Corporate climate change disclosure during the Trump administration: evidence from standalone CSR reports

ABSTRACT

In support of legitimacy theory arguments for corporate social responsibility disclosure, a variety of prior studies document that either increases or decreases in social and regulatory cost exposures appear to lead to corresponding changes in the level of information provided. However, none of these studies investigates a situation where social and regulatory cost exposure changes are not in sync; a situation we argue taking place with respect to the Trump presidency in the United States, and none considers the role that political polarization might play in company responses. Based on a sample of 170 large U.S. firms, we find that the extent of climate change disclosure in standalone CSR reports did not appear to change, on average, from the Pre-Trump (2014–2015) to the Trump (2017–2018) eras. However, cross sectional analysis indicates more negative changes in disclosure for companies headquartered in states strongly supporting Trump in the 2016 election relative to other firms and more positive changes for companies in carbon intensive industries. Where firms are subject to both factors, we provide evidence that, at least with respect to space allocated to climate change disclosure in the reports, the increased social concerns appear to dominate reduced regulatory exposures. Our results support legitimacy theory arguments, but illustrate the importance of disentangling potentially competing effects, and suggest attention needs to be paid to potential differing political climates in the locales of organizations’ headquarters.

KEYWORDS:

• Climate change
• environmental disclosure
• legitimacy theory
• Trump administration

Acknowledgements

The authors thank participants at the following conferences/seminars for helpful comments on earlier drafts of the paper: CSEAR International Congress on Social and Environmental Accounting Research 2019; Spanish Conference on Social and Environmental Accounting Research 2019; University of Exeter; Bangor University; University of Cincinnati; IESEG School of Management; Illinois State University. The authors also thank Associate Editor Chris van Staden and two anonymous reviewers for their helpful comments and suggestions. Carla Antonini acknowledges the support of the Spanish Ministry of Science & Innovation under Grant PID2019-104163RA-I00.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Throughout this paper we refer to the latter two sources combined as “regulatory cost” exposures.

2 As we identify in more detail below, we classify states as “Trump states” based on the voting outcomes from the 2016 U.S. presidential election.

3 Popovich, N., Albeck-Ripka, L., and Pierre-Louis, K. “67 Environmental Rules on the Way Out Under Trump.” Available at www.nytimes.com/interactive/2017/10/05/climate/trump-environment-rules-reversed.html.

4 As explained by Popovich et al. in their New York Times article (see preceding footnote), the old rule required factories emitting excess airborne pollutants to install technologies reducing emissions to the maximum level achievable and to maintain these controls even if emissions dipped below the threshold. The Trump administration removed the latter part of the requirement.

5 See, e.g., https://thehill.com/policy/energy-environment/425361-epa-referrals-for-criminal-prosecution-lowest-since-1988-report.

6 Zinke resigned his position in December of 2018.

7 Grandoni, D. “Environmental Donations are Up Thanks to Trump.” Available at: https://www.buzzfeednews.com/article/dinograndoni/environmental-trump-card.

8 Fandos, N. “Climate March Draws Thousands of Protesters Alarmed by Trump's Environmental Agenda.” Available at: https://www.newyorktimes.com/2017/04/29/US/politics/peoples-climate-march-trump.html.

9 Newport, F. “Americans Want Government to Do More on the Environment.” Available at: https://news.gallup.com/poll/232007/americans-want-government-more-environment.aspx.

10 Consistent with the results for the studies examining changes in environmental disclosure, a variety of additional investigations document increases in the provision of other types of CSR information in response to social and regulatory cost-inducing events. These include Coetzee and van Staden (2011) (mine safety disclosures), Blanc et al. (2019) (compliance and corruption-related disclosures), and Vourvachis et al. (2016) (health and safety disclosures).

11 Seyle and Newman (2006) identify that the color coding comes from the maps used by news agencies in reporting results of U.S. elections, with the specific color connotations for political party success taking hold with the protracted coverage of the highly contested 2000 presidential election (also see Klinker 2004).

12 We do not include companies where, for example, one report was a full CSR report but the report available for the other period was only a summary report. 

13 Available at database.globalreporting.org.

14 Results including these two firms are qualitatively consistent with the primary results reported below.

15 In non-tabulated sensitivity tests, we repeat all analyses including only companies with both 2015 and 2017 reports. All results remain qualitatively similar to those reported below.

16 A list of sample companies is available from the authors.

17 There was no consistent pattern with respect to increases or decreases in the size of the reports, and a t-test of differences in mean total pages in the reports across the Pre-Trump and Trump Era periods was not statistically significant at conventional levels.

18 The space devoted to environmental/climate change discussion in the CEO letters was not included as part of the space-based measures used to assess disclosure changes.

19 While we believe the metrics capturing the relative attention paid to climate change issues is less noisy due to the lack of consistency in overall report length and the length of the CEOs’ statements across periods and across firms, we run sensitivity tests using the raw space measures. See discussion below.

20 On the surface, our scenario looks like it might lend itself to a difference-in-differences (DID) design. However, Glaeser and Guay (2017, p. 310) identify that DID in theory requires what they refer to as the “perfect compliance assumption,” which they explain as “the assumption that no firms received the treatment in the pre-treatment period and that all firms in the treatment group – and only those firms – received the treatment in the post-treatment period.” As such, it is not clear what could be seen as the “treatment”. It's not Trump State as that doesn't change from pre-Trump to post-Trump, and it can't be Trump's election as that applies to all firms in the sample. As such, we do not employ a DID design for our study.

21 Classifying states’ primary political ideology based on presidential election voting is consistent with a variety of prior studies including Crouch and Abbot (2009), Grooms (2015), and Levendusky and Pope (2011), among others.

22 This includes firms from SIC codes 10xx-14xx, 26xx, 28xx, 29xx, 30xx-39xx, 45xx, and 49xx. However, following Patten (2002) and Cho et al. (2010) we do not include pharmaceutical companies (SIC 283x) within this group.

23 Given the significant correlations, we run collinearity statistics for all regression models. In no cases do variance inflation factor scores exceed 2.0. As such, multi-collinearity does not appear to be a concern with our analyses.

24 The BBC's Matt McGrath notes that the support actually increased as both Syria and Nicaragua signed on following Trump's decision to pull the U.S. out of the agreement. See bbc.com/news/science-environment-44330709.

25 Results remain qualitatively similar using the percentage of foreign sales in place of the classification variable.

26 This is consistent with the approach used by Rudin (2008) to capture home state political leanings.

27 Due to missing data from year 2015 for one firm, the models using the changes in financial control variables are based on 169 and 154 observations, respectively, for the carbon page percentage and CEO discussion models.