The results illustrate that the inclusion of social responsibility information in financial reports can have real effects. Dodd-Frank's requirement to include mine safety data in financial reports improves safety but reduces productivity. The real effects of mandated social responsibility disclosures in financial reports: Evidence from mine-.
Abstract: We examine the real effects of mandatory social responsibility disclosures that require SEC-registered mine owners to include their mine safety records in their financial reports. These safety records are already publicly available elsewhere, allowing us to isolate and estimate the incremental real effects of including this information in financial reports. Section 1503 of the Dodd-Frank Act requires SEC-registered firms to include information about mine safety in their financial reports.
A key feature of our approach is that the information provided through MSD is already publicly available on the Mine Safety and Health Administration's (MSHA) website; this allows us to isolate and estimate the size of the incremental real effects of including information in financial reports. Our paper contributes to the existing literature by documenting the magnitude of the real effects of including social responsibility information in financial reports.
ACCEPTED MANUSCRIPT
Real effects of MSD
In our analyzes of the actual effects of MSD, we focus on changes in safety citations, mining-related injuries, and labor productivity around the enactment of MSD. In this section, we assess the effect of MSD on the incidence rate of citations for violations of the Mining Act and mining-related injuries. Although the one-year incidence rates correspond to the length of the 10K reporting period, one year is a relatively short interval over which to measure infrequent outcomes such as citations and injuries.
An OLS regression does not effectively account for this concentration of observations, which can lead to biased estimates of the treatment effect (Wooldridge 2002). We find that the magnitude of the coefficient on the log of the conditional mean is close to one, indicating that overdispersion is not a serious problem. We estimate the magnitude of the bias this problem creates using a jackknife procedure (dropping each period in turn) and find that the bias is less than 5% of the treatment effects reported in the paper.
We include a relatively long preliminary period (six years) to allow a better assessment of the assumption of parallel trends (see section 4.3.2).9. In fact, 75% of companies that own coal mines have a primary industry that is not coal mining. In this section, we present the results of our analysis of the effect of MSD on citations.
We provide further evidence on the shared balance between MSD and non-MSD mines in relation to matching analysis in Appendix C. Overall, descriptive statistics show that MSD and non-MSD mines are similar in terms of citations they receive than to condition MSD. Given that our objective is to assess whether the MSD improves compliance, ideally we will examine actual Mine Act violations, rather than citations for violations. In an untabulated analysis, we find that the probability of closing a mine with citations per inspection hour above the highest decile of the 2009 citation distribution increases in the post-MSD period by about 1% for MSD relative to non-MSD mines .
The ideal setting to isolate and estimate the magnitude of the incremental real-world effects of including information in financial reports would be one in which the exact same safety reports included in financial reports were already publicly available elsewhere. Our setting falls short of this experimental ideal because the structure of the data on the MSHA website differs from the appendix in the 10K. Our assumption is that if the names of the legal owner of the mine and the SEC-registered parent company are the same, then the link between the MSHA website and the financial reports is straightforward.
In none of the three figures is there evidence of a differential response between MSD and non-MSD mines to the MINES ACT, suggesting that publicly and privately owned mines respond similarly to unobservables that precede regulations. 24. address a variety of other threats to our identification, including assessments of 1) the extent to which unobservables related to the Upper Big Branch disaster affect our results;14 2) the possibility that MSD and non-MSD firms respond differently to changes in macroeconomic conditions around the time of the adoption of Dodd-Frank; 3) the potentially confounding effects of two concurrent MSHA regulatory initiatives, which may affect MSD and non-MSD firms differently;.
MSD filings and awareness of safety violations
In this section, we compare stock returns with short windows following the public announcement of an imminent danger order (IDO) in the pre- and post-MSD period. In the post-MSD period, MSHA posts IDOs on their website and firms communicate them through an 8K file. If MSD-8Ks increase investor awareness of mine safety issues, we expect to observe a greater market response to the announcement of an IDO in the post-MSD period.
In the pre-MSD period, our event window covers the disclosure of the IDO on the MSHA website that occurs the morning after the IDO is issued. In the post-MSD period, the event window covers MSHA's website disclosure and the release of the MSD-8K, which must be filed within four business days of the IDO date. 17 In practice, most 8Ks are filed within two days of the IDO posting on the MSHA website, preventing us from separately examining market reactions to website postings and 8K filings in the post-MSD period.
In the pre-MSD period, when an IDO is only disclosed on the MSHA website, the mean and median CARs are close to zero and not statistically different from zero. Consistent with an increase in investor awareness in the post-MSD period, the difference in mean (mean) pre-period CAR and post-period CAR is statistically significant at the 1% level. Focusing on the post-minus pre-MSD average return differences in column (5) of Table 8, we find that the event window CAR is and 0.12% for firms in the coal, general and non-mining industries, respectively. .
In addition, there is potentially considerable heterogeneity among mutual funds in terms of sensitivity to security issues. 19 For example, one time-varying factor that could influence our results is the general increase in safety attention after the Upper Big Branch disaster in the post-MSD era. If it is true that the greater responses to IDO seen in the post-MSD period are attributable to greater safety concerns rather than MSD, we would expect to see similar increases in this period.
Using the Thomson Reuters Mutual Fund Database, we identify fund holdings for 111 of 151 companies subject to the MSD for the period from 2003, when quarterly holdings reports were required in the US, to 2014.20 The average firm owns approximately 31% of outstanding shares. The MSD is an indicator coded as one if the IDO is disclosed on the MSHA website and extended through 8K (ie, in the post-MSD period). In the post-MSD period, the coefficient on MSD × SRI × IDO is -0.097, indicating that the increased sensitivity of SRI assets to safety is further enhanced when IDO is also present.
Conclusion
First, the main threat to identification in our analyzes is violation of the assumption of parallel trends. These terms allow us to identify disclosures both in the 10K schedules (Exhibits 95 and 99 are often used) and in the body of the filing. We then compile a comprehensive list of MSD mines from these records and submit it to the MSHA database based on mine names and numbers.
Contractors are not involved in the operation of the mine and therefore have less influence on the safety of the mine. Second, for seven companies we were unable to match all MSD mines with the MSHA databases due to ambiguities in the disclosed names. Weights are then assigned to the control observations such that the representation of the control group in each tank matches that of the treatment group.
However, because the results of the matching procedure are similar across the samples, we only discuss detailed results for the citation sample. More importantly, when we apply the CEM weights, we observe a slight attenuation (increase) in the magnitude of the estimated coefficient on MSD in any of the three specifications. Notes: This table reports results from our analysis of the real effects of MSD before and after coarsened exact matching (CEM).
We provide a detailed description of the variables in the notes to Table 2 and describe our data collection procedures in Appendix B. Notes: This table presents descriptive statistics for issuers subject to section 1503 of the Dodd-Frank Act. Notes: This table reports results from our analysis of the effect of MSD on citation rates using both OLS and Poisson regression.
Notes: This table reports results from our analysis of the effect of MSD on severe and non-severe citation rates using Poisson regressions. Notes: This table reports results from our analysis of the effect of MSD on injury rates using both OLS and Poisson regressions. Notes: This table reports results from our analysis of the effect of MSD on labor productivity using an OLS regression and repeats the analyzes for citations and injuries from Tables 3 and 5, restricting the sample to coal mining using Poisson regression over two-year periods .
Notes: This table reports the results of our main analysis of the true effects of MSD, with the exception of the subset of MSD mines where the names of the SEC filer and legal owner are not nearly identical. We provide a detailed description of the variables in the notes to Table 2 and describe our data collection procedures in Appendix B.
