truth hurts? mandatory information disclosure and ...based tools, understanding the implications of...

52
Truth Hurts? Mandatory Information Disclosure and Regulatory Activities (Job Market Paper) Zhengyan Li O’Neill School of Public and Environmental Affairs Indiana University Abstract This paper examines the impacts of mandatory environmental information disclosure policy on the implementation of traditional environmental regulations in the context of the Toxics Release Inventory (TRI), which is the major environmental information disclosure program in the United States, and the Clean Air Act (CAA). Using quasi-experimental variation from TRI’s size-based reporting criteria and its expansion of industry coverage in 1998, I find that regulators significantly reduce the number of regulatory activities on facilities that disclose information in the TRI. Regulatory activities, however, do not seem to react to year-over-year change of disclosed TRI information. The results suggest strong intercorrelations between different forms of environmental policy. Information disclosure policy can complement the implementation of traditional environmental regulations as it provides new information for regulatory decision making and leverages actions from other information users to motivate improvement of environmental performance. At the same time, regulators’ reactions to information can serve as a mechanism for information disclosure policy to achieve its regulatory goals.

Upload: others

Post on 25-May-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

Truth Hurts? Mandatory Information Disclosure and Regulatory Activities

(Job Market Paper)

Zhengyan Li O’Neill School of Public and Environmental Affairs

Indiana University

Abstract This paper examines the impacts of mandatory environmental information disclosure policy on the implementation of traditional environmental regulations in the context of the Toxics Release Inventory (TRI), which is the major environmental information disclosure program in the United States, and the Clean Air Act (CAA). Using quasi-experimental variation from TRI’s size-based reporting criteria and its expansion of industry coverage in 1998, I find that regulators significantly reduce the number of regulatory activities on facilities that disclose information in the TRI. Regulatory activities, however, do not seem to react to year-over-year change of disclosed TRI information. The results suggest strong intercorrelations between different forms of environmental policy. Information disclosure policy can complement the implementation of traditional environmental regulations as it provides new information for regulatory decision making and leverages actions from other information users to motivate improvement of environmental performance. At the same time, regulators’ reactions to information can serve as a mechanism for information disclosure policy to achieve its regulatory goals.

Page 2: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

This Page is Intentionally Left Blank

Page 3: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

1

1 Introduction Information-based regulatory tools have increasingly been adopted by legislators and regulators to deal with thorny environmental challenges such as drinking water safety, toxic emissions, fracking pollution, and climate change. Embodying the widely shared democratic values of transparency and participation, mandatory environmental information disclosure policy, which requires private or public institutions to disclose factual information about their products, services, or operations in order to pursue a regulatory goal, offers a promising complement/remedy to costly and burdensome traditional environmental regulations (Durant, Fiorino, & O'Leary, 2004; Eisner, 2007; Fiorino, 2006; Kraft, Stephan, & Abel, 2011; Tietenberg, 1998), a powerful mechanism to empower citizens to hold polluters as well as bureaucrats accountable (Stephan, 2002; Tietenberg & Wheeler, 2001), and an effective instrument to improve the quality of environmental protection (Bennear & Olmstead, 2008; Foulon, Lanoie, & Laplante, 2002; Kleindorfer & Orts, 1998). Instead of directly mandating reduction of pollution, mandatory information disclosure policy uses information to empower intermediary actors, such as regulators, communities, the media, non-governmental organizations (NGOs), and the private sector to press polluters to improve their environmental performance. Thus, understanding the responses from these intermediary actors is critical to understand information disclosure policy’s effectiveness, equity, and its implications for environmental governance. In this paper, in the context of the Toxics Release Inventory (TRI), I examine how regulators react to information disclosure policy in their implementation of the Clean Air Act (CAA). The TRI is a mandatory information disclosure program established by the Environmental Protection Agency (EPA) under the provisions of the Emergency Planning and Community Right-to-Know Act (EPCRA) of 1986. Under this program, facilities meeting certain requirements must report their annual toxic emissions to the EPA, which in turn publishes these reports to the public. Extant literature has examined the reactions of many information users to the TRI, such as investors (Hamilton, 1995; Khanna, Quimio, & Bojilova, 1998; Konar & Cohen, 1997), home buyers (Bui & Mayer, 2003; Mastromonaco, 2015; Oberholzer-Gee & Mitsunari, 2006; Sanders, 2014), and the media (Campa, 2018; Hamilton, 1995; Saha & Mohr, 2013). Regulators claim to be important users of TRI information. According to the EPA (U.S. EPA, 1991b, 2013), TRI information has been extensively used by the Office of Enforcement and Compliance Assurance (OECA) among other EPA offices, as well as by state and local governments. For example, the OECA compares TRI information with other environmental information, such as air emissions data from the Air Facility System, to identify facilities that are potentially out of compliance with their permits (U.S. EPA, 2013). Lynn and Kartez (1994), in their survey of 313 state agencies, found that the three most frequently reported uses of TRI information were 1) comparing TRI data to permits, 2) source reduction efforts, and 3) comparing emissions patterns at similar facilities. Kraft et al. (2011) also found that both federal and state regulators used TRI information to assist regulation and enforcement, to understand

Page 4: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

2

facilities’ environmental performance, and to set environmental regulatory priorities. While survey evidence and anecdotes abound, systematic research on how regulators respond to the TRI is lacking. This paper fills the gap by examining the impacts of the TRI on CAA regulatory activities. Understanding regulatory responses to information disclosure policy is important for several reasons. First, responses from regulators are one of the key mechanisms that information disclosure policy works through to achieve its regulatory goals (Fung, Graham, & Weil, 2007; Kraft et al., 2011; Stephan, 2002; Tietenberg, 1998; Weil, Fung, Graham, & Fagotto, 2006). Enforceable traditional regulations are the leading motivator for environmental performance (M. A. Delmas & Toffel, 2008; Doonan, Lanoie, & Laplante, 2005; Khanna & Anton, 2002; May, 2005), and monitoring and enforcement activities under these regulations lead facilities to adjust their environmental behaviors (Deily & Gray, 2006; Gray & Deily, 1996; Gray & Shadbegian, 2005; Hanna & Oliva, 2010; Magat & Viscusi, 1990; Shimshack & Ward, 2005, 2008). Some scholars argue that the main reason that TRI works is because disclosed information has been quickly and strongly embedded in regulatory and administrative process (Fung et al., 2007; Kraft et al., 2011; Weil et al., 2006), which provides the incentives for facilities to improve their environmental performance (Graham, 2002; Graham & Miller, 2001). Thus, understanding regulators’ responses to information disclosure is critical to understand the efficacy of information disclosure policy. Second, answering this question will also shed light on the administrative decision making and bureaucratic behaviors in the implementation of traditional environmental regulations in the presence of information-based approaches. As a mainstay of U.S. environmental policy, despite their historical effectiveness, enforcement programs under traditional environmental regulations are becoming more controversial. The U.S. Office of Management and Budget assessed EPA’s enforcement performance as only “adequate” and urged the EPA for improvements (U.S. OMB, 2005). People in the policy communities regularly advocate the use of voluntary and information-based tools as complements to traditional enforcement (Alberini & Segerson, 2002; Arora & Cason, 1995; Eisner, 2007; Fiorino, 2006; Harrison, 1995; Harrison & Antweiler, 2003; Khanna, 2001; Weil, Graham, & Fung, 2013). The EPA has also identified information disclosure as one of the core features of its next-generation compliance initiative (Giles, 2013). As traditional enforcement programs embrace information-based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations is critical. Third, this question is also extremely policy relevant as the Trump Administration keeps rolling back a wide variety of traditional environmental regulations as well as information disclosure policy. Most notably, in late 2017 the Interior Department rescinded an Obama-era rule that would have required fracking facilities on federal lands to disclose the chemicals contained in fracking fluids, citing redundancy with existing regulations. Understanding the relationship between information disclosure and traditional environmental regulations will contribute new insights to this debate. To understand the relationship between information disclosure and the implementation of traditional environmental regulations, this paper investigates the impacts of the TRI on CAA regulatory activities. I conduct analyses at both state and facility levels. At the state level, I use

Page 5: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

3

difference-in-differences based approaches to exploit exogenous variation from TRI’s size-based reporting criteria and its expansion of industry coverage in 1998. Specifically, TRI disclosure requirements only apply to facilities that have 10 or more full-time employees, are in covered industry sectors, and manage TRI chemicals above threshold levels (need to satisfy all three conditions). In 1998, TRI expanded its industry coverage to include seven additional industry sectors (see section 2 for details). Facilities in these newly covered industries did not disclose in the TRI program before 1998; since 1998, they either have started to disclose in the TRI (if they have met the size-based criteria) or have continued to not disclose in the TRI (if they have not met the size-based criteria). The treatment group consists of facilities in the newly covered industries that have started to report to the TRI from 1998 due to the rule change and the control group consists of facilities in the newly covered industries that have continued to not report to the TRI because of not meeting the other two criteria. I use three quasi-experimental estimation strategies in the state level analysis. First, I begin with a basic difference-in-differences analysis by comparing the differences of the regulatory activities conducted on facilities in the treatment and control groups before and after the rule change to identify the impact of TRI disclosure on regulatory activities. Second, I expand the basic difference-in-differences design to an event study approach to analyze the impact of the rule change in 1998 on the treatment group over time. This approach allows me to explicitly examine the validity of the control group by examining differences in pretrends between the treatment and control groups leading up to the rule change. Third, I employ a triple difference-in-differences approach by introducing two more groups of facilities to control for the potential different trends of regulatory activities for the treatment and control groups that are due to different sizes of facilities in the treatment and control groups. The two added groups are 1) facilities in previously covered industries (SIC codes 20 -39) that report to the TRI and 2) facilities in previously covered industries (SIC codes 20 -39) that do not report to the TRI. The rule change in 1998 have not affected facilities in these two groups; they either reported to the TRI or did not report to the TRI throughout the study period. The size-based reporting criteria, however, apply to these two groups. Same as facilities in the treatment and control groups, facilities in these two groups also differ in their sizes. So I use the change in the differences between these two groups before and after the rule change to track the change in the differences between the treatment and control groups before and after the rule change that is due to the treatment and control groups consisting of facilities of different sizes. At facility level, I focus on facilities that disclose in the TRI and use facility fixed effects model to investigate how regulatory activities react to the marginal change of the information disclosed in the TRI (the amounts of toxic emissions). To summarize the key results, in the state level analysis I find that regulators significantly reduce regulatory activities on the group of facilities that report to the TRI. Regulators decrease inspection activities and enforcement activities on reporting facilities by 27% and 30% respectively. Results from the facility level analysis, however, suggest that regulatory activities do not react to the year-over-year change of information (emissions) disclosed in the TRI. Factors such as size of informational shock, quality of disclosed information, and reaction

Page 6: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

4

pattern of regulators and other information users, could potentially explain the different results. I discuss them in detail in the discussion section. Findings of this study suggest that there are strong intercorrelations between information disclosure policy and traditional environmental regulations. Information disclosure policy can be an important supplement to the implementation of traditional environmental regulations, and regulatory response to information disclosure program also can serve as a mechanism for information disclosure policy to achieve its regulatory goals. The findings also highlight important issues such as accuracy and novelty of disclosed information in the design of information disclosure policy. The remainder of the paper is organized as follows. In section 2, I provide the background information on the TRI and regulatory activities in the CAA. In section 3, I lay out several conceptual accounts to explain potential mechanisms through which information disclosure could affect the implementation of traditional environmental regulations. In section 4, I describe data, and in section 5, I explain my empirical strategy. Section 6 presents the results, and section 7 presents robustness and sensitivity analysis. In section 8, I discuss the findings, and in section 9, I conclude. 2 Policy Background 2.1 Toxics Release Inventory The TRI was established by the Emergency Planning and Community Right-to-Know Act (EPCRA) of 1986. Following the 1984 Union Carbide catastrophe in Bhopal (India), Congress passed, and President Reagan signed into law the EPCRA in response to growing public concerns about local preparedness for chemical emergencies and the availability of information on hazardous substances. Initially, the primary goal of the EPCRA is to support and promote community emergency planning (42 U.S.C. § 11001 et seq.). It requires communities to establish procedures to effectively prevent, respond to, and contain events of accidental release of toxic chemicals. This requires local agencies (firefighters, policy, and hospitals etc.) to have information on the types, the quantities, and the locations of toxic chemicals that are present in their communities so they can develop and coordinate capacities and resources to handle emergencies. Second, the TRI is also an attempt to reduce toxic releases without setting binding pollution-reduction goals (42 U.S.C. §13101 et seq.). By forcing firms to disclose information on their toxic releases and making that information available to the public, it hopes to leverage the power of individual choice, civil society, and participatory democracy to press firms to reduce toxic releases. Under Section 313 of the EPCRA, a facility must submit a form that details its release of a toxic chemical to different media (air, water, land) of the environment annually for each of the TRI-listed toxic chemicals if it satisfies the following three conditions: (1) it is in the manufacturing industry sectors (SIC codes 20 - 39); (2) it employs 10 or more full-time equivalent employees; (3) it manufactures, processes, or uses the chemical in quantities above threshold levels in a

Page 7: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

5

given year. A facility is required to submit each TRI form to both the EPA and the state in which the facility is located. The information submitted by facilities is compiled in the TRI and made available to the public by the EPA. The TRI program collects, processes, and analyzes TRI data on an annual cycle. Every year, facilities submit TRI forms for the previous calendar year to the EPA by July 1. A preliminary dataset is made available to the public shortly after the deadline. From July to October, the EPA conducts additional data quality checks, publishes the complete dataset, and begins analyzing the data. In January of the next year, the EPA publishes the TRI National Analysis.1 In addition to activities that occur at specific times of the year, the TRI program continually conducts data quality checks and provides analytical support for enforcement efforts led by EPA’s Office of Enforcement and Compliance Assurance (OECA). Since the inception of the TRI, reported toxic emissions have fallen dramatically. EPA’s analysis (U.S. EPA, 2001) shows that total releases in the United States decreased by 45.5% between 1988 and 1999. Emissions further dropped by approximately 30% between 2001 and 2010 (U.S. EPA, 2012). For the calendar year of 2016, more than 21,000 facilities submitted TRI data to the EPA, reporting 27.80 billion pounds of TRI-listed chemicals as production-related waste. Of this total, 87% was recycled, combusted for energy recovery, or treated. Only 13% (3.88 billion pounds) was disposed of or otherwise released into the environment. Total releases for 2016 were greater than the quantities reported for 2015, but 21% less than releases reported for 2006. In 2016, among the over 650 chemicals covered by the TRI Program, release quantities of 8 chemicals, including lead, zinc, nitrate compounds, arsenic, and ammonia, comprised 73% of the total releases. In terms of industry sector, the metal mining sector accounted for 44% of releases (1.52 billion pounds), followed by chemicals industry (14%) and electric utilities (10%). The TRI program has significant overlap with other environmental programs in terms of covered pollutants and facilities. For the calendar year 2016, among the more than 21,000 facilities that have reported to TRI, around 12,700 of them are also under the jurisdiction of the CAA, which represent 60% of TRI reporting facilities. These 12,700 facilities represent about 7% of the 173,000 facilities regulated by the CAA. In addition, among the 502 chemicals reported to the TRI, 185 (37%) of them are also regulated by the CAA. These CAA-regulated pollutants represent 59% of the reported air emissions and 51% of the reported total emissions. The mission of the TRI program is to provide information to the public. The EPA has made the data available through disk storage and on the Internet from the EPA and other environmental sources, such as the Right-To-Know Network (RTK-Net). In addition, the EPA has made some efforts in community outreach and education to facilitate access to the TRI information (U.S. EPA, n.d.-b), including web-based tools such as TRI Explorer, “myRTK” mobile application, TRI.NET, and Pollution Prevention (P2) search tool. Despite these efforts, the general public

1 This timeline reflects the current schedule of the TRI program. Before 2008, the TRI dataset was not available until the publication of the TRI National Analysis, and the TRI National Analysis was published about 10 months later after the submission deadline.

Page 8: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

6

seems to know little about the TRI program. In a survey of about 1,300 respondents in two counties, Atlas (2007b) found that only a very small percentage of the respondents knew about the TRI program and TRI facilities in their areas. In a telephone survey in three counties with significant amounts of toxic emissions, the U.S. General Accounting Office (1991) found that while about 50% of the respondents had read or heard about reports about local toxic chemical emissions, only 15-20% knew that the information was publicly available. While the raw data release is unlikely to directly reach households and individuals, TRI has drawn a significant amount of attention from the popular media (Bui, 2005; Campa, 2018; Hamilton, 1995; Saha & Mohr, 2013; Sanders, 2014). Bui (2005) found that more than 430 news reports on the TRI appeared in major newspapers between 1988 and 1995. Through a survey of news stories from LexisNexis, Sanders (2014) found that articles on the TRI occured with high frequency every year around the time when the EPA releases new data, with the largest spike occurring in May 2000 when TRI information on new, highly polluting sectors was available. Hamilton (1995) found that the higher the releases that were disclosed in the TRI, the more likely the media would cover facilities’ toxic releases. Saha and Mohr (2013) reached similar conclusions and they also found that media coverage of facilities’ toxic releases had no correlation with race or income of facilities’ neighborhood. Campa (2018) found that TRI facilities’ proximity to newspaper’s headquarters increased the likelihood of newspaper’s coverage of facilities’ toxic emissions. The literature also suggests that media attention has provided incentives for facilities to improve. With a difference-in-differences approach, Campa (2018) found that when newspapers covered the toxic emissions of consumer goods producers, these covered facilities reduced their emissions by 29% compared with those plants that were not covered, whereas there were no differential trends in the years leading up to the coverage. Saha and Mohr (2013) found a 40% decrease compared with pre-treatment years in releases after a facility was covered by the media with a similar difference-in-differences approach. Scorse (2000) exploited exogenous changes (inclusion of new industries in the TRI program) to facilities’ pollution rankings within states and found that removal from the list of “Top 10” led facilities to reduce their emissions by hundreds of thousands pounds less than they would otherwise if they had remained on the list. Media attention can also serve as a potential conduit to pass TRI information to households. Research that studies households’ reaction to the TRI focuses on the capitalization of TRI information in housing value (Bui & Mayer, 2003; Mastromonaco, 2015; Oberholzer-Gee & Mitsunari, 2006; Sanders, 2014) and the results are less conclusive. With a first-difference approach, Bui and Mayer (2003) examined the relationship between changes in disclosed TRI emissions and changes in housing prices in 231 zip codes in Massachusetts from 1987 to 1992 and found that TRI information had no statistical and substantial impacts on housing prices. Oberholzer-Gee and Mitsunari (2006) found inconsistent results regarding the impacts of the first disclosure of TRI data in 1989 on home prices across five Philadelphia counties; they found TRI information only affected homes with a distance of between a quarter and a half mile from TRI sites while they found no effects on homes closer to TRI sites. Mastromonaco (2015) exploited a strengthening of the reporting requirements for lead in 2001 as exogenous variation to study the

Page 9: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

7

changes in housing prices when nearby existing facilities started to report in the TRI. With a difference-in-differences design, he found that listing an existing firm in the TRI lowered housing prices by 11% within approximately 1 mile. Sanders (2014) used the inclusion of new industries in the TRI in 1998 as an exogenous shock to examine a similar question. With a difference-in-differences approach, he found that median housing prices dropped 2-3 percent for zip codes with facilities that started to report in the TRI. Using a change in reporting thresholds for persistent bioaccumulative toxins (PBTs) in 1999, Currie (2011) found a large reduction of white college-educated mothers living in areas close to PBTs emitting facilities when these facilities started to disclose their emissions due to the change. Another potential important user of TRI information are investors. Studies that focus on investor reactions to TRI information mostly use an event study approach to examine the relationship between TRI disclosure and abnormal returns in stock markets (Hamilton, 1995; Khanna et al., 1998; Konar & Cohen, 1997). Hamilton (1995) found that firms that reported TRI pollution figures experienced negative, statistically significant abnormal stock market returns upon the first release of the information. However, he found no statistically significant relationship between the amounts of disclosed emissions and the dollar values of abnormal returns. Konar and Cohen (1997) found similar results that the first release of TRI information led firms to have statistically significant negative abnormal returns. Khanna et al. (1998) found that repeated disclosure of TRI information, by allowing investors to benchmark the performance of firms, also led firms to have abnormal negative returns during the one-day period following each disclosure from 1990 to 1994. Studies also show that negative market returns provide incentives for firms to improve environmental performance. Both Konar and Cohen (1997) and Khanna et al. (1998) showed that firms with abnormal stock market returns subsequently reduced toxic emissions more than their industry peers. Information from the TRI program can also help government agencies to develop environmental policies, track environmental performance, and establish regulatory priorities (Fung et al., 2007; Kraft et al., 2011; Lynn & Kartez, 1994; Stephan, 2002; U.S. EPA, 1991b, 2013; Weil et al., 2006). Through a cross-sectional study, Decker (2005; 2009) found that in two of four frequently inspected manufacturing industries, state regulators undertook fewer inspections at plants that reported lower per unit output chemical releases in the TRI. Patten (1998) found that states with greater pollution problems, based on the 1988 TRI disclosure, increased more resources to environmental and natural resource programs in 1989 and 1990. In these studies, it is reasonable to use TRI emissions as a proxy for environmental performance. However, if the purpose is to evaluate regulators’ response to the TRI information itself, cross-sectional studies will probably suffer from omitted variable bias if unobserved facility or state characteristics is correlated with TRI emissions information. A TRI program rule change that came into effect in 1998, along with its size-based reporting criteria, provides good exogenous variation to explore regulators’ reactions to information disclosure policy. Specifically, the rule change added seven new industry sectors to the coverage of the TRI program starting in the 1998 reporting year. The seven industries are metal and coal

Page 10: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

8

mining (SIC codes 10 and 12), electric utilities that combust coal and/or oil (SIC codes 4911, 4931, and 4939), commercial hazardous waste treatment (SIC code 4953), solvent recovery (SIC 7389), petroleum bulk terminals (SIC code 5171), and wholesale chemical distribution (SIC code 5169). Prior to 1998, facilities in the seven industries did not need to report to the TRI. From 1998 on, a portion of facilities in these seven industries, those that employ more than ten full-time employees and meet the established thresholds for listed chemicals, have started to report to the TRI, and the rest of the facilities in the seven industries have continued to not report to the TRI. Figure 1 shows the locations of affected facilities (those that have started to report to the TRI) in the seven industries due to the rule change. Most states have facilities that have been affected by this rule change, except for Vermont. However, there is some variation in the number of affected facilities, with Ohio having the highest number of affected facilities (72) and the District of Columbia having the fewest number of affected facilities (1).

Figure 1. Locations of Affected Facilities by the 1998 TRI Rule Change

Source: Author’s Analysis Based on the TRI Database 2.2 Regulatory Activities in the Clean Air Act In this study, I evaluate the impacts of the TRI disclosure program on regulatory activities, specifically monitoring activities and enforcement activities, in the Clean Air Act (CAA). Enacted in 1970 and amended in 1977 and 1990, the CAA is the primary law that regulates air pollution in the United States. The CAA is jointly implemented by the federal EPA and states. While the EPA establishes and revises various air quality, emission, and technology standards, the responsibility of ensuring facilities’ compliance with these standards primarily falls on states. States are required to develop enforceable state implementation plans (SIPs), and the EPA reviews state plans to ensure that they comply with the requirements of the CAA. The EPA also issues federal implementation plans for states that fail to adopt and implement an adequate SIP. States assumes the primary responsibility to carry out the day-to-day activities, such as

Page 11: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

9

permitting, monitoring, and enforcement, following approved SIPs, while the EPA assists state efforts by providing technical and policy guidance and conducting a small amount of the monitoring and enforcement activities directly. During my analysis period from 1995 to 2003, the EPA carried out only 2.79% of the monitoring activities and 9.59% of the enforcement activities. Regulators use three types of monitoring activities (also referred to as inspections or evaluations) to determine the compliance status of a facility: Full Compliance Evaluations (FCEs); Partial Compliance Evaluations (PCEs); and Investigations (U.S. EPA, 2016). These monitoring activities can be conducted both on and off site and can include reviews of monitoring data (e.g., continuous emissions monitoring system (CEM) and continuous parameter monitoring reports, malfunction reports, excess emission reports, semiannual monitoring and periodic monitoring reports); reviews of permit, facility records, and operating logs; visual inspections of facility and equipment; and stack tests; among others (U.S. EPA, 2016). FCEs and PCEs differ in the scope of the monitoring activities. Investigation distinguishes from the other two types of activities in that it involves a more in-depth assessment of a particular issue usually based on issues discovered in FCEs/PCEs, or as a result of targeted industry, regulatory or statutory initiatives (U.S. EPA, 2016). The EPA sets monitoring activities frequency targets based on types of facilities (U.S. EPA, 2016). However, these targets are not binding requirements. The EPA monitoring guidelines allow states to take into consideration factors such as compliance history, location of facility, potential environmental impact, operational practices, use of control equipment, resources in the state’s compliance monitoring program, and participation in national enforcement initiatives to make monitoring decisions. Monitoring activities can also be triggered by citizen complaints, anonymous employee complaints and tips, or facility characteristics and behaviors that correlate with frequent violations or significant damages (U.S. EPA, 2016). When regulators find violations, they have authority to take enforcement actions. Enforcement options include informal administrative sanctions such as warning letters, telephone calls, and notices of violation, formal administrative sanctions such as administrative orders of compliance (with or without penalties), and civil judiciary actions filed through the Department of Justice or States’ Attorneys General (Congressional Research Service, 2014; U.S. EPA, 1991a). In rare cases, criminal judiciary actions are taken against the most serious violations, those that are willfully or knowingly committed. EPA enforcement guidelines generally require regulators to consider the following factors to determine the frequency and severity of the sanctions: actual or possible harm from the violation, violator’s economic benefit from the violation, facility’s compliance and enforcement history, violator’s ability to pay, violator’s intent, participation in supplemental environmental project, and litigation risk (U.S. EPA, 1991a). Figure 2 shows the number of monitoring/inspection activities and enforcement activities under the CAA over time.

Page 12: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

10

Panel A. Monitoring/Inspection Activities Panel B. Enforcement Activities

Figure 2. Trends of CAA Regulatory Activities

Source: Analysis of the Air Facility System Database 3 Conceptual Framework In this paper, I study the impacts of information disclosure on regulatory activities. In this section, I provide some conceptual accounts of why information disclosure can potentially affect regulatory activities. Regulators operate at the conjunction of multiple federal, state, and local institutions, and they carefully balance conflicting political demands and statutory requirements in their implementation of policies and regulations (Scholz & Wei, 1986). They develop regulatory routines that reflect their organizational structure and culture (Deily & Gray, 2006; Konisky & Reenock, 2013), that respond to task differences (Dion, Lanoie, & Laplante, 1998; Gray & Deily, 1996; Gray & Shadbegian, 2004; Kleit, Pierce, & Carter Hill, 1998; Oljaca, Keeler, & Dorfman, 1998; Scholz & Wei, 1986) and facility characteristics (Deily & Gray, 1991; Gray & Deily, 1996; Helland, 1998a), and that react to various political institutions (Atlas, 2007a; Innes & Mitra, 2015; Konisky, 2007; Scholz, Twombly, & Headrick, 1991; Scholz & Wei, 1986; Wood, 1992) and local community characteristics (Konisky, 2009; Konisky & Reenock, 2013; Konisky & Reenock, 2018; Scholz et al., 1991; Scholz & Wang, 2006). Information disclosure programs can potentially change regulatory activities by changing the regulatory tasks that regulators face and the equilibrium of multiple political influences. The classic economic models of enforcement generally assume a social welfare-maximizing enforcement system (Posner, 1974), in which regulators optimize enforcement strategy to maximize compliance. The most influential economic model that explains environmental monitoring and enforcement behaviors is a targeting model that is proposed by Harrington (1988) and refined by Harford and Harrington (1991) and Friesen (2003). In this targeting model, regulators achieve a high level of deterrence with relatively low levels of regulatory activities by adjusting the intensities of the inspection and sanction based on facilities’ historical

Page 13: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

11

environmental performance. Regulators implement a lax compliance regime on “good” facilities and a stringent regime on “bad” facilities. EPA’s compliance monitoring plans and enforcement response policies principally follow the targeting practice (Silverman, 1990), and numerous empirical studies have identified the pattern of targeting behaviors (Dion et al., 1998; Helland, 1998a). The TRI program, by disclosing facilities’ toxic emissions, can potentially change the probabilities that the disclosing facilities be in regulators’ “good” or “bad” groups (the regulatory tasks that regulators face). If so, and if regulators follow the targeting practice, then the TRI program will affect regulatory activities. TRI can affect the regulatory tasks faced by regulators in three ways. First, the TRI can provide regulators with information on facilities’ environmental performance that would otherwise need to be obtained through regulatory activities. If regulators are not sure if a facility is in compliance, they would use regulatory activities, especially monitoring/inspection activities, to collect more information. If TRI information can help them more efficiently target facilities in violation of the CAA, regulatory activities, especially monitoring/inspection activities, would decrease. In this sense, the TRI, by helping regulators sort facilities into the “good” or “bad” groups, is a substitute for some regulatory activities. Second, information disclosed in the TRI may be out of sync with regulators’ prior perception/expectation on facilities’ environmental performance (Stephan, 2002). The “shock” leads regulators to update their understanding of firms’ environmental performance and to change the regulatory regimes they adopt for these facilities. The impact through this mechanism, however, depends on whether the TRI program provides new information to regulators, whether the regulators trust the information, and how the new information changes perception of regulators. If the information provided by the disclosure program is not new or is irrelevant or untrustworthy to the regulators, we would not expect a change in the patterns of regulatory behaviors. If the information does have an impact, the direction of the impact will depend on how the new information alters prior expectations. Third, the TRI program also could potentially lead facilities to change their actual environmental performance, which will result in facilities moving between the “good” and “bad” groups. A large and growing theoretical literature suggests disclosure can impact performance (Loewenstein, Sunstein, & Golman, 2014; Weil et al., 2006). While the empirical evidence is mixed, there also exists a significant number of studies that illustrate the efficacy of information disclosure to improve performance (Bennear & Olmstead, 2008; Chatterji & Toffel, 2010; M. Delmas, Montes-Sancho, & Shimshack, 2010; Foulon et al., 2002; Jin & Leslie, 2003). In the TRI program, the significant reduction in reported toxic emissions over time suggests that facilities might have improved their environmental performance (Abel, Stephan, & Kraft, 2007; Doshi, Dowell, & Toffel, 2013; Hamilton, 2005; Kalnins & Dowell, 2015; Kraft, Abel, & Stephan, 2004; Kraft et al., 2011; Shapiro, 2005). While the reduction cannot be casually attributed to the disclosure program, some survey studies (Baram & Dillon, 1992; Santos, Covello, & McCallum, 1996) show that TRI facilities have adopted a wide variety of practices to reduce pollution since the passage of the program. In addition, researchers have found that states

Page 14: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

12

with programs that promulgate TRI information experience larger reductions in toxic emissions (Bae, Wilcoxen, & Popp, 2010; Grant & Downey, 1995), which suggests that TRI information could potentially improve performance. Moreover, the literature also has provided explicit evidence on how the TRI program has leveraged reactions from information users to press facilities to improve their environmental performance. Specifically, TRI disclosure has led to significant reactions from the media (Bui & Mayer, 2003; Campa, 2018; Hamilton, 1995; Saha & Mohr, 2013; Sanders, 2014) and the stock markets (Hamilton, 1995; Khanna et al., 1998; Konar & Cohen, 1997), and media attention and abnormal negative returns in stock markets have subsequently caused facilities to improve their environmental performance (Campa, 2018; Khanna et al., 1998; Konar & Cohen, 1997; Saha & Mohr, 2013; Scorse, 2000). If the TRI disclosure does lead facilities to have better environmental performance, we would expect regulators to reduce their monitoring and enforcement activities on these facilities. While regulators’ behaviors regularly conform to the prediction of the economic model, plenty of theoretical and empirical evidence also suggests regulators have political motivations when executing regulatory activities. Regulators can be captured by industrial groups (Peltzman, 1976; Stigler, 1971), attempt to minimize conflict and attention (Hilton, 1972; Joskow, 1974; Leaver, 2009), and be budget maximizers (Niskanen, 1971). Capture theory predicts that regulatory activities respond to industry groups. Conflict minimization theory predicts that regulators respond to the salience of issues. Budget maximizing theory predicts that regulatory activities will reflect the preference of the Congress and the executive branch. Empirical evidence shows that regulatory behaviors exemplify some characteristics of the predictions of all the theories. Regulators direct fewer enforcement actions towards facilities that are major local employers (Gray & Deily, 1996), and more corrupt states implement laxer environmental oversight after assuming enforcement primacy (Grooms, 2015). Regulatory activities also reflect preferences and party affiliations of Congressional representatives (Helland, 1998b; Innes & Mitra, 2015) and ideology and party affiliations of local elected officials (Atlas, 2007a; Scholz et al., 1991; Scholz & Wei, 1986). In addition, regulatory activities vary based on characteristics of local communities. Specifically, local economic and labor market conditions (Dion et al., 1998; Helland, 1998a; Scholz & Wei, 1986) and characteristics that are associated with political activism, such as racial composition, education, income, voter turnout, and environmental group membership (Helland, 1998b; Konisky, 2009; Konisky & Reenock, 2013; Konisky & Reenock, 2018; Scholz et al., 1991; Scholz & Wang, 2006), exert important influence on regulatory activities. If regulators respond to these political factors, information disclosure could potentially affect regulatory activities by changing the balances of these political forces. Some evidence shows that the TRI program has increased the salience of the issue of toxic emissions in reporting facilities by drawing media attention (Bui & Mayer, 2003; Campa, 2018; Hamilton, 1995; Saha & Mohr, 2013; Sanders, 2014). Some researchers also find households show concern about TRI information, as housing prices (Mastromonaco, 2015; Sanders, 2014) and demographic characteristics (Currie, 2011) react to TRI disclosure. If the increased salience and public

Page 15: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

13

concern translate into civic and political actions, regulators may face pressure to increase regulatory activities on TRI facilities. The impacts of community activism, however, may not always lead to more regulatory activities on TRI facilities. Regulators may view private scrutiny on TRI facilities as a substitute for their regulatory activities. Economic models predict that the impact of private enforcement on regulatory activities can go either way (Heyes & Rickman, 1999; Langpap, 2007). Empirically, Naysnerski and Tietenberg (1992) observed a negative correlation between public and private environmental enforcement activities, and Langpap and Shimshack (2010) found that private enforcement activities increased monitoring activities yet decreased sanctions. If regulators perceive facilities that face community activism to be less likely to violate regulations, scare enforcement resources could be deployed elsewhere. This scenario conforms with the economic theory where regulators try to maximize overall compliance or environmental quality. 4 Data The two main variables used in my analyses are regulatory activities and TRI disclosure. I derive the dependent variable—CAA regulatory activities—from EPA’s Air Facility System database. This database compiles CAA regulatory activities on stationary sources conducted by the federal EPA as well as state and local agencies and includes detailed information on these activities, including dates, types, penalties if any, facility identifiers, and facility location etc. I compile regulatory activities into two separate measures: inspection/monitoring activities and enforcement activities. The key independent variables—whether a facility reports to the TRI and the information (amounts of releases) disclosed in the TRI—are from EPA’s TRI database. This database compiles detailed information about reporting facilities’ management of the listed TRI chemicals, such as the amounts that they release to the environment via different media and the amounts they manage through recycling, energy recovery and treatment on an annual basis. My analyses explore how disclosure of environmental information affects regulatory activities. During the study period, TRI information for a certain calendar year was due to the EPA by July 1 of the next calendar year, and the information was available to the public roughly 10 months later after the submission due date, which makes public access to TRI information about 16 months after the end of the relevant calendar year. Since the TRI information about a certain calendar year does not “treat” regulatory activities in that calendar year, I need to assign appropriate information disclosure schedule for regulatory activities. In other words, I need to match the regulatory activities with a specific disclosure. I assign TRI disclosure schedule to regulatory activities as follows. The TRI disclosure for a certain calendar year will “treat” the period after the public-access date of this TRI disclosure and before the public-access date of the TRI information for the next calendar year. This assignment assumes that regulators react to the latest publicly available TRI information. It is a plausible assumption as information tends to have the largest effect when it is the newest.

Page 16: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

14

Researchers have found that other users of TRI information, specifically the media (Saha & Mohr, 2013; Sanders, 2014) and investors (Hamilton, 1995; Khanna et al., 1998; Konar & Cohen, 1997), respond right after the publication of new TRI information. In addition, studies (Bui & Mayer, 2003; Mastromonaco, 2015; Oberholzer-Gee & Mitsunari, 2006; Sanders, 2014) on how housing prices respond to TRI information use a similar assignment rule. One problem is that regulators could potentially access the TRI information before the public-access date as facilities submit TRI information to the EPA by an earlier date. This, however, is unlikely to cause major issues. State and local agencies and EPA’s Enforcement and Compliance Office carry out CAA regulatory activities. Facilities, however, submit TRI forms to EPA’s TRI office. It takes time for the TRI office to clean and organize the raw data to make them usable. By the time this is achieved, the TRI office would also make available the information to the public as the TRI office has attempted to make earlier public availability of TRI information possible (U.S. EPA, n.d.-a). My main results will be based on the public-access assignment rule, unless otherwise specified. As a sensitivity test, I also conduct analyses that follow a submission-deadline assignment rule. Under this assignment rule, the TRI information for a certain calendar year will “treat” the period after the submission deadline (July 1 of the next calendar year) for this TRI disclosure and before the submission deadline for the TRI information of the next calendar year. Table 1 illustrates the definition of reporting years based on the two approaches. Table 1. Illustration of Reporting Year Definition

Reporting Year Calendar Year

(Based on Public Access) Calendar Year

(Based on Submission Deadline) 1993 03/15/1995-06/15/1996 07/01/1994-07/01/1995 1994 06/15/1996-05/13/1997 07/01/1995-07/01/1996 1995 05/13/1997-12/15/1998 07/01/1996-07/01/1997 1996 12/15/1998-05/13/1999 07/01/1997-07/01/1998 1997 05/13/1999-05/11/2000 07/01/1998-07/01/1999 1998 05/11/2000-04/12/2001 07/01/1999-07/01/2000 1999 04/12/2001-05/23/2002 07/01/2000-07/01/2001 2000 05/23/2002-06/20/2003 07/01/2001-07/01/2002 2001 06/20/2003-06/23/2004 07/01/2002-07/01/2003 2002 06/23/2004-05/11/2005 07/01/2003-07/01/2004 2003 05/11/2005-04/12/2006 07/01/2004-07/01/2005 2004 04/12/2006-03/22/2007 07/01/2005-07/01/2006 2005 03/22/2007-02/21/2008 07/01/2006-07/01/2007

Source: public access dates and submission deadlines provided by the EPA 4.1 State Level Analysis Data For the state level analysis, I use data covering the period of reporting year from 1995 to 2003 since my analysis exploits the expansion of TRI’s industry coverage starting for reporting year 1998. I construct the dependent variable—regulatory activities—for the state level analysis as

Page 17: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

15

follows. First, from the TRI database, I obtain a list of CAA facilities in the newly covered seven industries that started to report to the TRI for reporting year 1998 due to the rule change; these facilities comprise the treatment group. Second, for each of the regulatory activities in EPA’s CAA monitoring and enforcement activities database, I mark whether it is conducted on a facility in the treatment group. If a regulatory activity was conducted on a facility in the newly covered industries but not in the treatment group, I mark it as an action on the control group. Third, I aggregate the number of regulatory activities by state, reporting year, and treatment/control status. The above variable—number of regulatory activities for the treatment/control groups at state-reporting year level—is the dependent variable for the state level difference-in-differences analysis. Following a similar approach, I also calculate the number of regulatory activities for another two groups: facilities in the previously covered industries (SIC codes 20-39) that report to the TRI and facilities in the previously covered industries (SIC codes 20-39) that do not report to the TRI. The rule change did not affect facilities in these two groups; they either reported to the TRI or did not report to the TRI throughout the analysis period. I organize regulatory activities into two categories: inspection/monitoring activities and enforcement activities, and I apply the above procedures for each category of regulatory activities respectively. Figure 3 presents the trends of regulatory activities conducted on these different groups of facilities over time. Panel A: Inspection Panel B: Enforcement

Figure 3. Trends of Regulatory Activity for Difference Groups

Source: Analysis of the Air Facility System Database and the TRI Database The state level analysis also includes several state-reporting year level control variables that are important determinants of regulatory behaviors. First, I take into consideration the preferences of state-elected officials. Researchers (Atlas, 2007a; Helland, 1998a; Innes & Mitra, 2015) have found that partisan control of governor’s office and state legislature have important influence on state environmental regulatory activities. Following Konisky (2007), I include three variables to measure the preferences of state-elected officials. Specifically, I include governor’s party affiliation (1= Republican, 0 = Democratic and Independent), the average percentages of Republicans in state Senate and House, and state government ideology. The data on the two

Page 18: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

16

partisan control variables are from Klarner’s (2013) “State Partisan Balance” dataset and the state ideology is from Berry, Fording, Ringquist, Hanson, and Klarner (2010). Second, I include controls for state economic and fiscal conditions. Some studies (Dion et al., 1998; Helland, 1998c; Scholz & Wei, 1986) find that regulatory behaviors respond to economic and labor market conditions. I use a state’s unemployment rate to measure its economic conditions and the data are from the U.S. Bureau of Labor Statistics. In addition, I use a “fiscal health” indicator to measure a state’s fiscal conditions. This indicator measures surplus (negative if deficit) as a share of total state expenditure. The state revenue and expenditure are from the U.S. Census Bureau’s Annual Survey of State Government Finances. Third, I include a few important state demographics and socioeconomic characteristics, specifically personal income and population. I compile these two variables from U.S. Bureau of Economic Analysis’ regional dataset. Table 2 presents summary statistics for the state level analysis. Table 2. Summary Statistics for the State Level Analysis

Variables N Mean Std. Dev. Min Max Inspection Actions (Treatment Group) 441 40 74 0 458 Enforcement Actions (Treatment Group) 441 6 12 0 113 Inspection Actions (Control Group) 441 82 122 0 717 Enforcement Actions (Control Group) 441 12 24 0 314 Inspection Actions (SIC 20-39, Reporting) 441 275 421 1 2954 Enforcement Actions (SIC 20-39, Reporting) 441 48 63 0 493 Inspection Actions (SIC 20-39, Non-reporting) 441 297 468 1 3430 Enforcement Actions (SIC 20-39, Non-reporting) 441 42 61 0 442 Unemployment Rate (%) 441 4.69 1.10 2.53 7.78 Fiscal Health 441 0.09 0.12 -0.25 0.90 Republican Governor 441 0.59 0.49 0 1 Republican Share of State Legislature 432 0.49 0.15 0.13 0.89 State Personal Income ($1,000) 441 30.40 5.50 19.69 51.43 State Population (Million) 441 5.81 6.27 0.49 35.90 State Government Ideology 441 46.37 12.88 21.31 71.67

4.2 Facility Level Analysis Data Facility level analysis focuses on facilities that report to the TRI and examines how regulatory activities react to information disclosed in the TRI. Specifically, I attempt to answer how would regulatory activities change when facilities disclose increase/decrease of toxic emissions in the TRI. For the facility level analysis, I use data covering the period of reporting year from 1998 to 2003. To construct the dependent variable, I aggregate regulatory activities to facility-reporting year level separately for inspection and enforcement activities. To construct the key independent variables—TRI information, I aggregate the amount of disclosed emissions by media (air, water, land, total) across all chemicals to facility-reporting year level. I primarily rely on fixed effects in the facility level analysis to control for potential confounders. Specifically, I include state-by-year fixed effects and facility fixed effects. State-by-year fixed

Page 19: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

17

effects control all the time-variant and time-invariant factors at the state and national level that influence regulators’ behaviors. Facility fixed effects control for all the time-invariant facility characteristics that influence regulators’ behaviors, such as industry sector, location, and most importantly the invariant/pre-existing part of regulators’ expectations. Besides fixed effects, I include population, personal income, and unemployment rate at county level to measure local economic conditions as they may influence the regulatory activities towards specific facilities (Dion et al., 1998). I compile county population and personal income from the Bureau of Economic Analysis and county unemployment rate from Bureau of Labor Statistic’s Local Area Unemployment Statistics Program. Another important factor that could affect regulatory activities on a facility is its environmental performance. Many scholars (Deily & Gray, 1991; Dion et al., 1998; Gray & Deily, 1996; Helland, 1998a; Kleit et al., 1998; Oljaca et al., 1998) have found that regulators target facilities with bad environmental performance. I include facilities’ High Priority Violator (HPV) status to measure its environmental performance. Regulators designate a facility as an HPV when the facility fails to meet core CAA obligations, usually emissions performance standards (Konisky & Reenock, 2013; U.S. EPA, 2014). I compile facilities’ HPV status from EPA’s Enforcement and Compliance History Online (ECHO) database. A facility’s HPV indicator takes one for a reporting year if the facility has been an HPV for any amount of time during the reporting year. Table 3 presents summary statistics for the facility level analysis. Table 3. Descriptive Statistics for the Facility Level Analysis

Variables N Mean Std. Dev. Min Max Air Releases (lb.) 66816 139,122 760,401 0 57,700,000 Land Releases (lb.) 66,816 224,416 6,360,275 0 487,000,000 Water Releases (lb.) 66,816 19,931 350,461 0 32,000,000 Total Releases (lb.) 66,816 404,051 6,461,065 0 487,000,000 Inspection Actions 66,816 1.30 3.88 0 212 Enforcement Actions 66,816 0.22 1.03 0 56 High Priority Violator 66,816 0.11 0.31 0 1 County Unemployment (%) 66,762 5.35 1.47 1.46 17.01 County Personal Income ($1,000) 66,807 30.97 7.59 14.61 101.38 County Population 66,807 627,519 1,335,427 1,469 9,790,460

5 Empirical Methods 5.1 State Level Analysis Method The research design for the state level analysis primarily employs difference-in-differences based approaches that exploit TRI’s size-based reporting requirements and a rule change that added seven new industry sectors to the coverage of the TRI program in 1998. The rule applied TRI reporting requirements to metal and coal mining facilities, electric power generators, commercial hazardous waste treatment operations, solvent recovery facilities, petroleum bulk terminals and wholesale chemical distributors. Before the rule change, facilities in these industries did not

Page 20: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

18

disclose their toxic emissions to the TRI. After the rule change, some facilities in these industries, those that employ more than ten full-time employees and meet the established thresholds for listed chemicals, have started to report to the TRI, and the rest of the facilities in the seven newly covered industries have continued to not report to the TRI. The basic model uses facilities in the newly covered industries that were not affected by the rule change (control group) as a comparison for facilities in the newly covered industries that started to report to the TRI from 1998 (treatment group). The control and treatment groups are in the same industry sectors, so that any industry specific technological changes, market shocks, and regulatory initiatives will affect both groups similarly. The basic model compares the differences of regulatory activities between the treatment group and the control group, before and after the rule change. By assuming all the other factors affect the treatment and control groups in the same way over time, any change in the differences of regulatory activities between these two groups before and after the rule change can be attributed to the rule change (the treatment group starting to disclose in the TRI). Specifically, I estimate the following DID model: 𝑌𝑌𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼 + 𝛿𝛿𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑡𝑡𝑖𝑖 + 𝛽𝛽 ∗ 𝑋𝑋𝑖𝑖𝑖𝑖 + 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑡𝑡𝑖𝑖 + 𝛾𝛾𝑖𝑖 + 𝛿𝛿𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 (1) Where 𝑌𝑌𝑖𝑖𝑖𝑖𝑖𝑖 are measures of regulatory activities (# of inspection activities, # of enforcement activities) for state 𝛿𝛿 in reporting year 𝑡𝑡 on group (treatment/control) 𝑖𝑖. 𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 is a dummy variable that equals one for the posttreatment period (1998 and after). 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑡𝑡𝑖𝑖 is a dummy variable that equals one for the treatment group. 𝑋𝑋𝑖𝑖𝑖𝑖 include state-year controls for population, personal income, unemployment rate, governor party affiliation, partisan composition of state legislature, state government ideology, and fiscal health indicator. I also include state fixed effects 𝛾𝛾𝑖𝑖 and year fixed effects 𝛿𝛿𝑖𝑖. Identification in the basic model requires that in the absence of the rule change, the control group would have the same regulatory activity trends with the treatment group. To explore the validity of the assumption, I extend the above basic DID analysis to an event study analysis. In practice, this means estimating equation (1) with a full set of year effects interacted with the treatment status 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑡𝑡𝑖𝑖 in lieu of 𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑡𝑡𝑖𝑖. Plot of the coefficients for the year-treatment interactions allows the examination of pretrends. If there was no difference in the trends of the regulatory activities for the treatment group and the control group leading up to the rule change, we would expect the coefficients for the year-treatment interactions to move around zero before the rule change. The second research design is a triple DID model. In the above DID model, while my control group (non-reporting facilities in newly covered industries) and treatment group (reporting facilities in newly covered industries) show no differential pretrends before the rule change, facilities in the two groups differ greatly in their sizes (sizes are the criteria whether a facility in covered industries report or does not report to the TRI). If there were confounding changes that affect facilities of different sizes differently at the time of and/or after the rule change, the basic DID model will attribute such difference to the rule change. To address this concern, I use a

Page 21: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

19

triple DID model, which adds two more groups of facilities to control for the potential change in the differences between the treatment group and the control group before and after the rule change that is due to different sizes of facilities in the treatment and control groups. The two added groups are 1) facilities in the previously covered industries (SIC codes 20 -39) since the establishment of the TRI program that report to the TRI and 2) facilities in the previously covered industries (SIC codes 20 -39) since the establishment of the TRI program that do not report to the TRI. The rule change in 1998 did not affect facilities in these two groups; they either report to the TRI or do not report to the TRI throughout the study period. The size-based reporting criteria are the same for these two groups as is for the treatment and control groups. Same as facilities in the treatment and control groups, facilities in these two groups also differ in their sizes, which are the reason that they reported or did not report in the first place. So I use the change in the differences of regulatory activities between these two groups before and after the rule change to track the change in the differences between the treatment group and control group before and after the rule change that is due to the different sizes of facilities in the treatment and control groups. I estimate the triple DID model as below (equation (2)). In addition, I use the event study approach as described for the basic DID model to inspect pretrends. Specifically, I replace interaction terms in equation (2) that contain 𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 with a series of year dummies interacted with other parts of these interaction terms. 𝑌𝑌𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼 + 𝛿𝛿𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 ∗ 𝑅𝑅𝑇𝑇𝑅𝑅𝛿𝛿𝑇𝑇𝑡𝑡𝑖𝑖 ∗ 𝑁𝑁𝑇𝑇𝑤𝑤𝑖𝑖 + 𝛽𝛽𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 ∗ 𝑅𝑅𝑇𝑇𝑅𝑅𝛿𝛿𝑇𝑇𝑡𝑡𝑖𝑖 + 𝛾𝛾𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 ∗ 𝑁𝑁𝑇𝑇𝑤𝑤𝑖𝑖 + 𝛿𝛿𝑅𝑅𝑇𝑇𝑅𝑅𝛿𝛿𝑇𝑇𝑡𝑡𝑖𝑖 ∗𝑁𝑁𝑇𝑇𝑤𝑤𝑖𝑖 + 𝜃𝜃 ∗ 𝑋𝑋𝑖𝑖𝑖𝑖 + 𝜇𝜇 ∗ 𝑅𝑅𝑇𝑇𝑅𝑅𝛿𝛿𝑇𝑇𝑡𝑡𝑖𝑖 + 𝜌𝜌 ∗ 𝑁𝑁𝑇𝑇𝑤𝑤𝑖𝑖 + 𝜎𝜎𝑖𝑖 + 𝜏𝜏𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 (2) Where 𝑌𝑌𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 are measures of regulatory activities (# of inspection activities, # of enforcement activities) for state 𝛿𝛿 in reporting year 𝑡𝑡 by industry sectors indicator 𝑗𝑗 and TRI reporting status 𝑖𝑖 (on and after the rule change). 𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡𝑖𝑖 is a dummy variable that equals one for the posttreatment period. 𝑅𝑅𝑇𝑇𝑅𝑅𝛿𝛿𝑇𝑇𝑡𝑡𝑖𝑖 is a dummy variable that that equals one for the groups of facilities that report to the TRI on and after 1998. 𝑁𝑁𝑇𝑇𝑤𝑤𝑖𝑖 is a dummy variable that equals one for the group of facilities that are in the newly covered industries due to the rule change. 𝑋𝑋𝑖𝑖𝑖𝑖 include state-year level controls for population, personal income, unemployment rate, governor party affiliation, partisan composition of state legislature, state government ideology, and fiscal health. I also include state fixed effects 𝜎𝜎𝑖𝑖 and year fixed effects 𝜏𝜏𝑖𝑖. 5.2 Facility Level Analysis Method The facility level analysis centers on whether regulatory activities react to disclosed TRI information on toxic emissions. It estimates the change of regulatory activities on the change of disclosed TRI information, which can be achieved through a facility fixed effects model. In this context, regulatory activities on facility 𝑖𝑖 and in reporting year 𝑡𝑡 can be represented as 𝑅𝑅𝑇𝑇𝑅𝑅𝑅𝑅𝑅𝑅𝑇𝑇𝑡𝑡𝛿𝛿𝑇𝑇𝑅𝑅 𝐴𝐴𝐴𝐴𝑡𝑡𝑖𝑖𝐴𝐴𝑖𝑖𝑡𝑡𝑖𝑖𝑇𝑇𝛿𝛿𝑖𝑖𝑖𝑖 = 𝛼𝛼0 + 𝛼𝛼1 ∗ 𝑇𝑇𝑅𝑅𝑇𝑇 𝑖𝑖𝑖𝑖𝑖𝑖𝛿𝛿𝑇𝑇𝑖𝑖𝑇𝑇𝑡𝑡𝑖𝑖𝛿𝛿𝑖𝑖 + 𝛼𝛼2 ∗ ℎ𝑅𝑅𝐴𝐴𝑖𝑖𝑖𝑖

+ 𝛼𝛼3 ∗ 𝐶𝐶𝛿𝛿𝑅𝑅𝑖𝑖𝑡𝑡𝑅𝑅 𝑇𝑇𝐴𝐴𝛿𝛿𝑖𝑖𝛿𝛿𝑖𝑖𝑖𝑖𝐴𝐴 𝑖𝑖𝑇𝑇𝐴𝐴𝑡𝑡𝛿𝛿𝑇𝑇𝛿𝛿𝑖𝑖𝑖𝑖 + 𝛾𝛾𝑖𝑖 + 𝜗𝜗𝑖𝑖𝑖𝑖 + 𝜖𝜖𝑖𝑖𝑖𝑖 (3)

Page 22: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

20

In equation (3), 𝑇𝑇𝑅𝑅𝑇𝑇 𝑖𝑖𝑖𝑖𝑖𝑖𝛿𝛿𝑇𝑇𝑖𝑖𝑇𝑇𝑡𝑡𝑖𝑖𝛿𝛿𝑖𝑖𝑖𝑖𝑖𝑖 is facilities’ aggregated air/water/land/total emissions for calendar year 𝑡𝑡, which is the information that the TRI discloses in reporting year 𝑡𝑡. I include TRI information on water and land releases as it also has the potential to provide new information on facilities’ environmental performance in the CAA. α1 measures how regulatory activities react to the disclosed information. ℎ𝑅𝑅𝐴𝐴𝑖𝑖𝑖𝑖 equals one if facility 𝑖𝑖 is a high priority violator at any time during the reporting year 𝑡𝑡. 𝐶𝐶𝛿𝛿𝑅𝑅𝑖𝑖𝑡𝑡𝑅𝑅 𝑇𝑇𝐴𝐴𝛿𝛿𝑖𝑖𝛿𝛿𝑖𝑖𝑖𝑖𝐴𝐴 𝑖𝑖𝑇𝑇𝐴𝐴𝑡𝑡𝛿𝛿𝑇𝑇𝛿𝛿𝑖𝑖𝑖𝑖 consists of three measures of local economic conditions: unemployment rate, personal income, and population. 𝛾𝛾𝑖𝑖 are facility fixed effects which control time invariant characteristics of facilities, such as location, industry sector, and the invariant part of regulators’ expectation. ϑst are state-by-year fixed effects and control for all time-invariant and time-variant factors at the state and national levels. 6 Results 6.1 State Level Analysis Results Difference-in-Differences Results I begin by presenting results for the basic difference-in-differences model based on TRI’s sized-based reporting criteria and the 1998 rule change that expanded its industry coverage. Following the rule change, the TRI for the first time started to cover seven new industry sectors. However, the reporting requirements only applied to facilities that met the size-based reporting criteria in the newly covered industries. The results from the basic DID model compares the difference of regulatory activities between the reporting and nonreporting facilities in the newly covered industries before and after the rule change to identify the effects of reporting to the TRI. I use two measures as the dependent variables: number of inspection activities and number of enforcement activities. Tables 4 and 5 present results for key variables from the DID estimation for inspection activities and enforcement activities respectively (Full results are in Table A1 and Table A2 in the appendix). Each column of tables 4 and 5 represents estimates from a separate regression. All standard errors are clustered at state level and in parentheses. My preferred models are columns (3) in both Tables 4 and 5. I can interpret the results as percentage change with the natural logarithm transformation of the dependent variables. Columns (3) in Tables 4 and 5 show that regulators reduce regulatory activities on the group of facilities that started to report to the TRI after the rule change. Specifically, on average regulators reduce the number of inspection activities on these facilities by about 27% and the number of enforcement activities by about 30%, compared with the scenario if this group of facilities did not report to the TRI. The results almost remain the same with the inclusion of different fixed effects and state-year controls (columns (1) and (2)). In addition, the results do not change substantively in specifications with the level forms of regulatory activities as the dependent variables (columns (4) – (6)).

Page 23: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

21

Table 4. DID Estimates for Inspection Activities (1) (2) (3) (4) (5) (6) Ln (Insp.) Ln (Insp.) Ln (Insp.) Insp. Insp. Insp.

Post x TRI -0.27*** -0.27*** -0.27** -25.53*** -26.03*** -25.53*** (0.07) (0.08) (0.10) (6.56) (6.71) (8.98)

TRI -0.64*** -0.62*** -0.64*** -25.28*** -25.08*** -25.28*** (0.09) (0.09) (0.12) (4.65) (4.76) (6.37)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 882 864 882 882 864 882 R2 0.79 0.8 0.93 0.70 0.70 0.90

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Inspection +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 5. DID Estimates for Enforcement Activities

(1) (2) (3) (4) (5) (6) Ln (Enf.) Ln (Enf.) Ln (Enf.) Enf. Enf. Enf.

Post x TRI -0.30*** -0.29*** -0.30** -4.44** -4.46** -4.44 (0.11) (0.11) (0.15) (1.97) (2.02) (2.70)

TRI -0.39*** -0.39*** -0.39** -2.78** -2.79* -2.78 (0.11) (0.11) (0.15) (1.37) (1.41) (1.88)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 882 864 882 882 864 882 R2 0.66 0.67 0.82 0.51 0.51 0.69

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Enforcement +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. A possible concern is that the regression results from Tables 4 and 5 could be driven by preexisting differential trends in regulatory activities for the treatment and control groups. To address this concern, I also show results from an event study approach. Particularly, I estimate a model similar to specifications in columns (3) in Tables 4 and 5, except that I replace 𝛿𝛿𝛿𝛿𝛿𝛿𝑡𝑡 ∗ 𝑇𝑇𝑅𝑅𝑇𝑇 with a full set of year dummies interacted with 𝑇𝑇𝑅𝑅𝑇𝑇. I plot the year by treatment interaction in Figure 4, where I normalize the coefficients to 0 in 1997, the year prior to the rule change. The figure suggests there was little to no pretrend before the rule change, validating the research design.

Page 24: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

22

Panel A. Inspection Activities Panel B. Enforcement Activities

Figure 4. Event Study Coefficients for the DID Model

Note: (1) Solid line is coefficient plot; dash lines are 95% confidence interval plot. (2) Standard errors are clustered at state level. Triple Difference-in-Differences Results In this part, I present the results from the triple difference-in-differences model. Tables 6 and 7 present the key results from the triple DID models (Tables A3 and A4 in appendix show full results). Each column in Tables 6 and 7 represent estimates from a different regression. All standard errors are clustered at state level and in parentheses. The term “Post x New x TRI” identifies the post-treatment regulatory activities on the group of the facilities that are in the newly covered industries and have started to report to TRI from 1998 due to the rule change (the treatment group in the post-treatment period). The coefficient of this term shows the effect of reporting to TRI on regulatory activities. The results from the triple DID models are consistent with the results of the DID models and the effects on both inspection and enforcement activities are larger compared with the DID models. Table 6. Triple DID Estimates for Inspection Activities

(1) (2) (3) (4) (5) (6) Ln (Insp.) Ln (Insp.) Ln (Insp.) Insp. Insp. Insp.

Post x New x TRI -0.44*** -0.45*** -0.44*** -91.40*** -93.20*** -91.40*** (0.09) (0.09) (0.10) (28.60) (29.20) (32.50)

Post x New 0.37*** 0.38*** 0.37*** -11.49 -12.5 -11.49 (0.05) (0.06) (0.06) (17.01) (17.38) (19.33)

Post x TRI 0.17*** 0.18*** 0.17** 65.87** 67.17** 65.87** (0.06) (0.06) (0.07) (27.63) (28.24) (31.40)

New x TRI -0.43*** -0.42*** -0.43*** 40.63 41.48 40.63 (0.11) (0.11) (0.13) (28.38) (29.02) (32.25)

New -1.37*** -1.37*** -1.37*** -206.88*** -209.26*** -206.88*** (0.12) (0.12) (0.14) (53.80) (55.00) (61.14)

TRI -0.21*** -0.20** -0.21** -65.91** -66.56** -65.91* (0.07) (0.08) (0.08) (29.09) (29.76) (33.06)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X

Page 25: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

23

State by Year Fixed Effects X X N 1764 1728 1764 1764 1728 1764 R2 0.81 0.81 0.89 0.56 0.56 0.64

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Inspection +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 7. Triple DID Estimates for Enforcement Activities

(1) (2) (3) (4) (5) (6) Ln (Enf.) Ln (Enf.) Ln (Enf.) Enf. Enf. Enf.

Post x New x TRI -0.43*** -0.44*** -0.43*** -10.35*** -10.66*** -10.35** (0.13) (0.13) (0.14) (3.69) (3.77) (4.20)

Post x New 0.34*** 0.36*** 0.34*** 1.52 1.64 1.52 (0.09) (0.09) (0.10) (3.15) (3.22) (3.58)

Post x TRI 0.13* 0.15** 0.13 5.91** 6.20** 5.91* (0.07) (0.07) (0.08) (2.71) (2.76) (3.08)

New x TRI -0.42*** -0.40*** -0.42** -4.73 -4.67 -4.73 (0.14) (0.14) (0.16) (6.17) (6.31) (7.01)

New -1.47*** -1.49*** -1.47*** -31.50*** -32.03*** -31.50*** (0.15) (0.15) (0.17) (6.48) (6.60) (7.36)

TRI 0.03 0.02 0.03 1.95 1.88 1.95 (0.09) (0.09) (0.11) (5.59) (5.72) (6.35)

Year Fixed Effects X X

X X

State Fixed Effects X X

X X

State by Year Controls

X

X

State by Year Fixed Effects

X

X

N 1764 1728 1764 1764 1728 1764 R2 0.75 0.75 0.81 0.52 0.53 0.61

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Enforcement +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Similarly, I conduct an event study analysis to inspect the pretrend based on the triple DID model. Figure 5 plots the coefficients from the event study approach, and it shows no significant pretrend leading up to the rule change.

Page 26: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

24

Panel A. Inspection Activities Panel B. Enforcement Activities

Figure 5. Event Study Coefficients for the Triple DID Model

Note: (1) Solid line is coefficient plot; dash lines are 95% confidence interval plot. (2) Standard errors are clustered at state level. 6.2 Facility Level Analysis The state level analysis examines whether reporting to TRI for the first time affects regulatory activities. The facility level analysis examines whether regulatory activities react to disclosed information (amounts of pollution releases) in the TRI program over the period from 1998 to 2003. The key independent variables are the amounts of pollution releases disclosed in the TRI. The coefficients on the amounts of disclosed releases measure the effects of TRI information on regulatory activities relative to expectations. Thus, insignificant coefficients might reflect either that regulators do not use the information in making regulatory decisions or that the information provides nothing new and was in line with their expectations. Tables 8 and 9 present the facility level fixed effects regression coefficients on TRI information (Tables A5 and A6 in appendix show full results). Each column represents results from a different regression. All standard errors are clustered at state level and in parentheses. The dependent variables are the number of inspection activities (Table 8) and the number enforcement activities (Table 9). The coefficients on Ln (Disclosed Air/Land/Water/Total Releases) measure the impact of disclosed TRI information on regulatory activities. The results suggest that CAA regulatory activities do not react much to information disclosed in the TRI. Specifically, TRI information on all types of releases (air/land/water/total) has no statistically and substantively significant impact on inspection activities. Information on air/land/total releases also has no impact on enforcement activities. While the information on water releases affects CAA enforcement actions, the effect is exceedingly small. A 10% increase of disclosed water releases will only increase 0.0009 enforcement activities (a 0.4% increase from 0.22, which is the mean number of enforcement activities on a facility in a given year). For all the specifications reported in Tables 8 and 9, I have run models without TRI information as

Page 27: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

25

explanatory variables and the explained variations (R2) do not change at all, which suggest TRI information offers little to explain regulatory activities. Table 8. Facility Level Estimates for Inspection Activities

(1) (2) (3) (4) Insp. Insp. Insp. Insp.

Ln (Disclosed Air Releases) 0.008 (0.006)

Ln (Disclosed Land Releases) -0.010 (0.011)

Ln (Disclosed Water Releases) 0.019 (0.015)

Ln (Disclosed Total Releases) 0.011 (0.007)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 66759 66759 66759 66759 R2 0.586 0.586 0.586 0.586

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 9. Facility Level Estimates for Enforcement Activities

(1) (2) (3) (4) Enf. Enf. Enf. Enf.

Ln (Disclosed Air Releases) 0.001 (0.002)

Ln (Disclosed Land Releases) 0.000 (0.003)

Ln (Disclosed Water Releases) 0.009** (0.004)

Ln (Disclosed Total Releases) 0.001 (0.002)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 66759 66759 66759 66759 R2 0.575 0.575 0.575 0.575

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. While the above analysis shows that regulators do not respond to this information, it could be that regulators do not use this type of information in their regulatory decision making or they have anticipated the disclosed information. To investigate this issue, I consider the possibility

Page 28: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

26

that regulators can observe contemporaneous toxic releases. Information on contemporaneous TRI emissions is not available to the regulator contemporaneously. However, regulators may have other channels to estimate such information such as other reports submitted by facilities and production of facilities. Tables 10 and 11 show the results with the addition of contemporaneous emissions as an independent variable. Each column in Tables 10 and 11 represents estimates from a different regression. All standard errors are clustered at state level and in parentheses. I have also estimated models in Tables 10 and 11 without disclosed releases in the TRI as independent variables in case of high collinearity between disclosed releases and contemporaneous releases but the results on contemporaneous emissions basically did not change. Table 10. Facility Level Estimates for Inspection Activities (Contemporaneous Releases)

(1) (2) (3) (4) Insp. Insp. Insp. Insp.

Ln (Disclosed Air Releases) 0.005 (0.007)

Ln (Contemp. Air Releases) 0.021* (0.011)

Ln (Disclosed Land Releases) -0.011 (0.011)

Ln (Contemp. Land Releases) (0.025) (0.015)

Ln (Disclosed Water Releases) 0.020 (0.015)

Ln (Contemp. Water Releases) (0.011) (0.025)

Ln (Disclosed Total Releases) 0.008 (0.008)

Ln (Contemp. Total Releases) 0.022** (0.008)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 65637 65637 65637 65637 R2 0.587 0.587 0.587 0.587

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 11. Facility Level Estimates for Enforcement Activities (Contemporaneous Releases)

(1) (2) (3) (4) Enf. Enf. Enf. Enf.

Ln (Disclosed Air Releases) 0.001 (0.002)

Ln (Contemp. Air Releases) 0.005**

Page 29: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

27

(0.002)

Ln (Disclosed Land Releases) 0.000 (0.003)

Ln (Contemp. Land Releases) 0.000 (0.003)

Ln (Disclosed Water Release) 0.008** (0.004)

Ln (Contemp. Water Releases) 0.015** (0.007)

Ln (Disclosed Total Release) 0.001 (0.002)

Ln (Contemp. Total Releases) 0.005** (0.002)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 65637 65637 65637 65637 R2 0.575 0.575 0.576 0.575

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Results in Tables 10 and 11 suggest that regulators react to contemporaneous emissions, especially to air and total emissions. Specifically, a 10% increase in contemporaneous TRI air emissions leads to a 0.002 increase in inspection actions (a 0.2% increase based on mean number of inspection activities 1.30) and a 0.0005 increase in enforcement actions (a 0.2% increase based on mean number of enforcement activities 0.22). The effects of contemporaneous TRI total emissions are almost the same as for those of the contemporaneous TRI air emissions. Even though the effects of contemporaneous emissions are very small, the results raises some concerns on the novelty of TRI information to regulators. Information on contemporaneous TRI emissions is not available to regulators, as the submission deadline for TRI emissions for a certain year is July 1 of the next calendar year and publication date for such information is on an even later date. Yet the results show that regulatory activities react to this unknow information. It looks like to some extent regulators have the ability to estimate facilities’ contemporaneous TRI emissions. When TRI program discloses such information on a later date, it will not be new to the regulators anymore. 7 Robustness and Sensitivity To examine the robustness of the findings, I conduct a few sensitivity analyses at both the state level and the facility level. First, I run models based on the definition of reporting year using the submission deadline assignment rule. Second, I limit the samples used for analysis. Specifically, for the state level analysis, I exclude states with relatively few treated facilities, and for the

Page 30: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

28

facility level analysis, I exclude facilities without any regulatory activities during the study period. Third, for the state level analysis, I use regulatory activities conducted by state regulators as the dependent variables, instead of the combined federal and state regulatory activities. Fourth, for the state level analysis, I use subcategories of regulatory activities—full inspection, partial inspection, formal enforcement, informal enforcement—as the dependent variables, instead of the two-type categorization of regulatory activities (inspection and enforcement). Fifth, for the state level analysis, I use different definitions of the treatment group. Sixth, for the facility level analysis, I use level forms of disclosed emissions instead of the natural logarithm transformation. Below I explain and present results for these sensitivity analyses. 7.1 State Level Sensitivity Analysis First, I examine the results based on the definition of reporting year using the submission deadline assignment rule. During the study period, TRI information for a certain calendar year was due to EPA by July 1 of the next calendar year and the information, after being processed, was disclosed to the public roughly 10 months later after the submission due dates. The definition of reporting year based on submission deadline assumes that regulators start to react to TRI information right after the submission deadline while the definition of reporting year based on public-access date assumes that regulators react to the TRI after TRI information is publicly available. The analysis so far has been based on the public-access assignment rule, and Tables 12 and 13 present results based on the submission deadline assignment rule for the DID model. Table 12. DID Estimates for Inspection Activities (Based on Submission Deadline)

(1) (2) (3) (4) (5) (6) Ln (Insp.) Ln (Insp.) Ln (Insp.) Insp. Insp. Insp.

Post x TRI -0.25*** -0.24*** -0.25*** -21.13*** -21.13*** -21.13*** (0.06) (0.06) (0.08) (4.99) (5.12) (6.84)

TRI -0.60*** -0.60*** -0.60*** -24.11*** -24.18*** -24.11*** (0.09) (0.10) (0.13) (4.66) (4.77) (6.37)

Constant 4.17*** -78.62 2.40*** 61.30*** 195.53 19.55*** (0.10) (49.87) (0.06) (6.32) (7431.45) (3.19)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 882 864 882 882 864 882 R2 0.78 0.79 0.92 0.68 0.68 0.90

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Inspection +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 13. DID Estimates for Enforcement Activities (Based on Submission Deadline)

(1) (2) (3) (4) (5) (6) Ln (Enf.) Ln (Enf.) Ln (Enf.) Enf. Enf. Enf.

Post x TRI -0.24* -0.22* -0.24 -4.44** -4.45** -4.44* (0.13) (0.13) (0.17) (1.91) (1.96) (2.62)

Page 31: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

29

TRI -0.39*** -0.38*** -0.39** -2.37* -2.38* -2.37 (0.11) (0.12) (0.15) (1.35) (1.39) (1.85)

Constant 0.71*** -42.1 0.89*** 0.30 -149.19 2.19** (0.09) (35.59) (0.08) (0.95) (510.92) (0.93)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 882 864 882 882 864 882 R2 0.66 0.66 0.81 0.53 0.53 0.69

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Enforcement +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Tables 12 and 13 shows that the results based on the submission deadline assignment rule do not change much from the results based on the public-access assignment rule. This is not surprising as we observe consistent lower levels of regulatory activities on the treatment group after the rule change in the event study analysis based on the public-access assignment rule. Shifting the treatment period by a few months won’t change the results much. Since the treatment status changed only once in the difference-in-differences analysis, identifying the precise time when regulators started to react to TRI does not matter much especially given that the analysis have included pretty long periods of time both before and after the rule change. Second, I examine the results by excluding states with relatively few treatment facilities from the analysis. The rule change in 1998 affected facilities located in every state and district, except for Vermont. However, the number affected facilities does vary significantly across states, with Ohio with the most affected facilities (72) and D.C. with the fewest affected facilities (1). In the main DID analysis, I have excluded Vermont and D.C. Here I exclude states with fewer than 10 affected facilities—a total of 17 states—from the analysis. Tables 14 and 15 present the results for this analysis. Table 14. DID Estimates for Inspection Activities (States with >10 affected Facilities)

(1) (2) (3) (4) (5) (6) Ln (Insp.) Ln (Insp.) Ln (Insp.) Insp. Insp. Insp.

Post x TRI -0.19** -0.19** -0.19 -34.57*** -34.57*** -34.57** (0.08) (0.08) (0.12) (9.31) (9.37) (12.72)

TRI -0.66*** -0.66*** -0.66*** -32.07*** -32.07*** -32.07*** (0.11) (0.11) (0.15) (6.48) (6.52) (8.85)

Constant 4.62*** -40.28 4.83*** 82.59*** -3425.01 107.04*** (0.13) (53.46) (0.07) (8.15) (9159.40) (4.43)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 594 594 594 594 594 594 R2 0.76 0.77 0.93 0.70 0.71 0.90

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Inspection +1).

Page 32: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

30

(2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 15. DID Estimates for Enforcement Activities (States with >10 affected Facilities)

(1) (2) (3) (4) (5) (6) Ln (Enf.) Ln (Enf.) Ln (Enf.) Enf. Enf. Enf.

Post x TRI -0.31** -0.31** -0.31 -5.93** -5.93* -5.93 (0.15) (0.15) (0.20) (2.90) (2.92) (3.96)

TRI -0.36** -0.36** -0.36* -3.13 -3.13 -3.13 (0.14) (0.14) (0.19) (1.99) (2.00) (2.72)

Constant 1.23*** -80.69** 1.53*** 6.54*** -634.70 4.57*** (0.12) (32.98) (0.10) (2.27) (472.33) (1.36)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 594 594 594 594 594 594 R2 0.63 0.64 0.81 0.50 0.50 0.68

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Enforcement +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Tables 14 and 15 show that the results do not substantively change from the main DID analysis. Regulators reduce regulatory activities on the group of facilities that start to report to TRI in states with relatively more affected facilities (larger treatment group) as well. The effects for inspection activities (natural logarithm transformed) are smaller compared with the full sample analysis and it is not significant at the 10% level for one of the model specifications. The effects for enforcement activities (natural logarithm transformed) are of similar magnitude with the full sample analysis but it is not significant at the 10% for one of the model specifications. Overall, however, the results are consistent with results from the full sample analysis. Third, I conduct analysis for different types of regulatory activities. I have examined the reactions of inspection activities and enforcement activities to the TRI in the main DID model. Here I further break down the two types of regulatory activities. Specifically, I examine the reactions of state inspection activities, state enforcement activities, full inspection activities, partial inspection activities, formal enforcement activities, and informal enforcement activities to the TRI. Table 16 present results from these analyses. Table 16. DID Estimates (Subcategories of Regulatory Activities)

(1) (2) (3) (4) (5) (6) Ln (State

Insp.) Ln (State

Enf.) Ln (Full

Insp.) Ln (Partial

Insp.) Ln (Formal

Enf.) Ln (Informal

Enf.) Post x TRI -0.25** -0.32** -0.48*** -0.26* -0.19** -0.23*

(0.10) (0.14) (0.09) (0.15) (0.09) (0.13)

TRI -0.68*** -0.43*** -0.47*** -0.26* -0.21** -0.38*** (0.12) (0.13) (0.13) (0.14) (0.10) (0.12)

Constant 2.69*** 1.85*** 2.82*** 0.13* 1.31*** 1.73***

Page 33: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

31

(0.06) (0.06) (0.06) (0.07) (0.07) (0.06)

State by Year Fixed Effects X X X X X X N 882 882 882 882 882 882 R2 0.93 0.84 0.93 0.96 0.80 0.80

Note: (1) The actual dependent variable used is Ln (Regulatory Activity +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Each of the column in Table 16 represents estimates from a different regression with its dependent variable on top of the column. For each dependent variable, I only present the results from the state-by-year fixed effects model specification as the results do not change much from other specifications. Table 16 shows that the results for subcategories of regulatory activities are consistent with the main models. It is not surprising to see that state regulatory activities react similarly in terms of magnitudes and statistical significance compared with reactions of the combined state and federal activities as state carried out predominant portions of the regulatory activities. In addition, all the other subcategories of regulatory activities react to the TRI in the same direction and in comparable magnitudes as the main results indicate. Fourth, I explore different definitions of the treatment group. In the main results, I define the treatment group as the group of facilities that started to report to the TRI as a result of the rule change in 1998. Here I narrow the definition and use three alternative definitions. Specifically, I define the treatment group as 1) facilities that report air emissions to the TRI as a result of the rule change, 2) facilities that report emissions of CAA regulated chemicals to the TRI as a result of the rule change, and 3) facilities that report air emissions of CAA regulated chemicals to the TRI as a result of the rule change. The control groups are the same as the main results, that is, facilities in the newly covered industries but do not report to the TRI. Table 17 presents results based on these alternative definitions of the treatment group. Table 17. DID Estimates (Alternative Definitions of Treatment Group)

(1) (2) (3) (4) (5) (6) Report Air Emissions Report CAA

Chemicals Emissions Report Air Emissions

of CAA Chemicals Ln (Insp.) Ln (Enf.) Ln (Insp.) Ln (Enf.) Ln (Insp.) Ln (Enf.)

Post x TRI -0.28*** -0.31** -0.22** -0.28* -0.22** -0.29* (0.10) (0.14) (0.11) (0.15) (0.11) (0.14)

TRI -0.67*** -0.39*** -0.74*** -0.44*** -0.78*** -0.45*** (0.12) (0.15) (0.12) (0.14) (0.12) (0.14)

Constant 2.92*** 2.07*** 2.95*** 2.09*** 2.97*** 2.09*** (0.06) (0.07) (0.06) (0.07) (0.06) (0.07)

State by Year Fixed Effects X X X X X X N 882 882 882 882 882 882 R2 0.93 0.82 0.93 0.82 0.93 0.82

Note: (1) The actual dependent variable used is Ln (Regulatory Activity +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01.

Page 34: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

32

For each alternative definition of the treatment group, I report estimates based on the state by year fixed effects model specification for the natural logarithm transformed inspection activities and enforcement activities respectively. Table 17 shows that results based on all the three alternative definitions of the treatment groups are similar in magnitudes and statistical significance compared with the main results. 7.2 Facility Level Sensitivity Analysis For the facility level sensitivity analysis, I first present the estimates based on the definition of reporting year that is based on submission deadline. Tables 18 and 19 show the results for inspection activities and enforcement activities respectively. While some estimates are statistically significant, the substantial impacts are all very small. For example, a 10% increase in total releases only expect to increase enforcement activities by 0.0009 (a 0.4% increase from the baseline enforcement activities of 0.22 per year). Table 18. Facility Level Estimates for Inspection Activities (Based on Submission Deadline)

(1) (2) (3) (4) Insp. Insp. Insp. Insp.

Ln (Disclosed Air Releases) 0.011 (0.007)

Ln (Disclosed Land Releases) 0.000 (0.011)

Ln (Disclosed Water Releases) 0.022 (0.015)

Ln (Disclosed Total Releases) 0.018** (0.008)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 66804 66804 66804 66804 R2 0.536 0.536 0.536 0.536

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 19. Facility Level Estimates for Enforcement Activities (Based on Submission Deadline)

(1) (2) (3) (4) Enf. Enf. Enf. Enf.

Ln (Disclosed Air Releases) 0.009** (0.003)

Ln (Disclosed Land Releases) 0.002 (0.003)

Ln (Disclosed Water Releases) 0.007** (0.003)

Ln (Disclosed Total Releases) 0.009***

Page 35: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

33

(0.003)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 66804 66804 66804 66804 R2 0.530 0.530 0.530 0.530

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Second, I exclude facilities that have not had any regulatory activities during the analysis period. There may be other unknown reasons that these facilities never had any regulatory activities. While excluding them may lead to sample selection bias, including them in the analysis may dampen the effects of TRI information on regulatory activities. Tables 20 and 21 shows results from this analysis. Again, the results suggest that regulatory activities do not react to TRI information. Table 20. Facility Level Estimates for Inspection Activities (Excluding Facilities without Regulatory Activities During the Analysis Period)

(1) (2) (3) (4) Insp. Insp. Insp. Insp.

Ln (Disclosed Air Releases) 0.012 (0.008)

Ln (Disclosed Land Releases) -0.013 (0.012)

Ln (Disclosed Water Releases) 0.021 (0.017)

Ln (Disclosed Total Releases) 0.015 (0.009)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 55353 55353 55353 55353 R2 0.582 0.582 0.582 0.582

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Table 21. Facility Level Estimates for Enforcement Activities (Excluding Facilities without Regulatory Activities During the Analysis Period)

(1) (2) (3) (4) Enf. Enf. Enf. Enf.

Ln (Disclosed Air Releases) 0.002 (0.003)

Ln (Disclosed Land Releases) 0.000 (0.003)

Page 36: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

34

Ln (Disclosed Water Releases) 0.010** (0.004)

Ln (Disclosed Total Releases) 0.001 (0.002)

Controls X X X X Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 55353 55353 55353 55353 R2 0.572 0.572 0.572 0.572

Note: (1) Controls include facilities’ High Priority Violator Status, county unemployment, county personal income, and county population. (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01. Lastly, I have also run regressions that are based on level forms of emissions information, and the results shows that regulatory activities do not respond to TRI information as well. 8 Discussion 8.1 Size of Effects In this paper, I study the impacts of information disclosure policy on the implementation of traditional environmental regulations in the context of the TRI and the CAA. With difference-in-differences based approaches that exploit TRI’s size-based reporting criteria and its industry coverage expansion in 1998, I find that regulators significantly reduce CAA regulatory activities on the group of new facilities that disclose emissions information in the TRI. However, through a facility level fixed effects model, I also find that regulatory activities have very limited response to the amounts of toxic emissions disclosed in the TRI. These findings are robust with several model specifications and sensitivity analyses. Specifically, in the state level analysis, I find that regulators reduce about 27% of the inspection activities and about 30% of the enforcement activities on the group of facilities that report to the TRI, compared with the scenario if these facilities did not report to the TRI. These are seemingly large effects. I put the size of the effects into context by comparing them with the results of other relevant studies. Since there is no previous study that directly examines the relationship between the TRI and regulatory activities, I mostly compare them with the reduction of toxic emissions given that researchers have found a strong correlation between the amounts of TRI releases and regulatory activities (Christopher S. Decker, 2005; Christopher S. Decker, Nielsen, & Sindt, 2005; Konisky & Reenock, 2018). First, these estimates are in line with the overall emissions trend of TRI reporting facilities. EPA’s 2003 TRI National Analysis (U.S. EPA, 2005) shows that total TRI releases decreased by 42 percent over the six years from 1998 to 2003 (the posttreatment period in my analysis), and these decreases disproportionally came from the newly covered industry sectors due to the 1998

Page 37: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

35

rule change. With this large drop in reported toxic emissions, we would expect the regulatory activities to drop significantly on these reporting facilities as well. Second, I compare my results with studies that examine the impacts of other information users’ actions on facilities’ toxic emissions. These studies primarily focus on the media and stock market. Saha and Mohr (2013) found that compared to the pre-treatment years, facilities with media attention reduced their toxic emissions by about 40% relative to facilities without media attention. Similarly, Campa (2018) found media coverage led to a 29% decrease in toxic emissions. On the impact of the stock market, Khanna et al. (1998) found that a mere 1 million dollar abnormal return resulted in a 16% decrease of on-site releases, and Konar and Cohen (1997) found that the 40 firms that experienced the largest abnormal return following the first release of TRI information subsequently reduced their TRI releases by about 30% more than their industry peers per $1,000 revenue. These studies compare TRI reporting facilities that experienced pressure from information users and those that did not. While this comparison cannot directly apply to the comparison made in this study, which is between TRI reporting and nonreporting facilities, it does show that information can have large effects on facilities’ reported emissions, which are highly correlated with CAA regulatory activities. 8.2 State Level Findings vs. Facility Level Findings While I find that regulators reduce regulatory activities on facilities that report to the TRI, I also find that their regulatory activities do not react to changes in the disclosed amount of toxic emissions. The results, while surprising, are consistent with previous studies, especially those on TRI’s impacts on housing prices. Mastromonaco (2015) exploited the strengthening of the reporting requirement for the chemical lead in 2001 with a difference-in-differences design and found that listing a facility in the TRI significantly lowered housing prices around the facility. Sanders (2014) utilized the same rule change used in this study, and with a difference-in-differences model, he found that listing an existing facility in the TRI caused significant decrease in home prices in the zip code with the newly listed facility. However, Bui and Mayer (2003) showed that TRI information had little impact on housing prices with a first-difference approach that exploits year-over-year change of toxic emission disclosed in the TRI. The pattern also exists in stock market response. For example, Hamilton (1995) found that firms experienced negative abnormal returns after being listed in the TRI, but the magnitude of abnormal returns was not correlated with the amounts of toxic emissions disclosed in the TRI. The two analyses in this paper employ different estimation strategies and exploit different variations. The state level analysis estimates the impacts of disclosing information in the TRI while the facility level analysis estimates the marginal effects of disclosed information. There are several reasons that could explain the different results from the two analyses. First, the magnitudes of the “shock” (new information in relative to prior expectation) in the variations I exploited could be very different for the two analyses. In the state level analysis, facilities in the treatment group started to report to the TRI for the first time ever following the rule change in 1998. The size of the informational “shock” is probably much greater than the

Page 38: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

36

year-over-year change of reported emissions. If regulators follow a non-marginal decision heuristic and only react when the “shock” is large enough, we would expect to see the different results from the two analysis. Some research supports the argument that the “shock” needs to be large enough to spur responses from information users. For example, Konar and Cohen (1997) found that facilities whose investors experienced the largest information shock in the TRI disclosure (those had the largest abnormal returns), reduced their emissions significantly. However, largest emitters did not significantly reduce their emissions as the TRI information was largely anticipated and not shocking (they did not experience large abnormal returns). Regulators could follow a similar pattern in their responses. Even if we believe regulators do not directly make decisions following such a rule, their responses may exemplify such a pattern as a result of the response patterns from other information users. For example, large shocks can lead facilities to improve their environmental performance due to pressure from investors, which further leads to fewer regulatory activities, while small changes in information cannot. Second, the quality of TRI emissions data might be another reason that regulators do not respond to the year-over-year change of disclosed emissions. Facilities self-report emissions data to the TRI. Rather than directly measured emissions, the reported emissions are based on estimates, for which facilities have quite large latitudes in terms of the estimation inputs and methods. While the EPA does conduct quality checks and can levy penalties for violation of the EPCRA, the accuracy of the disclosed emissions often encounters skepticism. For instance, Marchi and Hamilton (2006) have found inconstancy between emissions disclosed in the TRI and emissions measured by pollution monitors, among other abnormalities of TRI emissions data. Problems in the collection of TRI data makes it an imprecise measurement of emissions. While the cross-sectional difference of disclosed emissions between facilities is more or less valid, it is unreliable to use TRI disclosure as a measurement for precise change within a facility over time. This could also explain why regulators do not react to the year-over-year change of TRI information. Third, it could be that the impacts of information disclosure policy come from the reporting instead of reported information. Much research on the housing market and survey studies on facilities’ responses to the TRI reporting support such an argument. Kohlhase (1991) found that housing prices reacted to the presence of, but not to the severity of contamination at, Superfund sites. Mastromonaco (2015) found that the housing market was not reacting to the information disclosed in the TRI, but to other implications of a firm being required to report to the TRI. In addition, survey studies have found that reporting to the TRI has allowed facilities to better understand their management of production waste and pollution (Baram & Dillon, 1992; Kraft et al., 2011; Santos et al., 1996). If this is the case, we would expect that facilities that report to the TRI would improve their environmental performance, regardless of the level of their emissions. These patterns from facilities and other information users could lead regulatory responses to follow to some degree. In addition, regulators could also have a similar pattern of direct responses to information disclosure policy. They could have certain positive/negative stereotypes about facilities that disclose information, which would affect their regulatory activities, but they do not pay attention to actual disclosed information.

Page 39: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

37

8.3 Mechanisms In the conceptual framework section, I have laid out multiple pathways that information disclosure can influence regulatory activities. While I cannot perfectly isolate these mechanisms in this study, I find evidence from this analysis, along with evidence from the literature, that is consistent with the arguments that the TRI has provided new information to regulators and that the TRI has improved facilities’ environmental performance. First, results of this analysis are consistent with the argument that the TRI has provided new information to regulators. Regulatory activities, especially inspection activities, attempt to identify violations. If the TRI provides new information to help regulators better identify potential violations, regulatory activities can be more targeted and efficient, which leads to fewer of them. In this sense, TRI disclosure is itself a type of monitoring, which substitutes some CAA regulatory activities. The reduction in regulatory activities could also be due to TRI information showing that reporting facilities have had better environmental performance than regulators’ prior knowledge. The literature and documents from the EPA support this explanation. For example, the EPA has used TRI information in comparison with data from the Air Facility System to identify facilities potentially out of compliance with their permits (U.S. EPA, 2013), and survey studies have found that federal and state regulators regularly use TRI information to understand facilities’ emissions and to assist enforcement (Kraft et al., 2011; Lynn & Kartez, 1994). Second, my results are also consistent with the argument that the TRI has improved facilities’ environmental performance, which further leads to fewer regulatory activities. TRI facilities have drastically reduced their reported emissions over time (U.S. EPA, 2005). While there is no direct evidence that this is because of the disclosing requirements and skepticism exists on the quality of TRI information, the reduction in toxic emissions, given its magnitude, indicates improvement in environmental performance to some extent. Survey studies corroborate such a conclusion as facilities consistently state that TRI reporting has helped them identify needs for reduction, meet permits requirements, and achieve abatement goals (Baram & Dillon, 1992; Kraft et al., 2011; Santos et al., 1996). Third, my results do not support the arguments that information disclosure has impacted regulatory activities by leveraging political and community responses. If TRI information has led to more community activism which forces regulators to act, we would expect the regulatory activities to increase, instead of to decrease. More importantly, while the media has paid some attention to the TRI, previous work has found that communities and individuals barely incorporated the TRI information in their activities (Bui & Mayer, 2003; Fung et al., 2007; Weil et al., 2006) and that the public has little knowledge of the TRI program (Atlas, 2007b; Kraft et al., 2011; U.S. General Accounting Office, 1991). In addition, although some studies have found the stock market and housing market reacting to the TRI, passive reactions from these venues are unlikely to translate into increases in regulatory activities. On the other hand, the scenario that communities use TRI information for private enforcement to directly target reporting facilities

Page 40: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

38

and crowding out regulatory activities is unlikely. To use disclosed TRI information to identify if a facility is in violation of other environmental regulations and initiate private enforcement would be hard to achieve for communities (if anyone has the capacity, it would be the regulators). Little evidence shows that TRI information has been successfully used in private enforcement. Most legal cases related to the TRI in LexisNexis are about communities suing facilities for not reporting (Bui, 2005). Moreover, most facilities report they have very little interaction with communities and environmental groups with regard to TRI disclosure (Kraft et al., 2011). Given that the TRI had been in place for more than 10 years at the time of the rule change, regulators should be well aware of communities’ limited use of TRI information in private enforcement and are unlikely to perceive TRI disclosure as a substitute for its monitoring and enforcement activities. 9 Conclusions This paper studies the impacts of mandatory environmental information disclosure on the implementation of traditional environmental regulations in the context of the TRI and the CAA. Using exogenous variation from TRI’s size-based disclosing requirements and the expansion of its industry coverage in 1998, I find that regulators significantly decrease regulatory activities on newly covered facilities that disclose information in the TRI program. Results from the facility level analysis, however, suggest that regulators do not react to the marginal change of disclosed information over time. This study has important implications for both information-based and traditional environmental regulations. On the one hand, it suggests that information disclosure policy can play important roles in the implementation of traditional environmental regulations. It may provide new information to regulators to enhance the efficiency of regulatory activities in traditional environmental regulations. It may also provide opportunities and incentives for facilities to improve their environmental practice by leveraging the power of other information users and allowing facilities to better understand their environmental management. As resource constraints limit the deterrence of compliance programs in traditional environmental regulations and regulators search for initiatives to improve the efficacy and efficiency of these compliance programs, information-based tools could potentially supplement their efforts. One the other hand, regulators reacting to information disclosure also suggests that regulatory response and threat could be an important mechanism for information disclosure policy to achieve their regulatory goals. This study also highlights a few issues in the design of information disclosure policy. First, it emphasizes the importance of accuracy of reported information. If the information is not credible to information users, they will not respond to it. Skepticism on the accuracy of the TRI information may have contributed to the observation that regulators do not respond to the year-over-year change of disclosed information. Second, the results that regulators do not take into consideration the marginal change of disclosed information may limit the incentives for facilities

Page 41: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

39

to reduce their emissions. If a facility cannot make significant progress to make a difference to regulators, it may simply not make any effort. Third, the results suggest the potential peril of information disclosure as a tool of greenwashing. If regulators decrease their regulatory activities on reporting facilities because of a positive stereotype they have towards transparency, reporting facilities may increase pollution either by intentionally taking advantage of information disclosure programs or because of moral licensing. While the self-reported TRI information does not support such a scenario, future research needs to directly examine the impact of information disclosure on facilities’ environmental performance using more objective measurements.

Page 42: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

40

References 42 U.S.C. §13101 et seq. (2011ed). Pollution Prevention Act. In. 42 U.S.C. § 11001 et seq. (2011ed). Emergency Planning and Community-to-Know Act. In. Abel, T. D., Stephan, M., & Kraft, M. E. (2007). Environmental Information Disclosure and Risk

Reduction among the States. State and Local Government Review, 39(3), 153-165. Alberini, A., & Segerson, K. (2002). Assessing Voluntary Programs to Improve Environmental Quality.

Environmental and Resource Economics, 22(1), 157-184. Arora, S., & Cason, T. N. (1995). An Experiment in Voluntary Environmental Regulation: Participation

in EPA′s 33/50 Program. Journal of Environmental Economics and Management, 28(3), 271-286. Atlas, M. (2007a). Enforcement principles and environmental agencies: Principal-agent relationships in a

delegated environmental program. Law and Society Review, 41(4), 939-980. Atlas, M. (2007b). TRI to Communicate: Public Knowledge of the Federal Toxics Release Inventory*.

Social Science Quarterly, 88(2), 555-572. Bae, H., Wilcoxen, P., & Popp, D. (2010). Information disclosure policy: Do state data processing efforts

help more than the information disclosure itself? Journal of Policy Analysis and Management, 29(1), 163-182.

Baram, M. S., & Dillon, P. (1992). Managing Chemical RisksCorporate Response to Sara: Revised Edition: CRC Press.

Bennear, L. S., & Olmstead, S. M. (2008). The impacts of the “right to know”: Information disclosure and the violation of drinking water standards. Journal of Environmental Economics and Management, 56(2), 117-130.

Berry, W. D., Fording, R. C., Ringquist, E. J., Hanson, R. L., & Klarner, C. E. (2010). Measuring Citizen and Government Ideology in the U.S. States: A Re-appraisal. State Politics & Policy Quarterly, 10(2), 117-135.

Bui, L. T. (2005). Public disclosure of private information as a tool for regulating environmental emissions: firm-level responses by petroleum refineries to the Toxics Release Inventory: Bureau of the Census.

Bui, L. T., & Mayer, C. J. (2003). Regulation and Capitalization of Environmental Amenities: Evidence from the Toxic Release Inventory in Massachusetts. The Review of Economics and Statistics, 85(3), 693-708.

Campa, P. (2018). Press and leaks: Do newspapers reduce toxic emissions? Journal of Environmental Economics and Management, 91, 184-202.

Chatterji, A. K., & Toffel, M. W. (2010). HOW FIRMS RESPOND TO BEING RATED. Strategic Management Journal, 31(9), 917-945.

Congressional Research Service. (2014). Federal Pollution Control Laws: How Are They Enforced? Currie, J. (2011). Inequality at birth: Some causes and consequences. American Economic Review,

101(3), 1-22. Decker, C. S. (2005). Do Regulators Respond to Voluntary Pollution Control Efforts? A Count Data

Analysis. Contemporary Economic Policy, 23(2), 180-194. Decker, C. S. (2009). Voluntary pollution control and local economic conditions as determinants of

environmental monitoring and enforcement: evidence from the 1990s. Journal of Applied Economics & Policy, 28(1), 34.

Decker, C. S., Nielsen, D. A., & Sindt, R. P. (2005). Residential Property Values and Community Right-to-Know Laws: Has the Toxics Release Inventory Had an Impact? Growth and Change, 36(1), 113-133.

Deily, M. E., & Gray, W. B. (1991). Enforcement of pollution regulations in a declining industry. Journal of Environmental Economics and Management, 21(3), 260-274.

Deily, M. E., & Gray, W. B. (2006). Agency Structure and Firm Culture: OSHA, EPA, and the Steel Industry. The Journal of Law, Economics, and Organization, 23(3), 685-709.

Page 43: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

41

Delmas, M., Montes-Sancho, M. J., & Shimshack, J. P. (2010). INFORMATION DISCLOSURE POLICIES: EVIDENCE FROM THE ELECTRICITY INDUSTRY. Economic Inquiry, 48(2), 483-498.

Delmas, M. A., & Toffel, M. W. (2008). Organizational responses to environmental demands: opening the black box. Strategic Management Journal, 29(10), 1027-1055.

Dion, C., Lanoie, P., & Laplante, B. (1998). Monitoring of pollution regulation: do local conditions matter? In (Vol. 13, pp. 5-18).

Doonan, J., Lanoie, P., & Laplante, B. (2005). Determinants of environmental performance in the Canadian pulp and paper industry: An assessment from inside the industry. Ecological Economics, 55(1), 73-84.

Doshi, A. R., Dowell, G. W. S., & Toffel, M. W. (2013). How firms respond to mandatory information disclosure. Strategic Management Journal, 34(10), 1209-1231.

Durant, R. F., Fiorino, D. J., & O'Leary, R. (2004). Environmental governance reconsidered : challenges, choices, and opportunities. Cambridge, Mass.: MIT Press.

Eisner, M. A. (2007). Governing the environment: The transformation of environmental regulation: Lynne Rienner Publishers Boulder, CO.

Fiorino, D. J. (2006). The new environmental regulation: Mit Press. Foulon, J., Lanoie, P., & Laplante, B. t. (2002). Incentives for Pollution Control: Regulation or

Information? Journal of Environmental Economics and Management, 44(1), 169-187. Friesen, L. (2003). Targeting enforcement to improve compliance with environmental regulations.

Journal of Environmental Economics and Management, 46(1), 72-85. Fung, A., Graham, M., & Weil, D. (2007). Full disclosure: The perils and promise of transparency:

Cambridge University Press. Giles, C. (2013). Next generation compliance. Paper presented at the Environmental Forum. Graham, M. (2002). Democracy by disclosure: The rise of technopopulism: Brookings Institution Press. Graham, M., & Miller, C. (2001). Disclosure of toxic releases in the United States. Environment: Science

and Policy for Sustainable Development, 43(8), 8-20. Grant, D. S., & Downey, L. (1995). Regulation through Information: An Empirical Analysis of the

Effects of State-sponsored Right-to-know Programs on Industrial Toxic Pollution. Review of Policy Research, 14(3-4), 339-352.

Gray, W. B., & Deily, M. E. (1996). Compliance and Enforcement: Air Pollution Regulation in the U.S. Steel Industry. Journal of Environmental Economics and Management, 31(1), 96-111.

Gray, W. B., & Shadbegian, R. J. (2004). ‘Optimal’ pollution abatement—whose benefits matter, and how much? Journal of Environmental Economics and Management, 47(3), 510-534.

Gray, W. B., & Shadbegian, R. J. (2005). When and Why do Plants Comply? Paper Mills in the 1980s*. Law & Policy, 27(2), 238-261.

Grooms, K. K. (2015). Enforcing the Clean Water Act: The effect of state-level corruption on compliance. Journal of Environmental Economics and Management, 73, 50-78.

Hamilton, J. T. (1995). Pollution as News: Media and Stock Market Reactions to the Toxics Release Inventory Data. Journal of Environmental Economics and Management, 28(1), 98-113.

Hamilton, J. T. (2005). Regulation through revelation: the origin, politics, and impacts of the Toxics Release Inventory Program: Cambridge University Press.

Hanna, R. N., & Oliva, P. (2010). The Impact of Inspections on Plant-Level Air Emissions. In The B.E. Journal of Economic Analysis & Policy (Vol. 10).

Harford, J. D., & Harrington, W. (1991). A reconsideration of enforcement leverage when penalties are restricted. Journal of Public Economics, 45(3), 391-395.

Harrington, W. (1988). Enforcement leverage when penalties are restricted. Journal of Public Economics, 37(1), 29-53.

Harrison, K. (1995). Is cooperation the answer? Canadian environmental enforcement in comparative context. Journal of Policy Analysis and Management, 14(2), 221-244.

Page 44: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

42

Harrison, K., & Antweiler, W. (2003). Incentives for pollution abatement: Regulation, regulatory threats, and non-governmental pressures. Journal of Policy Analysis and Management, 22(3), 361-382.

Helland, E. (1998a). The Enforcement of Pollution Control Laws: Inspections, Violations, and Self-Reporting. The Review of Economics and Statistics, 80(1), 141-153.

Helland, E. (1998b). ENVIRONMENTAL PROTECTION IN THE FEDERALIST SYSTEM: THE POLITICAL ECONOMY OF NPDES INSPECTIONS. Economic Inquiry, 36(2), 305-319.

Helland, E. (1998c). The revealed preferences of state EPAs: Stringency, enforcement, and substitution. Journal of Environmental Economics and Management, 35(3), 242-261.

Heyes, A., & Rickman, N. (1999). Regulatory dealing – revisiting the Harrington paradox1We are grateful to Lars Hansen, Robert Cairns, Jeff Frank, two referees from this journal and seminar participants at the University of London for helpful comments. The usual disclaimer applies.1. Journal of Public Economics, 72(3), 361-378.

Hilton, G. W. (1972). The basic behavior of regulatory commissions. The American Economic Review, 62(1/2), 47-54.

Innes, R., & Mitra, A. (2015). PARTIES, POLITICS, AND REGULATION: EVIDENCE FROM CLEAN AIR ACT ENFORCEMENT. Economic Inquiry, 53(1), 522-539.

Jin, G. Z., & Leslie, P. (2003). The Effect of Information on Product Quality: Evidence from Restaurant Hygiene Grade Cards*. The Quarterly Journal of Economics, 118(2), 409-451.

Joskow, P. L. (1974). Inflation and environmental concern: Structural change in the process of public utility price regulation. The Journal of Law and Economics, 17(2), 291-327.

Kalnins, A., & Dowell, G. (2015). Community Characteristics and Changes in Toxic Chemical Releases: Does Information Disclosure Affect Environmental Injustice? Journal of Business Ethics.

Khanna, M. (2001). Non-Mandatory Approaches to Environmental Protection. Journal of Economic Surveys, 15(3), 291-324.

Khanna, M., & Anton, W. R. Q. (2002). Corporate environmental management: regulatory and market-based incentives. Land Economics, 78(4), 539-558.

Khanna, M., Quimio, W. R. H., & Bojilova, D. (1998). Toxics Release Information: A Policy Tool for Environmental Protection. Journal of Environmental Economics and Management, 36(3), 243-266.

Klarner, C. (2013). State Partisan Balance Data 1937-2011. In. Kleindorfer, P. R., & Orts, E. W. (1998). Informational Regulation of Environmental Risks. Risk

Analysis, 18(2), 155-170. Kleit, A. N., Pierce, M. A., & Carter Hill, R. (1998). Environmental Protection, Agency Motivations, and

Rent Extraction: The Regulation of Water Pollution in Louisiana. Journal of Regulatory Economics, 13(2), 121-137.

Kohlhase, J. E. (1991). The impact of toxic waste sites on housing values. Journal of Urban Economics, 30(1), 1-26.

Konar, S., & Cohen, M. A. (1997). Information as regulation: The effect of community right to know laws on toxic emissions. Journal of Environmental Economics and Management, 32(1), 109-124.

Konisky, D. M. (2007). Regulatory competition and environmental enforcement: Is there a race to the bottom? American Journal of Political Science, 51(4), 853-872.

Konisky, D. M. (2009). Inequities in enforcement? Environmental justice and government performance. Journal of Policy Analysis and Management, 28(1), 102-121.

Konisky, D. M., & Reenock, C. (2013). Compliance Bias and Environmental (In)Justice. Journal of Politics, 75(2), 506-519.

Konisky, D. M., & Reenock, C. (2018). Regulatory Enforcement, Riskscapes, and Environmental Justice. Policy Studies Journal, 46(1), 7-36.

Kraft, M. E., Abel, T. D., & Stephan, M. (2004). Information disclosure and risk reduction: The sources of varying state performance in control of toxic chemical emissions. Paper presented at the Conference on Corporate Environmental Behavior and the Effectiveness of Government Interventions, US EPA National Center for Environmental Economics, Washington, DC, April.

Page 45: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

43

Kraft, M. E., Stephan, M., & Abel, T. D. (2011). Coming clean: information disclosure and environmental performance: MIT Press.

Langpap, C. (2007). Pollution abatement with limited enforcement power and citizen suits. Journal of Regulatory Economics, 31(1), 57-81.

Langpap, C., & Shimshack, J. P. (2010). Private citizen suits and public enforcement: Substitutes or complements? Journal of Environmental Economics and Management, 59(3), 235-249.

Leaver, C. (2009). Bureaucratic Minimal Squawk Behavior: Theory and Evidence from Regulatory Agencies. American Economic Review, 99(3), 572-607.

Loewenstein, G., Sunstein, C. R., & Golman, R. (2014). Disclosure: Psychology Changes Everything. Annual Review of Economics, 6(1), 391-419.

Lynn, F. M., & Kartez, J. D. (1994). Environmental democracy in action: The Toxics Release Inventory. Environmental Management, 18(4), 511-521.

Magat, W. A., & Viscusi, W. K. (1990). Effectiveness of the EPA's regulatory enforcement: The case of industrial effluent standards. The Journal of Law and Economics, 33(2), 331-360.

Marchi, S. d., & Hamilton, J. T. (2006). Assessing the Accuracy of Self-Reported Data: an Evaluation of the Toxics Release Inventory. Journal of Risk and Uncertainty, 32(1), 57-76.

Mastromonaco, R. (2015). Do environmental right-to-know laws affect markets? Capitalization of information in the toxic release inventory. Journal of Environmental Economics and Management, 71, 54-70.

May, P. J. (2005). Regulation and Compliance Motivations: Examining Different Approaches. Public Administration Review, 65(1), 31-44.

Naysnerski, W., & Tietenberg, T. (1992). Private Enforcement of Federal Environmental Law. Land Economics, 68(1), 28-48.

Niskanen, W. (1971). Bureaucracy and representative democracy. Chicago und New York. Oberholzer-Gee, F., & Mitsunari, M. (2006). Information regulation: Do the victims of externalities pay

attention? Journal of Regulatory Economics, 30(2), 141-158. Oljaca, N., Keeler, A. G., & Dorfman, J. (1998). Penalty Functions for Environmental Violations:

Evidence from Water Quality Enforcement. Journal of Regulatory Economics, 14(3), 255-264. Patten, D. M. (1998). The impact of the EPA's TRI disclosure program on state environmental and natural

resource expenditures. Journal of Accounting and Public Policy, 17(4), 367-382. Peltzman, S. (1976). Toward a more general theory of regulation. The Journal of Law and Economics,

19(2), 211-240. Posner, R. A. (1974). Theories of Economic Regulation. Bell Journal of Economics and Management

Science, 5, 335. Saha, S., & Mohr, R. D. (2013). Media attention and the Toxics Release Inventory. Ecological

Economics, 93, 284-291. Sanders, N. J. (2014). The response to public information on environmental amenities: New evidence

housing markets care about the toxics release inventory. Unpublished manuscript, 6, 2014. Santos, S. L., Covello, V. T., & McCallum, D. B. (1996). Industry Response to SARA Title III: Pollution

Prevention, Risk Reduction, and Risk Communication. Risk Analysis, 16(1), 57-66. Scholz, J. T., Twombly, J., & Headrick, B. (1991). Street-Level Political Controls over Federal

Bureaucracy. American Political Science Review, 85(03), 829-850. Scholz, J. T., & Wang, C.-L. (2006). Cooptation or Transformation? Local Policy Networks and Federal

Regulatory Enforcement. American Journal of Political Science, 50(1), 81-97. Scholz, J. T., & Wei, F. H. (1986). Regulatory enforcement in a federalist system. American Political

Science Review, 80(04), 1249-1270. Scorse, J. (2000). Does being a “Top 10” worst polluter affect facility environmental releases? Evidence

from the US toxic release inventory. Atlantic Monthly. Shapiro, M. D. (2005). Equity and information: Information regulation, environmental justice, and risks

from toxic chemicals. Journal of Policy Analysis and Management, 24(2), 373-398.

Page 46: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

44

Shimshack, J. P., & Ward, M. B. (2005). Regulator reputation, enforcement, and environmental compliance. Journal of Environmental Economics and Management, 50(3), 519-540.

Shimshack, J. P., & Ward, M. B. (2008). Enforcement and over-compliance. Journal of Environmental Economics and Management, 55(1), 90-105.

Silverman, S. L. (1990). Federal enforcement of environmental laws. Massachusetts Law Review, 75(1), 95-98.

Stephan, M. (2002). Environmental Information Disclosure Programs: They Work, but Why? Social Science Quarterly, 83(1), 190-205.

Stigler, G. J. (1971). The theory of economic regulation. The Bell journal of economics and management science, 3-21.

Tietenberg, T. (1998). Disclosure Strategies for Pollution Control. Environmental and Resource Economics, 11(3), 587-602.

Tietenberg, T., & Wheeler, D. (2001). Empowering the community: Information strategies for pollution control. Frontiers of environmental economics, 85-120.

U.S. EPA. (1991a). Clean Air Act Stationary Source Civil Penalty Policy U.S. EPA. (1991b). Toxics in the Community: National and Local Perspectives. U.S. EPA. (2001). Toxics Release Inventory 1999. U.S. EPA. (2005). 2003 TRI National Analysis. U.S. EPA. (2012). 2010 TRI National Analysis. U.S. EPA. (2013). The Toxics Release Inventory in Action: Media, Government, Business, Community

and Academic Uses of TRI Data. Retrieved from www.epa.gov/toxics-release-inventory-tri-program/toxics-release-inventory-action-media-government-business

U.S. EPA. (2014). Revision of U.S. Environmental Protection Agency's Enforcement Response Policy for High Priority Violations of the Clean Air Act: Timely and Appropriate Enforcement Response to High Priority Violations- 2014.

U.S. EPA. (2016). Clean Air Act Stationary Source Compliance Monitoring Strategy. U.S. EPA. (n.d.-a). History of the Toxics Release Inventory (TRI) Program. Retrieved from

https://www.epa.gov/toxics-release-inventory-tri-program/history-toxics-release-inventory-tri-program-list

U.S. EPA. (n.d.-b). Inventory of TRI Community Outreach and Education Materials, 1989-2012. Retrieved from www.epa.gov/toxics-release-inventory-tri-program/inventory-tri-community-outreach-and-education-materials-1989

U.S. General Accounting Office. (1991). Toxic Chemicals: EPA’s Toxic Release Inventory is Useful But Can be Improved.

U.S. OMB. (2005). Program Assessment Rating Tool Review of EPA Enforcement of Environmental Laws.

Weil, D., Fung, A., Graham, M., & Fagotto, E. (2006). The effectiveness of regulatory disclosure policies. Journal of Policy Analysis and Management, 25(1), 155-181.

Weil, D., Graham, M., & Fung, A. (2013). Targeting Transparency. Science, 340(6139), 1410-1411. Wood, B. D. (1992). Modeling Federal-Implementation as a System - the Clean-Air Case. American

Journal of Political Science, 36(1), 40-67.

Page 47: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

45

Appendix

Table A1. DID Estimates for Inspection Activities (Full Results)

(1) (2) (3) (4) (5) (6) Ln (Insp.) Ln (Insp.) Ln (Insp.) Insp. Insp. Insp.

Post x TRI -0.27*** -0.27*** -0.27** -25.53*** -26.03*** -25.53*** (0.07) (0.08) (0.10) (6.56) (6.71) (8.98)

TRI -0.64*** -0.62*** -0.64*** -25.28*** -25.08*** -25.28*** (0.09) (0.09) (0.12) (4.65) (4.76) (6.37)

Unemployment (%) 0.03 -2.85 (0.06) (10.61)

Fiscal Health 0.18 6.50 (0.35) (28.08)

Repub. Governor -0.13 -48.35** (0.18) (22.63)

Repub. Legislature (%) 2.18 7.55 (1.42) (153.87)

Ln (Personal Inc.) -0.7 -8.03 (2.34) (311.13)

Ln (Population) 3.43 69.38 (2.69) (315.81)

Government Ideology -0.01 -2.33** (0.01) (0.98)

Constant 4.72*** -41.20 2.90*** 83.50*** -732.45 27.64*** (0.10) (47.23) (0.06) (5.52) (7429.57) (3.18)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 882 864 882 882 864 882 R2 0.79 0.8 0.93 0.70 0.70 0.90

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Inspection +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01.

Page 48: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

46

Table A2. DID Estimates for Enforcement Activities (Full Results) (1) (2) (3) (4) (5) (6) Ln (Enf.) Ln (Enf.) Ln (Enf.) Enf. Enf. Enf.

Post x TRI -0.30*** -0.29*** -0.30** -4.44** -4.46** -4.44 (0.11) (0.11) (0.15) (1.97) (2.02) (2.70)

TRI -0.39*** -0.39*** -0.39** -2.78** -2.79* -2.78 (0.11) (0.11) (0.15) (1.37) (1.41) (1.88)

Unemployment (%) 0.14*** 1.28 (0.05) (0.78)

Fiscal Health -0.36 -4.19 (0.35) (4.75)

Repub. Governor -0.20 -4.57 (0.17) (3.71)

Repub. Legislature (%) 0.21 -7.55 (1.13) (25.97)

Ln (Personal Inc.) 3.04* 21.52 (1.53) (16.18)

Ln (Population) 2.59 18.95 (1.59) (18.97)

Government Ideology 0.00 -0.22 (0.01) (0.23)

Constant 1.28*** -69.10** 2.06*** 5.73*** -487.06 11.89*** (0.09) (27.82) (0.07) (1.54) (377.92) (0.94)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 882 864 882 882 864 882 R2 0.66 0.67 0.82 0.51 0.51 0.69

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Enforcement +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01.

Page 49: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

47

Table A3. Triple DID Estimates for Inspection Activities (Full Results) (1) (2) (3) (4) (5) (6) Ln (Insp.) Ln (Insp.) Ln (Insp.) Insp. Insp. Insp.

Post x New x TRI -0.44*** -0.45*** -0.44*** -91.40*** -93.20*** -91.40*** (0.09) (0.09) (0.10) (28.60) (29.20) (32.50)

Post x New 0.37*** 0.38*** 0.37*** -11.49 -12.50 -11.49 (0.05) (0.06) (0.06) (17.01) (17.38) (19.33)

Post x TRI 0.17*** 0.18*** 0.17** 65.87** 67.17** 65.87** (0.06) (0.06) (0.07) (27.63) (28.24) (31.40)

New x TRI -0.43*** -0.42*** -0.43*** 40.63 41.48 40.63 (0.11) (0.11) (0.13) (28.38) (29.02) (32.25)

New -1.37*** -1.37*** -1.37*** -206.88*** -209.26*** -206.88*** (0.12) (0.12) (0.14) (53.80) (55.00) (61.14)

TRI -0.21*** -0.20** -0.21** -65.91** -66.56** -65.91* (0.07) (0.08) (0.08) (29.09) (29.76) (33.06)

Unemployment (%) 0.03 6.78 (0.06) (12.41)

Fiscal Health 0.01 14.75 (0.32) (49.76)

Repub. Governor -0.15 -110.88** (0.18) (53.01)

Repub. Legislature (%) 1.82 213.55 (1.27) (263.56)

Ln (Personal Inc.) -0.92 -111.67 (2.13) (460.53)

Ln (Population) 2.79 372.84 (2.28) (588.96)

Government Ideology -0.01 -5.66** (0.01) (2.58)

Constant 6.51*** -27.22 3.08*** 550.88*** -3764.99 136.24*** (0.10) (40.52) (0.07) (43.09) (10491.13) (38.05)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 1764 1728 1764 1764 1728 1764 R2 0.81 0.81 0.89 0.56 0.56 0.64

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Inspection +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01.

Page 50: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

48

Table A4. Triple DID Estimates for Enforcement Activities (Full Results) (1) (2) (3) (4) (5) (6) Ln (Enf.) Ln (Enf.) Ln (Enf.) Enf. Enf. Enf.

Post x New x TRI -0.43*** -0.44*** -0.43*** -10.35*** -10.66*** -10.35** (0.13) (0.13) (0.14) (3.69) (3.77) (4.20)

Post x New 0.34*** 0.36*** 0.34*** 1.52 1.64 1.52 (0.09) (0.09) (0.10) (3.15) (3.22) (3.58)

Post x TRI 0.13* 0.15** 0.13 5.91** 6.20** 5.91* (0.07) (0.07) (0.08) (2.71) (2.76) (3.08)

New x TRI -0.42*** -0.40*** -0.42** -4.73 -4.67 -4.73 (0.14) (0.14) (0.16) (6.17) (6.31) (7.01)

New -1.47*** -1.49*** -1.47*** -31.50*** -32.03*** -31.50*** (0.15) (0.15) (0.17) (6.48) (6.60) (7.36)

TRI 0.03 0.02 0.03 1.95 1.88 1.95 (0.09) (0.09) (0.11) (5.59) (5.72) (6.35)

Unemployment (%) 0.12** 3.43** (0.05) (1.64)

Fiscal Health -0.25 -4.95 (0.27) (8.32)

Repub. Governor -0.23 -6.70 (0.14) (4.64)

Repub. Legislature (%) 0.80 14.87 (0.94) (32.83)

Ln (Personal Inc.) 2.27 -14.88 (1.38) (31.70)

Ln (Population) 1.34 -3.13 (2.07) (40.51)

Government Ideology 0.00 -0.37 (0.01) (0.33)

Constant 3.42*** -40.27 2.53*** 48.71*** 248.66 23.95*** (0.09) (35.22) (0.09) (5.87) (707.97) (4.79)

Year Fixed Effects X X X X State Fixed Effects X X X X State by Year Controls X X State by Year Fixed Effects X X N 1764 1728 1764 1764 1728 1764 R2 0.75 0.75 0.81 0.52 0.53 0.61

Note: (1) The actual dependent variable used in regressions (1) – (3) is Ln (Enforcement +1). (2) All standard errors are clustered at state level and in parentheses. (3) * p < .10, ** p < .05, *** p < .01.

Page 51: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

49

Table A5. Facility Level Estimates for Inspection Activities (Full Results) (1) (2) (3) (4) Insp. Insp. Insp. Insp.

Ln (Disclosed Air Releases) 0.008 (0.006)

Ln (Disclosed Land Releases) -0.010 (0.011)

Ln (Disclosed Water Releases) 0.019 (0.015)

Ln (Disclosed Total Releases) 0.011 (0.007)

High Priority Violator 0.493*** 0.494*** 0.494*** 0.493*** (0.170) (0.170) (0.170) (0.170)

Unemployment (%) 0.080* 0.080* 0.080* 0.080* (0.046) (0.046) (0.046) (0.046)

Ln (County Personal Inc.) 1.493 1.498 1.499 1.490 (0.927) (0.927) (0.927) (0.926)

Ln (County Pop.) -1.272 -1.277 -1.280 -1.273 (0.918) (0.919) (0.919) (0.917)

Constant 0.816 0.893 0.882 0.836 (8.321) (8.361) (8.342) (8.324)

Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 66759 66759 66759 66759 R2 0.586 0.586 0.586 0.586

Note: (1) All standard errors are clustered at state level and in parentheses. (2) * p < .10, ** p < .05, *** p < .01.

Page 52: Truth Hurts? Mandatory Information Disclosure and ...based tools, understanding the implications of information disclosure for the implementation of traditional environmental regulations

50

Table A6. Facility Level Estimates for Enforcement Activities (Full Results) (1) (2) (3) (4) Enf. Enf. Enf. Enf.

Ln (Disclosed Air Releases) 0.001 (0.002)

Ln (Disclosed Land Releases) 0.000 (0.003)

Ln (Disclosed Water Releases) 0.009** (0.004)

Ln (Disclosed Total Releases) 0.001 (0.002)

High Priority Violator 1.075*** 1.075*** 1.075*** 1.075*** (0.083) (0.083) (0.083) (0.083)

Unemployment (%) 0.025 0.025 0.025 0.025 (0.021) (0.021) (0.021) (0.021)

Ln (County Personal Inc.) -0.038 -0.037 -0.036 -0.038 (0.189) (0.190) (0.188) (0.189)

Ln (County Pop.) -0.273* -0.273* -0.275* -0.273* (0.140) (0.140) (0.140) (0.140)

Constant 3.658 3.670 3.665 3.665 (2.526) (2.522) (2.527) (2.525)

Facility Fixed Effects X X X X State by Year Fixed Effects X X X X N 66759 66759 66759 66759 R2 0.575 0.575 0.575 0.575

Note: (1) All standard errors are clustered at state level and in parentheses. (2) * p < .10, ** p < .05, *** p < .01.