for 3162 rms. To scrape the tweets from these handles, we use a python library named Twint. The process created a dataset of 9.8 million tweets by 3162 companies for 6 years beginning from 2012. 9.2 Dictionaries: We rst lter out CSR-relevant tweets present within the dataset. Dictionary methods are very popular and established in natural language processing to label datasets or categorize the data present within them. Therefore, to obtain the CSR relevant tweets, we use the CSR dictionary built by Provalis Research. A tweet is identi ed as a CSR tweet if any word present within the CSR dictionary is present in the tweet. The Provalis Research CSR Dictionary consists of 1432 words across 4 dierent sections. The four sections are: Human Rights, Employee and Employment, Social and Community, and Environment. In addition, we remove the repeated words and commonly used verbs, adverbs, and adjectives which essentially by themselves cannot be assumed to be in relation to CSR activities. Upon applying this lter to the 9.8 million tweets present in our dataset, we identify close to 7.4% (730,000) as CSR-related. We then move to identify words related to product and charity to understand the distribution of the tweets. We use words and phrases taken from Buers (A social media tool to enhance engagement) website for identifying product-related words. The website identi es commonly used terms to market products and services on social media platforms. We build a dictionary relating words indicating charity and virtue using the words used by the charity and non-governmental organizations in their reports. We use online platforms such as \Merriam Websters thesaurus," \Related Words" , and \Your Dictionary" to identify the synonyms. https://provalisresearch.com/products/content-analysis-software/wordstat-dictionary/corporate-social- responsibility-text-analytics/ https://buer.com/library/words-and-phrases-that-convert-ultimate-list/ Electronic copy available at: https://ssrn.com/abstract=3909219 10 Alternative Explanation{Possible Change in Account- ing/Control Systems A skeptic might argue that the accounting and internal control systems relating to CSR changed after the mandate. For example, a change in the accounting/control system could mechanically aect our results if treatment rms reclassify certain CSR expenses from the pre-mandate period as non-CSR in the post-mandate period and the control rms behaved in exactly the opposite manner. We have carefully looked at the accounting/auditing standards applicable for CSR clas- si cation before and after the CSR spending mandate to address this concern. A rm is allowed to classify only those expenditures that are charitable in nature as CSR. Deliberate misclassi cation of business expenditure, say on research and development, as CSR is con- sidered as misrepresentation of books of accounts under the Indian Companies Act. None of the above regulations changed with the mandate. Therefore, a change in CSR classi cation in response to the mandate would be tantamount to an open admission of misrepresenta- tion, either during the pre-mandate or the post-mandate period. We believe that such an admission of misrepresentation is unlikely. Electronic copy available at: https://ssrn.com/abstract=3909219 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png ARN Conferences & Meetings SSRN http://www.deepdyve.com/lp/ssrn/does-mandated-corporate-social-responsibility-crowd-out-voluntary-hn3o0E4k1Z

Loading next page...

References (19)

D. Dhaliwal, Oliver Li, Albert Tsang, Yong Yang (2011)
Voluntary Nonfinancial Disclosure and the Cost of Equity Capital: The Initiation of Corporate Social Responsibility Reporting
The Accounting Review, 86
D. Greening, D. Turban (2000)
Corporate Social Performance As a Competitive Advantage in Attracting a Quality Workforce
Business & Society, 39
E. Dimson, Og̃uzhan Karakaş, Xi Li (2015)
Active Ownership
CSR & Management Practice eJournal
Holger Daske, Luzi Hail, C. Leuz, Rodrigo Verdi (2013)
Adopting a Label: Heterogeneity in the Economic Consequences Around IAS/IFRS Adoptions
Capital Markets: Asset Pricing & Valuation
Caroline Flammer (2013)
Does Corporate Social Responsibility Lead to Superior Financial Performance? A Regression Discontinuity Approach
ERN: Other Organizations & Markets: Firms & Other Social Institutions (Topic)
E. Dedman, Asad Kausar (2012)
The impact of voluntary audit on credit ratings: evidence from UK private firms
Accounting and Business Research, 42
Annita Florou, P. Pope (2012)
Mandatory IFRS Adoption and Institutional Investment Decisions
Banking & Insurance
Beiting Cheng, I. Ioannou, George Serafeim (2011)
Corporate Social Responsibility and Access to Finance
Harvard Business School Working Papers
Hans Christensen, E. Floyd, Lisa Liu, Mark Maffett (2016)
The Real Effects of Mandated Information on Social Responsibility in Financial Reports: Evidence from Mine-Safety Records
Capital Markets: Market Microstructure eJournal
Walter Blacconiere, W. Northcut (1997)
Environmental Information and Market Reactions to Environmental Legislation
Journal of Accounting, Auditing & Finance, 12
Dhammika Dharmapala, Vikramaditya Khanna (2016)
The Impact of Mandated Corporate Social Responsibility: Evidence from India's Companies Act of 2013
CESifo: Industrial Organisation (Topic)
Yi-Chun Chen, Mingyi Hung, Yongxiang Wang (2017)
The Effect of Mandatory CSR Disclosure on Firm Profitability and Social Externalities: Evidence from China
SRPN: Water Pollution (Pollution) (Topic)
Peranas Surabaya, Dosen Pengampu, Luciana Spica, Mingyi Hung, dan Subramanyam, dilihat keuangan, perusahaan menunjukkan, bahwa Roe, Rao Ato, nilai lebih, kecil perusahaan (2007)
Financial statement effects of adopting international accounting standards: the case of Germany
Review of Accounting Studies, 12
W. Davidson, Dan Worrell, Abuzar Eljelly (1995)
Influencing Managers to Change Unpopular Corporate Behavior through Boycotts and Divestitures
Business & Society, 34
Ing-Haw Cheng, Harrison Hong, K. Shue (2023)
Do Managers Do Good with Other People’s Money?
The Review of Corporate Finance Studies
Hans Christensen, Luzi Hail, C. Leuz (2019)
Adoption of CSR and Sustainability Reporting Standards: Economic Analysis and Review
SRPN: Sustainable Business (Topic)
J. Tirole, R. Bénabou (2010)
Individual and Corporate Social Responsibility
Microeconomics: Welfare Economics & Collective Decision-Making eJournal
Xin Deng, Jun-Koo Kang, Buen Low (2013)
Corporate Social Responsibility and Stakeholder Value Maximization: Evidence from Mergers
ERN: Other Organizations & Markets: Firms & Other Social Institutions (Topic)
W. Elliott, S. Grant, Kristina Rennekamp (2016)
How Disclosure Features of Corporate Social Responsibility Reports Interact with Investor Numeracy to Influence Investor Judgments
Corporate Governance & Accounting eJournal

Copyright: Copyright © 2021 Elsevier Inc.
eISSN: ARCH-4059
DOI: 10.2139/ssrn.3909219
Publisher site: See Article on Publisher Site

Abstract

Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India Shivaram Rajgopal Prasanna Tantri August 22, 2021 Shivaram Rajgopal can be reached at sr3269@gsb.columbia.edu. Prasanna Tantri can be reached at prasanna tantri@isb.edu. We thank the Editor, Christian Leuz, an anonymous Associate Editor, and an anonymous referee for helpful suggestions. We also thank Sakshi Gabba, Sharad Hotha, Gautham Kan- thasamy, Saiharsha Katuri, Aditya Murlidharan, Shradhey Prasad, and Nishka Sharma for excellent research assistance. We acknowledge helpful comments from Ray Ball and workshop participants at Booth School of Business, University of Chicago. We also acknowledge helpful comments from David Yermack and Confer- ence participants at NYU-IIM C India Conference. We are grateful to the Center for Analytical Finance, Indian School of Business for providing the data and the necessary nancial assistance for this project. Rajgopal thanks the Columbia Business School for nancial support. Any remaining errors are ours. Electronic copy available at: https://ssrn.com/abstract=3909219 Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility ? Evidence from India Abstract We investigate the implementation of a government of India mandate that requires rms to spend at least 2% of their pro ts on corporate social responsibility (CSR). We nd that mandated rms that voluntarily engaged in CSR before the mandate reduce their CSR spending signi cantly after the mandate. The erstwhile voluntary CSR spenders increase advertising expenditure plausibly to oset the lost signaling value of voluntary CSR. The 2% mandate negatively impacts valuations and operating performance. Our results show that regulatory intervention in CSR diminishes its signaling value and leads to a reduction in voluntary CSR spending. Electronic copy available at: https://ssrn.com/abstract=3909219 1 Introduction \I don't think you generate CSR by putting statutory requirements. I think there is enough social consciousness among the larger Companies to drive it on the basis of what they consider their responsibility."{ Azim Premji, Chairman of Azim Premji Foundation. Governments in several countries have begun playing an active role in corporate social responsibility (CSR). Some have nudged corporations to spend funds on CSR, while others have moved to mandate disclosures. A few countries, such as India (Manchiraju and Raj- gopal (2017), Dharmapala and Khanna (2018)) and Indonesia (Waagstein (2011)) have gone a step further and have promulgated laws that make not only disclosure but also spending on speci ed CSR activities mandatory. We study the Indian CSR mandate in this paper. A signi cant part of the current academic debate on CSR is centered around the rms' motives underlying voluntary CSR and the value consequences thereof (see Kitzmueller and Shimshack (2012), Orlitzky, Schmidt, and Rynes (2003), McWilliams, Siegel, and Wright (2006), Margolis, Elfenbein, and Walsh (2009), and Perrini, Russo, Tencati, and Vurro (2011)). Christensen, Hail, and Leuz (2019), however, point out that most extant stud- ies based on voluntary CSR are subject to selection issues. Furthermore, most CSR-related regulations worldwide mandate only disclosure. Therefore, an analysis of the impact of forced CSR spending in the Indian context is likely to shed light on the motives behind CSR spending. In particular, the government of India forced \eligible" rms to spend at least 2% of Source: http://economictimes.indiatimes.com/news/company/corporate-trends/azim-premji-against- law-on-mandatory-csr-spending-by-corporates/articleshow/7782555.cms. Wipro is one of the largest Infor- mation technology companies in India. Recently, the European Union member states have agreed to pass legislation requiring cor- porations to report their CSR activities in detail. Similar laws have been passed or are be- ing contemplated in countries such as China (Chen, Hung, and Wang (2018)) and Canada. Source: https://www.theguardian.com/sustainable-business/eu-reform-listed-companies-report- environmental-social-impact, http://corporatejustice.org/; http://corporatejustice.org/news/1174- getting-non- nancial-reporting-right-eu-commission-guidelines-clarify-expectations-towards- business; https://www.globalreporting.org/information/policy/Pages/EUpolicy.aspx; https://mastereia.wordpress.com/2014/04/10/mandatory-environmental-corporate-social-responsibility- can-canada-become-a-leader/; https://www.greenbiz.com/news/2009/01/07/mandatory-csr-reporting- denmarks-largest-companies Electronic copy available at: https://ssrn.com/abstract=3909219 3 average three years' pro ts on CSR. The eligibility threshold is either INR 50 million in pro t, INR 5 billion in net worth, or INR 10 billion in sales. Firms that exceed one or more of the above thresholds are subject to the 2% CSR spending rule. We focus on rms that voluntarily spent more than the minimum 2% limit before the law was passed. We can track CSR spending before the rule was enacted because Indian rms are required to disclose such spending as per the applicable accounting standards. The CSR law imposed by India is quite stringent in intent and enforcement. The mandate requires that the rm's board justify any failure to comply with the CSR requirement (\the comply or explain" model). However, the law does not specify what can be considered a reasonable explanation, creating room for ambiguous interpretation of such explanation by ocials of the federal ministry of corporate aairs (MCA). More importantly, the law elevates the responsibility of CSR spending from the management to the board. The violation of the CSR requirement is, therefore, considered a non-ful llment of directors' duties. If the law has the requisite bite, we expect rms that spent less than 2% before the mandate (labeled \low CSR rms") to spend close to 2% of pro ts after the mandate. We nd such a result. However, the expected reaction of rms that spent more than 2% (labeled \high CSR rms") in the pre-mandate period is not clear ex-ante. We expect high CSR rms to cut spending in the post mandate period if (i) if the mandate dilutes the underlying motive for voluntary CSR spending; or (ii) if alternative ways of satisfying such a motive become relatively more attractive; or (iii) if 2% is perceived as the society's new norm for CSR spending. Otherwise, the voluntary spending of high CSR rms should not change. Data on voluntary CSR spending are drawn from the Prowess database maintained by the Center for Monitoring Indian Economy (CMIE). In addition, we obtain data on mandated CSR spending from the federal Ministry of Corporate Aairs (MCA). In the post-mandate period, we nd that high CSR rms signi cantly reduce their CSR spending to around 2% of pro ts in univariate tests. On the other hand, as expected, low CSR rms increase their CSR spending to the 2% level. INR stands for Indian Rupee. Electronic copy available at: https://ssrn.com/abstract=3909219 The ideal identi cation strategy is unfortunately not available to us. Both high and low CSR rms are \treated" in an experimental sense as the 2% law applies to both these types of rms. Furthermore, comparing the entire mandated group as treated with the non- mandated control group does not work well because the post-mandate incentives of high and low CSR rms within the mandated group are not the same as explained next. As a compromise, in our baseline dierence-in-dierence (di-in-di, henceforth) analysis, the treated (control) rms are those that are (not) mandated to invest in CSR, within the sample of high CSR rms (see Figure 1). By construction, both the treatment and control samples in this experiment do not have to mechanically increase CSR spending to comply with the 2% law. Crucially, this sample includes voluntary high spenders who are not covered by the 2% mandate. The dependent variable in our research design is the magnitude of CSR spending at a rm-year level. Our focus is on the interaction between the post-mandate indicator variable and the indicator variable representing rms required to spend on CSR. We test for and rule out the existence of pre-trends. We acknowledge that the non-mandated group will react to the actions of the mandated group. Hence, we do not claim to measure the complete causal eect of the mandate. Instead, we document a dierence in the relative impact on CSR spending between the two groups We nd a statistically signi cant and economically meaningful 32.57% decline in CSR spending in the post-mandate period. Already compliant rms covered by the 2% rule cut CSR spending signi cantly. We expect low CSR rms covered by the mandate to spend more and nd such a result. These inferences hold when we use a triple interaction framework with all three margins (timing of the mandate, mandate eligibility, and pre-mandate CSR spending) considered together. In the second part of the paper, we try to understand why high CSR rms cut spending. We conjecture that the mandate dilutes strategic value of voluntary CSR which, in turn, leads to a negative impact on valuation and operating performance (Deng, Kang, and Low (2013)). Event study and other tests con rm that the imposition of mandatory CSR leads to a negative stock price reaction and lower operating performance for mandated rms within Electronic copy available at: https://ssrn.com/abstract=3909219 the universe of high CSR rms. This opens the question of whether a strategic motive suered from the 2% mandate and led high CSR rms to cut spending. We conjecture that voluntary CSR spending is a unique method of signaling both virtue and higher quality of their products simultaneously (Kausar, Shro, and White (2016), Kotler and Lee (2008)). The 2% mandate diluted such a special signal. We present a mosaic of evidence consistent with this conjecture. First, we examine the CSR sections of the annual reports and social media postings of rms. We nd a signi cant (i) decline in the number of such communications, in general; (ii) fall in the number of words that convey product or service quality; (iii) decline in words that signal virtue; and (iv) increase in standardization of CSR communication. Second, mandated high-CSR rms that reduce CSR expenditure, on account of its diluted signaling power, signi cantly increase advertising expenditure after the 2% rule. Moreover, the Rupee increase in advertising expenditure is very close to the Rupee decline in CSR expenditure for the mandated high-CSR rms. Despite such a one-to-one replacement, the mandated high-CSR rms suer a loss of value and operating performance. This fact pattern is consistent with the hypothesis that voluntary CSR, as opposed to advertising, is more eective at communicating product quality and virtue simultaneously. An important open question remains after presenting such evidence: why don't mandated high CSR rms continue to enjoy signaling bene ts by spending more than the mandated 2% level on CSR even after the mandate? We believe that compliance costs with the new law are a signi cant deterrent. CSR spending requires the board's approval post mandate. Additional approval lters and the requirement to spend at least 2% of pro ts every year also curtail exibility. Most importantly, in practice, CSR projects are not divisible into mandated and non-mandated parts. It may not be possible to run the mandated and non- mandated CSR projects in separate silos within an organization. Thus, compliance costs apply not only to the mandated part but increase in proportion to the entire CSR expendi- ture. Furthermore, the excess spending over 2% is likely viewed as a buer against a strict Electronic copy available at: https://ssrn.com/abstract=3909219 interpretation of CSR de nition by the government ocers, and hence, become less eective for signaling. Therefore, alternative ways of signaling, such as advertising, may become more attractive after the mandate. However, we acknowledge that we cannot cleanly disentangle the direct loss of signaling power of CSR from higher compliance costs. In follow-up work, we considered alternative strategic channels for CSR spending such as customers' willingness to overpay, workers' willingness to work for lower salaries, and the ability of CSR to deter value loss due to political interventions. None of the above channels seems to statistically explain the cut in CSR spending by high spenders, its replacement by advertisement, and change in CSR related communication. We also cannot rule out the possibility that voluntary CSR is partly motivated by (i) stakeholder altruism (Reinhardt, Stavins, and Vietor (2008), B enabou and Tirole (2010)); (ii) managerial moral hazard (Cheng, Hong, and Shue (2013), Masulis and Reza (2014)); or (iii) revelation of society's expectation related to CSR spending through the mandate. We in- terpret our four empirical ndings (a decline in voluntary CSR spending, negative impact on valuations and operating performance, change in CSR communication, and the replacement of CSR by advertisements) as tentative evidence that voluntary CSR is not driven solely by altruism, moral hazard or newly set social norms for CSR spending. Firms consciously used CSR for signaling quality and virtue, potentially in addition to other motives. Our paper contributes to the literature on the strategic value of CSR (Deng, Kang, and Low (2013), Cheng, Ioannou, and Serafeim (2014), Lins, Servaes, and Tamayo (2017), Dimson, Karaka s, and Li (2015), Blacconiere and Northcut (1997), Christensen, Floyd, Liu, and Maett (2017), Dhaliwal, Li, Tsang, and Yang (2011), Elliott, Grant, and Rennekamp (2017), Flammer (2015)). In their review of the CSR reporting literature, Christensen, Hail, and Leuz (2019) point out that most CSR spending is voluntary, which leads to selection- related issues. We identify a unique setting where the government mandates CSR spending. Manchiraju and Rajgopal (2017) and Dharmapala and Khanna (2018) also study the short-term stock price impact on the passage of the mandatory CSR rule in India. We focus on the impact of the mandate on actual CSR spending and show that erstwhile high spenders Electronic copy available at: https://ssrn.com/abstract=3909219 cut CSR spending post-mandate. More importantly, we examine plausible motives behind voluntary CSR spending and suggest that voluntary CSR, unlike advertising, uniquely signals both virtue and product quality. As in the case of literature dealing with the voluntary adoption of audits (Kausar, Shro, and White (2016), Dedman and Kausar (2012), Lennox and Pittman (2011)) and IFRS (Daske, Hail, Leuz, and Verdi (2013), Kim, Tsui, and Cheong (2011), Christensen, Hail, and Leuz (2013), Hung and Subramanyam (2007), Florou and Pope (2012)), we compare the implications of voluntary and mandated CSR spending. We propose that CSR spending due to a government mandate makes it a less valuable signaling tool. Despite a substitution of CSR by advertisement, the overall nancial performance of treated rms deteriorates, suggesting a unique role for voluntary CSR in signaling virtue and product quality. 2 Institutional Background and the Event Section 135 of the newly introduced Indian Companies Act of 2013 has mandated that rms above a threshold (de ned in terms of net worth, sales, and pro t) have to spend 2% of their average past three years' pro t on CSR activities. The eligibility threshold was de ned as either INR 50 million (USD 0.78 million) in pro t, INR 5 billion (USD 0.078 billion) in net worth, or INR 10 billion (USD 0.156 billion) in sales. Before the decree, rms were required only to disclose spending as per existing accounting standards. Every rm covered by the mandate is required to set up a CSR policy. Although the new Companies Act came into force on August 29, 2013, the CSR mandate was made eective from 2014-2015 (i.e., the year beginning April 1, 2014). We are not aware of any economically meaningful reason for xing the limit at 2%. It appears that the government picked a round number for simplicity. The law requires non-compliant rms to explain in their annual reports the reasons behind their non-compliance. However, the law does not specify guidelines to determine whether an explanation is valid, leaving room for regulatory discretion in interpretation. The Act de nes We assume an exchange rate of INR 63 to USD 1, which was prevalent when the Act was passed. Electronic copy available at: https://ssrn.com/abstract=3909219 CSR broadly but leaves the details to the boards of the individual rms (see Manchiraju and Rajgopal (2017) for information about the speci c provisions). Under the new law, the responsibility to manage CSR spending rests with the board and not the management. The ministry of corporate aairs (MCA) has issued show-cause notices to more than 1,000 rms, charging them with violations of the CSR law. Notices have been issued even when rms have preferred to explain rather than to comply on the grounds that the stated explanations are not satisfactory. Section 135, which imposes mandatory CSR, does not im- pose penal provisions if the spending mandate is violated. However, the MCA has charged the alleged non-compliant rms under a dierent section 134, which speci es directors' re- sponsibilities for nancial statements, and contains strict penal provisions. Although, in theory, mandatory CSR works on a \comply or explain" model, in practice, it is safer for the companies to comply than to explain. 3 Data, Variable De nition And Sample Construction We use the Prowess database maintained by the Center For Monitoring Indian Economy (CMIE) for the pre-mandate period. For the post-mandate period, we use the ministry of corporate aairs (MCA) data for rms covered by the ministry and the CMIE Prowess database in other cases. Both these databases source information from the annual reports of rms. Expectedly, both the data sets report the same Rupee numbers for CSR outlays th th for most cases. As shown in Table A1 of the online appendix, both the 5 and the 95 percentile thresholds of the dierence between reported CSR amounts in the two databases for a given rm are zeroes. We manually check the reasons behind the dierences between the two datasets in some extreme cases. The MCA data set is more accurate in all cases. However, we found data entry errors and incorrect round-os of decimals in the Prowess database in rare situations. 5https://www. nancialexpress.com/industry/government-issued-notices-to-1018- rms-for-csr-non- com- pliance/589099/ In one case, the Prowess database has left out one part of CSR expenditure reported in the annual report. In another case, the Prowess database used the budgeted CSR numbers presented in the annual Electronic copy available at: https://ssrn.com/abstract=3909219 Given these ndings, we use the MCA database for the post-mandate period for the rms covered by the ministry. Because the ministry does not maintain pre-mandate data and data for non-mandated rms, we have to rely on the Prowess dataset for (i) the pre-mandate period CSR spending of all rms; and (ii) pre and post-mandate period CSR spending of all non- mandated rms. The ministry data is available for years spanning 2014-2015 to 2018-2019. Compilation of the data thus gathered gives us CSR spending at a rm-year level. The variable thus created is labeled the \CSR amount." Because of data integrity issues at the extreme ends of the CSR spending distribution, we winsorize the variable at 1% and 99% in our primary analysis. We provide variable de nitions in Table 1. 3.1 Sample Construction We report the details underlying the construction of the sample in Table 2. The Prowess database contains information on 43,051 rms. Many of these are shell companies formed for money laundering. The ministry dataset covers 12,097 companies. We found 10,154 rms of these in the Prowess database using the unique corporate identity number (CIN) as the matching variable. Our sample starts from the nancial year 2009-2010. The years between 2009-2010 and 2013-2014 are labeled as the pre-regulation years. Years 2014-2015 and 2018- 2019 comprise the post-mandate period. Thus, we cover a sum total of 10 years. The table shows that the merged data set contains 39,309 rms and 236,044 rm-year observations with usable data in both pre and post-mandate periods. We employ two additional lters. We leave out observations where CSR information is missing instead of treating them as zeros because we are unsure why these data points are absent. We have data on CSR spending for 44,769 of the 236,044 rm-year observations. Of these, 16,251 (28,518) observations belong to rm-years where the average pre-period ratio of the amount spent on CSR and average past three year pro ts is more than 2% (less than report rather than the actual amount spent. We provide ve examples of the largest deviations between these two databases in Table A2 of the online appendix. http://www. rstpost.com/business/over-1-62-lakh-shell-companies-deregistered-over-half-from- mumbai-delhi-hyderabad-3907583.html Electronic copy available at: https://ssrn.com/abstract=3909219 or equal to 2%). The table also provides information related to the number of observations for which CSR to pro ts ratio is equal to or greater than the 5%, 7.5%, and 10% thresholds. 4 Empirical Strategy and Results 4.1 Univariate Tests We begin by calculating the CSR Ratio, de ned as the ratio of the CSR amount and average pro t before tax for the past three years, for each rm-year. We average the CSR ratio over the pre-regulation period. Firms with an average CSR Ratio of greater than 2% form the \high CSR" group. Firms below this threshold form the \low CSR" group. As a further robustness check, we use three other threshold cutos based on 5%, 7.5%, and 10% of the average three years' pro ts. In Table 3, we nd that the high CSR rms cut back their spending on CSR signi cantly post-mandate. However, the low CSR rms increase CSR spending up to 2% in the post- mandate period. For example, column 2 shows that rms that spent between 0% and 1% of pro ts on CSR before the mandate increase spending by 1.7 percentage points, on average, after the mandate. In column 4, we report that rms that used to spend between 2% and 3% before the mandate maintain almost the same level of spending even after the mandate. Interestingly, from column 5, we nd a signi cant decline in CSR spending after the mandate. In column 5, rms that used to spend between 3% and 4% in the pre-mandate period reduce spending to 3% in the post-mandate period. In column 7, rms that used to spend between 5% and 6% also reduce spending by 2.5 percentage points in the post-mandate period. A similar trend is seen in other columns where we consider rms that used to spend a higher proportion of pro ts on CSR in the pre-mandate period. Electronic copy available at: https://ssrn.com/abstract=3909219 4.2 Diculty With Comparing High And Low CSR Firms A potential empirical strategy is a di-in-di test within the set of rms subject to the 2% mandate where the high CSR rms (low CSR rms) are designated as the treated (control) group. One problem with this strategy is that both the treated and control groups are directly impacted by the mandate, although in opposite directions. The univariate results, presented in Table 3, clearly show that high CSR rms cut spending on CSR whereas low CSR rms increase spending. Therefore, it is dicult to tell whether our results are attributable to (i) an increase in spending by the control group; or (ii) a decline in spending by the treated group. Hence, we do not use this identi cation strategy. A second strategy of comparing the entire mandated group as the treated class and the entire non-mandated group as the control class does not work either because high and low CSR rms behave dierently within the mandated group. 4.3 Mandated Vs Non-Mandated- Within High CSR Firms To test the implication of the mandate on high CSR rms, we adopt an identi cation strategy that compares mandated and non-mandated rms within the high CSR group. The mandate applies only to those rms which satisfy at least one of the qualifying criteria described in Section 2. We consider such mandated rms as the treated group. We acknowledge that non- mandated rms will have to consider the reaction of the mandated rms while formulating their plans. Hence, our tests can only capture the incremental eect on the treated rms relative to the control sample and not the complete causal eect of the mandate. We discuss this issue in Section 6.3. We conduct a di-in-di comparison within high CSR rms between the mandated and non-mandated groups (depicted in Figure 1). As shown in the gure, we conduct a similar but separate di-in-di test within the sample of low CSR rms. The treated group within the set of low CSR rms will increase CSR spending mechanically to comply with the law, but the control group does not have a similar obligation. Electronic copy available at: https://ssrn.com/abstract=3909219 We estimate the following standard di-in-di regression equation by rst limiting the sample to the high CSR rms. (1) Y = + Post Treatment + X + + + it 1 t i 2 it 3 i 4 t it The dependent variable is the amount (in INR) of CSR expenditure made by a rm i in the year t. The variable Post is a dummy variable, which takes the value of one for post-regulation years and zero otherwise. Treatment is a dummy variable that takes the value of one if a rm satis es at least one of the three CSR mandate conditions based on average values during the pre-mandate years and zero otherwise. Firms that do not satisfy any of the 2% rule's qualifying conditions form the control group. The interaction between the above Post and Treatment variables is the explanatory variable of interest. Note that t i represents the rm xed eects, represents the year xed eects, and X represents i t it rm-level time-varying variables such as pro t and total net worth. Furthermore, we use total assets for scaling the dependent variable in subsequent tests. We use net worth as a proxy for size because we can nd more non-missing observations for net worth relative to that for sales. Finally, the standard errors are clustered at the industry level and adjusted for heteroskedasticity. The results are reported in Panels A and B of Table 4. The CSR amount is the dependent variable in Panel A. In columns 1 and 5, we use the 2% threshold to de ne high CSR rms. As expected, we nd a sharp decrease in the dierence between the treatment and control rms in the post-regulation period compared to the dierence in the pre-regulation period. The magnitude of the decline is INR. 7.87 million, which is 32.57% of the average CSR amount in the pre-mandate period. Therefore, the fall is economically meaningful. We present the results with higher thresholds of 5% (columns 2 and 6), 7.5% (columns 3 and 7), and 10% (columns 4 and 8) of past pro ts. The results strengthen with an increase in the threshold spending levels, suggesting that voluntary high spenders cut CSR spending signi cantly in response to the CSR mandate. In Panel B, we use the ratio between CSR amount and Electronic copy available at: https://ssrn.com/abstract=3909219 total assets as the dependent variable. Column 1 shows that the ratio decreased by 14 basis points. The decline represents an economically meaningful 51.9% of the pre-mandate levels. To test whether the treated rms drive the impact, we compare the pre and post-mandate periods within the sample of mandated high CSR rms. We estimate the following regression equation: (2) Y = + Post + X + + it 1 t 2 it 3 i it All the terms have the same de nitions as in equation 1. We cannot include year xed eects because the tests compare between pre and post period expenditure. The results are reported in Panels A and B of Table A3 of the online appendix. In column 1, where we limit the sample to rms spending at least 2% of their pro ts on CSR, the CSR spending declines by about INR 7.63 million after the mandate. The decline represents 31.56% of the average CSR amount in the pre-mandate period. We nd similar results even when we use the ratio between CSR amount and total assets as the dependent variable in Panel B. These results con rm that almost all of the decline in CSR spending documented in Table 4 is driven by the treated group. We perform the same di-in-di analysis as in equation 1 among low CSR rms. The comparison is between mandated and non-mandated rms, as before. We expect the man- dated rms to invest more in a di-in-di sense mechanically to comply with the legal requirement to hike spending. We nd such a result (reported in Table A4 of the online appendix). Next, we incorporate all three variations in one speci cation whereby the terms \Post-Pre," \Mandated-Non- mandated," and \High CSR-Low CSR," are considered in a triple interaction framework as shown at the bottom of Figure 1. We expect the triple inter- action term, the Post* Mandated* High CSR indicator variable, to be negative, and we nd that result (in Table A5 of the online appendix). Thus, our results stem from spending cuts initiated by the high CSR rms among the mandated group after the law was promulgated. The results taken together suggest that (i) the CSR mandate leads to a reduction in Electronic copy available at: https://ssrn.com/abstract=3909219 voluntary spending; (ii) the reduction stems from a cut in CSR spending post mandate by the erstwhile high spenders and (iii) the pre-mandate low spenders increase spending as required to by the law. 4.3.1 Dierence-In-Dierences Pre-Requisites And Robustness a. The Test of Pre-Trends: To mitigate the possibility that a mechanical continuation of a pre-existing trend drives our results, we plot the ratio of CSR amount and total assets for the treated and the control groups in Figure 2. We consider ve years before and ve years after the mandate. As shown in the gure, there does not appear to be any clear pre-trend. Further, we notice a sharp decline in CSR spending of the treated group in the post-mandate period. Second, in our baseline regression setup, we introduce indicators for individual years. Then, we interact each of the pre-year indicator variables with the treatment variable. We nd all the interaction terms, except the interaction between post-mandate and treatment indicator variables, to be statistically indistinguishable from zero. This result, reported in Table A6 of the online appendix, also helps us rule out the existence of pre-trends. b. Comparing Treated And Control Groups: We acknowledge that the treated rms are systematically larger than the control rms because the mandate applies to larger rms. While the non-existence of pre-trends helps ameliorate concerns in this regard, we conduct an additional test. We compare several operating and nancial ratios for the treated and control rms in Table 5. These ratios include operating margin, leverage, stock turnover, return on equity, advertisement to sales, and pro t per employee. None of the above ratios is signi cantly dierent between the treated and the control rms. This analysis shows that the two sets of rms are similar in terms of operating eciency, although they dier in size. c. Additional Robustness Tests We perform several additional robustness tests. First, to account for possible anticipation Electronic copy available at: https://ssrn.com/abstract=3909219 of the law on account of the discussion in the media before its implementation, we omit the years 2012-2013 and 2013-2014 from the sample, which is when the new Companies Act was discussed and introduced. The CSR provision came into force eective 2014-2015. We estimate equation 1 using this subsample and nd that the results are consistent with our hypotheses (reported in Table A7 of the online appendix). Note that we exclude the years 2012-2013 and 2013-2014 while arriving at the treated and control groups for the above test. Second, we conduct placebo tests with false treatment years within the pre-mandate period. We report the results using 2011-2012 as a false treatment year in Table A8 of the online appendix. We estimate regression equation 1. The coecient of the interaction term between the post-mandate indicator variable and the treatment indicator variable is statistically indistinguishable from zero. Furthermore, we get similar results when we use other false treatment years. Finally, control rms that lie very close to the cuto during the pre-mandate period may cross over during the post-mandate period and fall within the purview of the 2% rule. Thus, we potentially misclassify a few treated rms as control rms. To account for such a possibility, we leave out control rms very close to the cuto and re-estimate regression equation 1. We present the results in Table A9 of the online appendix. The results remain unchanged. 5 Why do high spenders reduce spending on CSR post the mandate? In the second part of the paper, we examine why the erstwhile high spenders cut CSR spending after the mandate. We hypothesize that voluntary CSR has a strategic value that gets diluted after the mandate (\the strategic CSR hypothesis"), which, in turn, negatively impacts rm value and operating performance. A decline in both operating performance and shareholder value would be consistent with a strategic CSR hypothesis. However, these Electronic copy available at: https://ssrn.com/abstract=3909219 results are insucient to establish a strategic motive at work and also cannot rule out non- strategic motives. In follow-up work, we explore speci c operational changes that plausibly explain the exact strategic motive underlying voluntary CSR. 5.1 Impact On Operating Performance We use return on assets (ROA) and return on equity (ROE) as measures of operating per- formance. Column 1 of Table 6 shows that ROE declines by 2.29 percentage points for the treated group in a di-in-di sense. Because the average ROE is 19.9% in the pre-mandate period, the decline represents an economically meaningful 11.51%. Similarly, in column 2, we nd a 0.93 percentage points decline in ROA or an economically meaningful 8.08% of the average ROA during the pre-mandate period. In columns 4 and 5, we include one-year lags of pro ts and net worth as control variables. Finally, in columns 5 and 6, we use the natural logarithm of sales as the dependent variable. We do not nd a signi cant change in sales. 5.2 Impact on Stock Valuations The dilution of strategic motivation of high CSR rms to cut spending should have clear shareholder value implications. Manchiraju and Rajgopal (2017) examine eight events lead- ing up to the CSR law and nd that the stock price reaction of the entire set of mandated rms when compared to non-mandated rms, is negative. Note that stock prices of man- dated rms could react negatively for two reasons: (i) they are forced to spend 2% of their average pro ts on CSR, which is akin to an increase of 2% in corporate taxes; and/or (ii) mandatory CSR makes a strategic motive behind voluntary CSR less valuable. Manchiraju and Rajgopal (2017) cannot distinguish between these explanations in their setting. We attempt to do so by comparing the stock market reaction of mandated and non- mandated rms' stocks but we limit the sample to high CSR rms. In this setting, any adverse price reaction of the mandated rms' stocks is likely attributable to the loss of strategic value and other compliance costs, rather than the increased CSR spending burden Electronic copy available at: https://ssrn.com/abstract=3909219 as high CSR sample rms already spend more than 2% of their pro ts on CSR. As noted by Manchiraju and Rajgopal (2017), the introduction of mandatory CSR was a part of new company legislation passed by the Parliament. Naturally, all the provisions of the bill were debated in detail. Hence, it is hard to use one date to conduct an event study. Instead, we rely on the sudden announcement made by the nance minister on July 17, 2019, that, henceforth, non-compliance with CSR provisions will constitute a criminal oense. Till then, non-compliance with CSR was considered a civil oense. The nance minister reversed her position on August 23, 2019. This event provides an excellent testing ground for the shareholder value implications of voluntary CSR for the following reasons. First, after the rst announcement by the minister, business executives faced the prospect of a jail term for non-compliance with the CSR provision. Therefore, the value of voluntary CSR should have signi cantly diminished after her rst announcement. Second, the relatively arbitrary reversal of her position mitigates other endogenous factors that might confound our event study results. It is hard to argue that some other unobservable factor moved precisely in line with the minister's announcements on both occasions. We present our results in Table 7. In Panel A, we consider the rst announcement which made non-compliance with CSR regulations a criminal oense. We compare all mandated with all non-mandated rms. We estimate a regression where the dependent variable is the excess return three days around the event of a stock compared to the market benchmark. In columns 1 and 2, the explanatory variable (\treatment") identi es mandated rms. We include industry- xed eects and cluster standard errors at an industry level. We nd that the mandated rms, in general, react negatively to the event (-1.2%), consistent with Manchiraju and Rajgopal (2017). In columns 3 and 4, we restrict the sample to high CSR rms and nd that the stock prices of mandated high CSR rms decline by about 0.8% relative to that of non-mandated high CSR rms. By construction, high CSR rms are generous spenders, and hence, none of the treated or the control rms needs to increase CSR spending to comply with the new law. We use Nifty 50, Indias most widely tracked market index, as the market benchmark. Electronic copy available at: https://ssrn.com/abstract=3909219 Therefore, it seems reasonable to attribute the decline in share prices to the loss of strategic value of CSR and additional compliance costs. In Panel B, we study the reaction to the reversal of position by the nance minister on August 23, 2019. The remaining details remain unchanged. Investors appear to reverse almost the entire under-performance of mandated high CSR stocks when the nance minister changed her mind. This pattern provides further support to our hypothesis that mandatory CSR negatively impacts a strategic channel that made voluntary CSR eective before the new law came into force. We explore what that strategic channel might be in the following section. 5.3 Possible Strategic Reasons In a quest for a potential strategic motivation behind voluntary CSR before the mandate, we begin with (i) a close examination of various forms of public communications related to CSR; and (ii) the types of expenditure that replace voluntary CSR spending after the mandate. 5.3.1 Loss Of Signaling Value? The literature has argued that the voluntary adoption of audits (Kausar, Shro, and White (2016), Dedman and Kausar (2012), Lennox and Pittman (2011)) has signaling value. Sim- ilarly, Kotler and Lee (2008, 2005) suggest that voluntary CSR positively impacts potential customers' views about the rm, its' strengths and product oerings. Therefore, we investi- gate the possibility that pre-mandate CSR was perceived as a credible signal about a rm's overall product quality. We extend this thesis to suggest that voluntary CSR provides a unique opportunity to signal both high quality and virtue. This bundle cannot be perfectly substituted away by conventional advertising. However, regulation and enforcement that circumscribes the exact nature of CSR potentially dilutes the power of such a signal. As noted in section 2, the ministry routinely issues show-cause notices to rms arguing that the CSR expenditure Electronic copy available at: https://ssrn.com/abstract=3909219 claimed by rms is not socially desirable but merely a commercial expenditure incurred in the ordinary course of business. Furthermore, even within the rm, CSR spending requires the approval of the board and has to comply with the CSR policy laid down by the board. Thus, both internal and external constraints could hamper the creative ability of the organization to use CSR for signaling purposes when compared to other means such as advertisements. CSR Communications- Social Media: We begin with a study of the rm's communication related to CSR. If the 2% rule dilutes CSR's signaling ability, we expect to observe (i) a reduction in CSR related communication; (ii) a reduction in number of words that convey positive CSR related product or service attributes; and (iii) an increase in the standardization of CSR communication. On the other hand, if the mandate makes no dierence to the signaling value or if there is no signaling value to CSR in the rst place, we expect no change in the communication relating to CSR. We identify the ocial Twitter handles of rms and scrape the tweets tweeted during the sample period. We label a tweet as CSR-related if any words related to CSR are mentioned in the tweet. In section 9 of the online appendix, we provide a detailed account of the process used to identify CSR related words. We ask whether the proportion of CSR related tweets comes down in a di-in-di sense. We estimate regression equation 1 with the proportion of CSR-related tweets among all tweets as the dependent variable and present the results in column 1 of Table 8. We nd a 2 percentage points decline in the proportion of CSR related tweets for the mandated rms. The general reduction in CSR-related tweets suggests that CSR-related communication becomes less valuable after the mandate. Next, we investigate the proportion of words within CSR tweets that signal the high quality of a rm's products or services. We use standard marketing dictionaries to identify such words, as explained in section 9.2 of the online appendix. As shown in column 2 of Table 8, we nd a 2 percentage points decline in the proportion of words that signal product or service quality within CSR related tweets in a di-in-di sense. Finally, we check whether virtue-signaling also reduces post the mandate. We identify words that signal charity or good intentions of the rms within CSR-related tweets (see Section 9.2 of the online appendix). Electronic copy available at: https://ssrn.com/abstract=3909219 As shown in column 3 of Table 8, we nd a one percentage point decline in the proportion of words that signal virtue. In the subsequent three columns, where we include rm-level control variables, we nd similar results. These results suggest that voluntary CSR was used to convey higher product quality and virtue, and the eectiveness of the signal reduces after the mandate. CSR Communications- Annual Reports: We then examine the CSR sections of the annual reports. We expect to nd higher standardization of CSR reporting in the annual reports in the post-mandate period. To test the above hypothesis, we calculate the cosine similarity of the text of the CSR section in the annual reports between rms every year. For each mandated (non-mandated) rm, we calculate the cosine similarity of the CSR section of its annual report with that of all other mandated (non-mandated) rms in a year and calculate the average of the cosine similarity score. The comparison is limited to rms whose annual reports are available. The average thus calculated is the dependent variable for each rm-year. Higher standardization is likely to make the reporting more similar. We nd an increase in the standardization in CSR reporting in a di-in-di sense for the mandated group in the post-mandate period. We report the above results in column 1 of Table 9. Increased standardization is consistent with reduced use of CSR for signaling purposes. In columns 2 and 3, we ask whether mandated rms reduce signaling their product quality and virtue. We follow the same methodology as in the case of social media posts. We nd a close to 8 percentage points (6 percentage points) reduction in the signaling of product quality (virtue). Increase in Advertising: A likely consequence of a reduction in signaling ability of CSR after the mandate and the consequent decrease in CSR spending is an increase in expenditure on advertisements. Therefore, we test whether mandated rms increase spending on advertising by estimating regression equation 1. The sample, as before, is restricted to high CSR rms. In Table 10, the level of advertisement expenditure (the ratio between advertisement spending and total We obtain annual reports from www.moneycontrol.com. Electronic copy available at: https://ssrn.com/abstract=3909219 assets) is the dependent variable in Panel A (B). The organization of the table exactly mimics Table 4. We nd a signi cant increase in spending on advertisements in all speci cations. The increase translates to 24.85% of the average advertisement expenditure in the pre- mandate period, and is therefore, materially signi cant. Next, we attempt to understand the relative magnitude of advertisement expenditure in- curred by rms to replace CSR. To this end, we sum the CSR expenditure and advertisement expenditure and use the newly created variable as the dependent variable in the regression equation 1. An increase (decrease) in a di-in-di sense indicates that the increase in adver- tisement expenditure is higher (lower) than the reduction in CSR spending. No change would mean close to a one to one replacement. We present the results in Table A10 of the online appendix. We nd that that the di-in-di coecient is statistically indistinguishable from zero. In other words, it appears that the replacement of CSR expenditure by advertisement expenditure is almost one-to-one. Note that in section 5.2, we nd that mandatory CSR leads to a loss of shareholder value. It seems reasonable to infer that voluntary CSR spending is more valuable than advertisements as a signaling device. Hence, even a close to one-to-one replacement of CSR by advertising does not prevent the loss of shareholder value. The question of why rms do not increase advertisement beyond a one-for-one replace- ment remains. We conjecture that nancial constraints could be a part of the explanation. Perhaps some customers respond only to voluntary CSR as opposed to advertising. The mandate might have also shut the rms' ability to signal virtue. Hence, even a higher level of advertisement spending cannot replace the value loss due to reduced CSR. Finally, rms could have potentially increased spending on other types of signaling such as certi cations and quality control. Such extra spending could plausibly explain the decline in operating performance. Unfortunately, we lack detailed, granular data to verify these conjectures. Electronic copy available at: https://ssrn.com/abstract=3909219 5.3.2 Other Strategic Reasons- Customer Overpayment? Kitzmueller and Shimshack (2012) suggest that some customers are willing to pay more for the products of rms engaged in CSR-related activities. We ask whether the reduction in CSR expenditure by erstwhile high spenders re ects consumers' unwillingness to pay more for products sold by rms whose CSR activities are mandated by the government. The proxy we use is the ex-CSR pro t margin (Panel A of A11). Note that the change in the CSR spending post-mandate does not mechanically in uence this variable. If consumers are unwilling to pay more for a product with mandatory CSR, the ex-CSR margin (pro t margin without considering CSR) should decline. We conduct a di-in-di test of the form of equation 1 with ex-CSR margin as the dependent variable and report the results in Panel A of A11 of the online appendix. We nd no change in a di-in-di sense. The result implies no change in consumers' willingness to pay for products in response to mandated CSR. 5.3.3 Other Strategic Reasons- Labor Donations? We next consider the \labor donation" argument, which posits that some employees are willing to accept lower wages to work for socially responsible rms (Greening and Turban (2000)). We ask whether the dilution of the labor donations channel in response to the mandate leads to reduced spending on CSR by the mandated rms within the high CSR rms. If the labor donation channel works, we would expect wages for the mandated rms within the high CSR group to increase in the post-mandate period. However, we do not nd such an increase as reported in Table A12 of the online appendix. 5.3.4 Other Strategic Reasons- Politics The politics view of CSR posits that CSR deters interventions by activists (Davidson III, Worrell, and El-Jelly (1995)) and governments (Khanna and Anton (2002)). Such interven- In Panel B, we use the margin before CSR and salaries as the dependent variable and nd similar results. The variable accounts for any change in salaries due to the mandate. We elaborate more on this point in Section 5.3.3. Electronic copy available at: https://ssrn.com/abstract=3909219 tions can make a rm's product or service unpopular. However, it is challenging to estimate in advance the level of CSR spending activists and governments expect from the rm. An explicit 2% rule clearly signals what governments and activists expect in terms of CSR. Therefore, rms that invested in CSR with a political motive likely converge their spending to the 2% limit. Those who invested more (less) reduce (increase) CSR expenditure. We rely on highly polluting rms as a sub-sample of companies that likely invest in CSR to deter governments and activists from unwanted intervention. We obtain data relating to highly polluting rms from the database maintained by the Ministry of Environment. Our focus is on the triple interaction between Treatment, Post, and PollutingFirms. In the results reported in Table A13 of the online Appendix, the triple interaction term is statistically indistinguishable from zero. Therefore, reduction in CSR spending is likely not driven by politically sensitive rms. 5.3.5 Revelation of Social Norms We now consider the more general possibility of learning about society's expectations from the mandate. Before the 2% mandate, we observe heterogeneity in CSR spending plausibly because rms have diculty in assessing the social norms for CSR spending. Thus, we would expect to observe a reduction (increase) in spending by high (low) CSR rms after the mandate. In an analogous situation, Rose and Wolfram (2002) nd that an attempt to restrict CEO pay to USD one million using tax policy had the perverse eect of treating USD 1 million as a new expected norm whereby CEO pay increased in cases where it was less than USD 1 million before. If realization of a new social norms is the dominant explanation, both mandated and non- mandated voluntary high spenders should have moved symmetrically towards the 2% limit as the norms apply to all rms. Our results show that, even among high spenders, mandated rms cut CSR spending more. While 43% of non-mandated high spenders cut spending after the mandate, the proportion is more than 80% for the mandated high spenders. Mandated Source: http://pib.nic.in/newsite/PrintRelease.aspx?relid=137373 Electronic copy available at: https://ssrn.com/abstract=3909219 high spenders also cut more of the actual Rupees spent. In sum, we believe the dilution of signaling value of CSR likely explains fall in CSR spending over and above the revelation of societal expectations. Our belief is based on a combined assessment of the following empirical observations (i) excess reduction of CSR spending by mandated high spenders; (ii) the replacement of CSR by advertising; (iii) the loss of shareholder value and operating performance linked to the 2% rule; and (iv) the higher rates of standardized CSR communications. 6 Discussion 6.1 Signaling Impact on Spend Above 2% Thus far, we have interpreted the cut in voluntary CSR spending by the erstwhile high spenders as a dilution in the ability of CSR to signal both product quality and virtue. A skeptic can ask what stops high CSR rms from deriving signaling bene ts on voluntary CSR spending above the mandated 2% level? We believe that regulatory intervention has increased compliance costs and reduced the signaling bene ts of CSR across the board, for both the 2% mandated part and any voluntary spending above that 2% threshold for four reasons. First, CSR spending is not easily divisible into strict 2% and above 2% compartments. As an illustration, consider a mandated rm that intends to spend 10% of pro ts in a year on a CSR mission with signaling bene ts. Assume that the future spending on this project is contingent on the perceived strategic bene ts. Before the mandate, the marketing department could have worked out and executed a CSR project. Post mandate, it is hard to design a CSR budget such that 2% falls within compliance with the law, and the remaining budget is intended to be spent to signal virtue and product quality. Second, before the mandate the rm did not have to worry about any externally imposed de nition of CSR that is also subject to ex-post regulatory interpretation. As mentioned Electronic copy available at: https://ssrn.com/abstract=3909219 in Section 5.3.1, the ministry of corporate aairs now has discretionary powers to question what constitutes CSR for a particular rm. Hence, additional spending over and above 2% level potentially constitutes slack that the rm can use to protect itself against regulatory questioning. Such slack-based spending is unlikely to have the same signaling value as unfettered CSR outlays. Third, the board of directors did not have to get involved in the activity earlier, and their approval was not explicitly required. Relatedly, the rm could exibly continue or stop CSR projects in future years. Now, the 2% will have to be spent every year. Hence, the marketing department may have to commit spending for a certain number of years to obtain board approval. It might be dicult for the board to nd a worthy CSR opportunity at the last minute if a prior project were discontinued. Fourth, the issue of tax-deductibility of the entire CSR expenditure is sub-judice. The revenue department claims that the entire CSR expenditure is not tax-deductible. Because of such additional costs, rms may look for a exible alternative way of signaling. Accordingly, mandated high CSR rms increasingly use advertisements as a replacement for CSR spending. These rms are likely to view CSR as a tax and hence reduce product quality and virtue signaling in CSR communication. As noted in Section 5.3.1, we do nd signi cant changes in CSR communication that show de-emphasize on signaling product quality and virtue. 6.2 Compliance And Signaling The discussion in section 6.1 shows that mandated high CSR rms could be impacted both due to loss of direct signaling value and also higher cost of compliance, making CSR a less preferred way of signaling. We do not have a de nitive way of disentangling the two costs partly because these costs reinforce one another. With that caveat in mind, we test whether larger rms behave any dierently when compared to smaller rms. Larger rms should be able to absorb the xed costs of compliance better. If compliance costs were the prime Electronic copy available at: https://ssrn.com/abstract=3909219 driver behind reduced CSR spending documented in Table 4, larger rms should be less impacted by the mandate. We estimate a triple interaction speci cation (Treated X Post X Large Firms) and report the results in Table A14 of the online appendix. In our main speci cation, where we use the CSR Amount as the dependent variable, larger rms cut CSR spending more than the smaller rms. When we consider the ratio of CSR Amount and assets, we nd that the large rms cut CSR spending as much as the small rms. These ndings are inconsistent with a simple compliance cost explanation. 6.3 Are Control Group Firms Also Aected? A change in the marketing and CSR strategy of one set of rms in an industry impacts all rms in the ecosystem. Even non-mandated rms are likely to change their CSR, commu- nications, and marketing strategy. Therefore, our identi cation strategy of comparing the mandated and non-mandated rms can only capture the relative dierence in the new rules impact and not the complete causal eect. First, as discussed above in Section 6.1, compliance costs are higher for mandated rms. Several restrictions such as the need to obtain the board of directors' approval, scrutiny by the ocials of the ministry of corporate aairs, and others apply only to mandated rms. As noted in Section 2, the ministry has issued over 1,000 notices rejecting claimed expenses as CSR. Non-mandated rms do not face any of these hurdles. As long as they follow accounting standards, no regulatory questions can be asked about the nature of such expenditure. Second, as noted in section 2, mandated rms lose exibility concerning CSR spending as they must spend at least 2% of pro ts every year and strictly adhere to the law. For instance, a strategy of spending a large amount on CSR once in many years and nothing in between becomes impossible for the mandated rms. The non-mandated rms can continue with such a strategy. Third, we nd a higher impact of the law on stock price reaction and operating perfor- mance of mandated set among the high CSR rms. Fourth, mandated rms increase ad- Electronic copy available at: https://ssrn.com/abstract=3909219 vertising expenditure in proportion to the decline in CSR expenditure in the post-mandate period. Finally, textual analysis suggests that the mandated set among the high CSR rms reduces the usage of words that convey product quality and virtue through CSR compared to the non-mandated group among high CSR rms. 6.4 Non Strategic Motives The existence of strategic outcomes alone is not sucient to rule out the existence of non-strategic motives such as stakeholder altruism (Reinhardt, Stavins, and Vietor (2008), B enabou and Tirole (2010)) and manager moral hazard (Cheng, Hong, and Shue (2013), Masulis and Reza (2014)). For instance, a rm may derive signaling bene ts even when altruism motivates its CSR spending. Therefore, we cannot design sharp tests to cleanly separate strategic and non-strategic motives. Our results can only reject non-strategic motives as the sole explanation for voluntary CSR. However, we do not believe that the data support the claim that non-strategic motives primarily drove voluntary CSR spenders, and the signaling bene t was only an unintended consequence. If that were indeed the case, impacted rms would not have increased adver- tisement spending and changed communication strategy in response to the mandate. 7 Conclusion In this study, we examine the impact of a regulatory edict related to minimum CSR spending on the actual CSR spending of rms. We rely on the recent law passed in India that requires all rms above a certain threshold to spend at least 2% of their average three years' pro ts on CSR. We examine the impact of this law on rms that were voluntarily engaged in CSR before the regulation was passed relative to those that were not. Within voluntary high spenders in the pre-mandate period, we compare rms mandated to spend on CSR and those that are not. Electronic copy available at: https://ssrn.com/abstract=3909219 We nd that voluntary spenders reduce their CSR spending signi cantly after the man- date to the legally prescribed limit of 2% of the average three years' pro ts. On the other hand, rms that spent less on CSR during the pre-regulation period increase their spending slightly to meet the new requirement. Our ndings suggest that the imposition of mandatory CSR crowds out voluntary spending on CSR. In the second part of the paper, we attempt to understand the channel at work behind such cuts. The negative share price reaction of the mandated rms that used to spend more than 2% of their pro ts in the pre-mandate period indicates that some strategic motive be- hind CSR becomes less eective after the mandate. We nd that aected rms increase their advertising expenditure. They also change the tone of the CSR communication signi cantly away from signaling product quality and virtue to one of compliance. These results indicate that voluntary CSR signals product quality and virtue to stakeholders. Regulators might want to consider the possible impact of a proposed intervention on the CSR spending of rms that voluntarily engage in pro-social behavior. In the short run, a mandate may plausibly lead to increased total CSR spending because the edict brings a larger number of rms into the mandatory CSR net. In fact, a recent KPMG report shows that the total CSR spending increased following the regulation. However, if the compulsion to spend on CSR crowds out voluntary spending, then such a mandate may lead to a reduction in CSR spending in the long run. Firms that would have voluntarily spent on CSR, with some persuasion by NGOs, may not do so when regulation is imposed. In such a case, the magnitude of CSR spending in the pre-regulation period may not serve as the appropriate counterfactual. Many Indian rms would have potentially initiated CSR investments voluntarily in the absence of a mandate in response to growing prosperity, given that India is among the fastest-growing large economies in the world. We acknowledge that we cannot make de nitive welfare claims. The overall welfare depends on several factors such as (i) the utility of increased spending when the country Source-KPMG Report can be found here: https://assets.kpmg.com/content/dam/kpmg/in/pdf/2018/02/CSR- Survey-Report.pdf Electronic copy available at: https://ssrn.com/abstract=3909219 is underdeveloped relative to spending at a higher stage of development; (ii) the relative eciency of private spending compared with that of government spending on social issues; (iii) the impact of CSR on tax compliance; and (iv) other factors. We do not have credible evidence on these fronts. Further, as acknowledged before, our identi cation strategy allows us to estimate only the dierential impact between the mandated and non-mandated rms as opposed to the complete causal impact of the mandate. Finally, we do not have high-quality survey or eld evidence to test what kind of stakeholders were impacted by dierent types of signaling. We rely instead on carefully constructed archival proxies for drawing inferences on signaling. Our study opens up interesting follow-up questions. First, how does the government of India intend to assess the social impact of mandatory CSR spending? Second, does the imposition of a mandate reduce the responsiveness of rms to local disasters? Top management may feel that they have discharged their social responsibility by complying with the regulation and, hence, do not need to do more. Answers to these questions may reveal additional costs of mandating CSR over and above those identi ed in this study. Electronic copy available at: https://ssrn.com/abstract=3909219 References Benabou, R., and J. Tirole (2010): \Individual and corporate social responsibility," Economica, 77(305), 1{19. Blacconiere, W. G., and W. D. Northcut (1997): \Environmental information and market reactions to environmental legislation," Journal of Accounting, Auditing & Fi- nance, 12(2), 149{178. Chen, Y.-C., M. Hung, and Y. Wang (2018): \The eect of mandatory CSR disclosure on rm pro tability and social externalities: Evidence from China," Journal of Accounting and Economics, 65(1), 169{190. Cheng, B., I. Ioannou, and G. Serafeim (2014): \Corporate social responsibility and access to nance," Strategic Management Journal, 35(1), 1{23. Cheng, H., H. Hong, and K. Shue (2013): \Do managers do good with other people's money?," Discussion paper, National Bureau of Economic Research. Christensen, H. B., E. Floyd, L. Y. Liu, and M. Maffett (2017): \The real eects of mandated information on social responsibility in nancial reports: Evidence from mine- safety records," Journal of Accounting and Economics, 64(2-3), 284{304. Christensen, H. B., L. Hail, and C. Leuz (2013): \Mandatory IFRS reporting and changes in enforcement," Journal of Accounting and Economics, 56(2-3), 147{177. (2019): \Adoption of CSR and Sustainability Reporting Standards: Economic Analysis and Review," Discussion paper, National Bureau of Economic Research. Daske, H., L. Hail, C. Leuz, and R. Verdi (2013): \Adopting a label: Heterogeneity in the economic consequences around IAS/IFRS adoptions," Journal of Accounting Research, 51(3), 495{547. Davidson III, W. N., D. L. Worrell, and A. El-Jelly (1995): \In uencing managers to change unpopular corporate behavior through boycotts and divestitures: A stock market test," Business & Society, 34(2), 171{196. Dedman, E., and A. Kausar (2012): \The impact of voluntary audit on credit ratings: evidence from UK private rms," Accounting and Business Research, 42(4), 397{418. Deng, X., J.-k. Kang, and B. S. Low (2013): \Corporate social responsibility and stake- holder value maximization: Evidence from mergers," Journal of Financial Economics, 110(1), 87{109. Dhaliwal, D. S., O. Z. Li, A. Tsang, and Y. G. Yang (2011): \Voluntary non nancial disclosure and the cost of equity capital: The initiation of corporate social responsibility reporting," The Accounting Review, 86(1), 59{100. Dharmapala, D., and V. Khanna (2018): \The impact of mandated corporate social responsibility: evidence from Indias Companies Act of 2013," International Review of Law and Economics, 56, 92{104. Electronic copy available at: https://ssrn.com/abstract=3909219 Dimson, E., O. Karakas , and X. Li (2015): \Active ownership," The Review of Financial Studies, 28(12), 3225{3268. Elliott, W. B., S. M. Grant, and K. M. Rennekamp (2017): \How disclosure fea- tures of corporate social responsibility reports interact with investor numeracy to in uence investor judgments," Contemporary Accounting Research, 34(3), 1596{1621. Flammer, C. (2015): \Does corporate social responsibility lead to superior nancial perfor- mance? A regression discontinuity approach," Management Science, 61(11), 2549{2568. Florou, A., and P. F. Pope (2012): \Mandatory IFRS adoption and institutional in- vestment decisions," The Accounting Review, 87(6), 1993{2025. Greening, D. W., and D. B. Turban (2000): \Corporate social performance as a com- petitive advantage in attracting a quality workforce," Business & society, 39(3), 254{280. Hung, M., and K. Subramanyam (2007): \Financial statement eects of adopting in- ternational accounting standards: the case of Germany," Review of Accounting Studies, 12(4), 623{657. Kausar, A., N. Shroff, and H. White (2016): \Real eects of the audit choice," Journal of Accounting and Economics, 62(1), 157{181. Khanna, M., and W. R. Q. Anton (2002): \What is driving corporate environmentalism: Opportunity or threat?," Corporate Environmental Strategy, 9(4), 409{417. Kim, J.-B., J. S. Tsui, and H. Y. Cheong (2011): \The voluntary adoption of Interna- tional Financial Reporting Standards and loan contracting around the world," Review of Accounting Studies, 16(4), 779{811. Kitzmueller, M., and J. Shimshack (2012): \Economic perspectives on corporate social responsibility," Journal of Economic Literature, 50(1), 51{84. Kotler, P., and N. Lee (2005): \Best of breed: When it comes to gaining a market edge while supporting a social cause,corporate social marketing leads the pack," Social Marketing Quarterly, 11(3-4), 91{103. (2008): Corporate social responsibility: Doing the most good for your company and your cause. John Wiley & Sons. Lennox, C. S., and J. A. Pittman (2011): \Voluntary audits versus mandatory audits," The Accounting Review, 86(5), 1655{1678. Lins, K. V., H. Servaes, and A. Tamayo (2017): \Social capital, trust, and rm per- formance: The value of corporate social responsibility during the nancial crisis," The Journal of Finance. Manchiraju, H., and S. Rajgopal (2017): \Does Corporate Social Responsibility (CSR) Create Shareholder Value? Evidence from the Indian Companies Act 2013," Journal of Accounting Research. Electronic copy available at: https://ssrn.com/abstract=3909219 Margolis, J. D., H. A. Elfenbein, and J. P. Walsh (2009): \Does it pay to be good... and does it matter? A meta-analysis of the relationship between corporate social and nancial performance," Working Paper. Masulis, R. W., and S. W. Reza (2014): \Agency problems of corporate philanthropy," The Review of Financial Studies, 28(2), 592{636. McWilliams, A., D. S. Siegel, and P. M. Wright (2006): \Corporate social respon- sibility: Strategic implications," Journal of Management Studies, 43(1), 1{18. Orlitzky, M., F. L. Schmidt, and S. L. Rynes (2003): \Corporate social and nancial performance: A meta-analysis," Organization Studies, 24(3), 403{441. Perrini, F., A. Russo, A. Tencati, and C. Vurro (2011): \Deconstructing the rela- tionship between corporate social and nancial performance," Journal of Business Ethics, 102(1), 59{76. Reinhardt, F. L., R. N. Stavins, and R. H. Vietor (2008): \Corporate Social Re- sponsibility Through an Economic Lens," Review of Environmental Economics and Policy, 2(2), 219{239. Rose, N. L., and C. Wolfram (2002): \Regulating executive pay: Using the tax code to in uence chief executive ocer compensation," Journal of Labor Economics, 20(S2), S138{S175. Waagstein, P. R. (2011): \The mandatory corporate social responsibility in Indonesia: Problems and implications," Journal of Business Ethics, 98(3), 455{466. Electronic copy available at: https://ssrn.com/abstract=3909219 Figure 1: Comparison Between Mandated And Non-Mandated Firms Within Sub Samples This gure depicts our identi cation strategy. High (low) CSR spenders are those that spend more (less) than 2% of their average three-year pro ts on CSR in the pre-mandate period. The rms that are required by law to spend on CSR form the mandated group and those that are not so required form the non-mandated group. Electronic copy available at: https://ssrn.com/abstract=3909219 Figure 2: Pre And Post Trend- Comparison between Mandated and Non- Mandated Firms Within High CSR Category This gure depicts the pre and post trend between the treated and the control groups in terms of CSR spending. The sample is restricted to high CSR spenders in the pre-mandate period. In other words, all rms in the sample spend more than 2% of their average three- year pro ts on CSR in the pre-mandate period. The rms that are required by law to spend on CSR form the treated group and those that are not so required form the control group. The vertical axis plots the ratio of the CSR amount and the total assets and the horizontal axis denotes the years. The blue line represents the treated group and the orange line the control group. 2010 2012 2014 2016 2018 2020 Year Treatment Control Electronic copy available at: https://ssrn.com/abstract=3909219 CSR_Asset .002 .003 .004 .005 .006 34 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 1: Variable Definition In this table, we de ne the key variables. Variable De nition High CSR Firms Firms that spent more than a threshold, usually 2%, in terms of CSR-to-pro t ratio in the pre-regulation period. We also use other thresholds. Pro t here refers to the average pro ts of the previous three years. Low CSR Firms Firms that spent less than a threshold, usually 2%, in terms of CSR-to-pro t ratio in the pre-regulation period. Post-regulation period Financial years 2014-2015 and after. Mandated Firms Firms that breach in any one or more of the criteria speci ed by Section 135 of the Companies Act. These are INR 50 million in pro ts; INR 5 billion in net worth; INR 10 billion in sales. The values are arrived at based on annual averages in the pre-mandate period. CSR- Amount INR spending on CSR. Pre-regulation period data comes from Prowess; Post-regulation period data comes from the ministry for rms covered by the ministry and from Prowess database for other rms. CSR Ratio The ratio between CSR and average pro ts after tax in the previous three years. Pro ts Pro t After Tax at a rm-year level. Net worth Book value of equity at a rm-year level. TABLE 2: Sample Construction In this table, we report details about the sample used. Variable Value Firms in Prowess Database 43,051 Firms in Ministry Database 12,097 Firms in MCA that could be merged with Prowess 10,154 Firms in both the pre and post-mandate Period 39,309 Number of sample years 10 Total observations in the merged data set 236,044 Total observations with non-missing average CSR numbers in the pre-mandate period 44,769 Observations with CSR more than 2% of average three year pro ts 16,251 Observations with CSR less than 2% of average three year pro ts 28,518 Observations with CSR more than 5% of average three year pro ts 9,108 Observations with CSR more than 7.5% of average three year pro ts 6,873 Observations with CSR more than 10% of average three year pro ts 5,523 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 3: Univariate Comparison Between High And Low CSR Firms In this table, we present univariate comparisons between the high CSR and the low CSR rms. CSR Ratio, as de ned in Table 1, is the dependent variable. The sample is restricted to rms that earn positive pro ts across all the sample years. Firms are grouped based on their pre-regulation spending on CSR. In column 1, we consider rms that spent nothing on CSR in the pre-mandate period. In column 2, we consider rms that spent between 0 to 1 percent of previous three years' average pro ts in the pre-regulation period. Similarly, in each column, we consider a progressively higher range. We compare the dierence between pre-regulation and post-regulation expenditure and also report the t-statistics. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. Group-Based Pre-Period CSR Ratio 0 0-1 1-2 2-3 3-4 4-5 5-6 6-7 7-10 More than 10 Pre 0.000 0.004 0.015 0.025 0.036 0.046 0.056 0.066 0.087 0.276 Post 0.024 0.021 0.024 0.026 0.030 0.036 0.031 0.038 0.050 0.103 Di(post-pre) -0.024 -0.017 -0.009 -0.001 0.006 0.010 0.025 0.029 0.037 0.173 T-Stat -11.124 -28.556 -8.074 -0.697 2.491 2.672 6.807 5.098 8.004 24.777 No. of observations 2878 15040 5718 3079 1975 1501 1031 676 1230 4857 Electronic copy available at: https://ssrn.com/abstract=3909219 37 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 4: Comparison Between Mandated and Non-Mandated Firms Within High CSR Firms In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In panel A, the CSR Amount, as de ned in Table 1, is the dependent variable. In Panel B, the ratio between CSR amount and total assets is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the last three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is a mandated rm and zero otherwise. The rm-year level controls included in columns 5 to 8 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. Dependent Variable Panel A: CSR Amount Treatment X Post-Mandate -7.87*** -18.58*** -24.63*** -27.80*** -7.49*** -18.98*** -25.17*** -29.00*** [-5.84] [-7.60] [-7.51] [-7.87] [-5.33] [-8.13] [-7.98] [-7.93] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,678 8,708 6,544 5,251 15,673 8,704 6,541 5,248 R-squared 0.71 0.66 0.67 0.68 0.72 0.68 0.68 0.68 Dependent Variable Panel B: CSR/Total Assets Treatment X Post-Mandate -0.0014*** -0.0040*** -0.0056*** -0.0067*** -0.0014*** -0.0041*** -0.0057*** -0.0067*** [-3.1038] [-5.4022] [-5.6622] [-5.5534] [-3.2415] [-5.6697] [-6.1635] [-6.3933] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,073 8,288 6,216 4,971 15,070 8,285 6,213 4,968 R-squared 0.56 0.57 0.58 0.59 0.56 0.57 0.58 0.59 TABLE 5: Summary Comparison In this table, we compare the treated and the control groups in terms of various accounting ratios. Mandated (Non-mandated) rms form the treated (control) groups. The sample is restricted to rms that spend more than 2% of their average three-year pro ts on CSR in the pre-mandate period. Variable Control Treatment Dierence T- Stat Margin 0.12 0.13 -0.01 -0.95 Leverage 1.13 0.63 0.50 1.24 Turnover 1.40 0.91 0.48 0.81 Return On Equity 0.41 0.35 0.06 0.25 Advertisement To Sales Ratio 0.02 0.01 0.00 0.82 Pro t Per Employee 0.44 0.28 0.16 0.10 Electronic copy available at: https://ssrn.com/abstract=3909219 39 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 6: Comparison Between Mandated and Non-Mandated Firms Within High CSR Firms In Terms of Oper- ating Performance In this table, we compare the operating performance before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The return on equity (ROE) (the return on assets (ROA))(natural logarithm of sales) is the dependent variable in columns 1 and 4 (2 and 5) (3 and 6). The sample is restricted to rms investing above 2% of last three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls such as one year lagged pro ts and net worth in columns 4 and 5. In column 6, we use the current year pro ts and net worth as control variables. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) Dependent Variable ROE ROA log Sales ROE ROA log Sales Treatment x Post-mandate -0.0229* -0.0093*** 0.0720 -0.0263* -0.0099*** 0.0605 [-1.7267] [-2.8758] [1.4981] [-1.7999] [-2.8863] [1.4045] Lag Pro t After Tax 0.0000 0.0000 [0.8915] [1.6313] Net Worth -0.0000 0.0000 0.0000 [-1.2108] [1.6392] [1.0428] Pro t After Tax 0.0000 [1.5784] Firm Fixed Eects Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Observations 23,629 23,629 23,521 22,389 22,387 23,520 R-squared 0.3768 0.4606 0.8850 0.3927 0.4774 0.8854 TABLE 7: The Value Impact Of Mandatory CSR In this table, we present the results relating to stock price reaction to mandatory CSR. We use the July 17, 2019 announcement that made the violation of mandatory CSR a criminal oense as the event in Panel A and the announcement of the withdrawal of the above provision on August 23, 2019, as the event in Panel B. The three-day excess return around the event is the dependent variable. Treatment is a dummy variable that takes the value of one if the rm under consideration is mandated to invest in CSR and zero otherwise. In columns 1 and 2, we consider all rms. In columns 3 and 4, the data are restricted to rms that used to spend more than 2% of their average three-year pro ts on CSR in the pre-mandate period. To account for liquidity, we include turnover as a control variable in columns 2 and 4. We employ industry xed eects and cluster the errors at an industry level. Robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. Panel A: CSR violation to be a criminal oence (1) (2) (3) (4) Dependent Variable Excess Return Mandatory -0.005*** -0.012*** -0.005* -0.008* (0.002) (0.003) (0.003) (0.004) Turnover 0.002 0.008*** 0.008*** 0.013*** (0.002) (0.003) (0.002) (0.003) Industry F.E Yes Yes Yes Yes Sample All Firms High CSR Firms Observations 1,994 1,994 824 824 R-squared 0.094 0.125 0.047 0.061 Panel B: CSR violation not to be a criminal oence Dependent Variable Excess Return Mandatory 0.012*** 0.014*** 0.013*** 0.017*** (0.003) (0.004) (0.004) (0.006) Turnover -0.000 0.001 -0.003 0.000 (0.003) (0.003) (0.003) (0.003) Industry F.E Yes Yes Yes Yes Sample All Firms High CSR Firms Observations 1,994 1,994 824 824 R-squared 0.091 0.093 0.068 0.059 Electronic copy available at: https://ssrn.com/abstract=3909219 41 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 8: Test of Signaling Hypothesis- Social Media Communication In this table, we examine the social media communications of rms through their ocial Twitter handles. Each observation represents a rm-year. In columns 1 and 4, the proportion of CSR related tweets over all tweets is the dependent variable. In columns 2 and 5 (3 and 6), the proportion of words signaling product or service quality (virtue) within CSR related tweets is the dependent variable. The procedure used to identify CSR related tweets, words signaling quality of products and services, and words signaling virtue is detailed in Section 9 of the online appendix. The sample is restricted to rms investing above 2% of the proportion of the previous three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. The rm-year level controls included in columns 4 to 6 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) Dependent Variable prop csr prop product prop charity prop csr prop product prop charity Treatment x Post -0.02* -0.02* -0.01** -0.03* -0.02* -0.01** [-1.87] [-1.99] [-2.24] [-1.99] [-1.89] [-2.20] Firm Level Controls No No No Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Observations 704 500 409 704 500 409 R-squared 0.58 0.39 0.61 0.58 0.39 0.61 42 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 9: Test of Signaling Hypothesis- Annual Reports In this table, we examine the CSR sections of annual reports of rms. Each observation represents a rm-year. In columns 1 and 4, average cosine similarity at a rm-year level is the dependent variable. For each mandated (non-mandated) rm, we calculate the cosine similarity of the CSR section of the annual report with every other mandated (non-mandated) rm in a year and calculate the average of the cosine similarity score. The average so calculated is the dependent variable. In columns 2 and 5 (3 and 6), the proportion of words signaling product or service quality (virtue) within CSR section of the annual report is the dependent variable. The procedure used to identify words signaling quality of products and services, and words signaling virtue is detailed in Section 9 of the online appendix. The sample is restricted to rms investing above 2% of the proportion of the previous three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. The rm-year level controls included in columns 4 to 6 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) Dependent Variable Similarity Product Charity Similarity Product Charity Treatment x Post-mandate 0.21*** -0.19*** -0.08*** 0.22*** -0.19*** -0.09*** [3.78] [-6.73] [-6.54] [3.86] [-6.21] [-6.04] Firm Level Controls No No No Yes Yes Yes Firm Fixed Eect Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Observations 242 242 242 242 242 242 R-squared 0.70 0.66 0.78 0.70 0.66 0.78 43 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE 10: Test of Signaling Hypothesis- Advertisement In this table, we compare advertising expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In Panel A (B), the INR amount spent on advertising (the ratio between advertisement expenditure and total assets) is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. The rm-year level controls included in columns 5 to 8 are pro t after tax and net worth. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Panel A: Advertising Expenditure Treatment X Post-mandate 31.97*** 32.64*** 36.79*** 38.66** 31.56*** 33.55*** 37.03*** 38.00*** [4.97] [2.80] [2.76] [2.60] [5.25] [3.18] [2.92] [2.80] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 16,300 8,964 6,711 5,397 16,292 8,961 6,708 5,394 R-squared 0.87 0.89 0.89 0.88 0.87 0.90 0.89 0.88 Dependent Variable Panel B: Advertisement/Total Assets Treatment X Post-mandate 0.0052** 0.0071* 0.0050 0.0081* 0.0051** 0.0070* 0.0049 0.0079* [2.1987] [1.8668] [1.0296] [1.7337] [2.1703] [1.8227] [0.9889] [1.6888] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,657 8,554 6,398 5,135 15,652 8,552 6,396 5,133 R-squared 0.59 0.59 0.57 0.62 0.59 0.59 0.57 0.62 Online Appendix Does Mandated Corporate Social Responsibility Reduce Intrinsic Motivation? Evidence from India Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A1: Comparison of Difference Between CSR Amount As Per Two Databases In this table, we present the distribution of the dierence between the CSR amount as per the prowess database and the ministry database for the same rm. The dierence is normalized using pro ts. Statistic Value Mean 0 Median 0 Q1 0 Q3 0 5th Percentile 0 95th Percentile 0 1st Percentile -10.7 99th Percentile 0.15 Electronic copy available at: https://ssrn.com/abstract=3909219 46 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A2: Comparison of Difference Between CSR Amount As Per Two Databases-Examples In this table, we provide ve examples of large deviations between the Prowess and MCA database. Company Name Year Actual CSR (INR mill) Ministry CSR (INR mill) Prowess CSR (INR mill) Reliance Industries Limited 2015 7,610 7,610 15,660 Reliance Industries Limited 2016 6,520 6,520 13,280 NTPC Limited 2016 4,918 4,918 8,272 South Eastern Coal elds Limited 2015 404 404 1,313.3 Power Finance Corporation 2015 516.8 516.8 1,178.3 47 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A3: Pre And Post Comparison Within Mandated High CSR Firms In this table, we compare CSR expenditure before and after the government mandate for mandated high CSR rms. Each observation represents a rm-year. In Panel A, CSR Amount, as de ned in Table 1, is the dependent variable. In Panel B, the ratio between CSR amount and total assets is the dependent variable. The sample is restricted to mandated rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The rms that are required to spend on CSR as per law are considered mandated. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value of one for years after the regulation change, and zero otherwise. The rm-year level controls included in columns 5 to 8 are pro t after tax and net worth. We also include rm xed eects in all 8 columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Panel A: CSR Amount post-mandate -7.63*** -18.58*** -24.84*** -28.06*** -7.48*** -18.93*** -25.14*** -28.97*** [-5.60] [-7.61] [-7.52] [-7.73] [-5.31] [-8.09] [-7.95] [-7.90] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 5,869 2,486 1,633 1,215 5,867 2,485 1,633 1,215 R-squared 0.68 0.62 0.62 0.63 0.70 0.64 0.64 0.64 Dependent Variable Panel B: CSR To Assets Ratio post-mandate -0.0039*** -0.0077*** -0.0099*** -0.0114*** -0.0039*** -0.0077*** -0.0099*** -0.0114*** [-9.7106] [-10.5799] [-9.5238] [-9.0795] [-10.9308] [-11.5832] [-11.3180] [-11.2556] Firm-Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 5,755 2,426 1,588 1,181 5,755 2,426 1,588 1,181 R-squared 0.46 0.51 0.53 0.56 0.46 0.51 0.54 0.56 48 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A4: Comparison Between Mandated and Non-Mandated Firms Within Low CSR Firms In this table, we compare CSR expenditure before and after the government mandate for rms that used to spend less than 2% of average three years' pro ts before the mandate. CSR Amount, as de ned in Table 1, is the dependent variable in columns 1, 2, 3 and 4. The ratio between CSR amount and the total assets is the dependent variable in columns 5, 6, 7, and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. In columns 1, 2, 5, and 6, we compare the post and pre-mandate spending by rms that are required to spend on CSR by law but were spending less than the stipulated level in the pre-mandate period. Each observation represents a rm-year. In columns 3, 4, 7, and 8 we conduct di-in-di tests. Treatment and Control are as de ned in Table 4. We include rm-year level controls in columns 2, 4, 6, and 8. We include rm xed eects in all columns and year xed eects in columns 3, 4, 7, and 8. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR/Total Assets Post-Mandate 9.23*** 8.81*** 0.0004*** 0.0004*** [9.84] [10.04] [4.2066] [4.2267] Post-Mandate X Treatment 12.98*** 12.49*** 0.0006*** 0.0006*** [9.97] [10.16] [5.2417] [5.2892] Firm-Level Controls No Yes No Yes No Yes No Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects No No Yes Yes No No Yes Yes Observations 27,932 27,913 27,932 27,913 27,527 27,509 27,527 27,509 R-squared 0.68 0.69 0.69 0.70 0.5533 0.5547 0.5568 0.5582 49 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A5: Impact of the CSR Mandate- Triple Interaction In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms in a triple interaction framework. Each observation represents a rm-year. CSR Amount, as de ned in Table 1, is the dependent variable in columns 1 and 2. The ratio between CSR amount and total assets is the dependent variable in columns 3 and 4. High CSR is a dummy variable that takes the value of one for rms investing above 2% of the previous three years' average pro ts in the pre-mandate period and zero otherwise. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in columns 2 and 4. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) Dependent Variable CSR Amount CSR/Total Assets Treatment x Post x High CSR -20.91*** -20.51*** -0.0020*** -0.0020*** [-12.09] [-12.41] [-4.6661] [-4.6519] Mandate x Post 13.00*** 12.54*** 0.0006*** 0.0006*** [9.96] [10.26] [5.3128] [5.3554] High CSR x Post -0.79*** -0.79*** -0.0025*** -0.0025*** [-3.31] [-3.29] [-10.3162] [-10.3050] Firm-Level Controls No Yes No Yes Firm Fixed Eects Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Observations 43,610 43,586 42,600 42,579 R-squared 0.70 0.70 0.5841 0.5846 50 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A6: Comparison Between Mandated and Non-Mandated Firms Within High CSR Firms- Test of Pre- mandate Trend In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The ratio between CSR Amount, as de ned in Table 1, and total assets is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in column 1, 5% in column 2, 7.5% in column 3, and 10% in column 4. Post- mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Pre Year 1 is a dummy variable taking the value one for the year just before the mandate and zero otherwise. Pre Year 2 is a dummy variable taking the value one for the year two years before the mandate and zero otherwise. Pre Year 3 is a dummy variable taking the value one for the year three years before the mandate and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is a mandated rm and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) Dependent Variable CSR/Total Assets post-mandate X Treatment -0.0014*** -0.0043*** -0.0058*** -0.0066*** [-2.6571] [-4.6457] [-4.4503] [-4.2774] Pre Year 3 X Treatment 0.0001 -0.0006 -0.0003 0.0000 [0.1501] [-0.6293] [-0.2653] [0.0248] Pre Year 2 X Treatment 0.0000 -0.0002 0.0001 0.0005 [0.0710] [-0.1855] [0.0988] [0.2709] Pre Year 1 X Treatment -0.0001 -0.0002 0.0002 0.0005 [-0.2374] [-0.2557] [0.1583] [0.2980] Firm Level Controls Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Observations 15,070 8,285 6,213 4,968 R-squared 0.56 0.57 0.58 0.59 51 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A7: Comparison Between Mandated and Non-Mandated Firms- High CSR- Dropping Years In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In columns 1 to 4 (5 to 8) CSR amount (the ratio between CSR amount and total assets) is the dependent variable. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. We leave out years 2012-2013 and 2013-2014. We do not consider the above years while identifying the treated and control groups in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR/Total Assets post-mandate X Treatment -6.63*** -19.29*** -24.43*** -29.09*** -0.0011** -0.0037*** -0.0052*** -0.0065*** [-3.46] [-5.55] [-5.68] [-5.95] [-2.0871] [-3.8028] [-4.7241] [-4.6288] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 10,074 5,576 4,131 3,322 9,657 5,276 3,901 3,124 R-squared 0.72 0.68 0.68 0.69 0.5662 0.5760 0.5820 0.5936 52 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A8: Comparison Between Mandated and Non-Mandated Firms- High CSR-Placebo Test In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. We use a false treatment year of 2011-2012. We restrict the sample to pre-mandate period. Each observation represents a rm-year. CSR amount (the ratio between CSR amount and total assets), as de ned in Table 1, is the dependent variable in columns 1 to 4 (5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-placebo treatment period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the placebo treatment year, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR /Total Assets Treatment X Post-mandate -2.02 -3.65 -3.77 -3.76 -0.0002 -0.0008 -0.0012 -0.0013 [-1.52] [-1.44] [-1.04] [-1.19] [-0.4258] [-0.8285] [-0.8319] [-0.8894] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 5,855 3,508 2,664 2,177 5,583 3,310 2,509 2,046 R-squared 0.87 0.88 0.87 0.87 0.6824 0.6699 0.6676 0.6604 53 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A9: Comparison Between Mandated and Non-Mandated Firms- High CSR-Excluding Firms Close to the Threshold In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. CSR amount (the ratio between CSR amount and total assets), as de ned in Table 1, is the dependent variable in columns 1 to 4 (5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. We exclude control group rms close to the threshold. We rst calculate the distance of a rm from the threshold following Manchiraju and Rajgopal (2017) and exclude those that are less than one standard deviation of such distances for all rms from the threshold. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR /Total Assets Treatment X Post-mandate -5.67*** -12.54*** -16.47*** -21.20*** -0.0016*** -0.0040*** -0.0055*** -0.0068*** [-6.41] [-7.41] [-6.71] [-6.90] [-3.2012] [-4.9494] [-5.1719] [-5.4145] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 14,346 8,242 6,232 5,021 13,759 7,829 5,909 4,743 R-squared 0.56 0.58 0.58 0.59 0.5738 0.5844 0.5921 0.6011 54 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A10: Test of Signaling Hypothesis- CSR And Advertising Expenditure In this table, we compare the sum of advertising and CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The sum of the amount spent on advertising and CSR (the ratio between the sum of the amount spent on advertising and CSR and total assets) is the dependent variable in columns 1 to 4 (5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Advertisement and CSR Amount Advertisement and CSR to Assets Ratio Treatment x Post-mandate 1,053.01 869.55 1,302.59 1,751.16 -0.09 869.55 1,302.59 1,751.16 [1.38] [1.00] [0.99] [1.00] [-0.32] [1.00] [0.99] [1.00] Firm Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 20,001 11,326 8,573 6,898 19,217 11,326 8,573 6,898 R-squared 0.10 0.11 0.11 0.11 0.10 0.11 0.11 0.11 55 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A11: Comparison Between Mandated and Non-Mandated Firms Within High CSR- The Consumer Channel In this table, we compare the ex-CSR margin before and after the government mandate between treatment and control rms. In Panel A (B), margin before CSR (margin before CSR and salary) is the dependent variable. Each observation represents a rm-year. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in columns ve to eight. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Panel A: Ex CSR Margin Treatment X Post-mandate 0.19 -0.84 -1.62 -2.39 -0.64 -1.37 -2.04 -2.72 [0.55] [-0.86] [-0.95] [-1.05] [-1.14] [-1.12] [-1.08] [-1.07] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Observations 23,519 13,672 10,440 8,418 23,518 13,671 10,439 8,417 R-squared 0.13 0.14 0.13 0.13 0.13 0.14 0.13 0.13 Dependent Variable Panel B: Margin including CSR Exp and Wages Treatment X Post-mandate -0.88 -1.81 -2.54 -3.33 -0.88 -1.81 -2.54 -3.33 [-1.28] [-1.17] [-1.09] [-1.09] [-1.28] [-1.17] [-1.09] [-1.09] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 23,254 13,571 10,387 8,369 23,254 13,571 10,387 8,369 R-squared 0.14 0.14 0.13 0.13 0.14 0.14 0.13 0.13 56 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A12: Comparison Between Mandated and Non-Mandated Firms Within High CSR- Labor Donations Chan- nel In this table, we compare the wage expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in columns 1 and 5, 5% in columns 2 and 6, 7.5% in columns 3 and 7, and 10% in columns 4 and 8. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. We include rm-year level controls in columns 5 to 8. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Wages Treatment X Post-mandate 0.0006 0.0011 0.0015 0.0018 -0.0000 0.0001 0.0002 0.0004 [1.1855] [1.1761] [1.1001] [1.0992] [-0.1142] [0.2620] [0.2686] [0.3335] Firm Level Controls No No No No Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eect Yes Yes Yes Yes Yes Yes Yes Yes Observations 22,006 12,752 9,724 7,836 22,005 12,751 9,723 7,835 R-squared 0.2175 0.2154 0.1964 0.1881 0.2172 0.2150 0.1958 0.1873 TABLE A13: Preempting Public Or Private Politics In this table, we compare the CSR expenditure before and after the government mandate between treatment and control rms in a triple interaction framework. CSR Amount, as de ned in Table 1 (the ratio between CSR amount and total assets), is the dependent variable in columns 1 to 4 (columns 5 to 8). The sample is restricted to rms investing above a threshold in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. The threshold used is 2% in column 1, 5% in column 2, 7.5% in column 3, and 10% in column 4. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. Red is a dummy variable that takes the value of one for rms belonging to polluting industries. We include rm-year level controls in all columns. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable CSR Amount CSR/Total Assets Treatment X Post-mandate -7.77*** -17.91*** -22.47*** -26.04*** -0.0011** -0.0037*** -0.0051*** -0.0055*** [-5.03] [-6.42] [-6.33] [-6.78] [-2.3399] [-5.0419] [-5.6535] [-5.3400] Red X Post 0.38 0.39 0.46 0.58 0.0008* 0.0012* 0.0014 0.0020** [1.46] [0.81] [0.69] [0.73] [1.6826] [1.7018] [1.6311] [2.1020] Post X Red X Treatment 0.28 -2.76 -6.72 -7.15 -0.0009 -0.0011 -0.0018 -0.0034 [0.10] [-0.59] [-1.12] [-1.01] [-1.2113] [-0.8661] [-1.0080] [-1.6414] Firm-Level Controls Yes Yes Yes Yes Yes Yes Yes Yes Firm Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Eects Yes Yes Yes Yes Yes Yes Yes Yes Observations 15,519 8,598 6,452 5,166 14,953 8,205 6,147 4,906 R-squared 0.72 0.68 0.68 0.68 0.5645 0.5786 0.5872 0.5981 Electronic copy available at: https://ssrn.com/abstract=3909219 58 Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A14: Comparison Between Mandated and Non-Mandated Firms- Compliance And Signaling In this table, we compare CSR expenditure before and after the government mandate between treatment and control rms. Each observation represents a rm-year. In columns 1 and 2 (3 and 4) CSR amount (the ratio between CSR amount and total assets) is the dependent variable. The sample is restricted to rms investing above a 2% in terms of the proportion of the previous three years' average pro ts in the pre-mandate period. Post-mandate is a dummy variable taking the value one for years after the regulation change, and zero otherwise. Treatment is a dummy variable that takes the value of one if the rm under consideration is required to spend on CSR as per law and zero otherwise. Large rms is a dummy variable that takes the value of one for rms above the medium in terms asset values in the pre-mandate period and zero otherwise. We include rm-year level controls in columns 2 and 4. We include rm and year xed eects in all columns. Errors are clustered at the industry level and robust t-statistics are reported in parentheses. ***, **, and * represent signi cance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) Dependent Variable CSR Amount CSR Amount CSR Asset CSR Asset Post x Mandate -3.26*** -3.19*** -0.0025** -0.0025** [-3.06] [-3.00] [-2.1755] [-2.1596] Post x Large Firms -0.49 -0.50 0.0001 0.0001 [-1.15] [-1.18] [0.2097] [0.2180] Large Firms x Mandate x Post -4.85*** -4.68*** 0.0012 0.0012 [-2.89] [-2.85] [1.0260] [1.0201] Controls No Yes No Yes Year Fixed Eect Yes Yes Yes Yes Firm Fixed Eect Yes Yes Yes Yes Observations 15,678 15,673 15,073 15,070 R-squared 0.71 0.72 0.5638 0.5638 8 Industry Level CSR Spending Electronic copy available at: https://ssrn.com/abstract=3909219 TABLE A15: Industry Level CSR Spending: In this table, we calculate CSR expenditure by rms before the mandate at an industry level. Here Industry classi cation follows the three digit National Industrial Classi cation (NIC- 2008) by Ministry of Corporate Aairs (MCA). Column 1 presents the total CSR Amount, that is aggregate value of CSR expenditure by an Industry. Column 2 presents average CSR Amount per rm, that is aggregate CSR expenditure (Column 1) by total rm years in an industry. Column 3 presents average CSR ratio, that is aggregate CSR ratio, de ned as per table1, by total rm years in an industry. NIC Code Industry Total CSR Average CSR Average CSR Amount(INR Amount(INR Proportion Million) Million) 11 Growing of non-perennial crops 561.60 4.07 0.0026 12 Growing of perennial crops 31.70 0.79 0.0020 13 Plant propagation 0.10 0.05 0.0029 14 Animal production 92.10 1.00 0.0024 16 Support activities to agriculture and post-harvest crop activities 23.40 0.90 0.0021 23 Gathering of non-wood forest products 155.00 6.46 0.0016 24 Support services to forestry 24.10 2.01 0.0106 32 Aquaculture 123.50 3.63 0.0073 51 Mining of hard coal 7848.20 115.41 0.0095 52 Mining of lignite 339.60 84.90 0.0031 61 Extraction of crude petroleum 716.50 65.14 0.0008 71 Mining of iron ores 2598.60 66.63 0.0040 72 Mining of non-ferrous metal ores 1335.90 46.07 0.0071 81 Quarrying of stone, sand and clay 81.70 1.00 0.0021 89 Mining and quarrying n.e.c. 1267.60 37.28 0.0061 101 Processing and preserving of meat 2142.80 52.26 0.0156 102 Processing and preserving of sh, crustaceans and molluscs 139.00 3.31 0.0073 103 Processing and preserving of fruit and vegetables 1.60 0.15 0.0034 104 Manufacture of vegetable and animal oils and fats 398.30 2.36 0.0012 105 Manufacture of dairy products 58.10 0.95 0.0027 106 Manufacture of grain mill products, starches and starch products 83.50 0.98 0.0017 107 Manufacture of other food products 1687.80 2.96 0.0028 108 Manufacture of prepared animal feeds 30.20 0.86 0.0026 110 Manufacture of beverages 964.80 4.14 0.0026 120 Manufacture of tobacco products 1192.60 31.38 0.0035 131 Spinning, weaving and nishing of textiles 1753.90 2.40 0.0017 139 Manufacture of other textiles 80.80 0.64 0.0013 141 Manufacture of wearing apparel, except fur apparel 266.70 1.91 0.0029 143 Manufacture of knitted and crocheted apparel 36.40 0.81 0.0006 151 Tanning and dressing of leather; manufacture of luggage, handbags, sad- 111.80 3.73 0.0126 dlery and harness; dressing and dyeing of fur 152 Manufacture of footwear 134.10 2.16 0.0041 161 Sawmilling and planing of wood 0.40 0.20 0.0011 162 Manufacture of products of wood, cork, straw and plaiting materials 119.10 1.89 0.0016 170 Manufacture of paper and paper products 750.20 2.57 0.0023 181 Printing and service activities related to printing 287.40 13.06 0.0015 191 Manufacture of coke oven products 0.80 0.11 0.0025 192 Manufacture of re ned petroleum products 4191.60 38.11 0.0016 201 Manufacture of basic chemicals, fertilizer and nitrogen compounds, plastics 4929.90 7.90 0.0062 and synthetic rubber in primary forms 202 Manufacture of other chemical products 6597.70 12.15 0.0042 203 Manufacture of man-made bres 1004.80 23.92 0.0018 210 Manufacture of pharmaceuticals, medicinal chemical and botanical prod- 7637.30 10.31 0.0033 ucts 221 Manufacture of rubber products 215.30 1.84 0.0015 222 Manufacture of plastics products 637.60 1.54 0.0014 231 Manufacture of glass and glass products 155.20 3.23 0.0011 239 Manufacture of non-metallic mineral products n.e.c. 3867.30 10.62 0.0019 241 Manufacture of basic iron and steel 2926.90 3.35 0.0010 242 Manufacture of basic precious and other non-ferrous metals 4496.20 27.58 0.0032 243 Casting of metals 415.10 1.63 0.0014 251 Manufacture of structural metal products, tanks, reservoirs and steam 1335.80 12.97 0.0012 generators Electronic copy available at: https://ssrn.com/abstract=3909219 NIC Code Industry Total CSR Average CSR Average CSR Amount(INR Amount(INR Proportion Million) Million) 252 Manufacture of weapons and ammunition 0.10 0.10 0.0029 259 Manufacture of other fabricated metal products; metalworking service ac- 259.70 1.06 0.0016 tivities 261 Manufacture of electronic components 46.90 0.72 0.0029 262 Manufacture of computers and peripheral equipment 20.30 1.56 0.0008 263 Manufacture of communication equipment 269.40 5.99 0.0024 264 Manufacture of consumer electronics 562.90 43.30 0.0005 265 Manufacture of measuring, testing, navigating and control equipment; 448.70 4.67 0.0019 watches and clocks 266 Manufacture of irradiation, electromedical and electrotherapeutic equip- 32.00 1.19 0.0027 ment 271 Manufacture of electric motors, generators, transformers and electricity 355.80 2.34 0.0012 distribution and control apparatus 272 Manufacture of batteries and accumulators 583.60 18.24 0.0021 273 Manufacture of wiring and wiring devices 292.20 2.40 0.0016 274 Manufacture of electric lighting equipment 46.10 2.00 0.0007 275 Manufacture of domestic appliances 59.70 1.57 0.0024 279 Manufacture of other electrical equipment 171.90 1.87 0.0019 281 Manufacture of general purpose machinery 1749.70 5.07 0.0021 282 Manufacture of special-purpose machinery 841.00 3.05 0.0020 291 Manufacture of motor vehicles 1438.70 49.61 0.0005 292 Manufacture of bodies (coachwork) for motor vehicles; manufacture of 442.20 2.70 0.0010 trailers and semi-trailers 293 Manufacture of parts and accessories for motor vehicles 552.70 1.42 0.0015 301 Building of ships and boats 521.40 18.62 0.0005 302 Manufacture of railway locomotives and rolling stock 18.70 0.89 0.0017 303 Manufacture of air and spacecraft and related machinery 0.00 0.00 0.0000 309 Manufacture of transport equipment n.e.c. 989.00 20.60 0.0014 310 Manufacture of furniture 22.90 2.86 0.0055 321 Manufacture of jewellery, bijouterie and related articles 941.90 4.40 0.0012 323 Manufacture of sports goods 5.30 0.53 0.0017 325 Manufacture of medical and dental instruments and supplies 20.70 1.15 0.0017 329 Other manufacturing n.e.c. 3.00 0.43 0.0009 351 Electric power generation, transmission and distribution 14008.10 25.66 0.0023 360 Water collection, treatment and supply 42.10 14.03 0.0002 370 Sewerage 0.20 0.10 0.0053 382 Waste treatment and disposal 1.10 1.10 0.0202 410 Construction of buildings 4400.20 4.99 0.0022 421 Construction of roads and railways 2018.90 13.20 0.0018 422 Construction of utility projects 1521.00 6.98 0.0015 429 Construction of other civil engineering projects 2571.90 6.26 0.0025 431 Demolition and site preparation 139.10 4.64 0.0017 432 Electrical, plumbing and other construction installation activities 0.50 0.50 451 Sale of motor vehicles 177.80 1.57 0.0018 453 Sale of motor vehicle parts and accessories 37.40 1.04 0.0008 454 Sale, maintenance and repair of motorcycles and related parts and acces- 0.50 0.10 0.0012 sories 461 Wholesale on a fee or contract basis 801.60 2.17 0.0048 462 Wholesale of agricultural raw materials and live animals 294.50 2.45 0.0045 463 Wholesale of food, beverages and tobacco 176.90 1.47 0.0045 464 Wholesale of household goods 585.80 1.41 0.0028 465 Wholesale of machinery, equipment and supplies 705.40 2.20 0.0035 466 Other specialized wholesale 2229.50 3.60 0.0023 469 Non-specialized wholesale trade 1217.00 2.28 0.0047 471 Retail sale in non-specialized stores 9.20 0.58 0.0027 472 Retail sale of food, beverages and tobacco in specialized stores 65.10 10.85 0.0174 Electronic copy available at: https://ssrn.com/abstract=3909219 NIC Code Industry Total CSR Average CSR Average CSR Amount(INR Amount(INR Proportion Million) Million) 474 Retail sale of information and communications equipment in specialized 38.50 1.75 0.0009 stores 475 Retail sale of other household equipment in specialized stores 21.10 0.50 0.0013 476 Retail sale of cultural and recreation goods in specialized stores 4.30 1.08 0.0008 477 Retail sale of other goods in specialized stores 199.90 2.30 0.0024 479 Retail trade not in stores, stalls or markets 1.30 0.16 0.0057 491 Transport via railways 777.10 33.79 0.0030 492 Other land transport 160.50 1.56 0.0024 493 Transport via pipeline 179.60 8.98 0.0008 501 Sea and coastal water transport 49.40 1.34 0.0023 511 Passenger air transport 64.60 5.38 0.0009 521 Warehousing and storage 1885.00 15.45 0.0016 522 Support activities for transportation 2383.60 7.24 0.0028 532 Courier activities 24.90 1.92 0.0050 551 Short term accommodation activities 610.00 1.89 0.0023 563 Beverage serving activities 1.70 0.34 0.0037 581 Publishing of books, periodicals and other publishing activities 625.00 7.18 0.0023 591 Motion picture, video and television programme activities 87.20 1.25 0.0013 601 Radio broadcasting 0.00 0.00 0.0000 602 Television programming and broadcasting activities 186.70 7.47 0.0005 611 Wired telecommunications activities 31.20 1.42 0.0029 612 Wireless telecommunications activities 1267.60 57.62 0.0003 619 Other telecommunications activities 9.80 0.58 0.0015 620 Computer programming, consultancy and related activities 2613.80 7.97 0.0030 631 Data processing, hosting and related activities; web portals 5.60 0.40 0.0030 639 Other information service activities 372.70 4.10 0.0019 641 Monetary intermediation 3678.20 3.56 0.0028 643 Trusts, funds and other nancial vehicles 3714.40 5.79 0.0041 649 Other nancial service activities, except insurance and pension funding 4705.60 5.63 0.0025 activities 651 Insurance 5.50 5.50 0.0378 661 Activities auxiliary to nancial service activities, except insurance and 1166.60 3.61 0.0019 pension funding 663 Fund management activities 29.80 0.56 0.0007 682 Real estate activities on a fee or contract basis 0.30 0.15 0.0004 691 Legal activities 0.30 0.30 0.0010 702 Management consultancy activities 903.10 6.02 0.0058 711 Architectural and engineering activities and related technical consultancy 540.00 3.83 0.0039 731 Advertising 67.90 1.89 0.0047 732 Market research and public opinion polling 0.80 0.10 0.0069 749 Other professional, scienti c and technical activities n.e.c. 0.00 0.00 0.0000 781 Activities of employment placement agencies 38.90 1.62 0.0098 791 Travel agency and tour operator activities 34.70 0.72 0.0013 822 Activities of call centres 0.20 0.10 0.0010 823 Organization of conventions and trade shows 32.90 6.58 0.0064 829 Business support service activities n.e.c. 66.40 1.90 0.0064 841 Administration of the State and the economic and social policy of the 190.80 4.65 0.0031 community 851 Primary education 5.20 0.74 0.0006 853 Higher education 205.40 4.28 0.0091 854 Other education 0.60 0.20 0.0007 861 Hospital activities 469.50 2.50 0.0041 862 Medical and dental practice activities 2.50 0.31 0.0006 869 Other human health activities 7.50 0.75 0.0029 900 Creative, arts and entertainment activities 3.90 1.95 0.0013 932 Other amusement and recreation activities 117.90 3.47 0.0051 941 Activities of business, employers and professional membership organiza- 13.40 2.68 0.0136 tions 949 Activities of other membership organizations 392.50 39.25 0.0195 Electronic copy available at: https://ssrn.com/abstract=3909219 9 A Note On Dictionaries Used 9.1 Twitter Scraping We use the prowess database to get a list of all rm names. Then, the names of the com- panies are processed to remove unnecessary spaces within the names, which were present in the original list. These processed names are used as input to search the Twitter handles associated with those names. To download these tweets, we use Tweepy, a python library that is used to interface with the Twitter API. To lter out the Twitter handles with more precision, we use a string similarity metric. It generates similarity scores between 0 and 1, 1 being the same and 0 being no similarity present. We use a threshold of 0.8 to match prowess names with names mentioned in the Twitter handles. After ltering based on this threshold, we obtain approximately 8000 handles for 3162 rms. To scrape the tweets from these handles, we use a python library named Twint. The process created a dataset of 9.8 million tweets by 3162 companies for 6 years beginning from 2012. 9.2 Dictionaries: We rst lter out CSR-relevant tweets present within the dataset. Dictionary methods are very popular and established in natural language processing to label datasets or categorize the data present within them. Therefore, to obtain the CSR relevant tweets, we use the CSR dictionary built by Provalis Research. A tweet is identi ed as a CSR tweet if any word present within the CSR dictionary is present in the tweet. The Provalis Research CSR Dictionary consists of 1432 words across 4 dierent sections. The four sections are: Human Rights, Employee and Employment, Social and Community, and Environment. In addition, we remove the repeated words and commonly used verbs, adverbs, and adjectives which essentially by themselves cannot be assumed to be in relation to CSR activities. Upon applying this lter to the 9.8 million tweets present in our dataset, we identify close to 7.4% (730,000) as CSR-related. We then move to identify words related to product and charity to understand the distribution of the tweets. We use words and phrases taken from Buers (A social media tool to enhance engagement) website for identifying product-related words. The website identi es commonly used terms to market products and services on social media platforms. We build a dictionary relating words indicating charity and virtue using the words used by the charity and non-governmental organizations in their reports. We use online platforms such as \Merriam Websters thesaurus," \Related Words" , and \Your Dictionary" to identify the synonyms. https://provalisresearch.com/products/content-analysis-software/wordstat-dictionary/corporate-social- responsibility-text-analytics/ https://buer.com/library/words-and-phrases-that-convert-ultimate-list/ Electronic copy available at: https://ssrn.com/abstract=3909219 10 Alternative Explanation{Possible Change in Account- ing/Control Systems A skeptic might argue that the accounting and internal control systems relating to CSR changed after the mandate. For example, a change in the accounting/control system could mechanically aect our results if treatment rms reclassify certain CSR expenses from the pre-mandate period as non-CSR in the post-mandate period and the control rms behaved in exactly the opposite manner. We have carefully looked at the accounting/auditing standards applicable for CSR clas- si cation before and after the CSR spending mandate to address this concern. A rm is allowed to classify only those expenditures that are charitable in nature as CSR. Deliberate misclassi cation of business expenditure, say on research and development, as CSR is con- sidered as misrepresentation of books of accounts under the Indian Companies Act. None of the above regulations changed with the mandate. Therefore, a change in CSR classi cation in response to the mandate would be tantamount to an open admission of misrepresenta- tion, either during the pre-mandate or the post-mandate period. We believe that such an admission of misrepresentation is unlikely. Electronic copy available at: https://ssrn.com/abstract=3909219

Journal

ARN Conferences & Meetings – SSRN

Published: Aug 22, 2021

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India

Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India

Does Mandated Corporate Social Responsibility Crowd Out Voluntary Corporate Social Responsibility? Evidence from India

References (19)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies