Financial Constraints and Corporate
Environmental Policies
Taehyun Kim
Chung-Ang University
This paper documents evidence that financial constraints increase firms’ toxic emissions
given that firms actively trade off abatement costs against potential legal liabilities.
Exploring three quasi-natural experiments in which firms’ financial resources are likely
exogenously affected, we find that relaxing financial constraints reduces U.S. public firms’
toxic releases. The effects of financial constraints on toxic releases are amplified when
regulatory enforcement and external monitoring weaken. Overall, our evidence highlights
the real effects of financial constraints in the form of environmental pollution, which is a
costly negative externality imposed on society and public health. (JEL G32, G38, K32,
Q50)
Received March 25, 2019; editorial decision February 19, 2021 by Editor Wei Jiang. Authors
have furnished an Internet Appendix, which is available on the Oxford University Press
Web site next to the link to the final published paper online.
In modern production processes, firms often generate byproducts that have
adverse impacts on the environment and public health. In 2015 alone, U.S. firms
produced 27.24 billion pounds of toxic chemicals generated in productionrelated processes. Researchers have documented costly adverse outcomes
from exposure to pollutants and toxicants, including infant mortality and
neurodevelopment disorders, lower education attainment, reduced labor force
participation, and lower earnings in later life (for a review, see Currie et al.
2014). A better understanding of firms’ environmental decisions and how they
We are grateful for the thoughtful comments from two anonymous referees and the guidance of the editor, Wei
Jiang. We thank Kee-Hong Bae, Paulo Fulghieri, Nandini Gupta, April Knill, and Paul Schultz and seminar
participants at the KCMI-KAFA Symposium, the University of Notre Dame, the 2017 Wabash River Finance
Conference, the 2017 HKUST Finance Symposium, and the 2018 Western Finance Association Conference
for valuable comments and suggestions. We also thank Andriy Bodnaruk, Gerald Hoberg, Tim Loughran, Max
Maksimovic, and Bill McDonald for making their textual financial-constraint measures available to us; Dong
Lou for sharing mutual fund flow-induced trade data; Justin Mohr, David Yeh, and Pete Pietraszewski for
assistance with the data; and Tim Antisdel from the EPA for answering questions about the TRI Program. Send
correspondence to Qiping Xu, qipingxu@illinois.edu.
The Review of Financial Studies 35 (2022) 576–635
© The Author(s) 2021. Published by Oxford University Press on behalf of The Society for Financial Studies.
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
doi:10.1093/rfs/hhab056
Advance Access publication May 5, 2021
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 576
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Qiping Xu
University of Illinois Urbana-Champaign
Financial Constraints and Corporate Environmental Policies
1 In 2005, the last year for which we can access official data, U.S. manufacturers spent over $26.57 billion on
pollution abatement, which is approximately 1% of the manufacturing sector’s shipment value, or more than
20% of total capital expenditure.
2 Karpoff, Lott, and Wehrly (2005) examine the size of losses in market value in companies that violate
environmental regulations and show that the losses are similar in magnitude to the legal liabilities imposed.
577
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 577
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
connect to financial market frictions and regulatory settings will inform more
fruitful discussions of environmental protections.
Environmental abatement is exorbitantly expensive because it requires
substantial inputs of energy, labor, contracted services, and raw materials
in a process deeply integrated into every aspect of corporate decisionmaking.1 Under the current U.S. regulatory regime, firms are required
by law to partially internalize environmental costs by allocating resources
for environmental protection. The U.S. Environmental Protection Agency
(EPA) works with federal, state, and local authorities to ensure compliance
with environmental regulations by enforcing penalties and sanctions upon
confirmation of violations.
We derive firms’ optimal environmental decisions within a value-maximizing
net present value (NPV) framework, under which firms actively trade off the
present value of abatement expenditures against the present value of expected
legal liabilities. Conditional on production output, the total volume of toxic
releases captures the pollution emission intensity, measured as pollution emitted
per unit of output (Copeland and Taylor 2003; Shapiro and Walker 2018).
Abatement expenditures reduce pollution emission intensity and consequently
firms’ expected legal liabilities. The optimal environmental abatement
expenditures presuppose that the marginal cost of abatement equals the
marginal reduction in expected legal liabilities (Shapira and Zingales 2017).2
In this paper, we apply this fundamental NPV framework to examine
how financial frictions, in particular financial constraints, affect corporate
environmental policies. As financial constraints unveil and drive up the
cost of financing, the marginal cost of environmental abatement increases
correspondingly. Holding other factors constant, financial constraints reduce
firms’ abatement activities and consequently increase total toxic releases.
Exploring the Toxics Release Inventory (TRI) establishment-level microdata
from the EPA for the period running from 1990 through 2014, we first
show economically large and statistically significant correlations between
volumes of toxic chemicals released and two text-based financial-constraint
measures for public firms in the United States recently developed by
Hoberg and Maksimovic (2014) and Bodnaruk, Loughran, and McDonald
(2015). These text-based measures extract qualitative information about
financial constraints from corporate disclosure documents and interpret
this vast information source using well-defined algorithms. Our results are
robust to controls for production level, overall capital expenditures, firm
financial characteristics, commonly used accounting-based financial-constraint
The Review of Financial Studies / v 35 n 2 2022
578
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 578
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
measures, and alternative categorizations of toxic releases under various EPA
regulations. In terms of economic magnitude, a one-standard-deviation increase
in the textual financial-constraint measure is associated with a 4% increase in
total toxic releases for an average establishment in the sample.
To establish the causal impact of financial constraints on firms’
environmental policies, we explore three quasi-natural experiments to generate
plausibly exogenous shocks to firms’ available financial resources. In the first
experiment, we use the 2004 American Jobs Creation Act (AJCA), which
created a positive cashflow windfall by lowering the repatriation tax rate for
firms that repatriated foreign earnings previously held by foreign subsidiaries
(Faulkender and Petersen 2012). In the second experiment, we exploit the
collateral value of firms’ real estate assets, where higher collateral value reduces
lending frictions and therefore facilitates financing through higher debt capacity
(Chaney, Sraer, and Thesmar 2012). In the third experiment, we use mutual fund
flow-induced price pressure (FIPP), where large inflows generate temporary
price appreciations and induce seasoned equity offerings (SEOs) (Lou 2012;
Khan, Kogan, and Serafeim 2012). Given the exogenous nature of these shocks,
they are unrelated to firm fundamentals and hence unlikely to be associated
with firms’ environmental policies other than through financing costs. We find
consistent results across the three experiments that relaxing financial constraints
reduces discharges of toxic chemicals. These three experiments result in total
toxic release reductions ranging from 8% to 18%, with the average effect being
approximately 14%.
After establishing causality, we turn to establishing the link between toxic
releases and legal liabilities, and investigate the impact of toxic releases
on firm value. We study legal liabilities by compiling information on
administrative, civil, and judicial cases filed by government agencies under
various environment statues. Our results reveal that higher total toxic releases
make government agency investigations more likely, increase the likelihood
of legal liabilities imposed consequently, and make legal liabilities (including
federal and local penalties, compliance and recovery costs, and the costs of
supplemental environmental projects) costlier. Legal enforcement activities
represent additional operating costs, reducing firms’ net income. Our analysis
shows that greater toxic releases predict worse operating performance (lower
returns on assets (ROA) and smaller profit margins) as well as lower market
valuation (Tobin’s q). The reported evidence substantiates the underlying tradeoffs between abatement spending and potential legal liabilities faced by firms
when making environmental decisions.
Pollution is a costly negative externality imposed on society and public
health. The natural environment is a public good given that clean air, water,
and land are shared by all. The absence of clearly defined private ownership
determines market failure with environmental abatement, because firms that
pollute do not bear the full costs associated with pollution under the current
U.S. environmental regulatory scheme. Therefore, the marginal costs of
Financial Constraints and Corporate Environmental Policies
579
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 579
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
environmental abatement that firms incur are significantly lower than the
marginal social cost, and this difference drives a wedge between the level of
abatement that is optimal for firms and the level preferred by society. Reduction
in pollution abatement, driven by financial constraints, thus generates further
resource misallocation from society’s point of view.
We conduct a back-of-the-envelope calculation to show the additional costs
that were imposed on society by firms’ abatement reduction when driven by
financial constraints, using the average effect identified across our three quasinatural experiments (14%). The EPA prospective report on the Clean Air Act
(CAA, 1990–2020) estimates that every dollar by which firms cut back on
environmental abatement generates a welfare loss of $60 to the public and
society. We apply a linear extrapolation assuming that abatement expenditures
are cut by 14%, representing a $7.14 million reduction for an average firm-year
in our sample. For toxic chemicals governed by the CAA alone, this assumption
imposes an additional $428 (=$7.14 million*60) million cost on society. This
estimate translates to around an $8 billion additional welfare loss to society
for all establishments during our sample period. The total welfare costs across
all other environmental regulations and statues and across all other firms and
establishments in the United States would be significantly higher. Our estimates
indicate that financial constraints exacerbate the costly negative externalities
of environmental pollution.
Lastly, we explore a number of cross-sectional variations in the inner
workings of financial constraints and corporate environmental policies. In
our first set of cross-sectional tests, we focus on heterogeneity in regulatory
environments. If a certain geographic region is designated as a “nonattainment”
zone by the EPA, environmental laws and regulations mandate enhanced
monitoring and have costly ramifications. Our empirical analysis confirms
that more stringent regulations in nonattainment counties increase legal
liabilities through a higher likelihood of legal actions and more costly
enforcement resolutions. We show that regulatory strictness affects firms’
ongoing environmental abatement as well as the sensitivity of toxic releases to
financial constraints. Establishments located in nonattainment areas on average
reduce toxic releases by about 30% relative to their intrafirm peers located in
attainment areas. Furthermore, as financial constraints increase the marginal
cost of abatement expenditures, firms shift financial resources to nonattainment
areas where additional toxic releases might incur large regulatory penalties
and further reduce abatement in attainment areas with looser environmental
regulations. In our second set of cross-sectional tests, we study polluter sizes
where large polluters are typically the focus of EPA monitoring and enforcement
activities. We find that small polluters are much more responsive to financial
constraints, whereas large polluters display lower sensitivity given the presence
of higher expected legal liabilities. Overall, the managers in our sample are
opportunistic: they are keenly aware of the costs and benefits associated with
environmental abatement and strategically choose when and where to pollute.
The Review of Financial Studies / v 35 n 2 2022
3 See Margolis, Elfenbein, and Walsh (2009), Bénabou and Tirole (2010), and Kitzmueller and Shimshack (2012)
for a comprehensive review of the literature.
580
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 580
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Our paper contributes to the literature on the real effects of financial
constraints. Our evidence complements and extends earlier work focusing on
outcomes within the scope of a firm, such as investment and employment
activities (e.g., Baker, Stein, and Wurgler 2003; Almeida and Campello 2007;
Campello, Graham, and Harvey 2010; Chaney, Sraer, and Thesmar 2012;
Chodorow-Reich 2013). In contrast, our paper focuses on environmental
pollution as the key outcome variable, which by its nature is a costly
negative externality imposed on society and public health. We demonstrate
that firms carefully evaluate the private costs and benefits associated with the
implementation of environmental policies. As a result, financial constraints
amplify the negative externality of firms’ abatement decisions and impose
additional costs on society. In addition, our paper relates to the discussion of
how financial market frictions affect corporate environmental activities (e.g.,
Masulis and Reza 2015; Cheng, Hong, and Shue 2016; Shapira and Zingales
2017; Fernando, Sharfman, and Uysal 2017; Starks, Venkat, and Zhu 2017;
Akey and Appel 2021).
Our paper complements the work on corporate environmental decisions.3
The vast majority of these studies exploit firm-level variation using the
Kinder, Lydenberg, and Domini (KLD) ratings, which cannot precisely capture
intrafirm resource allocations for environmental protection. In comparison, we
focus on establishment-level toxic releases and utilize a set of well-defined and
high-quality performance metrics using a large panel. Having such granular
data available enables us to exploit establishment-level variation to gauge the
effects of important factors, such as regulatory mechanisms, which is essential
for a better understanding of firms’ environmental decisions.
Our paper also contributes to the debate over “doing well by doing good”
versus “doing good by doing well.” The formal view juxtaposes corporate
environmental policies and corporate financial performance. The idea is
that when firms act as “responsible” partners with the environment and
other nonfinancial stakeholders, their bottom lines benefit (e.g., Baron 2001;
Hong and Kacperczyk 2009; Edmans 2011; Eccles, Ioannou, and Serafeim
2014; Dimson, Karakaş, and Li 2015; Dunn, Fitzgibbons, and Pomorski
2017, Lins, Servaes, and Tamayo 2017; Hoepner et al. 2018; Albuquerque,
Koskinen, and Zhang 2019). Implementing environmental protection is
costly, however, and “doing good” may imply sacrificing shareholder value
(Barber, Morse, and Yasuda 2021). We show that, especially when regulatory
forces are weak, actions that can be construed as “doing well” seem to dominate
the motives underlying “doing good” (Hong, Kubik, and Scheinkman 2012;
Cohn and Wardlaw 2016; Andersen 2017).
Financial Constraints and Corporate Environmental Policies
1. Institutional Background and Data
4 For a complete list of laws and executive orders, please see https://www.epa.gov/laws-regulations.
5 The EPA provides sector-by-sector environmental statutes and regulations (https://www.epa.gov/regulatory-
information-sector).
581
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 581
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
1.1 Institutional background
Born amid elevated concern about environmental pollution, the EPA was
founded in 1970 to consolidate a variety of federal research, monitoring,
standard-setting, and enforcement activities into one agency. A number of
laws serve as the EPA’s foundation for protecting the environment and public
health, with several presidential executive orders also playing a central role.
Congress authorizes the EPA to develop and enact regulations as well as to
explain the critical details regarding the steps that are necessary to implement
environmental laws. Panel A in Figure 1 illustrates how EPA rules regulate
some products used and produced in the manufacturing process.4 For instance,
the CAA focuses on hazardous air pollutants, including lead, sulfur oxides,
and nitrogen oxides, and chemicals capable of harming the stratospheric ozone
layer, such as halons and methyl chloroform; the Clean Water Act (CWA)
focuses on sources of water contamination, such as fertilizers and pesticides,
and other naturally occurring chemicals and minerals (i.e., radon, uranium); and
the Comprehensive Environmental Response, Compensation, and Liability Act
(CERCLA), commonly known as the Superfund, was enacted to fund efforts to
clean up abandoned or uncontrolled hazardous waste sites. Other government
agencies also coordinate with the EPA on environmental regulatory issues.
For example, the Occupational Safety and Health Administration (OSHA)
has separate enforcement standards to ensure workplace safety and health,
targeting known or suspected carcinogens, such as vinyl chloride, benzene,
and formaldehyde.
A manufacturing facility typically uses multiple chemicals and emits toxic
chemicals through many types of media, such as air, water, and land, so it
needs to abide by multiple federal and local regulations. Facilities across
industries might have industry-specific guidelines based on the nature of the
production process.5 In addition, state and local environmental regulations take
overlapping forms with the federal ones. The EPA works closely with state and
local authorities to ensure compliance with environmental regulations through
conducting routine inspections and investigations, and enforcing penalties and
sanctions on confirmation of violations. These agencies have the authority
to pursue civil administrative (nonjudicial enforcement) actions directing
responsible entities to come into compliance or clean up a site, either with
or without penalties. For entities that have failed to comply with regulatory
requirements or administrative orders, civil trials and penalties are sought
by filing charges through the Department of Justice. For the most serious
violations that are committed knowingly, criminal actions are pursued where
The Review of Financial Studies / v 35 n 2 2022
A
C
Figure 1
Environmental Protection Agency
Panel A showcases several Environmental Protection Agency (EPA) regulations that govern toxic releases in the
United States. Panel B illustrates EPA waste management guidelines, with disposal or other releases being the
least preferred method. Panel C presents abatement expenditure categories from the 2005 Pollution Abatement
Costs and Expenditures (PACE) survey (the most recent available) conducted by the U.S. Census Bureau in the
manufacturing sector.
Source: United States Environmental Protection Agency.
court convictions can result in fines or even incarceration. In summary, legal
enforcement actions can conclude with settlements through administrative or
judicial actions, civil and criminal penalties, or injunctive relief (requirements to
perform designated actions). In addition, supplemental environmental projects,
where the violator voluntarily agrees to provide tangible environmental or
public health benefits to affected communities or the environment, also can
be involved in an enforcement settlement.
Under the current U.S. environmental regulatory system, legal liabilities
represent an important factor that firms consider in their environmental
decisions. Of the 197,476 investigations initiated by government agencies
582
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 582
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
B
Financial Constraints and Corporate Environmental Policies
6 Coal prices differ by rank and grade. According to the U.S. Energy Information Administration, in 2019 the
average annual sale price of coal per ton (2,000 pounds) ranged from $19.86 for lignite coal to $58.93 for
bituminous coal to $102.22 for anthracite coal (https://www.eia.gov/coal/annual/). As one might expect, lignite
and bituminous coal generate much more sulfur dioxide and smog than anthracite coal.
7 The definition excludes the use of these materials as fuel substitutes or for energy production (National Recycling
Coalition 1995).
8 Energy recovery is often associated with electricity generation, although it can also offset fossil fuels used at
industrial sites.
583
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 583
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
during the period of 1990–2014, 111,808 resulted in an average legal liability
of $6.75 million per case. When applicable, the average penalty amounted
to $270,000, with an average cost of recovery (to stabilize and/or clean up
Superfund sites) of $21 million, compliance costs (the sum of injunctive relief
and the physical or nonphysical costs of returning to compliance) of $14 million,
and supplemental environmental project costs of $842 thousand.
Panel B of Figure 1 presents the so-called “Waste Management Hierarchy,”
which intuitively outlines EPA’s waste management guidelines. Source
reduction involves maximizing or reducing the use of natural resources at
the beginning of an industrial process, thereby reducing the waste produced
by the process. Source reduction is the EPA’s preferred method of waste
management. Consider a typical coal-fired plant. To minimize its environmental
impact, the plant can use higher-grade (or cleaner) coal that reduces pollutants
from the beginning, but doing so significantly increases production costs.6
Recycling involves a series of activities through which discarded materials
are collected, sorted, processed, and converted into raw materials and used in
the production of new products.7 Energy recovery is the process of generating
energy from the combustion of wastes, including at waste-to-energy combustion
facilities and landfill-gas-to-energy facilities.8 Treatment involves the use of
various processes, such as incineration or oxidation, to alter the properties
or composition of hazardous materials. Direct disposal is the least preferred
method of waste management. Unsurprisingly, direct disposal is often the least
costly method from a firm’s perspective. The tension between environmental
protection and a firm’s bottom line lies precisely in the fact that the EPApreferred waste management methods are more costly, while less preferred
methods are more harmful to the environment.
In panel C of Figure 1, we decompose the cost categories for abatement
expenditures according to the 2005 EPA Pollution Abatement Costs and
Expenditures (PACE) survey summary for the manufacturing sector. Pollution
abatement operating costs amounted to $20,677.6 million in 2005 across all
industries, of which $2,848.4 million (14%) was attributed to depreciation.
All new capital expenditures amounted to $128,325.2 million, of which
only $5,907.8 million (4.6%) was attributed to pollution abatement capital
expenditures. In contrast, expenditures associated with energy, contract work,
labor, and materials and supplies make up the vast majority of abatement
The Review of Financial Studies / v 35 n 2 2022
1.2 Data
Our main source of data is the establishment-level TRI program administered
by the EPA. Any facility in the United States that falls within a TRIreportable industry sector, has ten or more employees, and cross a certain
threshold in manufactured or processed TRI-listed chemicals is required
9 The data come from PACE: 2005, published by the U.S. Census Bureau and available from
https://www.census.gov/prod/2008pubs/ma200-05.pdf.
10 The only comprehensive establishment-level data information available on waste management activities is the
Pollution Abatement Costs and Expenditures (PACE) survey, which records abatement costs and expenditures
for the manufacturing sector in the United States. The PACE survey was conducted annually between 1973 and
1994 (with the exception of 1987) but was discontinued after 1994 by the U.S. Census Bureau. In 1999, a single
PACE survey was conducted, but it differed in many ways from previous surveys, making longitudinal analysis
difficult. The last available PACE survey was administered in 2005. The PACE survey does not overlap with the
vast majority of the TRI sample period (1990–2014), and inconsistency in the data across vintages makes the
survey unsuitable for our study, which heavily relies on time-series variation.
11 Ambient air pollution is measured by EPA pollution monitors that take hourly and/or daily readings. The choice
and management of monitoring location is not subject to local authorities.
584
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 584
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
costs.9 This simple decomposition emphasizes that environmental policies
not only are a sideshow to regular corporate capital investment but also run
much deeper along many dimensions of operations in modern corporations.
Understanding the impact of financial constraints on corporate environmental
policies is important in its own right and deserves careful investigation. In all
our empirical analyses, we control for firms’ overall capital expenditures to
emphasize impacts that extend beyond the regular investment channel.10
In some of our empirical tests, we use a county’s attainment or nonattainment
status to identify regulatory strictness. Under the Clean Air Act Amendments
of 1977 (1977 CAAA), each year every county is classified by the EPA as
attainment or nonattainment of the national standards for criterion pollutants.
The threshold for excessive pollution is applied uniformly across the United
States.11 In any given year, some counties generate pollution that cross these
thresholds while others do not. Figure 2 presents the September 2017 version
of the nonattainment map from the EPA. The EPA applies both mandatory
and discretionary sanctions to nonattainment areas. For example, the EPA can
impose a mandatory sanction for highway funding through the Federal Highway
Administration. Discretionary sanctions mandate that local plants emitting a
pollutant must adopt “lowest achievable emission rates” (LAER) technologies,
which requires the installation of the cleanest available technologies regardless
of the costs. Furthermore, if any new plants plan to locate in a nonattainment
county, the EPA forces them to reduce their releases from another polluting
source within the county. In contrast, for designated “attainment” areas, large
polluters are required only to use the “best available control technology”
(BACT), which is significantly less costly than LAER technology. In summary,
nonattainment status results in more stringent regulations to reduce toxic
releases without regard to cost (Becker and Henderson 2000; Walker 2013).
Financial Constraints and Corporate Environmental Policies
to report information about the release of such toxins. Section 313 of the
Emergency Planning and Community Right-to-Know Act (EPCRA) created
the TRI Program, and specifies that chemicals covered by the TRI Program
cause one or more of the following: (1) cancer or other chronic human health
effects, (2) significant adverse acute human health effects, and (3) significant
adverse environmental effects. The current TRI toxic chemical list contains
over 600 individually listed chemicals and chemical categories. This long list
is compiled for the numerous environmental categories, including air pollution,
clean energy, acid rain, hazardous waste, and safe drinking water. These topics
correspond to over 40 environmental laws and presidential EOs where each
focuses on a subset of the chemicals and compound categories, with potential
overlap across topics.
Section 1101 of Title 18 of the U.S. Code makes it a criminal offense to falsify
information given to the U.S. Government (including intentionally falsifying
records maintained for inspection). Section 325(c) authorizes civil and
administrative penalties for noncompliance with TRI reporting requirements.
The EPA also conducts an extensive quality analysis of TRI reporting data
585
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 585
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Figure 2
EPA nonattainment status
This figure displays a map of counties designated as “nonattainment” zones for National Ambient Air Quality
Standards (NAAQS) pollutants as of September 2017. The EPA publishes attainment status for each county in
its Green Book publication each year.
Source: https://www.epa.gov/green-book
The Review of Financial Studies / v 35 n 2 2022
12 Please refer to Barnatchez, Crane, and Decker (2017) for a detailed comparison between these two data sets.
Notice that we exclude establishments with fewer than 10 employees, which is the size class where NETS have
large imputation rates, as documented by Barnatchez, Crane, and Decker (2017).
586
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 586
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
and provides analytical support for enforcement efforts led by its Office of
Enforcement and Compliance Assurance (OECA). The EPA first identifies TRI
forms containing potential errors and contacts the facilities that submitted them.
If errors are confirmed, the facilities must then submit corrected reports.
We cross-check the EPA’s list of priority pollutants under several EPA
regulations (published through the Code of Federal Regulations (CFR) Title
40) with the TRI list. For example, of 33 hazardous air pollutants listed in
CFR Title 40 61.01 under CAA section 122(b), 29 are included on the TRI list.
Regarding hazardous waste treatment, storage, and disposal facilities (CFR title
264.94), 12 of the 14 priority chemicals are included on the TRI list. Under the
CWA (CFR 40 401.15), 54 of 65 toxic pollutants are listed by the TRI. The
comparison confirms that the TRI list contains chemicals that are particularly
important with respect to their environmental and public health impact as well
as regulatory concerns.
We obtain information pertaining to government agency investigations and
enforcement activities through the EPA’s comprehensive Enforcement and
Compliance History Online (ECHO) database. For each investigation started by
the EPA or state and local agencies, ECHO provides exact filing dates, detailed
violation information, milestone dates, and final enforcement actions settled. It
reports the dollar amounts for federal and local penalties assessed, compliance
actions, cost recovery, and supplemental environmental projects. We aggregate
across all these items to evaluate the total legal liability for each case.
We extract facility information from the National Establishment Time-Series
(NETS) database produced by Walls & Associates, which is a continuous annual
compilation of different vintages of the Dun & Bradstreet (D&B) Million
Dollar Directory database. The organizational structure of the NETS database
shares many similarities with that of the Longitude Business Database (LDB)
maintained by the U.S. Census Bureau.12
We draw firm-level accounting information from the Compustat database
and stock market information from the Center for Research in Security Prices
(CRSP) database. We also use flow-induced price pressure (FIPP) from Lou
(2012), the metropolitan statistical area (MSA)-level Home Price Index (HPI)
from the Federal Housing Finance Agency (FHFA), MSA-level local housing
supply elasticity (Saiz 2010), and the 30-year U.S. fixed mortgage rate to
identify the causal link between financial constraints and toxic releases. To
analyze the effects of the AJCA, we hand-collect data from firms’ 10-K filings
from 2001 through 2007 and determine whether a firm mentioned the AJCA
in its 10-K filings and whether the firm repatriated foreign earnings under the
AJCA. For seasoned equity issuance, we follow the literature (Khan, Kogan,
and Serafeim 2012) and obtain data from the SDC database. We retain only
Financial Constraints and Corporate Environmental Policies
1.3 Financial constraint measures
Financial constraints are difficult to measure (Farre-Mensa and Ljungqvist
2016). In our empirical analysis, we made a conscious choice to avoid using
accounting-based measures of financial constraints because they tend to be
highly correlated with production levels, which is a key factor in determining
the volume of total toxic releases. Instead, we rely mainly on two text-based
financial-constraint measures developed by Hoberg and Maksimovic (2014)
and Bodnaruk, Loughran, and McDonald (2015).
To construct the financial-constraint measure, Bodnaruk, Loughran,
and McDonald (2015) first define several words that describe financial
constraints. They classify a firm-year as more constrained if the list of
financial constraint words, such as “required,” “obligations,” “requirements,”
“permitted,” “comply,” and “imposed,” occur more often.13 Equipped with
such a dictionary, they search the entire 10-K archive and use a simple “bagof-words” approach to delineate the “tone” of management discussion and
13 See Bodnaruk, Loughran, and McDonald (2015) for detailed examples using the New York Times’ 10-K filed
on February 26, 2008, where the constraining count was provoked by discussions concerning debt, legal issues,
and employees. In its 10-K, the company notes that 47% of its workers are unionized (“As a result, we are
required to negotiate the wages, salaries, benefits, staffing levels . . .”); the document also includes discussions
of credit agencies (“To maintain our investment-grade ratings, the credit rating agencies require us to meet
certain financial performance ratios”); a mandatory contract with a major paper supplier (“The contract requires
us to purchase annually the lesser of a fixed number of tons . . .”); obligations (“The Company would have to
perform the obligations of the National Edition printers under the equipment and debt guarantees if the National
Edition printers defaulted under the terms of their equipment leases or debt agreements”); and underfunded
defined benefit pension plans (“As of December 30, 2007, our postretirement obligation was approximately $229
million, representing the unfunded status of our postretirement plans”) (all constraint words in italics).
587
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 587
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
common stock issues traded on the NYSE, the AMEX, or NASDAQ, and
exclude real estate investment trusts (REITs), American Depository Receipts
(ADRs), utilities (SIC codes 4910–4939), and secondary offerings in which no
new shares are issued. Lastly, we use the expected default frequency (EDF)
provided by Moody’s Analytics to measure the probability that a firm will
default (fail to make scheduled debt payments) over the next year.
Without common and consistent linking keys connecting the EPA TRI
report, NETS, and Compustat/CRSP databases, linking these databases poses
a challenge. We first link the EPA TRI report with the NETS database at
the facility level, using a link file provided by the EPA with facility-level
D&B numbers (also known as “DUNS numbers”). In the second step, we link
EPA TRI parent company information with the Compustat /CRSP databases
using a historical name-matching algorithm. It is crucial to use historical name
information, which is time-varying to plant openings, closings, and ownership
changes. We obtain historical company names from CRSP, supplemented by
historical name and address information obtained from 10K, 10-Q, and 8-K
filings using the SEC Analytical Package provided by the Wharton Research
Data Service (WRDS). Please refer to Appendix Section A.1 for a detailed
description of our company name-matching process.
The Review of Financial Studies / v 35 n 2 2022
1.4 Summary statistics
Our final sample includes 8,294 establishments operating in 1,544 U.S. public
firms over the sample period running from 1990 through 2014. A total of 92,803
establishment-year observations is included. In Table 1, we present summary
statistics for the firm-level observations in our sample (panel A) and compare
these with statistics for all Compustat nonfinancial firms during the sample
period (panel B). Establishment-level summary statistics for the key variables
14 Not every firm completes the liquidity and capitalization resource subsections in the MD&A section. Hoberg and
Maksimovic (2014) show that these firms are generally healthy firms that have few liquidity issues to disclose.
588
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 588
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
disclosure. They validate their measure by its predictive power over events, such
as future dividend omissions, pension underfunding, and other related events
generally described as syndromes of financial constraints. In contrast, many
accounting-based financial-constraint measures have no predictive power.
Hoberg and Maksimovic (2014) take a different approach. They focus on
mandated disclosures regarding each firm’s liquidity as well as a discussion
of the financing source each firm intends to use. Based on disclosures in
the Management’s Discussion and Analysis (MD&A) section of the 10-K,14
Hoberg and Maksimovic (2014) evaluate financial constraints by counting
instances when a firm was constrained from raising capital. In our empirical
analysis, we focus on the debt-market constraint measure, which describes a
firm’s intention to issue debt to solve its liquidity problems.
We first cross-check our text-based measures against a set of commonly used
accounting-based variables: interest coverage, credit rating, expected default
frequency (EDF), leverage, a dividend dummy, and the payout ratio. Please refer
to Table A.1 in the appendix for variable definitions. These variables have been
commonly used in the literature to approximate financial constraints. We regress
our two text-based measures on these accounting-based variables and report
the results in Table A.2 in the appendix for the full Compustat sample (panel
A) and for our EPA sample firms (panel B). We include firm fixed effects to
capture the correlation within each firm. In addition, we include several financial
characteristics: log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s q,
which served as controls in our main tests later. Across both panels, we see a
strong correlation between the textual-FC measures and the accounting-based
variables that line up in the expected direction: more financially constrained
firms show lower interest coverage, are less likely to pay dividends, and show
lower payout ratios. In addition, more financially constrained firms tend to
have worse credit ratings and higher leverage, and are closer to default as
indexed by higher EDF. The evidence shows that our text-based measures
do match the more traditional and tangible accounting-based measures well,
and capture financial constraints in an intuitive manner. In Section 2 we also
conduct robustness checks on our baseline results with these accounting-based
measures.
Financial Constraints and Corporate Environmental Policies
Table 1
Summary statistics
A. Sample firm characteristics
Mean
Median
SD
N
11,421
8,788
18,519
18,513
17,993
18,513
17,448
17,817
9,350
16,806
18,460
18,519
18,482
18,149
18,504
N
Text FC
HM debt
Assets(mil)
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
Interest coverage
Rating
EDF
Leverage
Dividend dummy
Payout ratio
ROA
Profit margin
0.70
0.02
6,027.26
0.09
0.21
0.33
1.57
22.63
9.70
2.36
0.41
0.65
0.18
0.04
0.09
0.69
0.01
838.56
0.04
0.17
0.29
1.32
6.70
10.00
0.24
0.39
1.00
0.10
0.05
0.08
0.20
0.06
25,989.75
0.11
0.16
0.18
0.81
64.03
3.67
6.50
0.27
0.48
0.26
0.09
0.10
B. Compustat firm characteristics
Mean
Median
SD
89,738
74,296
190,436
190,243
168,289
190,093
167,213
157,697
34,737
147,733
186,742
206,371
188,336
174,653
174,581
0.69
0.00
2,504.21
0.20
0.45
0.30
2.95
3.36
10.70
6.20
0.37
0.31
0.11
−0.24
−1.62
0.67
−0.01
99.31
0.09
0.21
0.21
1.46
3.75
11.00
0.84
0.31
0.00
0.00
0.02
0.05
0.19
0.06
14,609.50
0.25
0.79
0.27
5.29
174.48
3.82
10.58
0.34
0.46
0.28
0.91
8.77
C. Establishment characteristics
N
Mean
Median
Total release
CAA release
CWA release
CERCLA release
OSHA release
Health effects release
No health effects release
Air release
Water release
RSEI hazard
log(total release)
log(CAA release)
log(CWA release)
log(CERCLE release)
log(OSHA release)
log(health effects release)
log(no health effects release)
log(air release)
log(water release)
log(RSEI hazard)
log(sales)
Pr(investigation)%
Pr(legal_liab>0)%
Legal liabilities
log(legal_liab)
92,803
92,803
92,803
92,803
92,803
88,290
36,402
91,178
91,178
86,934
92,803
79,266
85,512
89,420
53,999
88,290
36,402
91,178
91,178
86,934
92,790
92,803
92,803
92,803
92,803
114.90
59.33
77.72
101.54
10.15
80.19
17.71
53.06
3.79
13,925.70
1.67
1.18
1.36
1.53
-0.28
2.31
1.51
1.96
0.25
16.49
3.74
4.32
2.92
725,511
0.32
7.86
2.75
4.46
6.30
0.01
6.47
1.79
3.82
0.00
10.99
2.06
1.73
1.79
1.96
0.13
2.01
1.03
1.57
0.00
16.21
3.72
0.00
0.00
0.00
0.00
P25
P75
0.55
−0.02
250.11
0.01
0.11
0.19
1.05
3.30
7.00
0.09
0.22
0.00
0.00
0.01
0.04
0.82
0.05
3,116.86
0.12
0.25
0.43
1.80
14.51
13.00
1.00
0.58
1.00
0.24
0.09
0.13
P25
P75
0.55
−0.04
17.09
0.02
0.10
0.08
1.05
−0.92
8.00
0.17
0.03
0.00
0.00
−0.15
−0.09
0.81
0.03
627.35
0.29
0.44
0.47
2.49
11.63
14.00
6.04
0.61
1.00
0.10
0.07
0.12
SD
P25
P75
345.22
177.48
241.47
302.64
33.60
233.94
47.86
154.98
21.24
64,764.91
3.22
3.38
3.24
3.29
3.56
1.97
1.53
1.92
0.90
5.02
1.37
20.34
16.84
30,227,928
1.92
0.75
0.07
0.25
0.44
0.00
0.56
0.13
0.13
0.00
0.39
−0.28
−0.84
−0.66
−0.46
−2.14
0.44
0.12
0.12
0.00
12.87
2.77
0.00
0.00
0.00
0.00
48.18
24.85
30.41
40.69
2.24
38.21
12.10
26.51
0.00
644.14
3.88
3.55
3.59
3.79
2.38
3.67
2.57
3.31
0.00
20.28
4.65
0.00
0.00
0.00
0.00
589
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 589
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Text FC
HM debt
Assets(mil)
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
Interest coverage
Rating
EDF
Leverage
Dividend dummy
Payout ratio
ROA
Profit margin
The Review of Financial Studies / v 35 n 2 2022
Table 1
(Continued)
D. Correlation of various toxic release measures
Total release
CAA
CWA
CERCLE
Total release
CAA
CWA
CERCLE
OSHA
1
0.904
0.935
0.971
0.736
1
0.878
0.910
0.796
1
0.932
0.765
1
0.747
OSHA
1
Figure 3
Toxic releases time series
This figure presents the time series of toxic releases for establishments held by U.S. public firms in our sample
for a period running from 1990 through 2014. We include the total toxic release volumes (in thousands of tons)
and toxic releases under the Clean Air Act (CAA) (in thousands of tons).
are presented in panel C. Compared with firms in the overall Compustat universe
during the same period, our median asset size is $838.56 million, while the
Compustat median asset size is $99.31 million. Our firms also have more
tangible assets (the Compustat median tangible ratio is 21%, while for our
sample it is 29%) and less cash (Compustat median cash-to-asset ratio is 9%,
while our sample median is only 4%). The differences are mainly driven by the
fact that our sample overweights the manufacturing sector.
Figure 3 presents the time-series plot of aggregate toxic releases of our sample
establishments by year.15 The total volume of toxic releases declines over time,
and toxic releases under the CAA account for over half of total toxic releases,
15 A major expansion in the industries required to report toxic releases occurred in 1998. Seven new industry
sectors were required to report to TRI, including metal mining, coal mining, electric utilities, chemical wholesale
distributors, petroleum bulk storage and terminals, hazardous waste management facilities, and solvent recovery
facilities. To populate this figure with data, we remove some new sectors that were added to the TRI program
in 1998 to keep the number of sectors constant throughout the time period. In subsequent empirical analysis,
590
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 590
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Panel A presents firm-level summary statistics for our sample of U.S. public firms during the 1990–2014 period.
Panel B presents summary statistics for Compustat nonfinancial firms during the 1990–2014 period. Panel C
provides summary statistics for establishment-level data. Panel D presents the correlation matrix of toxic release
volumes administered under various EPA regulations.
Financial Constraints and Corporate Environmental Policies
however, we include those sectors. A number of smaller expansions in the reporting requirements were made
between 2000 and 2014. Most of the expansion was related to newly added carcinogenic toxins based on the
National Toxicology Program (NTP) in their Report on Carcinogens (ROC). Using the gradual introduction of
newly identified carcinogens as events that increase corporate liabilities, Gormley and Matsa (2011) explore
managerial responses.
591
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 591
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
displaying a proportional decreasing trend. Some important factors behind the
decline in toxic releases include more stringent environmental regulations and
the higher pollution tax that firms are paying (Henderson 1996; Levinson 2009;
Shapiro and Walker 2018), migration of heavily polluting industries to other
countries (Copeland and Taylor 2004), and introduction of greener technologies
(Levinson 2015). Because of the above time-series attributes, we include year
fixed effects in all of our specifications.
In Table A.3 in the appendix, we summarize total toxic releases using the
Fama-French 48-industry classification. As one might expect, the chemicals,
construction materials, steel, machinery, and auto industries have the largest
number of facilities discharging toxic chemicals, followed by the consumer
products, food products, rubber and plastic petroleum, and public utility
industries. Another feature of the data is that, for certain industries, such
as precious metal, metal mining, and utilities, the average emissions per
establishment are much higher than for other industries.
While the TRI program includes over 600 chemicals, most establishments
emit several chemicals on the list. The average establishment-year emits around
five distinct chemicals, with the median being three. Over 25% of our sample
establishments release only one chemical, and establishments in the 99th
percentile release 28 chemicals. The number of chemicals emitted within each
establishment is highly persistent in our sample. The correlation between the
number of chemicals reported in year t and those reported in year t-1 is 0.973 for
all establishments, 0.971 for establishments emitting more than one chemical,
and 0.967 for establishments emitting more than three chemicals. We further
manually check a randomly selected 10% of our sample establishments to
confirm that establishments tend to emit a consistent list of chemicals without
drastic changes in composition over time.
In addition, chemicals emitted across industries and establishments overlap
only to a certain degree. In Table A.4 in the appendix, we first list the top-30
chemicals ranked by the percentage of establishments emitting these chemicals.
Even the most commonly observed chemicals, such as toluene, xylene, and
ammonia, are emitted by only around 20% of sample establishments. In
addition, the top chemicals ranked by aggregated volume differ from the list
of chemicals ranked by coverage of establishments, indicating that volume and
coverage do not correspond in a one-to-one manner. Furthermore, in Table
A.5 in the appendix, we tabulate the top-five chemicals ranked by aggregated
volume across the Fama-French 48 industries, and the top-five chemicals vary
largely across industries. These summary statistics highlight the need to account
for establishment or industry characteristics when analyzing the data.
The Review of Financial Studies / v 35 n 2 2022
2. Baseline
In this section, we describe our baseline ordinary least squares (OLS)
regression model that relates firms’ total toxic releases to financial-constraint
measures. The purpose of the correlation analysis is to establish some empirical
regularities and benchmark cases. The baseline regression is as follows:
Toxic Releasesi,c,t = α +βFinancial Constraintsc,t−1 +γ Firm Controlsc,t−1
+κEstablishment Controli,c,t +FEs+i,c,t ,
(1)
592
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 592
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Given the importance of the TRI list in capturing firms’ overall negative
environmental and public health impacts, we use total toxic releases of all
TRI chemicals as our key outcome variable, with an equal weight assigned
to each chemical. Considering the complex regulatory structures and the
large number of specific regulations involved, total toxic releases provide the
most comprehensive coverage with respect to both establishments on record
and chemicals on the list, capturing the variation in establishments’ overall
environmental footprint. Given that detailed establishment-level abatement
expenditures are not available, we rely on total toxic releases to infer firms’
abatement spending, assuming total toxic releases are a decreasing function of
abatement spending conditional on the volume of production. We cross-check
our total TRI releases with toxic releases administrated under the most relevant
environmental regulations, namely, the CAA, the CWA, the CERCLA, and the
OSHA.
Panel C of Table 1 present summary statistics at the establishment level. The
average total TRI toxic releases per establishment-year is approximately 115
tons, and the average releases per establishment-year based on the CAA, CWA,
and CERCLA definitions range from 60 to 102 tons. Chemicals associated with
health effects account for a majority of the total releases, whereas a smaller
fraction of establishments emit chemicals associated with no health effects and
the volumes are generally lower. Consistent with Figure 3, air releases account
for a major part of the total toxic releases. Among our sample establishments,
4.3% have been subjects of investigations initiated by government agencies,
with 2.92% of these cases leading to positive legal liabilities. The average
legal liability is around $24.8 million conditional on cases with positive legal
liabilities and $725,511 across all sample observations. Both toxic releases and
legal liabilities are highly skewed. In our analysis, we use the natural logarithm
of toxic releases (in tons) to address skewness in the data. Panel D tabulates the
correlation across toxic releases under various EPA regulations. Across various
regulations, the release amounts are highly correlated with the total TRI toxic
releases, with the correlation ranging above 0.9 for the CAA, the CWA, and
the CERCLA.
Financial Constraints and Corporate Environmental Policies
Table 2
Total toxic releases
Text FC
(1)
(2)
(3)
0.221∗∗∗
0.195∗∗∗
0.160∗∗
(0.070)
(0.065)
log(assets)
0.042
(0.039)
−0.063
(0.352)
0.043
(0.119)
−0.344
(0.387)
0.057
(0.036)
−0.001
(0.029)
51,413
.83
Yes
Yes
No
No
0.030
(0.039)
−0.135
(0.343)
0.063
(0.114)
−0.374
(0.378)
0.063∗
(0.035)
0.002
(0.027)
51,155
.84
Yes
No
Yes
No
0.031
(0.033)
0.193
(0.297)
−0.060
(0.110)
−0.199
(0.296)
0.057∗
(0.033)
0.022
(0.028)
51,361
.84
Yes
No
No
Yes
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Establishment FE
Year FE
State-year FE
Industry-year FE
0.654∗
(0.360)
0.039
(0.039)
0.194
(0.296)
0.008
(0.130)
0.012
(0.369)
0.082∗∗
(0.032)
0.010
(0.038)
36,562
.86
Yes
Yes
No
No
(5)
0.631∗∗
(0.312)
0.030
(0.039)
0.116
(0.295)
0.002
(0.124)
0.003
(0.364)
0.093∗∗∗
(0.033)
0.017
(0.034)
36,422
.86
Yes
No
Yes
No
(6)
0.635∗∗
(0.303)
0.006
(0.039)
0.303
(0.292)
−0.085
(0.124)
0.148
(0.315)
0.075∗∗
(0.035)
0.027
(0.034)
36,511
.86
Yes
No
No
Yes
This table presents results of OLS regressions of total toxic releases (measured by tons in logarithm) on two textbased financial-constraint measures. Firm-level controls include lagged log(assets), Cash/assets, CAPEX/PPE,
Tangible, and Tobin’s q. The establishment-level control is contemporaneous log (sales). Standard errors are
clustered at the firm level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
where i denotes an establishment, c denotes a firm, and t denotes a year. Firm
controls include log(asset), Cash/assets, Tangible, and Tobin’s q, capturing
several aspects of firms’ growth and financial positions. We also include total
capital investment, CAPEX/PPE, to control for the impact of overall capital
investment on our outcome variables. Establishment-level control is log(sales)
to account for production scale.
Table 2 presents estimates deriving from regressing financial constraints on
total toxic releases (measured by tons in logarithm) as our main outcome
variable. The regressions for columns 1 through 3 include the financialconstraint measure Text FC, which is the text-based measure developed in
Bodnaruk, Loughran, and McDonald (2015). The regressions for columns 4
through 6 include the measure HM debt, which is the debt-market constraint
measure taken from Hoberg and Maksimovic (2014). We hypothesize the
coefficient β to be positive, that is, the more financially constrained firms are
likely to release more toxins. Standard errors are clustered at the firm level.
We impose establishment fixed effects to account for time-invariant
unobservable establishment attributes that might affect total toxic releases.
Across both textual measures and all specifications, financial constraints show
significantly positive effect on firms’ toxic releases. The key coefficient of 0.221
(column 1) implies that a one-standard-deviation (0.20) increase in Text FC is
associated with approximately a 4.4% (0.2*0.22) increase in the log tons of
593
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 593
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
(0.073)
HM debt
(4)
The Review of Financial Studies / v 35 n 2 2022
16 In a log-liner model LogY = α +β ∗X + , a one-unit increase in X leads to an expected increase in LogY of β̂
i
i i
units. For small values of β̂ , we use the approximation eβ ≈ 1+ β̂ for a quick interpretation of the coefficients,
finding that 100*β̂ is the expected percentage change in Y for a unit increase in X.
17 The RSEI hazard score includes about 400 of the TRI chemicals and chemical categories. However, an RSEI
hazard score is calculated according to toxicity weights solely based on human health effects associated with
long-term exposure to chemicals, while short-term exposure and ecological effects are not considered.
18 The volume of chemicals released through other disposal media, including underground and landfills, is zero for
the majority of our sample, leaving too few observations for analysis.
594
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 594
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
total toxic releases.16 Similarly, the results reported in column 4 of Panel panel
A in Table 2 show that a one-standard-deviation (0.06) increase in HM Debt is
associated with a 4% (0.6*0.65) increase in total toxic releases. All of the point
estimates across regression specifications are of similar economic magnitudes
at around 4%.
Appendix Table A.6 in the appendix presents estimates of regressing
financial constraints on toxic chemical releases categorized under various EPA
regulations. We include emissions under the CAA, the CWA, and the CERCLA
to examine robustness across disposal methods. We also include emissions
covered under the OSHA to demonstrate robustness against emissions that
represent the highest known threats to work place safety. The coefficients of
the two text-based financial-constraint measures remain quantitatively similar
across categories of toxic chemical emissions.
To further understand which chemicals are driving the variation of total
toxic releases, we group the TRI chemicals by their per-plant or economywide magnitudes, relative human health effects, and physical properties, and
examine the relationship between various toxic releases and firms’ financial
constraints. For per-plant or economy-wide magnitudes, we rank chemicals
according to their volume either within establishments or in our full sample.
For relative human health effects, we consider whether certain chemicals are
associated with human health effects identified by the EPA, and include the
EPA’s Risk-Screening Environmental Indicators (RSEI) hazard score which
uses relative toxicity weights that describe each chemical’s toxicity relative to
that of other chemicals.17 For physical properties, we consider toxic chemicals
simultaneously released through the air and water.18
We first group chemicals by their per-plant magnitudes and conduct our
baseline tests. In Table A.7 in the appendix, we present regression estimates
with the outcome variables being the volume of the number one (panel A)
and the top-three (panel B) chemicals ranked within establishments. For each
panel, we further separate our sample establishments based on the number of
chemicals released to better understand which groups are giving the identifying
power. The results presented in Table A.7 in the appendix demonstrate fairly
consistent elasticity across establishments that emit fewer versus those that emit
more chemicals. Taking panel A with the outcome variable being the volume of
the number one chemical within establishments as an example, the coefficient
estimates for Text FC is 0.220 for groups that emit one or two chemicals, 0.193
Financial Constraints and Corporate Environmental Policies
595
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 595
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
for groups that emit more than two chemicals, and 0.190 for our full sample,
and all these results are statistically significant at the conventional level and
within close proximity to each other. A similar pattern is observed in panel B
with the outcome variable being the volume of the top-three chemicals.
Next, we group chemicals by their economywide magnitudes and examine
toxic releases by the top-five, the top-30, and the rest of the TRI chemicals
according to aggregated volume (the top-30 chemicals are listed in Table A.4
in the appendix). in the appendix. As shown in Table A.4 in the appendix,
no individual chemical can provide comprehensive coverage across our
sample establishments given the differences in production processes across
establishments and industries. When we focus only on the top-five chemicals
by aggregated volume, our sample size is reduced sharply in columns 1 and 2
of panel C. Both coefficient estimates are positive but nonsignificant, and the
lack of statistical power can result from a significantly small sample size. In
contrast, when we expand our chemical list to the top-30 chemicals and report
the results in columns 3 and 4, or to the remaining chemicals on the TRI list,
we observe results showing effects close to that of our baseline results reported
in Table 2. Overall, Table A.7 in the appendix demonstrates that the effects
of financial constraints on total toxic releases are driven mainly by the top
chemicals within establishments rather than by a few top chemicals ranked by
aggregated volume, and similar effects are observed across establishments that
emit varying numbers of chemicals.
Last, we group chemicals by their relative health effects and physical
properties and present the results in Table A.8 in the appendix. The results
reported in panel A show that chemicals associated with adverse human health
impacts are the key drivers of the effects of financial constraints on total toxic
releases (columns 1 and 2), and the documented effects are more profound if we
weight chemicals by their relative toxicity (columns 3 and 4). In contrast, we
do not observe a similar effect for chemicals associated with no health effects
(columns 5 and 6). The different results reported in columns 1 and 2, on the one
hand, and columns 5 and 6, on the other hand, can be explained by the fact that
chemicals are selected into the TRI program based on their adverse acute health
effects or environmental effects. Chemicals with health impacts are the major
component of total toxic releases. TRI chemicals with negative health effects,
which comprise 77% of the TRI list, account for roughly 70% (80.2/114.9
tons) of the total release volume based on summary statistics listed in panel C
of Table 1. On the other hand, fewer than 40% of sample establishments emit
chemicals associated with no health effects, and the smaller sample size could
contribute to the lack of statistical power associated with the results reported
in columns 5 and 6. The results presented in panel B show that air releases are
the main component within total releases driving the correlation with firms’
financial constraints, and chemicals released through water are observed in a
much smaller fraction of the sample and do not display a similar correlation.
Overall, the results reported in Table A.8 in the appendix show that the effects of
The Review of Financial Studies / v 35 n 2 2022
3. Identification: Three experiments
In Section 2 we report a positive correlation between firms’ toxic releases
and financial-constraint measures. It remains challenging, however, to identify
the causal impact of financial constraints on environmental policies. The key
concern lies in omitted variable issues. For example, some unobservables might
affect firms’ financial health and environmental decisions, which could bias the
OLS coefficients in either direction. Alternatively, there is a reverse-causality
concern: poorer environmental performance is associated with worse financial
performance (Margolis, Elfenbein, and Walsh 2009) and higher financing costs
(Chava 2014), which likely makes firms more financially constrained. To
596
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 596
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
financial constraints on total toxic releases concentrate on chemicals associated
with adverse human health effects, highlighting the negative externality they
impose on public health.
In Table A.9 in the appendix, we rerun our baseline results with several
accounting-based financial-constraint measures as the variable of interest:
interest coverage, credit rating, EDF, leverage, a dividend dummy, and the
payout ratio using the same specification as that used to derive the results
reported in Table 2. More financially constrained firms, that is, firms with
worse credit rating, shorter distance to default, higher leverage, and no or lower
payouts, emit a greater volume of toxic chemicals. The coefficient estimates
are statistically significant at the 1% level for credit rating and leverage,
whereas the coefficient estimates of EDF, dividend dummy, and payout ratio
deliver a similar message but are statistically insignificant at the conventional
level. The coefficient estimate for interest coverage is positive but statistically
insignificant. For a one-standard-deviation increase in financial constraints
measured by these accounting-based variables, our estimates in Table A.9 in
the appendix imply an increase in total toxic releases by approximately 1%-2%
for leverage, dividend dummy, and payout ratio, and 3%–4% for credit rating
and EDF. The estimated impact is close to the 4% economic magnitude derived
using two text-based financial-constraint measures.
Table A.10 in the appendix also presents the results of a horse race between
the text-based and accounting-based financial-constraint measures. Including
accounting-based financial-constraint measures does not change the economic
magnitude or statistical significance of the textual financial-constraint
measures. In addition, accounting-based financial-constraint measures do
not seem to affect total toxic releases once the textual financial-constraint
measures are included. In our setting, accounting-based measures of financial
constraints have disadvantages because they tend to be closely correlated with
production levels, which is a key factor in determining the volume of total
toxic releases. This particular concern makes text-based financial-constraint
measures preferable for our analysis. In summary, we recognize that no financial
constraint metric is perfect. We hope that these two text-based measures capture
some relevant aspects of financial constraints in an intuitive manner, and rely
on quasi-natural experiments to generate causal interpretations.
Financial Constraints and Corporate Environmental Policies
3.1 Experiment 1: Tax holiday
The 2004 AJCA provided a temporary tax break for firms by lowering the
repatriation tax rate from 35% to 5.25% for multinational U.S. firms that had
earnings held by foreign subsidiaries. It was implemented to boost domestic
investment and employment by incentivizing U.S. firms to bring back their
stockpiles of cash that had been “trapped” overseas because of repatriation
taxes due. The U.S. Department of the Treasury specified how the repatriated
earnings could be spent, such as on expenditures for plants and equipment,
employment, and acquisitions. To qualify for the tax break, companies need
to have “domestic reinvestment plans” that outline the uses of the repatriated
funds. In a nutshell, the AJCA introduced a positive cashflow windfall for
firms that repatriated foreign earnings, which is well-suited to capturing the
temporary and significant changes in financing cost for our sample firms located
in the United States.
Collecting data from 10-K statements from 2000 through 2007, we are able
to identify 350 unique firms that repatriated foreign earnings under the AJCA,
of which 132 are matched to our EPA-Compustat-merged sample. We exploit a
differences-in-differences (DID) analysis as in Faulkender and Petersen (2012).
We classify firms into three categories: (1) firms with little or no likelihood
of repatriating foreign earnings, (2) firms with a reasonable likelihood of
repatriating foreign earnings that choose not to repatriate such earnings under
the AJCA, and (3) firms with a reasonable likelihood of repatriating foreign
earnings that choose to repatriate under the AJCA. To properly account for the
heterogeneous ex ante likelihood of repatriation and the actual repatriation of
foreign earnings, the empirical design explicitly considers both the predicted
probability of repatriation and the residual from actual repatriation. Our focus
is on the residual term, which identifies differential reactions between treatment
(category 3) and control firms (category 2).
597
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 597
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
establish the causal link, we need to generate an exogenous shock to financial
constraints, while the shock should be unrelated to firms’ environmental
abatement decisions.
In this section, we exploit three quasi-natural experiments to generate an
exogenous shock. These three experiments include the American Jobs Creation
Act (AJCA) of 2004 (Faulkender and Petersen 2012), the collateral value of
firms’ real estate assets (Chaney, Sraer, and Thesmar 2012), and mutual fund
flow-induced price pressure (FIPP) (Lou 2012; Khan, Kogan, and Serafeim
2012). These experiments affect the cost of financing through cash flow
windfalls from foreign income repatriation, higher debt capacity driven by
collateral values, and seasoned equity issuance as a result of short-term price
appreciation. These three shocks are completely independent of each other and
generate exogenous shocks to firms’ financial constraints through orthogonal
channels. We study toxic releases in response to these three shocks to generate
causal inferences regarding how financial constraints affect toxic releases.
The Review of Financial Studies / v 35 n 2 2022
Table 3
Tax holidays, AJCA
(1)
Residual*FC
Pr(Repatriates)
Residual
Preinvestment profit
log(assets)
Cash/assets
CAPX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Establishment FE
Year FE
Industry-year FE
State-year FE
Firm FE
(3)
−0.731∗∗∗
(0.243)
−0.128
(0.265)
0.209∗∗
(0.099)
−0.206∗∗
(0.100)
0.449
(0.325)
0.118∗∗
(0.055)
0.633∗∗
(0.321)
−0.183
(0.125)
0.549
(0.373)
0.054
(0.036)
0.033
(0.037)
19,255
.91
Yes
−0.500∗∗
(0.222)
−0.196
(0.246)
0.147
(0.095)
−0.114
(0.092)
0.750∗∗
(0.318)
0.131∗∗
(0.053)
0.381
(0.290)
−0.185
(0.129)
0.581∗
(0.347)
0.036
(0.034)
0.030
(0.034)
19,170
.91
Yes
Dom inv
(4)
0.038∗∗∗
(0.015)
−0.027
(0.017)
−0.017∗∗
(0.007)
−0.018∗∗∗
(0.006)
0.183∗∗∗
(0.035)
−0.037∗∗∗
(0.006)
0.227∗∗∗
(0.040)
0.033∗∗
(0.016)
0.009
(0.037)
0.006
(0.004)
3,024
.32
Yes
Yes
Yes
Yes
This table presents regression estimates of the effects of the AJCA. For columns 1 through 3, we use the total
amount of toxic releases (measured by tons in logarithm) at the establishment level as outcome variables. The
residual is defined as the dummy variable Repatriation minus Pr(Repatriate), where Pr(Repatriate) is estimated
from the cross-sectional logistic regression as in Table A.11 in the appendix. FC is defined as the proportion of years
during which a firm had insufficient after-taxes earnings to fully finance capital expenditures in the four-year period
prior to the AJCA (Faulkender and Petersen 2012). Firm-level controls include lagged log(assets), Cash/assets,
CAPEX/PPE, Tangible, and Tobin’s q. The establishment-level control is contemporaneous log(sales). Standard
errors are clustered at the firm level. Standard errors appear in parentheses. *p <.1; **p <. 05; ***p <.01.
Specifically, we first estimate the probability of foreign earnings repatriation
(P r(Repatriate)c,t ) based on firm characteristics prior to 2004 and present
the results in Table A.11 in the appendix. Firms with fewer investment
opportunities, higher unrepatriated foreign earnings, and larger tax breaks
generated by the AJCA are more likely to repatriate, consistent with findings
reported in previous studies. We then generate the residual term (Residualc,t =
Repatriatec,t −P r(Repatriate)c,t ), which captures the effects of repatriation
under the AJCA, holding firm characteristics and the probability of repatriation
constant (i.e., foreign earnings in low-tax jurisdictions).
The financial assumption behind the AJCA is that a significant portion
of firms with profitable overseas subsidiaries are financially constrained in
their domestic operations. Firms that were not constrained before the AJCA
should have reached desired levels of environmental activities, and therefore
the AJCA would change only the source of financing without affecting the
598
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 598
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
FC
−0.544∗∗
(0.234)
−0.157
(0.250)
0.150
(0.096)
−0.091
(0.095)
0.790∗∗
(0.311)
0.140∗∗∗
(0.053)
0.480
(0.299)
−0.136
(0.125)
0.639∗
(0.342)
0.034
(0.034)
0.022
(0.035)
19,269
.91
Yes
Yes
log(total release)
(2)
Financial Constraints and Corporate Environmental Policies
3.2 Experiment 2: The collateral channel
In our second identification strategy, we exploit the collateral channel of firms’
real estate assets, which represents a major part of their tangible assets. An
19 We use the Compustat segment file to compute domestic capital expenditures and domestic R&D.
599
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 599
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
abatement level. In contrast, for financially constrained firms that previously
underfunded their environmental activities, the AJCA significantly reduced
the cost of internal capital for their domestic segments. Repatriating income
under the AJCA enables constrained firms to fund environmental abatement
activities that otherwise would have been forgone, resulting in a reduction in
toxic releases in their domestic segments.
We examine the effects of foreign earnings repatriation on corporate
environmental policies as follows:
Toxic ReleasesI,c,t = α +β1 Residualc,t ∗FCc,t +β2 Pr(Repatriate)c,t
(2)
+β3 Residualc,t +β4 FCc,t +γ Controls+F E +i,t ,
where i denotes an establishment, c denotes a firm, and t denotes a year, with
firm controls including log(assets), Preinvestment earnings, CAPEX/PPE, and
Cash/assets, Tangible, and Tobin’s q. The establishment control is log(sales).
In this experiment, we follow Faulkender and Petersen (2012) and use FC to
identify the financially constrained group, where FC is defined as the proportion
of years during which a firm’s investment expenditures exceed its internal cash
flow from 2000 through 2003. We choose this measure for easy replication
of their findings regarding the effects of the AJCA on domestic investment,
therefore lending support to our empirical settings. Our main variable of interest
is Residual*FC, which captures the effects of foreign earnings repatriation on
financially constrained firms.
Table 3 presents the regression estimates of Equation (2). The negative
coefficient for Residual*FC suggests that repatriating firms that were
previously constrained (i.e., unable to fully fund abatement activities)
significantly reduce toxic chemical releases after the repatriation. In terms
of economic magnitude, a one-standard-deviation increase in the repatriation
shock (0.27) leads to a drop of roughly 15% (0.27*0.544) in total toxic
releases (based on coefficient estimates reported in column 1 with the loglinear specification) for US establishments owned by financially constrained
firms, relative to their nonconstrained peers.
To put these estimates in perspective, we further examine the effects of the
AJCA on domestic investment and present the results in column 4 of Table
3. The dependent variable Domestic Inv is defined as the sum of domestic
capital expenditures, domestic R&D, advertising expenses, and acquisitions
scaled by assets.19 We find that financially constrained firms disproportionately
increased domestic investment relative to unconstrained firms after foreign
earnings repatriation (Faulkender and Petersen 2012).
The Review of Financial Studies / v 35 n 2 2022
H P Ii,t = α +β ×Elasticityi ×Mortgage Ratet +MSA F E +Y ear F E +i,t .
(3)
In the second stage, we run the following regression specification:
Yk,j,i,t = α +β ×RE V aluej,i,t +γ H P Ii,t +
κn I nit. Condj
n
×H P Ii,t +controls +k,j,i,t ,
(4)
where k denotes an establishment, j denotes a firm, i denotes an MSA, t
denotes a year, and Firm j ’s real estate value in year t (RE V aluej,i,t ) is
20 Several steps are involved in this process. First, the asset values provided by Compustat in 1993 are book values
instead of market values. To generate market values in 1993, we approximate the average age of these assets
using the ratio of accumulated depreciation of buildings to the historic costs of buildings (assuming a 40-year
depreciation schedule) and then calculate the historical cost using the CPI before 1975 and state-level HPI after
1975. Second, after obtaining the 1993 market value, we use headquarters MSA-level HPI as the price inflater
to generate real estate values for 1993–2007. Chaney, Sraer, and Thesmar (2012) collect information about real
estate assets using firms’ 10-K filings and confirm that facilities are clustered near headquarters in the same MSA
areas and that a major portion of corporate real estate assets are located at the headquarters.
600
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 600
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
increase in the value of real estate assets reduces external financing frictions, and
has been documented to generate more debt issuance and investment activities
(Chaney, Sraer, and Thesmar 2012). We therefore use this setting to identify the
effects of financial constraints on environmental policies. This experiment uses
variations in a firm’s real estate value driven by local real estate prices. Before
1994, Compustat provided a detailed decomposition of property, plant, and
equipment (PPE), which includes three major categories of real estate assets:
building, land and improvement, and construction in progress. We retrieve the
market value of these three categories for firms in 1993 and then inflate their real
estate value using local MSA-level Housing Price Index (HPI) data provided
by the Federal Housing Finance Agency (FHFA) for 1993–2007.20 Real estate
value calculated this way has the advantage of not being affected by decision
to acquire additional real estate assets after 1993, which helps us to obtain a
cleaner identification.
One may have concerns about the potential for omitted variables associated
with both HPI and firms’ environmental decisions that could contaminate
the collateral channel that we try to establish. To address these endogeneity
concerns, we use the MSA-level HPI instrumented by the MSA-level supply
elasticity (Saiz 2010) interacted with 30-year mortgage rates in the United
States (Mian and Sufi 2011; Chaney, Sraer, and Thesmar 2012). The intuition
is that, as housing demand increases based on changes in the mortgage rate,
home prices are less sensitive to demand in elastically supplied markets because
these areas capitalize demand shocks into quantities rather than prices, whereas
in areas with very inelastic supplies, demand shocks translate into prices instead
of quantities (e.g., San Francisco or Boston). The first-stage specification is as
follows:
Financial Constraints and Corporate Environmental Policies
3.3 Experiment 3: Mutual fund flow-induced price pressure
In our third experiment, we take advantage of one key observation from
the mutual fund literature: when investors move capital to or away from
mutual funds, the inflows and outflows force mutual fund managers to
proportionally scale their existing stock positions up or down. Consequently,
their trades induce price pressure that pushes stock prices up (down) with
capital inflows (outflows) (Coval and Stafford 2007; Lou 2012). As FIPP slowly
dissipates, a subsequent return reversal occurs. A common interpretation of the
temporary price pressure is that it represents a source of “nonfundamental”
shocks (Edmans, Goldstein, and Jiang 2012; Hoberg and Maksimovic 2014).
Khan, Kogan, and Serafeim (2012) examine the effects of temporary price
pressure resulting from flow-driven trading and find that positive price pressure
induces SEOs. In other words, large inflow-induced price pressure generates
exogenous positive variation in external financing that reflects equity market
601
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 601
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
calculated by inflating the starting value using local MSA-level HPI in year t
(RE V aluej,i,1993 ×H P Ii,t ), normalized by the lagged PPE. The regressions
include MSA-level HPI to control for the direct impact of real estate prices
on toxic releases. It also includes firm-level initial conditions (Init. Cond) ,
which are the five quintiles of age, assets, and return on assets interacted with
MSA-level HPI to control for the heterogeneous ownership decisions and their
potential impact on the sensitivity that we measure. Establishment fixed effects
are included in all specifications.
We present the first-stage regression estimates in Table A.12 in appendix,
where the interaction terms are significantly positively related to MSA HPI.
Table 4, panel A, presents the regression estimates for firms’ debt issuance,
overall debt level, and financial-constraint measures on the instrumented
real estate value. An increase in real estate value translates into more debt
issuance and higher total debt, confirming that the increase in real estate value
facilitates debt issuance through the collateral channel. Higher debt capacity
and consequently more debt issuance relaxes firms’ financial constraints, as
shown in columns 3 and 4 of panel A, where both coefficient estimates are
negative and statistically significant.
We then examine the effects of higher real estate value on firms’ toxic releases
and report the results in panel B of Table 4. Column 1 uses real estate value
inflated by HPI, and column 2 uses real estate value inflated by the instrumented
HPI. Across both specifications, higher real estate value leads to reduced toxic
releases. The coefficient estimates reported in column 2 show that, for a onestandard-deviation increase (1.47) in the instrumented real estate value (RE
value IV ), total toxic releases drop by around 8% (1.47*0.053) based on the
log-linear specification. In column 3, we report the results of examining the
effects of total debt on toxic releases. The coefficient estimate on total debt
is negative and statistically significant, demonstrating that firms issuing more
debt are those that reduce toxic releases.
The Review of Financial Studies / v 35 n 2 2022
Table 4
The collateral channel: Real estate shock
A. Real estate value, debt issuance, and financial constraints
(1)
Debt issuance
RE value IV
0.061∗∗
0.440∗∗∗
(0.025)
(0.082)
18,941
18,237
.28
.45
Yes
Yes
Yes
Yes
Yes
Yes
B. Real estate value and toxic releases
(1)
RE value
(3)
Text FC
−0.046∗∗
(0.018)
RE value IV
−0.419∗∗
(0.167)
12,691
.34
Yes
Yes
Yes
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Establishment FE
Year FE
Init characteristics
−0.047
(0.054)
0.536
(0.450)
0.020
(0.144)
0.067
(0.350)
−0.047
(0.045)
0.039
(0.034)
25,161
.82
Yes
Yes
Yes
−0.219∗∗∗
(0.049)
10,523
.48
Yes
Yes
Yes
(2)
−0.053∗∗∗
(0.017)
Total debt
log(assets)
(4)
HM debt
−0.117∗∗
(0.051)
0.552∗∗
(0.279)
0.030
(0.117)
0.023
(0.272)
−0.079∗∗∗
(0.030)
0.064∗∗∗
(0.025)
20,565
.83
Yes
Yes
Yes
(3)
−0.299∗∗
(0.137)
−0.015
(0.051)
0.274
(0.427)
0.041
(0.138)
0.283
(0.356)
−0.044
(0.045)
0.039
(0.036)
25,796
.82
Yes
Yes
Yes
This table presents regression estimates of the effects of firms’ real estate value. Firms’ real estate value is
calculated as the 1993 market value inflated by the MSA-level House Price index (HPI), where MSA-level
HPI is instrumented by the interaction between MSA-level local housing supply elasticity and U.S. 30-year fixed
mortgage rates. In panel A, we demonstrate the effects of real estate value on new debt issuance, total debt amount,
and textual financial-constraint measures, while controlling for MSA-level HPI, initial firm characteristics (the
five quintiles of age, assets, and return on assets) interacted with MSA-level HPI, Tobin’s q, and cash position. In
panel B, we tabulate the effects of real estate value on total toxic releases. Firm-level controls include MSA-level
HPI, initial firm characteristics (the five quintiles of age, assets, and return on assets) interacted with MSA-level
HPI, lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s q. The establishment-level control is
contemporaneous log(sales). Standard errors are clustered at the firm level. *p <.1; **p <.05; ***p <.01.
valuation, relaxing firms’ financial constraints. We explore this setting to test
the associated impact on corporate environmental policies.
Specifically, we use a quarterly measure of FIPP from Lou (2012):
Sharesi,j,t−1 ×P erc F lowi,t ∗P SFi,t−1
F I P Pj,t = i
,
(5)
i Sharesi,j,t−1
where Sharesi,j,t−1 is the number of firm j ’s shares held by mutual fund i at
the end of quarter t −1, P erc F lowi,t denotes capital flow to mutual fund
i in quarter t as a fraction of its total net assets at the end of quarter t −
602
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 602
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Observations
Adj. R-squared
Firm FE
Year FE
Controls
(2)
Total debt
Financial Constraints and Corporate Environmental Policies
21 The DGTW data are available via http://terpconnect.umd.edu/ wermers/ftpsite/Dgtw/coverpage.htm.
603
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 603
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
1, and P SFi,t−1 captures mutual funds’ trading responses to capital flows.
More details of the construction of the FIPP measure are included in Appendix
Section A.2. Intuitively, FIPP measures flow-induced trading by the aggregate
mutual fund industry. One key feature of our FIPP calculation is that, instead
of using the actual transactions, we use hypothetical trades based on mutual
funds’ previous quarter-end portfolio weights. The uninformed nature of the
trading arises as it captures the change in a mutual fund’s positions that are
mechanically induced by fund flows instead of fund managers’ discretionary
trading that is potentially driven by information about firms’ fundamentals.
Another advantage of the FIPP measure is that it excludes contemporaneous
dollar volume in the construction, which isolates the direct impact of fund flows
from the potential confounding effect of returns on dollar volume documented
in Wardlaw (2020).
To further validate the transitory nature of the FIPP, we trace DGTW
characteristic adjusted abnormal returns (Daniel et al. 1997) for each quintile
and the highest decile of FIPP from the event quarter (the portfolio-formation
quarter) to 12 quarters afterward and report the results in Table A.13.21 We first
note that the event quarter abnormal returns line up nicely with the magnitude
of FIPP in quarter 0: quintile 1 (outflow) shows negative abnormal returns
while quintile 5 (inflow sample) shows positive abnormal returns. In addition,
return reversal, where the outflow sample displays positive abnormal returns
and the inflow sample displays negative abnormal returns from quarters 1-12
post-event, shows that prices recover to fundamental values after around three
years. This pattern is consistent with the idea that FIPP drives stock prices
away from the fundamentals only temporarily (Lou 2012; Gredil, Kapadia, and
Lee 2019). Decile 10, our large-inflow sample, shows the highest FIPP and
abnormal returns in quarter 0, and the largest reversal post-event. Figure 4,
panel A, plots DGTW abnormal returns for decile 10 and extends the event
window to quarter 16. The inflow-induced price appreciation reverses around
quarter 12 and does not continue the downward trend, consistent with transitory
price pressure instead of potential selection bias with underlying stocks.
We apply a DID analysis to examine the effects of FIPP on firms’ SEOs,
financial constraints, and total toxic releases. For each specific inflow event,
we pair the treatment (decile 10 of FIPP) firm with a control firm from the
same industry-year and track both the treatment and control groups from three
years before to three years after a large inflow shock. In particular, the control
firm is chosen from those that experience large inflow shocks outside the [-3y,
3y] event window of the specific inflow shock. For example, for a treatment
firm that experienced a large inflow shock in year t, we choose a control firm
that experienced a large inflow shock in year t-10. In other words, a firm can
serve as a “control” outside its designated [-3y, 3y] event window of being
The Review of Financial Studies / v 35 n 2 2022
A
Figure 4
Flow-induced price pressure
Panel A presents cumulative abnormal returns (DGTW-adjusted) from quarter 1 through quarter 16 for sample
firms that experience large flow-induced price pressure (FIPP), defined as the highest decile of FIPP in the event
year t (but not in year t-1). Panel B shows coefficient estimates βk of the dynamic difference-in-differences
analysis of the likelihood that treatment and control firms conduct seasoned equity offerings (SEOs). For every
treatment firm, the control firm is defined as a firm within the same industry-year that experiences large inflow
shocks outside of the [-3,3] event window. The omitted benchmark year is year t-3. Firm-level controls include
lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s q. Standard errors are clustered at the firm
t+3
level: Yi,t = α + t+3
k=t−2 βk ·T reatedi ×1[Y ear = k]+ k=t−2 θk ×1[Y ear = k]+γ ·Xi,t +ηi +νt +i,t .
604
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 604
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
B
Financial Constraints and Corporate Environmental Policies
Table 5
Flow-induced price pressure
A. FIPP, SEO, and financial constraints
(1)
(2)
SEO dummy
Text FC
Post
Inflow
Observations
Adj. R-squared
Firm FE
Year FE
Controls
−0.242∗∗∗
(0.094)
0.010
(0.007)
0.014
(0.012)
8,125
.43
Yes
Yes
Yes
−0.058∗
(0.032)
0.001
(0.002)
0.003
(0.004)
6,989
.55
Yes
Yes
Yes
B. FIPP, SEO, and total toxic releases
(1)
Inflow*Post
(2)
−0.167∗∗
(0.082)
−0.177∗∗
(0.086)
0.252∗∗
(0.099)
0.244∗∗
(0.099)
0.089
(0.082)
0.101
(0.082)
0.114
(0.070)
0.155
(0.364)
−0.117
(0.174)
0.778
(0.552)
0.163∗∗∗
(0.060)
−0.002
(0.071)
10,336
.87
Yes
Yes
SEO*Post
Inflow
SEO
Post
log(assets)
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Establishment FE
Year FE
10,558
.87
Yes
Yes
(3)
(4)
−0.213∗∗
(0.101)
−0.163∗
(0.097)
0.676∗
(0.380)
0.078
(0.072)
0.595∗
(0.312)
0.088
(0.066)
0.038
(0.121)
−0.194
(0.444)
0.202
(0.176)
−1.589∗∗
(0.732)
0.085
(0.063)
0.067
(0.074)
8,428
.84
Yes
Yes
8,656
.84
Yes
Yes
This table presents regression estimates of the effects of large flow-induced price pressure (FIPP), defined as
the highest decile of FIPP in the event year t (but not in year t-1), on firms’ seasoned equity offerings (SEOs)
and the total amount of toxic releases. For every treatment firm, the control firm is defined as a firm within the
same industry-year that experiences large inflow shocks outside of the [-3,3] event window. Firm-level controls
include lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s q. The establishment-level control is
contemporaneous log(sales). Standard errors are clustered at the firm level. Standard errors appear in parentheses.
*p <.1; **p <.05; ***p <.01.
“treated,” and the treatment dummy of a same firm can switch between zero
and one depending on which role the firm plays at a particular point in time.
This control selection and matching process is designed to address the concerns
that firms experiencing large inflow shocks might differ fundamentally from
firms that never experience flow shocks along major dimensions that might
605
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 605
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
0.020∗∗
(0.008)
−0.012∗∗
(0.005)
−0.006
(0.009)
13,868
.12
Yes
Yes
Yes
Inflow*Post
(3)
HM debt
The Review of Financial Studies / v 35 n 2 2022
affect investment opportunities and subsequent decisions, such as SEOs (Berger
2019).
In the following DID specification, our coefficient of interest is β3 , which
captures the differential changes in outcome variables between the treatment
and control groups after large inflow shocks:
In addition to using the usual year fixed effects to capture aggregate timeseries variation, we include firm fixed effects and establishment fixed effects
to absorb time-unvarying characteristics at the firm or establishment level.
Because a firm can be in the “treatment” or “control” groups at different points
in time, firm or establishment fixed effects do not fully absorb the treatment
dummy.
Table 5, panel A, presents the effects of large inflow shocks on firms’ SEOs
(column 1) and the coefficient estimate shows a 2% increase in SEO likelihood,
which represents a 51.2% increase relative to the sample average (3.9%). The
estimate lies within the range of 30%–84% documented in Khan, Kogan, and
Serafeim (2012), confirming that managers seize the window of opportunity
for equity issuance when the stock price temporarily appreciates relative to
the fundamentals. Figure 4, panel B, also plots the coefficient estimates of a
dynamic DID specification to formally examine pretrend and post-shock SEO
behaviors. Prior to the inflow pressure, the estimated difference between the
treatment and control groups is not statistically different from zero, indicating
parallel pretrends. As the FIPP unfolds for the treated group from year 0, the
treatment group displays a sharp increase in the likelihood of SEOs compared
with the control group. The increase in SEO likelihood remains statistically
positive through years 1 and 2 and reverts back in year 3, which matches the
reversal in abnormal returns demonstrated in panel A. The parallel pretrends
and the sharp increase in the treatment group’s SEOs from the large inflow
year suggests that the FIPP is responsible for the movement. Table 5, panel
A, columns 2 and 3, report that SEOs driven by the large inflow shock also
reduce firms’ textual financial-constraint measures correspondingly, validating
the relaxation in financial constraints post-large inflows.
Table 5 reports DID regression estimates of firms’ toxic releases based on
Equation (6). Columns 1 and 2 present estimates of the large inflow dummy on
toxic releases, with the coefficient for the interaction term T reated ∗P ost
being negative and statistically significant. Treatment establishments show
significant reductions in toxic releases relative to control establishments. In
terms of economic magnitude, large inflow shocks lead to an approximate
18% drop in toxic releases. To obtain the results reported in columns 3 and
4, we apply the same matching procedure as in the FIPP test to pair the SEO
(treatment) and corresponding control groups. The results indicate that firms
that conduct SEOs display greater toxic release reductions than control firms,
606
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 606
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Yj,t = α +β1 Postj,t +β2 Treatedj,t +β3 Post*Treatedj,t +γ Controls+F E +j,t .
(6)
Financial Constraints and Corporate Environmental Policies
with the economic magnitude similar to that indicated in columns 1 and 2.
Overall, Table 5 illustrates that short-term price appreciations driven by FIPP
lead to SEOs and accompanying relaxation of financial constraints, which result
in lower toxic release volumes.
4. Toxic Releases, Firm Value, and Social Costs
22 For a linear-log model Y = α +β ∗logX + , the expected change in Y associated with a p% increase in X can
i
i i
be calculated as β̂ ∗log([100+p]/100). For small values of p, we approximate the expected change in Y using
β̂ ∗p as log([100+p]/100) ≈ p%.
607
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 607
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
In this section, we examine the relationship between toxic releases, firm
value, and the potential environmental costs imposed on society. In particular,
we provide detailed evidence linking toxic releases to government agency
enforcement actions and legal liabilities, substantiating the trade-offs faced by
firms while making environmental decisions. We further extend the discussion
to examine the extra costs imposed on society, highlighting the costly negative
externality of additional toxic releases driven by financial constraints.
We first connect total toxic releases to the likelihood of enforcement actions
and the dollar amount of legal liabilities. Given that legal enforcement activities
are rare events in the sample, with only about 4.3% of the observations
ever experiencing investigations regarding environmental issues, we explore
the cross-sectional variation within each industry-year first. We regress total
toxic releases on the likelihood of an investigation being initiated against an
establishment as well as the likelihood that an investigation results in positive
legal liabilities (including federal and local penalties, compliance and recovery
costs, and the costs of supplemental environmental projects). Panel A of Table
6, columns 1 and 3, present regression estimates obtained with a logistic model,
while columns 2 and 4 present results obtained with an OLS model, all with
industry-year fixed effects included while controlling for production volume
(log(sales)).
The results reported in panel A reveal that a greater volume of total
toxic releases significantly increases the likelihood of legal investigation
initiated by government agencies and eventually legal costs imposed. OLS and
logistic models generate quantitatively similar estimates. We use the average
effects of financial constraints on total toxic releases identified across our
three quasi-natural experiments ((8%+15%+18%)/3=14%) to help quantify
the magnitudes. A 14% increase in total toxic releases corresponds to an
approximately 3% increase in the likelihood of investigation in the current
year relative to the sample average based on the estimates reported in column 2
(=0.800*14%/4.3), and a 3% increase in the likelihood of positive enforcement
costs relative to the sample average based on the estimates reported in column
4 (=0.547*14%/2.92).22 Column 5 presents the results of the log of total dollar
amount of legal liabilities, with estimated elasticity of 0.062 and statistically
The Review of Financial Studies / v 35 n 2 2022
Table 6
Toxic releases, legal liabilities, and firm value
A. Toxic releases and legal liabilities
Pr(investigation)%
Pr(legal_liab>0)%
(1)
(2)
(3)
(4)
log(total release)
log(sales)
log(legal_liab)
(5)
0.800∗∗∗
(0.040)
0.601∗∗∗
(0.088)
0.793∗∗∗
(0.048)
0.249∗∗∗
(0.071)
0.547∗∗∗
(0.032)
0.315∗∗∗
(0.068)
0.062∗∗∗
(0.004)
0.038∗∗∗
(0.007)
85,096
92,746
.05
Yes
OLS
78,012
92,746
.04
Yes
OLS
92,746
.05
Yes
OLS
Observations
Adj. R-squared
Industry-year FE
Model
Yes
Logit
Yes
Logit
B. Toxic releases and firm value
ROA(%)
(1)
Lagged.log(total release)
Profit margin(%)
(2)
Tobin’s q
(3)
−0.121∗∗
(0.055)
−0.751∗∗
(0.320)
−0.004∗∗
(0.002)
0.656∗∗
(0.292)
10.209∗∗∗
(1.317)
8.576∗∗∗
(0.486)
−8.811∗∗∗
(1.308)
0.764∗∗
(0.298)
−24.402∗∗∗
(1.711)
−22.458∗∗∗
(7.728)
21.739∗∗∗
(2.875)
−16.266∗∗
(7.591)
44.126∗∗∗
(1.742)
−0.142∗∗∗
(0.010)
1.022∗∗∗
(0.045)
0.449∗∗∗
(0.017)
0.106∗∗
(0.045)
0.030∗∗∗
(0.010)
Lead.log(total release)
log(assets)
Cash/assets
CAPEX/PPE
Tangible
Lagged.log(sales)
Lead.log(sales)
Observations
Adj. R-squared
Year FE
Firm FE
17,558
.38
Yes
Yes
17,544
.40
Yes
Yes
16,906
.63
Yes
Yes
(4)
−0.007∗∗∗
(0.002)
0.002
(0.002)
−0.349∗∗∗
(0.013)
1.284∗∗∗
(0.050)
0.414∗∗∗
(0.018)
0.152∗∗∗
(0.049)
−0.097∗∗∗
(0.012)
0.405∗∗∗
(0.013)
13,749
.66
Yes
Yes
In this table, we report the effects of total toxic releases on government agency enforcement actions as well as
firms’ operating performance and value. Panel A links total toxic releases to the likelihood of investigation, the
likelihood of positive legal liabilities imposed (including federal and local penalties, compliance and recovery
costs, and the costs of supplemental environmental projects), and the log dollar amount of legal liabilities.
Coefficient estimates of the OLS regressions and margin effects of the logistic regressions are presented in panel
A. Panel B shows the effects of total toxic releases on firms’ ROA, Profit margin, and Tobin’s q. log(sales) is
included in both panels to control for production volume at the establishment level. Firm-level controls used
in panel B include log(assets), Cash/assets, CAPEX/PPE, and Tangible. Standard errors are clustered at the
establishment level in panel A and at the firm level in panel B. Standard errors appear in parentheses. *p <.1;
**p <.05; ***p <.01.
significant at the 1% level. To confirm that our results are also robust to withinestablishment variation, we report OLS regressions with establishment and year
fixed effects in Table A.14, panel A, in the appendix.23 The positive effects of
toxic releases on legal liabilities that we document in Table 6, panel A, remain
statistically significant across all specifications.24
23 A logistic regression specification will not suitably incorporate a large number of fixed effects.
24 In untabulated results, using the full EPA data generates estimates very similar to those reported for establishments
belonging to Compustat firms, for either cross-sectional or within-establishment specifications.
608
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 608
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
1.064∗∗∗
(0.055)
0.453∗∗∗
(0.083)
Financial Constraints and Corporate Environmental Policies
609
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 609
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Notice that the effects of total toxic releases on legal enforcement actions
reported in panel A of Table 6 likely present a lower bound. Given the limited
resources of government agencies, the complexity of the legal framework, and
the time required to comply with legal requirements, it might take government
agencies several years to detect, pursue, and finalize legal enforcement actions.
Our data set does not provide information that allows us to precisely link toxic
releases in a specific firm-year to the designated enforcement actions, making
it difficult to capture the full impact on current and future legal liabilities. The
results reported in Table A.14, panel B, in the appendix, reflect regressions of
total toxic releases from the previous year and current year on legal liabilities in
the current year, controlling for production in both years. Total toxic releases
from both years display a significant positive loading on all three outcome
variables, with the coefficient estimates being slightly smaller in the previous
year. This analysis reflects potential delays in government agency actions. We
thus interpret our results as providing informative lower-bound estimates that
reflect the additional legal liabilities faced by firms when increasing their toxic
releases.
Next, we investigate firms’ operating performance and value. Legal
enforcement activities represent additional operating costs imposed on firms,
reducing net income. Panel B of Table 6 reports the results of testing for the
effects of total toxic releases on firms’ ROA, profit margin, and Tobin’s q
by aggregating across all establishments within each firm-year. As shown in
columns 1 and 2 of panel B, the coefficient estimates for both ROA and profit
margins are negative and statistically significant at the conventional level. While
ROA and profit margins reflect the impact on firm value in a concurrent manner,
Tobin’s q captures the present value of all future cash flows by incorporating
current and future expected legal liabilities. We use Tobin’s q to refer the effects
on overall firm value in column 3, where a higher volume of total toxic releases
is associated with a lower Tobin’s q and is statistically significant at the 5%
level.
Although we focus on within-firm variations and control for other observable
financial characteristics, the estimated effects of total toxic releases on firm
value are not fully identified as “causal.” Omitted variables at firm-year
frequency might be driving both total toxic releases and worse market
performance. To further address this concern, we include the next year’s total
releases in the regressions together with the lagged version and present the
results of this lead-lag structure in column 4 (Cohn and Wardlaw 2016). While
the coefficient for lagged total release remains stable, we do not observe
statistically significant loading of the lead total releases. The evidence is
consistent with a higher tonnage of toxic releases predicting lower future firm
value, but not vice versa. The average firm in our sample has a book asset
value of $6 billion with Tobin’s q of 1.57. Therefore, a 14% increase in total
toxic releases translates into a decrease in market value of around $9.2 million
The Review of Financial Studies / v 35 n 2 2022
5. Cross-Sectional Analysis
In Section 3, we illustrate the causal impact of financial constraints on firms’
environmental policies. Important questions remain unanswered regarding
financial constraints and environmental protection. For example, how do
financial constraints interact with regulatory enforcement? In this section, we
explore cross-sectional settings for the regulatory environment and external
monitoring, leading to the heterogeneous impacts of financial constraints on
firms’ corporate environmental policies.
As discussed in Section 1.1, under the CAA the EPA applies the
“nonattainment” county label when the air contains a specified amount of any
of the common air pollutants for which the EPA has established a National
Ambient Air Quality Standard. Designation as a nonattainment county triggers
air quality planning and control requirements under which corrective actions
must be taken. LAER, (supposedly) without cost consideration, is required
610
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 610
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
(14%*0.007*1.57*$6 billion) based on estimates reported in column 4 for an
average firm-year.
Lastly, we conduct a back-of-the-envelope calculation to gauge the additional
costs imposed on society associated with discretionary toxic releases. We draw
references from the EPA’s prospective reports on the Clean Air Act from 1990
through 2020, which provides a comprehensive quantification of the costs and
benefits of pollution controls. The economic welfare associated with pollution
abatement occurs because cleaner air leads to better health and productivity
as well as savings on medical expenses, both of which are projected to more
than offset expenditures for pollution abatement. The central benefits estimate
exceeds the costs estimate by a factor of more than 30 to 1, and the high benefits
estimate exceeds the costs estimate by 90 to 1. In other words, for every dollar
that firms cut environmental abatement, there is a welfare loss of $60 to the
public and society, using the middle range of the estimated benefits and costs
factor.
The PACE survey (2005) specifies that, on average, abatement costs account
for approximately 20% of total CAPEX. Within our sample the average
CAPEX is $256 million per firm-year, and therefore the average abatement
costs are about $51 million. For a 14% increase in total toxic releases
driven by financial constraints, we apply a linear extrapolation assuming that
abatement expenditures are cut by 14%, representing a $7.14 million reduction
in abatement spending. This reduction imposes an additional $428 million
($7.14 million*60) cost on society for toxic chemicals regulated by the CAA,
using the estimated costs and benefits factor of 60 from EPA prospective reports.
Aggregating across all 18,519 firm-years in our sample results in a social cost
of approximately $8 billion under the CAA. Total welfare costs across all
other environmental regulations and statutes, and across all other firms and
establishments in the United States, would be significantly higher.
Financial Constraints and Corporate Environmental Policies
611
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 611
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
for major new or modified emission sources located in nonattainment areas.
Additionally, any new emissions are required to be offset by an existing
emission source within the same county. This set of environmental regulations
generates cross-county variation in the strictness of regulatory monitoring and
enforcement.
For firms that are actively trading off the present value of abatement
costs and legal liabilities, regulatory strictness is an important factor. While
greater volumes of toxic releases increase expected legal liabilities, the effect
is expected to be much stronger in nonattainment counties given stringent
regulatory statutes. We test the sensitivity of expected legal liabilities across
attainment and nonattainment counties and present the results in Table A.15
in the appendix using both toxic releases administered under the CAA (panel
A) and total toxic releases (panel B). Coefficients for the interaction terms
between toxic releases and nonattainment status are positive and statistically
significant, confirming that polluting in nonattainment areas triggers higher
legal liabilities, through both the likelihood of legal enforcement actions and
more costly resolutions.
We investigate how nonattainment status, associated with costly legal
consequences, affects firms’ ongoing abatement-expenditure decisions. We
start by examining average toxic releases based on nonattainment status and
report the results in columns 1 and 2 of Table 7. The regression specifications
include firm-year fixed effects to narrow the comparison to the same firmyear, absorbing any firm-year-level movements that might affect environmental
decisions. Furthermore, we include industry fixed effects using either FamaFrench 48 industry categories or four-digit SIC codes at the establishment
level to control for cross-industry differences in the production and polluting
process. The reported economic magnitude is major: the nonattainment dummy
is associated with a 25%–31% lower volume of toxic releases. This evidence
reveals firms’ strategic behavior with respect to their ongoing environmentalabatement activities: within the same firm and industry, establishments
located in attainment counties (facing weak regulations) devote significantly
fewer resources to environmental protection compared with their peers in
nonattaintment countries (facing strong regulations). The results confirm that
nonattainment status imposes significantly higher average abatement costs on
establishments located in those areas (Becker 2005).
The results reported in columns 3 and 4 of Table 7 answer a different
question: when firms are more financially constrained, how do establishments
from counties with either attainment status manage environmental abatements?
We interact the textual financial-constraint measures with the attainment status
of the county where an establishment is located. To ensure that differences
in other observable characteristics are not driving a differential response,
we also include the interaction terms between observable characteristics
and two financial constraint measures in the regressions. The coefficient
estimates for the interaction terms involving nonattainment status and financial
The Review of Financial Studies / v 35 n 2 2022
Table 7
Nonattainment status
Nonattainment=1
(1)
(2)
(3)
−0.306∗∗∗
−0.253∗∗∗
(0.074)
(0.072)
0.159
(0.106)
−0.215∗∗
(0.107)
Nonattainment=1 × Text FC
Nonattainment=1 × HM debt
HM debt
log(assets)
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Firm-year FE
FF48 FE
SIC4 FE
Establishment FE
Year FE
Interaction terms
0.403∗∗∗
(0.038)
86,224
.36
Yes
Yes
0.368∗∗∗
(0.033)
86,206
.41
Yes
0.017
(0.070)
−0.867∗
(0.481)
−0.026
(0.045)
−0.499
(0.697)
−0.461
(0.284)
−1.684∗∗∗
(0.402)
0.066
(0.077)
0.092∗∗
(0.046)
51,023
.84
1.616
(1.564)
0.042
(0.034)
0.181
(0.236)
0.019
(0.100)
-0.051
(0.262)
0.086∗∗∗
(0.029)
0.013
(0.024)
36,562
.86
Yes
Yes
Yes
Yes
Yes
Yes
Yes
In this table, we report results pertaining to relations between total toxic releases and the nonattainment status of the
county where an establishment operates. Firm-level controls include lagged log(assets), Cash/assets, CAPEX/PPE,
Tangible, and Tobin’s q. The establishment-level control is contemporaneous log(sales). Columns 1 and 2 include
firm, year, and industry fixed effects (defined at the establishment level). Columns 3 and 4 include establishment and
year fixed effects. Interaction terms between financial constraint measures and control variables are also included.
Standard errors are clustered at the firm level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
constraint measures indicate that the effects of financial constraints are
significantly weaker in nonattainment countries. When a firm experiences
financial constraints, establishments in nonattainment counties (facing strong
regulations) show smaller changes in toxic releases, while their peers in
attainment counties (facing weak regulations) display a sharp spike in toxic
releases. Our results suggest a spillover effect driven by regulatory strictness
when firms are financially constrained. Financially constrained firms reallocate
abatement expenditures from attainment counties to nonattainment counties
because of the higher expected legal liabilities in the latter areas.
In addition to exploring nonattainment status, we explore polluters’ size
as another gauge of regulatory and external monitoring forces. In a world of
limited regulatory resources and fixed costs for investigating establishments,
large polluters are the focus of EPA monitoring and enforcement activities—
they pose the largest threat to public health and the environment and their higher
visibility makes these establishments attractive targets for regulators seeking
612
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 612
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
−0.978∗∗∗
(0.376)
Text FC
(4)
Financial Constraints and Corporate Environmental Policies
Table 8
Large polluters
70th percentile
(1)
(2)
Large polluter=1 × Text FC
−0.283∗∗∗
(0.086)
Large polluter=1 × HM debt
Text FC
HM debt
log(assets)
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Year FE
Establishment FE
Interaction terms
−0.027
(0.042)
−0.283
(0.627)
−0.552∗∗
(0.271)
−1.658∗∗∗
(0.384)
0.084
(0.061)
0.042
(0.033)
50,774
.87
Yes
Yes
Yes
2.010
(1.383)
0.041
(0.039)
0.236
(0.275)
−0.152
(0.115)
−0.404
(0.350)
0.074∗∗
(0.034)
0.009
(0.020)
36,029
.89
Yes
Yes
Yes
−0.247∗∗∗
(0.087)
−0.697∗∗
(0.340)
−0.017
(0.041)
−0.455
(0.663)
−0.546∗∗
(0.272)
−1.717∗∗∗
(0.377)
0.074
(0.063)
0.056
(0.035)
50,867
.86
Yes
Yes
Yes
−0.728∗
(0.398)
1.974
(1.452)
0.030
(0.038)
0.202
(0.269)
−0.113
(0.110)
−0.435
(0.323)
0.081∗∗
(0.033)
0.008
(0.021)
36,171
.88
Yes
Yes
Yes
80th percentile
(5)
(6)
−0.283∗∗∗
(0.091)
−0.917∗∗
(0.357)
−0.030
(0.042)
−0.514
(0.652)
−0.493∗
(0.271)
−1.940∗∗∗
(0.377)
0.012
(0.067)
0.061∗
(0.037)
51,003
.85
Yes
Yes
Yes
−0.770∗
(0.411)
1.700
(1.498)
0.018
(0.037)
0.179
(0.262)
−0.080
(0.108)
−0.500
(0.308)
0.076∗∗
(0.032)
0.009
(0.021)
36,250
0.87
Yes
Yes
Yes
In this table, we report the heterogeneous effects of financial constraints on total toxic releases for large and
small polluters. Large polluter indicates establishments with total toxic releases above the 70th, 75th, and 80th
percentiles of the sample distribution. Firm-level controls include lagged log(assets), Cash/assets, CAPEX/PPE,
Tangible, and Tobin’s q. The establishment-level control is contemporaneous log(sales). Interaction terms
between financial constraint measures and control variables are also included. Standard errors are clustered
at the firm level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
to send signals to other firms and the public (Becker 2005). In addition, large
polluters also face greater public scrutiny of their environmental activity and
performance. For example, a number of environmental advocacy groups, public
interest groups, and news media routinely scrutinize so-called “superpolluters.”
We predict that smaller establishments with loose external monitoring forces
will increase toxic releases to a greater extent when firms experience financial
constraints than large establishments facing higher expected legal liabilities
will.
To derive the results reported in Table 8, we examine the impact of financial
constraints on toxic releases across polluter size, where large polluters are
defined as establishments with total toxic releases above the 70th, 75th, and
80th percentiles of the sample distribution. Again, the interaction terms between
observable characteristics and two financial constraint measures are included in
the regressions to ensure that heterogeneity in other observable characteristics
is not driving a differential response. Our most robust finding reported in Table
8 is that financial constraints have a much smaller impact on large polluters,
and similar differences between large and small polluters are observed across
the specifications reported in Table 8.
613
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 613
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
−0.726∗∗
(0.331)
−0.913∗∗
(0.401)
75th percentile
(3)
(4)
The Review of Financial Studies / v 35 n 2 2022
6. Conclusion
Exploiting several novel establishment-level data sets, we provide evidence
that financial constraints have direct impacts on corporate environmental
25 While our matching-sample design can ensure that firm-level characteristics remain the same, county- or
establishment-level unobservable characteristics may exist that can confound the interpretation of the interaction
terms between our cross-sectional features and financial constraint measures.
614
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 614
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
For a robustness check, we conduct a matching-sample test where we
create matched samples of establishments under the same parent firms
but with different cross-sectional features. Our matching process ensures
that the matched establishments are exposed to exactly the same firmlevel characteristics, and therefore firm-level observable or unobservable
characteristics cannot drive a differential response to financial constraints. This
specification compares the sensitivity of toxic releases to financial constraints
between establishments in attainment counties and matched establishments
located in nonattainment counties, or between large polluters and matched
smaller polluters, all operated by the same firms. We present results for the
robustness tests in Table A.16 in the appendix. Columns 1 and 2 include
results for nonattainment status. Consistent with the results presented Table
7, establishments in nonattainment counties display less sensitivity to financial
constraints than their peers located in attainment counties. Columns 3–8 include
the results of robustness checks for large polluters, and these results are
largely consistent with the results presented in Table 8, where large polluters
display less sensitivity to financial constraints than peer smaller polluters. The
coefficients reported in columns 3, 5, and 7 are statistically significant at the
5%–10% levels, and the coefficients reported in columns 4, 6, and 8 are negative
but statistically nonsignificant. The decrease in statistical power in columns 4,
6, and 8 is likely caused by the sharp drop of around 50% in the sample size
through the matching process.25
Overall, the cross-sectional results for nonattainment status and polluters’
size highlight firms’ active management of toxic releases: firms are keenly
aware of the external regulatory environment when trading off legal liabilities
and abatement expenditures to maximize value. As financial constraints
increase the marginal cost of abatement expenditures, firms shift resources
to establishments where additional toxic releases might incur large regulatory
penalties and reduce abatement in areas with looser environmental regulations.
It is important to recognize that intrafirm resource adjustment is possible
because of the composition of abatement costs. Heavy equipment is not the only
option for processing industrial waste and reducing toxic chemical releases.
Instead, the vast majority of abatement costs are variable costs, involving factors
such as labor and contract work, which can be modified across locations fairly
easily.
Financial Constraints and Corporate Environmental Policies
A.1. Company Name Matching Process
We use a string-matching algorithm to link TRI establishments to Compustat firms. The TRI
database records the ultimate parent company name for each establishment every year, which
can change over time following such incidents as ownership changes and parent company name
changes. Compustat company information provides only the most up-to-date parent company
names. To ensure matching accuracy, for each Compustat company identified by a GVKEY
we obtain historical company name from CRSP, supplemented by historical name and address
information from the 10K, 10-Q, and 8-K filings using the SEC Analytical Package provided by
the Wharton Research Data Service.
For each company name in the TRI or Compustat/CRSP data, a time stamp is generated
to indicate the effective period of the identifier. We remove all punctuation marks, delete
corporate designators, such as “Corporation,” “Company,” “INC,” or “LLC,” standardize the
most common words to a consistent format, and then generate a similarity score between the
deduplicated TRI parent names and Compustat/CRSP company names using a string-matching
command in STATA.26 We further narrow potential matches to cases in which time stamps from
Compustat/CRSP and the TRI coincide. This time stamp filter further reduces false positives
as public companies that share the same names in different time periods will not be matched
incorrectly.
We use the example of Skyworks Solution, Inc., to demonstrate the importance of using historical
name information in our matching process. Skyworks Solution, Inc., was formed as a result of the
merger of Alpha Industries and the wireless communications division of Conexant in June 2002.27
Skyworks continued using the GVKEY 1327 after the merger in the Compustat database. From the
Compustat company information, we can obtain only “Skyworks solutions” as the current company
name. The CRSP historical name specifies that between January 1980 and May 2002, GVKEY
1327 corresponded to “Alpha Industries, Inc.,” which is also the parent company name presented
in the TRI database. The final matches link establishments to GVKEY 1327 with historical parent
26 For instance, “United States” is simplified to “US,” “Manufacturing” to “MFG,” and “Internationals” to “INTL.”
27 http://investors.skyworksinc.com/news-releases/alpha-and-conexant-announce-plan-close-skyworks-merger-
today
615
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 615
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
policies. Treating toxic chemicals generated during the manufacturing process
is costly and consumes significant financial resources. Firms reduce abatement
expenditures when facing financial constraints because their environmental
protection costs increase correspondingly. Given that firms’ private costs of
pollution abatement are in general much smaller than the social cost, the
additional toxic chemicals released impose costs on the environment, society,
and public health. Collaborative cross-sectional test results also illustrate that
the documented impacts are amplified by weak regulatory enforcement and
external monitoring. These results consistently point to the externalities of
financial constraints in the form of environmental pollution.
In terms of policy implications, our evidence suggests that temporal
variations in toxic releases are closely tied to a firm’s financial strength. When
regulatory oversight and enforcement are resource constrained, our results
suggest the wisdom of adopting a nonrandom auditing policy that focuses on
scenarios in which violations of environmental regulations are most likely to
occur.
The Review of Financial Studies / v 35 n 2 2022
names “Alpha Industries, Inc.,” during 1990 through 2001 and “Skyworks Solutions, Inc.,” from
2002 through 2014.
All unique matches with similarity scores equal to 100% are kept. For cases with multiple
matches and similarity scores below 100%, we drop any matches with similarity scores below
25%, which is a very conservative lower bound. We then rank all potential matches according to
their similarity scores and manually check for the correct matches. In a few cases, the EPA parent
information misses one or two years of parent names while all the other years display the same
parent names. The parent names from other years are then filled in the missing years.
The survivorship-bias-free CRSP mutual fund database provides mutual funds’ total net assets, net
monthly returns, expense ratios, and other fund characteristics. Monthly fund returns are calculated
as net returns plus 1/12 of annual fees and expenses. For mutual funds with multiple share classes,
total net assets (TNA) are summed up across all share classes, net returns and expense ratios are
weighted by TNA across all share classes, and the investment objective code of the share class with
the largest TNA is used. MFLinks file is then used to merge the CRSP mutual fund database with
the Thompson Financial CDA/Spectrum holding database, which contains quarterly mutual fund
holdings data.
To exclude all fixed-income funds, international funds, and previous metal funds, the investment
objective code reported by CDA/Spectrum is limited to aggressive growth, growth, growth and
income, balanced, unclassified, or missing. Since some mutual funds misreport their investment
objective codes, the ratio of equity holdings is set to between 0.75 and 1.2. Additional requirements
include a minimum fund size of $1 million and that the TNAs reported by CDA/Spectrum and
CDA
CRSP do not differ by more than a factor of two (i.e., 0.5< T N ACRSP <2).
T NA
The investment flow to fund i in quarter t is defined as the following:
f lowi,t =
T N Ai,t −T N Ai,t−1 ∗(1+RETi,t )−MGNi,t
,
T N Ai,t−1
(A1)
where MGNi,t is the increase in TNA following fund mergers in quarter t. To estimate the merger
date, we use the latest net asset value (NAV) report date of the target fund. A target fund is matched
to its acquirer starting one month before its last NAV report date until five months after, and then
the month in which the acquirer has the smallest absolute percentage flow is designated as the event
month. Mutual funds that are initiated have inflows equal to their initial TNA, and funds that are
liquidated have outflows equal to their terminal TNA.
The following equation is estimated to gauge the effect of trading costs and other constraints
on the degree of partial scaling:
tradei,j,t = β0 +β1 f lowi,t +γ2 X +γ3 f lowi,t X +i,t .
The dependent variable, tradei,j,t =
sharesi,j,t
splitadj
sharesi,j,t−1
(A2)
−1 is the percentage of trading in stock j by fund
i in quarter t, with split adjustments. The key independent variable is f lowi,t , which is the capital
flow to fund i in quarter t as a fraction of the fund’s TNA at the end of the previous quarter. X
includes variables that capture trading costs: (a) the ownership share of mutual fund i in stock j
(defined as
sharesi,j,t−1
shroutj,t−1 ),
and (b) the effective bid-ask spread of stock j derived from the basic
market-adjusted model (Hasbrouck 2009). The portfolio-weighted average ownership share and
liquidity cost are included in the X vector to examine the effects of portfolio-level constraints on
managers’ decisions to invest capital inflows in their existing positions as opposed to initiating new
positions. The regression specification corresponds to columns 1 and 7 of table 2 in Lou (2012) for
the outflow and inflow samples, respectively. The estimated β1 is extracted as the partial scaling
factors (PSF) and used as an input in the final step of the FIPP calculation.
616
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 616
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
A.2. Flow-Induced Price Pressure
Financial Constraints and Corporate Environmental Policies
Table A.1
Variable definitions
Firm
Text FC
HM debt
Interest coverage
Rating
EDF
Leverage
Dividend dummy
Payout ratio
Dom inv
RE value
Debt issuance
Total debt
FIPP
Inflow
SEO
ROA
Profit margin
Establishment
log(total release)
log(CAA release)
log(CWA release)
log(CERCLA release)
log(OSHA release)
log(health effects release)
log(no health effects release)
log(air release)
log(water release)
log(RSEI hazard)
Nonattainment
Pr(investigation)%
Pr(legal_liab)%
log(legal_liab)
log(sales)
Logarithm of tons of total toxic releases administered under the TRI
program
Logarithm of tons of toxic release administered under the Clean Air Act
Logarithm of tons of toxic release administered under the Clean Water
Act
Logarithm of tons of toxic release administered under the Comprehensive
Environmental Response, Compensation, and Liability Act
Logarithm of tons of toxic release administered by the Occupational Safety
and Health Administration
Logarithm of tons of toxic release associated with health effects
Logarithm of tons of toxic release not associated with health effects
Logarithm of tons of toxic release through air
Logarithm of tons of toxic release through water
Logarithm of the EPA’s Risk-Screening Environmental Indicators (RSEI)
hazard score
A dummy that equals one if an establishment resides in a county with
nonattainment status
The likelihood of government agencies filing an environmental investigation against an entity
The likelihood of positive legal liabilities( including federal and local
penalties, compliance and recovery cost, and the costs
of supplemental environmental projects) imposed on an entity
Logarithm of dollar amount of legal liabilities imposed on an entity
Logarithm of number of sales dollar amount (inflation adjusted) at the
establishment level
617
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 617
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
log(assets)
Cash/assets
CAPEX/ PPE
Tangible
Tobin’s q
Textual financial-constraint measure by Bodnaruk, Loughran, and
McDonald (2015)
Debt focus financial-constraint measured by Hoberg and Maksimovic
(2014)
log number of total assets
Cash and Short-term investment/L.Assets
Capital expenditures/ L.PPENT
PPENT/L.Assets
(Total asset + Common shares outstanding × Closing price (Fiscal year)
− Common equity − Deferred taxes)/Asset
Operating income before depreciation/Interest payment
S&P domestic long-term issuer credit rating
The probability that a company fails to make a scheduled debt payment
within a year
(Debt in current liabilities + Long-term debt)/ Assets
A dummy that equals to one if a firm-year pays dividends
(dividend payment + share repurchases)/operating income before
depreciation
(Domestic Capital Expenditure+Domestic R&D+Advertising expenses,
and Acquisitions)/Assets
1993 market value of real estate assets inflated by the MSA-level House
Price index (HPI)
New debt issued /L.PPENT
Total amount of debt/L.PPENT
Flow-induced price pressure defined in Lou (2012)
A dummy variable that equals one for firms with FIPP in the highest decile
in year t (but not in year t-1)
A dummy variable that equals one if a firm conducts pure seasoned equity
offerings in a year
Income before extraordinary items/L.Assets
EBIT/sales
618
Observations
Adj. R-squared
Firm FE
Year FE
Tobin’s q
Tangible
CAPEX/PPE
Cash/assets
log(assets)
HM debt
Text FC
(1)
13.680∗∗∗
(1.961)
−6.372
(11.146)
−1.953
(1.500)
−1.398
(9.509)
1.166∗∗∗
(0.341)
66,673
.36
Yes
Yes
−18.857∗∗∗
(4.095)
−78.718∗∗∗
(19.772)
16.839∗∗∗
(2.279)
−2.984
(12.538)
−1.708
(1.785)
9.108
(11.878)
2.221∗∗∗
(0.479)
54,554
.38
Yes
Yes
(2)
(3)
−1.033∗∗∗
(0.074)
0.381
(0.346)
−0.180∗∗∗
(0.051)
−0.733∗
(0.386)
−0.189∗∗∗
(0.051)
17,031
.89
Yes
Yes
0.808∗∗∗
(0.102)
(4)
1.482∗∗∗
(0.392)
−1.087∗∗∗
(0.072)
0.282
(0.353)
−0.217∗∗∗
(0.056)
−0.411
(0.419)
−0.169∗∗∗
(0.045)
14,255
.87
Yes
Yes
Rating
(5)
−2.689∗∗∗
(0.094)
−7.134∗∗∗
(0.343)
−1.036∗∗∗
(0.052)
2.076∗∗∗
(0.564)
−0.394∗∗∗
(0.017)
73,056
.62
Yes
Yes
2.153∗∗∗
(0.222)
EDF
8.335∗∗∗
(0.827)
−2.513∗∗∗
(0.100)
−6.670∗∗∗
(0.359)
−1.047∗∗∗
(0.060)
3.102∗∗∗
(0.640)
−0.383∗∗∗
(0.019)
63,668
.62
Yes
Yes
(6)
A. Full Compustat sample
(7)
−0.041∗∗∗
(0.003)
−0.383∗∗∗
(0.012)
−0.018∗∗∗
(0.002)
0.116∗∗∗
(0.018)
0.004∗∗∗
(0.000)
79,441
.66
Yes
Yes
0.102∗∗∗
(0.007)
(8)
0.394∗∗∗
(0.025)
−0.041∗∗∗
(0.004)
−0.367∗∗∗
(0.013)
−0.018∗∗∗
(0.002)
0.124∗∗∗
(0.020)
0.004∗∗∗
(0.001)
65,597
.68
Yes
Yes
Leverage
0.019∗∗∗
(0.004)
0.024∗
(0.014)
0.003
(0.002)
−0.001
(0.019)
0.001∗∗∗
(0.000)
80,347
.63
Yes
Yes
−0.058∗∗∗
(0.010)
(9)
−0.129∗∗∗
(0.039)
0.012∗∗∗
(0.004)
0.040∗∗∗
(0.015)
0.007∗∗∗
(0.002)
−0.002
(0.022)
0.002∗∗∗
(0.001)
66,265
.63
Yes
Yes
(10)
Dividend dummy
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Interest coverage
Table A.2
Textual financial constraint measures
−0.175∗∗∗
(0.029)
0.006∗∗
(0.003)
0.022∗
(0.011)
−0.006∗∗∗
(0.001)
−0.014
(0.014)
−0.001∗∗∗
(0.000)
66,127
.23
Yes
Yes
(12)
(Continued)
0.009∗∗∗
(0.002)
0.020∗∗
(0.009)
−0.006∗∗∗
(0.001)
−0.013
(0.011)
−0.001∗∗∗
(0.000)
80,187
.24
Yes
Yes
−0.022∗∗∗
(0.008)
(11)
Payout ratio
The Review of Financial Studies / v 35 n 2 2022
Page: 618
576–635
619
Page: 619
576–635
−1.008
(2.814)
74.978∗∗∗
(19.691)
18.400∗∗
(7.692)
27.412
(19.446)
18.494∗∗∗
(3.782)
10,253
.49
Yes
Yes
−83.971∗∗∗
(20.895)
−1.035
(3.730)
43.648∗
(25.925)
19.115∗∗
(9.636)
7.691
(28.228)
19.687∗∗∗
(4.099)
7,824
.50
Yes
Yes
(2)
(3)
−1.251∗∗∗
(0.131)
0.103
(0.587)
−1.427∗∗∗
(0.263)
−1.533∗∗
(0.673)
−0.722∗∗∗
(0.080)
5,913
.90
Yes
Yes
0.848∗∗∗
(0.134)
Rating
1.669∗∗
(0.679)
−1.156∗∗∗
(0.136)
0.611
(0.640)
−1.792∗∗∗
(0.306)
−0.966
(0.868)
−0.686∗∗∗
(0.095)
4,406
.89
Yes
Yes
(4)
(5)
−1.505∗∗∗
(0.320)
−2.666∗∗
(1.067)
−3.451∗∗∗
(0.595)
2.122
(1.837)
−1.015∗∗∗
(0.148)
10,157
.53
Yes
Yes
1.227∗∗∗
(0.356)
EDF
3.768∗∗
(1.763)
−1.167∗∗∗
(0.369)
−1.504
(1.292)
−4.359∗∗∗
(0.749)
1.442
(2.243)
−1.188∗∗∗
(0.168)
8,131
.53
Yes
Yes
(6)
B. EPA sample
(7)
0.016
(0.012)
−0.303∗∗∗
(0.049)
−0.072∗∗∗
(0.017)
0.006
(0.059)
−0.023∗∗∗
(0.006)
10,657
.71
Yes
Yes
0.083∗∗∗
(0.013)
(8)
0.239∗∗∗
(0.056)
0.013
(0.013)
−0.202∗∗∗
(0.050)
−0.080∗∗∗
(0.017)
0.043
(0.070)
−0.028∗∗∗
(0.006)
8,178
.73
Yes
Yes
Leverage
0.126∗∗∗
(0.019)
0.039
(0.072)
−0.041
(0.032)
−0.023
(0.082)
0.007
(0.010)
10,677
.72
Yes
Yes
−0.061∗∗∗
(0.022)
(9)
−0.211∗
(0.108)
0.102∗∗∗
(0.020)
−0.033
(0.080)
−0.016
(0.035)
−0.061
(0.095)
0.015
(0.012)
8,191
.73
Yes
Yes
(10)
Dividend dummy
−0.018∗
(0.010)
0.056
(0.059)
−0.023
(0.024)
−0.060
(0.056)
−0.006
(0.007)
10,664
.31
Yes
Yes
−0.015
(0.016)
(11)
−0.139∗∗
(0.064)
−0.032∗∗∗
(0.012)
−0.026
(0.068)
−0.008
(0.027)
−0.196∗∗∗
(0.074)
−0.011
(0.010)
8,176
.31
Yes
Yes
(12)
Payout ratio
In this table, we report the results of examining the relationship between two textual financial-constraint measures and several accounting-based measures, including interest coverage, credit
rating, expected default frequency (EDF), leverage, a dividend dummy, and the payout ratio. Firm-level controls include lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s
q. Standard errors are clustered at the firm level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
Observations
Adj. R-squared
Firm FE
Year FE
Tobin’s q
Tangible
CAPEX/PPE
Cash/assets
log(assets)
HM debt
−7.924∗∗
(3.179)
(1)
Interest coverage
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Text FC
Table A.2
(Continued)
Financial Constraints and Corporate Environmental Policies
The Review of Financial Studies / v 35 n 2 2022
N
Agriculture
Food products
Candy & soda
Beer & liquor
Tobacco products
Recreation
Printing and publishing
Consumer goods
Apparel
Medical equipment
Pharmaceutical products
Chemicals
Rubber and plastic products
Textiles
Construction materials
Construction
Steel works, etc.
Fabricated products
Machinery
Electrical equipment
Automobiles and trucks
Aircraft
Shipbuilding, railroad equipment
Defense
Precious metals
Nonmetallic and industrial metal mining
Coal
Petroleum and natural gas
Utilities
Personal services
Business services
Computers
Electronic equipment
Measuring and control equipment
Business supplies
Shipping containers
Transportation
Wholesale
Retail
Almost nothing
Total
Mean
179
161.18
3,953
82.33
204
15.56
255
35.10
162
153.70
595
59.30
160
8.52
4,314
52.41
374
23.00
1,077
33.73
1,590
101.88
10,806
152.23
3,135
78.70
1,121
51.73
7,271
46.77
364
95.63
7,119
208.82
2,689
25.57
6,766
21.92
3,335
35.31
73.19
4,927
2,224
37.46
650
43.47
466
37.80
152 1,079.73
363
586.05
62
242.26
2,623
180.09
2,773
746.93
314
55.28
1,910
87.47
576
43.53
3,342
12.81
1,116
14.22
4,104
263.02
2,211
113.45
247
88.08
2,512
67.93
745
54.31
395
362.62
87,181
119.37
Median
6.67
13.66
9.26
11.50
42.38
37.71
4.48
9.95
10.21
6.01
9.12
10.25
9.55
8.60
7.04
1.45
8.66
3.87
2.98
3.60
10.88
6.91
20.20
6.30
654.31
39.67
68.17
33.79
430.55
1.01
5.90
5.00
1.55
1.37
29.58
39.50
3.89
2.23
4.86
16.22
8.42
SD
426.08
209.21
24.13
52.61
211.58
76.29
17.49
116.32
32.54
118.25
294.00
415.81
217.41
150.91
149.83
344.72
540.37
80.53
98.88
139.81
166.31
113.10
71.85
126.26
1,015.07
848.03
363.19
339.30
771.09
257.38
254.39
197.31
44.21
48.12
445.00
253.47
249.68
244.89
155.53
691.44
351.30
P25
P75
0.38
4.14
0.44
3.92
9.67
5.60
0.73
0.65
1.81
0.55
1.00
1.70
1.29
1.61
0.43
0.20
0.73
0.35
0.25
0.23
1.01
0.59
4.67
0.59
10.60
2.55
24.10
2.25
108.06
0.13
0.73
0.14
0.13
0.13
4.38
7.13
1.16
0.55
0.72
1.08
0.85
78.00
65.00
19.10
41.67
215.64
87.00
9.76
52.22
28.50
21.28
52.96
66.44
44.35
37.72
29.89
34.98
71.69
21.62
14.39
21.12
53.00
27.78
51.47
31.81
2,248.90
1,014.77
284.00
198.52
1,219.48
12.98
39.07
21.17
7.72
7.50
341.45
116.61
24.97
17.20
23.24
259.53
52.20
The table presents summary statistics indicating total toxic releases (tons) by Fama-French 48 industry
classifications. Industries with fewer than 50 observations have been dropped.
620
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 620
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.3
Summary statistics by industry
Financial Constraints and Corporate Environmental Policies
Rank By coverage of establishments
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
TOLUENE
XYLENE (MIXED ISOMERS)
AMMONIA
LEAD COMPOUNDS
ZINC COMPOUNDS
METHANOL
SULFURIC ACID
HYDROCHLORIC ACID
LEAD
CERTAIN GLYCOL ETHERS
COPPER
1,1,1-TRICHLOROETHANE
METHYL ETHYL KETONE
NICKEL
CHROMIUM
MANGANESE
CHROMIUM COMPOUNDS
COPPER COMPOUNDS
MANGANESE COMPOUNDS
ETHYLBENZENE
ETHYLENE GLYCOL
NICKEL COMPOUNDS
ACETONE
1,2,4-TRIMETHYLBENZENE
N-BUTYL ALCOHOL
POLYCYCLIC AROMATIC COMPOUNDS
BARIUM COMPOUNDS
NITRATE COMPOUNDS
NITRIC ACID
CHLORINE
Percentage (%) By aggregated volume
21.82
20.55
19.76
18.68
17.92
17.88
16.98
15.17
15.16
14.55
14.36
13.56
13.43
13.35
12.71
11.71
11.09
10.70
10.45
9.92
9.78
9.42
8.90
8.58
8.55
8.54
8.28
7.99
7.90
7.88
Tonnage
HYDROCHLORIC ACID
2,076,614.00
ZINC COMPOUNDS
2,021,955.00
ARSENIC COMPOUNDS
1,252,983.00
COPPER COMPOUNDS
1,195,326.00
MANGANESE COMPOUNDS 1,137,704.00
METHANOL
1,050,810.00
NITRATE COMPOUNDS
929,225.80
AMMONIA
849,445.10
BARIUM COMPOUNDS
621,526.70
SULFURIC ACID
525,227.80
LEAD COMPOUNDS
517,826.20
TOLUENE
370,201.50
XYLENE (MIXED ISOMERS) 274,026.00
CHROMIUM COMPOUNDS 228,378.10
HYDROGEN FLUORIDE
208,405.60
ZINC (FUME OR DUST)
186,306.00
METHYL ETHYL KETONE 174,344.60
N-HEXANE
166,820.90
NITRIC ACID
161,019.40
CERTAIN GLYCOL ETHERS 153,482.60
VANADIUM COMPOUNDS 142,625.20
DICHLOROMETHANE
121,415.30
N-BUTYL ALCOHOL
119,302.40
CARBONYL SULFIDE
116,372.10
PHOSPHORIC ACID
110,929.10
ETHYLENE
108,518.60
NICKEL COMPOUNDS
108,077.50
FORMALDEHYDE
99,570.78
STYRENE
98,966.32
ACETONE
94,531.52
The table lists the top-30 chemicals ranked by aggregated volume and coverage of sample establishments.
621
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 621
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.4
Top chemicals by coverage of establishments
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 622
576–635
DICHLOROMETHANE
N-HEXANE
AMMONIA
NITRATE COMPOUNDS
HYDROCHLORIC ACID
1,1,1-TRICHLOROETHANE
CERTAIN GLYCOL ETHERS
METHANOL
ACETONE
FREON 113
DICHLOROMETHANE
HYDROCHLORIC ACID
TOLUENE
HYDROCHLORIC ACID
AMMONIA
ACETONITRILE
MANGANESE COMPOUNDS
METHYL ETHYL KETONE
XYLENE (MIXED ISOMERS)
LEAD COMPOUNDS
CERTAIN GLYCOL ETHERS
1,1,1-TRICHLOROETHANE
N-HEXANE
DICHLOROMETHANE
MANGANESE COMPOUNDS
ZINC COMPOUNDS
BARIUM COMPOUNDS
AMMONIA
BARIUM COMPOUNDS
AMMONIA
HYDROCHLORIC ACID
TOLUENE
FREON 113
1,1,1-TRICHLOROETHANE
HYDROCHLORIC ACID
CERTAIN GLYCOL ETHERS
DICHLOROMETHANE
METHANOL
AMMONIA
LEAD COMPOUNDS
MANGANESE COMPOUNDS
METHANOL
ZINC COMPOUNDS
ZINC COMPOUNDS
ZINC COMPOUNDS
ZINC (FUME OR DUST)
MANGANESE COMPOUNDS
XYLENE (MIXED ISOMERS)
METHYL ETHYL KETONE
XYLENE (MIXED ISOMERS)
1,1,1-TRICHLOROETHANE
ARSENIC COMPOUNDS
COPPER COMPOUNDS
AMMONIA
NITRATE COMPOUNDS
HYDROCHLORIC ACID
METHANOL
TOLUENE
METHYL ETHYL KETONE
1,1,1-TRICHLOROETHANE
COPPER
METHANOL
METHANOL
AMMONIA
AMMONIA
HYDROCHLORIC ACID
ZINC (FUME OR DUST)
2
NITRATE COMPOUNDS
NITRATE COMPOUNDS
NITRATE COMPOUNDS
AMMONIA
AMMONIA
STYRENE
1,1,1-TRICHLOROETHANE
TOLUENE
TOLUENE
TOLUENE
METHANOL
AMMONIA
STYRENE
1
AMMONIA
N-BUTYL ALCOHOL
TOLUENE
HYDROCHLORIC ACID
NITRATE COMPOUNDS
LEAD
SULFURIC ACID
MANGANESE COMPOUNDS
METHANOL
FREON 113
ZINC COMPOUNDS
METHANOL
MANGANESE COMPOUNDS
TOLUENE
HYDROCHLORIC ACID
LEAD COMPOUNDS
MANGANESE COMPOUNDS
NITRATE COMPOUNDS
ZINC COMPOUNDS
METHANOL
NITRATE COMPOUNDS
CERTAIN GLYCOL ETHERS
MANGANESE
ZINC COMPOUNDS
ZINC COMPOUNDS
TRICHLOROETHYLENE
N-BUTYL ALCOHOL
ZINC COMPOUNDS
AMMONIA
MANGANESE COMPOUNDS
HYDROCHLORIC ACID
ZINC COMPOUNDS
ACETONE
TOLUENE
XYLENE (MIXED ISOMERS)
METHYL ETHYL KETONE
XYLENE (MIXED ISOMERS)
ACETONE
NITRATE COMPOUNDS
METHANOL
3
SULFURIC ACID
HYDROCHLORIC ACID
CYCLOHEXANE
N-HEXANE
SULFURIC ACID
ZINC COMPOUNDS
HYDROGEN FLUORIDE
ZINC COMPOUNDS
SULFURIC ACID
METHANOL
DICHLOROMETHANE
ZINC (FUME OR DUST)
ZINC COMPOUNDS
PROPYLENE
COPPER
ZINC COMPOUNDS
LEAD COMPOUNDS
METHANOL
FORMALDEHYDE
AMMONIA
COPPER COMPOUNDS
XYLENE (MIXED ISOMERS)
STYRENE
XYLENE (MIXED ISOMERS)
METHYL ISOBUTYL KETONE
TOLUENE
MANGANESE COMPOUNDS
N-HEXANE
HYDROCHLORIC ACID
HYDROCHLORIC ACID
N-BUTYL ALCOHOL
NICOTINE AND SALTS
TOLUENE
COPPER COMPOUNDS
AMMONIA
CHROMIUM COMPOUNDS
DICHLORODIFLUOROMETHANE
TOLUENE
MANGANESE COMPOUNDS
CARBON DISULFIDE
4
MANGANESE COMPOUNDS
AMMONIA
N-HEXANE
DICHLOROMETHANE
METHANOL
COPPER COMPOUNDS
MANGANESE COMPOUNDS
HYDROCHLORIC ACID
N-HEXANE
1,1,1-TRICHLOROETHANE
AMMONIA
N,N-DIMETHYLFORMAMIDE
LEAD COMPOUNDS
SULFURIC ACID
FREON 113
COPPER COMPOUNDS
SULFURIC ACID
ETHYLENE GLYCOL
XYLENE (MIXED ISOMERS)
MANGANESE COMPOUNDS
LEAD COMPOUNDS
TOLUENE
1,1,1-TRICHLOROETHANE
NITRATE COMPOUNDS
N-BUTYL ALCOHOL
TETRACHLOROETHYLENE
STYRENE
N,N-DIMETHYLFORMAMIDE
BARIUM COMPOUNDS
BROMOMETHANE
CERTAIN GLYCOL ETHERS
NITRATE COMPOUNDS
XYLENE (MIXED ISOMERS)
METHANOL
METHYL ETHYL KETONE
CERTAIN GLYCOL ETHERS
CERTAIN GLYCOL ETHERS
NITRATE COMPOUNDS
METHANOL
METHYL ETHYL KETONE
5
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
622
The table lists the top-five chemicals ranked by aggregated volume for each of the Fama-French 48 industries.
Agriculture
Food products
Candy soda
Beer liquor
Tobacco products
Recreation
Printing and publishing
Consumer goods
Apparel
Medical equipment
Pharmaceutical products
Chemicals
Rubber and plastic
products
Textiles
Construction materials
Construction
Steel works etc
Fabricated products
Machinery
Electrical equipment
Automobiles and trucks
Aircraft
Shipbuilding, railroad
equipment
Defense
Precious metals
Nonmetallic and
industrial metal
mining
Coal
Petroleum and natural
gas
Utilities
Personal services
Business services
Computers
Electronic equipment
Measuring and control
equipment
Business supplies
Shipping containers
Transportation
Wholesale
Retail
Almost nothing
Industry
Table A.5
Top chemicals by industry
The Review of Financial Studies / v 35 n 2 2022
Financial Constraints and Corporate Environmental Policies
(1)
CAA
Text FC
(2)
CWA
(3)
CERCLA
0.260∗∗∗ 0.187∗∗∗
(0.059)
(0.055)
0.197∗∗∗
(0.055)
HM debt
log(assets)
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Year FE
Establishment FE
0.040
0.062∗∗
(0.033)
(0.030)
−0.023
0.083
(0.283)
(0.250)
0.097
0.092
(0.104)
(0.098)
∗
−0.427
−0.131
(0.244)
(0.250)
∗∗
0.085
0.066∗∗
(0.035)
(0.030)
0.044∗
0.007
(0.026)
(0.023)
45,143
47,360
.83
.84
Yes
Yes
Yes
Yes
0.054∗
(0.031)
−0.166
(0.269)
0.066
(0.098)
−0.267
(0.256)
0.077∗∗
(0.032)
0.008
(0.024)
49,410
.84
Yes
Yes
(4)
OSHA
(5)
CAA
(6)
CWA
(7)
CERCLA
(8)
OSHA
0.333∗∗∗
(0.077)
0.918∗∗∗ 0.546∗∗
0.865∗∗∗ 0.749∗∗
(0.273)
(0.245)
(0.254)
(0.352)
0.018
0.057
0.096∗∗∗ 0.069∗∗
0.055
(0.041)
(0.035)
(0.034)
(0.032)
(0.048)
−0.158
0.264
0.300
0.252
0.194
(0.315)
(0.246)
(0.236)
(0.236)
(0.305)
∗∗∗
0.432
0.048
0.000
0.044
0.043
(0.155)
(0.114)
(0.103)
(0.102)
(0.171)
0.007
−0.109
0.102
−0.022
0.094
(0.266)
(0.250)
(0.270)
(0.261)
(0.299)
∗∗∗
∗∗∗
∗∗∗
0.065
0.093
0.075
0.087
0.092∗∗
(0.043)
(0.031)
(0.029)
(0.029)
(0.043)
0.027
0.058∗∗ 0.027
0.031
0.027
(0.029)
(0.027)
(0.023)
(0.026)
(0.031)
33,143
31,627
33,359
35,275
23,411
.81
.86
.87
.87
.83
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
This table presents the results of OLS regressions of toxic releases under various EPA regulations on two textbased financial-constraint measures. The toxic release measures include the log tons of toxic releases administered
under the Clean Air Act(CAA), the Clean Water Act (CWA), the Comprehensive Environmental Response,
Compensation and Liability Act (CERCLA), and the Occupational Safety and Health Act (OSHA). Firm-level
controls include lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s q. The establishment-level
control is contemporaneous log(sales). Standard errors are clustered at the firm level. Standard errors appear in
parentheses. *p <.1; **p <.05; ***p <.01.
623
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 623
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.6
Toxic releases under various EPA Regulations
The Review of Financial Studies / v 35 n 2 2022
Table A.7
Chemicals grouped by per-plant and economywide magnitude
Text FC
HM debt
Observations
Adj. R-squared
Year FE
Establishment FE
Controls
# chemicals <=2
(1)
(2)
0.220∗∗
(0.109)
0.342
(0.570)
22,241
16,593
.79
.83
Yes
Yes
Yes
Yes
Yes
Yes
# chemicals >2
(3)
(4)
0.193∗∗∗
(0.062)
0.628∗
(0.330)
28,251
19,304
.87
.89
Yes
Yes
Yes
Yes
Yes
Yes
Full sample
(5)
(6)
0.190∗∗∗
(0.069)
0.562∗
(0.333)
51,390
36,546
.83
.86
Yes
Yes
Yes
Yes
Yes
Yes
B. Top-three chemicals ranked by per-plant magnitude
Text FC
HM debt
Observations
Adj. R-squared
Year FE
Establishment FE
Controls
# chemicals <=2
(1)
(2)
0.220∗∗
(0.109)
0.342
(0.570)
22,241
16,593
.79
.83
Yes
Yes
Yes
Yes
Yes
Yes
# chemicals >2
(3)
(4)
0.193∗∗∗
(0.062)
0.628∗
(0.330)
28,251
19,304
.87
.89
Yes
Yes
Yes
Yes
Yes
Yes
Full sample
(5)
(6)
0.190∗∗∗
(0.069)
0.562∗
(0.333)
51,390
36,546
.83
.86
Yes
Yes
Yes
Yes
Yes
Yes
C. Chemicals ranked by economywide magnitude
Top 5
(1)
Text FC
16,224
.87
Yes
Yes
Yes
(3)
0.154∗∗∗
(0.051)
0.087
(0.086)
HM debt
Observations
Adj. R-squared
Year FE
Establishment FE
Controls
Top 30
(2)
0.130
(0.402)
10,817
.89
Yes
Yes
Yes
38,516
.86
Yes
Yes
Yes
(4)
0.486∗∗
(0.237)
27,282
.88
Yes
Yes
Yes
Below top 30
(5)
(6)
0.139∗
(0.071)
38,615
.79
Yes
Yes
Yes
0.591∗
(0.344)
26,943
.82
Yes
Yes
Yes
This table presents our baseline results for the volume of the number one (panel A) and top-three (panel B)
chemicals (measured by tons in logarithm) within each establishment. Columns 1 and 2 present results for
establishments emitting only one or two chemicals; columns 3 and 4 present results for establishments emitting
more than two chemicals; and columns 5 and 6 present results for the full sample. Panel C presents our baseline
results for the volume of the number one, the top-30, and the rest of the chemicals ranked by aggregated total
release in the sample. Firm-level controls include lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and
Tobin’s q. The establishment-level control is contemporaneous log(sales). Standard errors are clustered at the
firm level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
624
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 624
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
A. Number one chemical ranked by per-plant magnitude
Financial Constraints and Corporate Environmental Policies
Health effects
(1)
(2)
Text FC
0.192∗∗∗
(0.072)
0.563∗
HM debt
Observations
Adj. R-squared
Year FE
Establishment FE
Controls
(0.336)
35,632
.87
Yes
Yes
Yes
50,034
.85
Yes
Yes
Yes
(1)
Text FC
44,531
.84
Yes
Yes
Yes
0.307∗∗∗
(0.091)
49,829
.82
Yes
Yes
Yes
B. Physical properties
Air
(2)
0.233∗∗∗
(0.077)
HM debt
Observations
Adj. R-squared
Year FE
Establishment FE
Controls
A. Health effects
Hazard score
(3)
(4)
0.759∗
(0.419)
31,126
.87
Yes
Yes
Yes
0.985∗∗∗
(0.343)
35,483
.86
Yes
Yes
Yes
No health effects
(5)
(6)
0.138
(0.163)
0.537
(0.556)
12,789
.86
Yes
Yes
Yes
19,622
.84
Yes
Yes
Yes
Water
(3)
(4)
0.151
(0.128)
13,698
.88
Yes
Yes
Yes
0.268
(0.463)
9,616
.90
Yes
Yes
Yes
Panel A presents our baseline results based on the health effects of chemicals. Columns 1 and 2 present results
for chemicals associated with adverse human health impacts; columns 3 and 4 present results for the EPA’s
Risk-Screening Environmental Indicators (RSEI) hazard score with toxic chemicals weighted by their relative
human health effects; and columns 5 and 6 present results for chemicals that are not associated with human health
impacts. In panel B, we group chemicals by their physical properties: columns 1 and 2 present results for chemicals
released through the air, and columns 3 and 4 present results for chemicals released through water. Firm-level
controls include lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s q. The establishment-level
control is contemporaneous log(sales). Standard errors are clustered at the firm level. Standard errors appear in
parentheses. *p <.1; **p <.05; ***p <.01.
625
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 625
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.8
Chemicals grouped by health effects and physical properties
The Review of Financial Studies / v 35 n 2 2022
(1)
Interest coverage
0.001
(0.000)
Rating
(2)
(3)
(4)
(5)
0.009∗∗∗
(0.003)
EDF
0.006
(0.007)
Leverage
0.075∗∗∗
(0.029)
−0.038
(0.045)
Dividend dummy
Payout ratio
log(assets)
Cash/assets
CAPEX/PPE
Tangible
Tobin’s q
log(sales)
Observations
Adj. R-squared
Year FE
Establishment FE
(6)
0.011
(0.028)
−0.206
(0.232)
−0.031
(0.083)
−0.283
(0.196)
−0.053∗∗
(0.027)
0.028
(0.020)
85,096
.78
Yes
Yes
−0.030
(0.039)
−0.666∗∗
(0.299)
0.012
(0.122)
−0.275
(0.233)
−0.060∗
(0.033)
0.026
(0.023)
64,674
.80
Yes
Yes
0.009
(0.027)
−0.052
(0.227)
−0.030
(0.082)
−0.293
(0.198)
−0.019
(0.026)
0.021
(0.020)
80,895
.79
Yes
Yes
0.010
(0.027)
−0.132
(0.223)
0.007
(0.081)
−0.295
(0.195)
−0.038
(0.026)
0.030
(0.020)
86,134
.78
Yes
Yes
0.014
(0.027)
−0.157
(0.222)
0.003
(0.081)
−0.300
(0.195)
−0.040
(0.026)
0.030
(0.020)
86,201
.78
Yes
Yes
−0.031
(0.035)
0.010
(0.027)
−0.149
(0.223)
−0.005
(0.081)
−0.307
(0.195)
−0.039
(0.026)
0.030
(0.020)
86,140
.78
Yes
Yes
This table presents results of OLS regressions of total toxic releases (measured by tons in logarithm) on several
accounting-based financial-constraint measures: interest coverage, credit rating, expected default frequency
(EDF), leverage, a dividend dummy, and the payout ratio. Firm-level controls include lagged log(assets),
Cash/assets, CAPEX/PPE, Tangible, and Tobin’s q. The establishment-level control is contemporaneous
log(sales). Standard errors are clustered at the firm level. Standard errors appear in parentheses. *p <.1; **p
<.05; ***p <.01.
626
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 626
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.9
Accounting-based financial constraint measures
627
Page: 627
576–635
50,805
.83
Yes
Yes
Yes
0.000
(0.000)
40,828
.84
Yes
Yes
Yes
−0.002
(0.017)
(3)
51,142
.83
Yes
Yes
Yes
−0.000
(0.004)
0.219∗∗∗
(0.073)
(4)
51,364
.83
Yes
Yes
Yes
−0.002
(0.112)
0.221∗∗∗
(0.072)
(5)
51,413
.83
Yes
Yes
Yes
0.032
(0.064)
0.224∗∗∗
(0.073)
(6)
−0.069
(0.043)
51,386
.83
Yes
Yes
Yes
0.220∗∗∗
(0.073)
35,945
.86
Yes
Yes
Yes
0.564
(0.363)
0.000
(0.000)
(7)
28,504
.86
Yes
Yes
Yes
−0.011
(0.018)
0.986∗∗
(0.417)
(8)
36,237
.86
Yes
Yes
Yes
−0.001
(0.004)
0.658∗
(0.364)
(9)
36,514
.86
Yes
Yes
Yes
−0.042
(0.119)
0.679∗
(0.359)
(10)
36,562
.86
Yes
Yes
Yes
−0.024
(0.064)
0.651∗
(0.359)
(11)
−0.093∗
(0.053)
36,532
.86
Yes
Yes
Yes
0.647∗
(0.360)
(12)
In this table, we report the results of testing the robustness of our baseline results against additional accounting-based financial-constraint measures: interest coverage, credit rating,
expected default frequency (EDF), leverage, a dividend dummy, and the payout ratio. Firm-level controls include lagged log(assets), Cash/assets, CAPEX/PPE, Tangible, and Tobin’s
q. The establishment-level control is contemporaneous log(sales). Standard errors are clustered at the firm level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
Observations
Adj. R-squared
Establishment FE
Year FE
Controls
Payout ratio
Dividend dummy
Leverage
EDF
Rating
Interest coverage
HM debt
(2)
0.349∗∗∗
(0.083)
(1)
0.217∗∗∗
(0.073)
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Text FC
Table A.10
Robustness: Accounting-based financial constraint measures
Financial Constraints and Corporate Environmental Policies
The Review of Financial Studies / v 35 n 2 2022
(1)
Rep(yes/no)
log(market value of assets)
Tobin’s q
Preinvestment earnings/BVA
0.378∗∗∗
(0.023)
−0.088∗∗
(0.036)
2.829∗∗∗
(0.480)
log(1+ foreign earnings (3 yr))
Foreign earnings (3 yr)>0 (1 if yes)
log(1+perm reinvested earnings)
Perm reinvested earnings>0 (1 if yes)
Estimated repatriation tax/MVA
Tax loss carryforward/MVA
Observations
Adj. R-squared
6,032
.13
(2)
Rep(yes/no)
0.187∗∗∗
(0.041)
−0.083∗
(0.049)
2.026∗∗∗
(0.729)
0.065
(0.045)
1.983∗∗∗
(0.216)
0.086
(0.069)
0.127
(0.410)
28.231
(22.683)
−1.365∗∗∗
(0.510)
5,358
.27
(3)
Rep consider
0.185∗∗∗
(0.020)
−0.059∗∗∗
(0.023)
1.009∗∗∗
(0.294)
0.003
(0.034)
1.799∗∗∗
(0.141)
0.077
(0.052)
0.720∗∗
(0.287)
15.310
(15.497)
−0.317∗∗
(0.138)
5,358
.19
This table presents the results of estimating the probability of repatriation using cross-sectional logistic regressions.
For columns 1 and 2, the dependent variable is an indicator variable that takes the value of one when a firm
repatriated foreign income under the AJCA in 2004 or later and zero otherwise. The independent variables are
based on values in 2003 or earlier. Ordered logit model estimation results are shown in column 3. The dependent
variable takes the value of two if a firm repatriated foreign earnings under the AJCA, the value of one if it discussed
repatriation of foreign earnings under the AJCA in its 10-K but did not repatriate, and zero otherwise. Standard
errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
628
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 628
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.11
Tax holidays (AJCA): First stage
Financial Constraints and Corporate Environmental Policies
MSA HPI
(1)
MSA supply elasticity * Mortgage rate
0.027∗∗∗
(0.005)
1 quartile elasticity * Mortgage rate
2 quartile elasticity * Mortgage rate
3 quartile elasticity * Mortgage rate
Observations
Adj. R-squared
Year FE
MSA FE
1,246
.94
Yes
Yes
(2)
-0.062∗∗∗
(0.008)
-0.046∗∗∗
(0.008)
-0.014∗∗
(0.007)
1,246
.94
Yes
Yes
In this table we report the first stage regression estimates of MSA-level HPI on the interaction between local
housing supply elasticity and U.S. 30-year fixed mortgage rates for the period running from 1993 through 2007
(Chaney, Sraer, and Thesmar 2012). For column 1, we use MSA-level local housing supply elasticity (Saiz 2010),
and, for column, 2 we use quartiles of the elasticity. MSA and year fixed effects are included in all specifications
and standard errors are clustered at the MSA level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p
<.01.
629
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 629
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.12
The collateral channel: Real estate shock first stage
The Review of Financial Studies / v 35 n 2 2022
Table A.13
Flow-induced price pressure and abnormal returns
DGTW-adjusted returns (%)
Qtr 0
Qtr 1-12
Quintiles
−5.229
-0.665
1.906
5.299
17.535
−1.112
−0.819
−0.155
0.757
2.762
0.978
1.077
0.608
−0.348
−1.128
Decile 10
Inflow
27.919
3.359
-2.551
This table reports the average FIPP, DGTW-adjusted returns for event quarter 0 (the portfolio-formation quarter),
and DGTW-adjusted cumulative abnormal returns from quarter 1 through quarter 12 for each quintile and the
highest decile (inflow) of the mutual fund flow sample.
Table A.14
Toxic releases and legal liabilities
A. Toxic releases and legal liabilities
Pr(investigation)%
Pr(legal_liab>0)%
(1)
(2)
log(total release)
log(sales)
Observations
Adj. R-squared
Year FE
Establishment FE
0.153∗∗∗
(0.038)
−0.000
(0.150)
91,725
.15
Yes
Yes
0.117∗∗∗
(0.030)
−0.202
(0.132)
91,725
.11
Yes
Yes
0.011∗∗∗
(0.003)
−0.023
(0.014)
91,725
.10
Yes
Yes
B. Lagged toxic releases and legal liabilities
Pr(investigation)%
Pr(legal_liab>0)%
(1)
(2)
(3)
(4)
Lagged.log(total release)
log(total release)
Lagged.log(sales)
log(sales)
Observations
Adj. R-squared
Industry-year FE
Model
0.415∗∗∗
(0.120)
0.744∗∗∗
(0.118)
0.453∗∗
(0.184)
−0.028
(0.177)
70,412
Yes
Logit
0.313∗∗∗
(0.071)
0.561∗∗∗
(0.071)
0.602∗∗∗
(0.215)
−0.018
(0.208)
77,133
.05
Yes
OLS
0.339∗∗∗
(0.112)
0.542∗∗∗
(0.111)
0.297∗
(0.166)
−0.062
(0.162)
64,686
Yes
Logit
log(legal_liab)
(3)
0.238∗∗∗
(0.059)
0.373∗∗∗
(0.059)
0.396∗∗
(0.181)
−0.081
(0.178)
77,133
.04
Yes
OLS
log(legal_liab)
(5)
0.026∗∗∗
(0.006)
0.043∗∗∗
(0.006)
0.045∗∗
(0.022)
−0.008
(0.021)
77,133
.05
Yes
OLS
Panel A reports the regression estimates of the effects of total toxic releases on government agency enforcement
actions, including the likelihood of investigation, the likelihood of positive legal liabilities being imposed
(including federal and local penalties, compliance and recovery costs, and the costs of supplemental environmental
projects), and the log dollar amount of legal liabilities. Coefficient estimates of the OLS regressions and marginal
effects of the logistic regressions are presented. log(sales) is included to control for production volume at the
establishment level. Standard errors are clustered at the establishment level. Standard errors appear in parentheses.
*p <.1; **p <.05; ***p <.01.
630
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 630
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
FIPP(%)
Low (outflow)
2
3
4
High (inflow)
Financial Constraints and Corporate Environmental Policies
A. Toxic releases under the CAA
Pr(investigation)%
Pr(legal_liab>0)%
(1)
(2)
(3)
(4)
Nonattain*log(CAA)
log(CAA)
Nonattainment
log(sales)
Observations
Adj. R-squared
Industry-year FE
Model
0.486∗∗∗
(0.096)
0.800∗∗∗
(0.064)
0.044
(0.354)
0.514∗∗∗
(0.093)
72,449
Yes
Logit
0.497∗∗∗
(0.078)
0.541∗∗∗
(0.039)
0.900∗∗∗
(0.224)
0.675∗∗∗
(0.099)
79,202
.05
Yes
OLS
0.346∗∗∗
(0.084)
0.621∗∗∗
(0.056)
0.171
(0.300)
0.281∗∗∗
(0.080)
66,108
Yes
Logit
0.366∗∗∗
(0.065)
0.373∗∗∗
(0.031)
0.695∗∗∗
(0.182)
0.360∗∗∗
(0.076)
79,202
.04
Yes
OLS
log(legal_liab)
(5)
0.037∗∗∗
(0.007)
0.045∗∗∗
(0.003)
0.067∗∗∗
(0.020)
0.043∗∗∗
(0.008)
79,202
.05
Yes
OLS
B. Total toxic releases
Pr(investigation)%
(1)
(2)
Nonattain*log(total release)
log(total release)
Nonattainment
log(sales)
Observations
Adj. R-squared
Industry-year FE
Model
0.323∗∗∗
(0.081)
0.944∗∗∗
(0.057)
0.257
(0.342)
0.422∗∗∗
(0.082)
85,096
Yes
Logit
0.463∗∗∗
(0.077)
0.644∗∗∗
(0.038)
0.682∗∗∗
(0.180)
0.586∗∗∗
(0.087)
92,746
.05
Yes
OLS
Pr(legal_liab>0)%
(3)
(4)
0.226∗∗∗
(0.071)
0.709∗∗∗
(0.050)
0.341
(0.288)
0.224∗∗∗
(0.071)
78,012
Yes
Logit
0.354∗∗∗
(0.065)
0.428∗∗∗
(0.029)
0.506∗∗∗
(0.144)
0.303∗∗∗
(0.067)
92,746
.04
Yes
OLS
log(legal_liab)
(5)
0.035∗∗∗
(0.007)
0.050∗∗∗
(0.003)
0.049∗∗∗
(0.015)
0.036∗∗∗
(0.007)
92,746
.05
Yes
OLS
In this table, we report the results obtained by examining the relationship between toxic releases and legal liabilities
across attainment and nonattainment counties. Outcome variables include the likelihood of government agency
investigations, the likelihood of positive legal liabilities being imposed (including federal and local penalties,
compliance and recovery costs, and the costs of supplemental environmental projects), and the log dollar amount
of legal liabilities. Coefficient estimates of the OLS regressions and marginal effects of the logistic regressions
are presented. log(sales) is included to control for production volume at the establishment level. Standard errors
are clustered at the establishment level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
631
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 631
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Table A.15
Legal liabilities and nonattainment status
The Review of Financial Studies / v 35 n 2 2022
Table A.16
Robustness: Nonattainment status and large polluters
Large polluter
Nonattainment status
(1)
Nonattainment=1 × Text FC
Nonattainment=1 × HM debt
Large polluter=1 × HM debt
Text FC
HM debt
log(assets)
Tobin’s q
Tangible
Cash/assets
CAPEX/PPE
log(sales)
Observations
Adj. R-squared
Year FE
Establishment FE
−1.393∗
(0.770)
70th percentile
(3)
−0.210∗
(0.116)
(4)
75th percentile
(5)
(6)
80th percentile
(7)
(8)
−0.218∗
(0.114)
−0.272∗∗
(0.117)
−0.228
−0.078
−0.081
(0.483)
(0.486)
(0.491)
0.312∗∗
0.121
0.088
0.066
(0.133)
(0.088)
(0.086)
(0.084)
1.604∗∗
0.696∗
0.627∗
0.648∗
(0.642)
(0.382)
(0.380)
(0.376)
0.047
0.039
−0.040 −0.000 −0.039 −0.008 −0.010
0.005
(0.045)
(0.053)
(0.034) (0.043) (0.035) (0.044) (0.037) (0.046)
0.079∗
0.059
0.081∗∗ 0.079∗∗ 0.077∗∗ 0.080∗∗ 0.082∗∗ 0.099∗∗∗
(0.047)
(0.043)
(0.034) (0.038) (0.032) (0.037) (0.033) (0.038)
0.182
0.196
−0.259
0.280 −0.107
0.400 −0.087
0.474
(0.326)
(0.324)
(0.251) (0.287) (0.251) (0.294) (0.255) (0.318)
0.319
0.481
−0.147
0.062 −0.183 −0.164 −0.076 −0.023
(0.359)
(0.322)
(0.279) (0.266) (0.298) (0.286) (0.312) (0.305)
0.226
0.156
0.168
0.198
0.170
0.138
0.067
0.041
(0.169)
(0.188)
(0.120) (0.132) (0.125) (0.139) (0.127) (0.137)
−0.036
−0.007
−0.013
0.026 −0.010
0.022 −0.006
0.020
(0.036)
(0.034)
(0.022) (0.024) (0.022) (0.024) (0.021) (0.025)
21,599
14,772
34,54523,100 31,984 21,311 29,321 19,131
.86
.87
.86
.88
.86
.88
.87
.89
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
In this table, we report results where we match establishments operated by the same parent firms but with different
nonattainment status and size. Large polluter indicates establishments with total toxic releases above the 70th,
75th, and 80th percentiles of the sample distribution. Firm-level controls include lagged log(assets), Cash/assets,
CAPEX/PPE, Tangible, and Tobin’s q. The establishment-level control is contemporaneous log(sales). Standard
errors are clustered at the firm level. Standard errors appear in parentheses. *p <.1; **p <.05; ***p <.01.
References
Akey, P., and I. Appel. 2021. The limits of limited liability: Evidence from industrial pollution. Journal of Finance
76:5–55.
Albuquerque, R., Y. Koskinen, and C. Zhang. 2019. Corporate social responsibility and firm risk: Theory and
empirical evidence. Management Science 65:4451–69.
Almeida, H., and M. Campello. 2007. Financial constraints, asset tangibility, and corporate investment. Review
of Financial Studies 20:1429–60.
Andersen, D. C. 2017. Do credit constraints favor dirty production? theory and plant-level evidence. Journal of
Environmental Economics and Management 84:189–208.
Baker, M., J. C. Stein, and J. Wurgler. 2003. When does the market matter? stock prices and the investment of
equity-dependent firms. Quarterly Journal of Economics 118:969–1005.
Barber, B. M., A. Morse, and A. Yasuda. 2021. Impact investing. Journal of Financial Economics 139:162–85.
Barnatchez, K., L. Crane, and R. Decker. 2017. An assessment of the national establishment time series (NETS)
database. Working Paper, Board of the Governor of the Federal Reserve System.
632
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 632
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Large polluter=1 × Text FC
−0.298∗
(0.172)
(2)
Financial Constraints and Corporate Environmental Policies
Baron, D. P. 2001. Private politics, corporate social responsibility, and integrated strategy. Journal of Economics
& Management Strategy 10:7–45.
Becker, R., and V. Henderson. 2000. Effects of air quality regulations on polluting industries. Journal of political
Economy 108:379–421.
Becker, R. A. 2005. Air pollution abatement costs under the clean air act: evidence from the pace survey. Journal
of environmental economics and management 50:144–69.
Bénabou, R., and J. Tirole. 2010. Individual and corporate social responsibility. Economica 77:1–19.
Bodnaruk, A., T. Loughran, and B. McDonald. 2015. Using 10-k text to gauge financial constraints. Journal of
Financial and Quantitative Analysis 50:623–46.
Campello, M., J. R. Graham, and C. R. Harvey. 2010. The real effects of financial constraints: Evidence from a
financial crisis. Journal of Financial Economics 97:470–87.
Chaney, T., D. Sraer, and D. Thesmar. 2012. The collateral channel: How real estate shocks affect corporate
investment. American Economic Review 102:2381–409.
Chava, S. 2014. Environmental externalities and cost of capital. Management Science 60:2223–47.
Cheng, I.-H., H. G. Hong, and K. Shue. 2016. Do managers do good with other peoples’ money? Working Paper,
Dartmouth College.
Chodorow-Reich, G. 2013. The employment effects of credit market disruptions: Firm-level evidence from the
2008–9 financial crisis. Quarterly Journal of Economics 129:1–59.
Cohn, J. B., and M. I. Wardlaw. 2016. Financing constraints and workplace safety. Journal of Finance 71:2017–58.
Copeland, B., and S. Taylor. 2003. Trade and the environment: Theory and evidence. Princeton, NJ: Princeton
University Press.
—– 2004. Trade, growth, and the environment. Journal of Economic literature 42:7–71.
Coval, J., and E. Stafford. 2007. Asset fire sales (and purchases) in equity markets. Journal of Financial Economics
86:479–512.
Currie, J., J. G. Zivin, J. Mullins, and M. Neidell. 2014. What do we know about short- and long-term effects of
early-life exposure to pollution? Annual Review of Resource Economics 6:217–47.
Daniel, K., M. Grinblatt, S. Titman, and R. Wermers. 1997. Measuring mutual fund performance with
characteristic-based benchmarks. Journal of finance 52:1035–58.
Dimson, E., O. Karakaş, and X. Li. 2015. Active ownership. Review of Financial Studies 28:3225–68.
Dunn, J., S. Fitzgibbons, and L. Pomorski. 2017. Assessing risk through environmental, social and governance
exposures. Journal of Investment Management.
Eccles, R. G., I. Ioannou, and G. Serafeim. 2014. The impact of corporate sustainability on organizational
processes and performance. Management Science 60:2835–57.
Edmans, A. 2011. Does the stock market fully value intangibles? employee satisfaction and equity prices. Journal
of Financial Economics 101:621–640. ISSN 0304-405X.
Edmans, A., I. Goldstein, and W. Jiang. 2012. The real effects of financial markets: The impact of prices on
takeovers. Journal of Finance 67:933–71.
Farre-Mensa, J., and A. Ljungqvist. 2016. Do measures of financial constraints measure financial constraints?
Review of Financial Studies 29:271–308.
Faulkender, M., and M. Petersen. 2012. Investment and capital constraints: Repatriations under the american
jobs creation act. Review of Financial Studies 25:3351–88.
633
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 633
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Berger, E. 2019. Selection bias in mutual fund flow-induced fire sales: Causes and consequences. Working Paper,
Cornell University.
The Review of Financial Studies / v 35 n 2 2022
Fernando, C. S., M. P. Sharfman, and V. B. Uysal. 2017. Corporate environmental policy and shareholder value:
Following the smart money. Journal of Financial and Quantitative Analysis 52:2023–51.
Gormley, T. A., and D. A. Matsa. 2011. Growing out of trouble? corporate responses to liability risk. Review of
Financial Studies 24:2781–821.
Gredil, O., N. Kapadia, and J. H. Lee. 2019. Do rating agencies deserve some credit? evidence from transitory
shocks to credit risk. Working Paper, Tulane University.
Henderson, J. V. 1996. Effects of air quality regulation. The American Economic Review 86:789–813.
Hoberg, G., and V. Maksimovic. 2014. Redefining financial constraints: A text-based analysis. Review of Financial
Studies 28:1312–52.
Hoepner, A. G., I. Oikonomou, Z. Sautner, L. T. Starks, and X. Zhou. 2018. ESG shareholder engagement and
downside risk. Working Paper, University College Dublin.
Hong, H., and M. Kacperczyk. 2009. The price of sin: The effects of social norms on markets. Journal of Financial
Economics 93:15–36.
Hong, H., J. D. Kubik, and J. A. Scheinkman. 2012. Financial constraints on corporate goodness. Working Paper,
Columbia University.
Karpoff, J. M., J. R. Lott, and E. W. Wehrly. 2005. The reputational penalties for environmental violations:
Empirical evidence. Journal of Law and Economics 48:653–75.
Khan, M., L. Kogan, and G. Serafeim. 2012. Mutual fund trading pressure: Firm-level stock price impact and
timing of seos. Journal of Finance 67:1371–95.
Kitzmueller, M., and J. Shimshack. 2012. Economic perspectives on corporate social responsibility. Journal of
Economic Literature 50:51–84.
Levinson, A. 2009. Technology, international trade, and pollution from us manufacturing. American economic
review 99:2177–92.
———. 2015. A direct estimate of the technique effect: changes in the pollution intensity of us manufacturing,
1990–2008. Journal of the Association of Environmental and Resource Economists 2:43–56.
Lins, K. V., H. Servaes, and A. Tamayo. 2017. Social capital, trust, and firm performance: The value of corporate
social responsibility during the financial crisis. Journal of Finance 72:1785–824.
Lou, D. 2012. A flow-based explanation for return predictability. Review of Financial Studies 25:3457–89.
Margolis, J. D., H. Anger Elfenbein, and J. P. Walsh. 2009. Does it pay to be good ... and does it matter? A
meta-analysis of the relationship between corporate social and financial performance. Working Paper, Harvard
University.
Masulis, R. W., and S. W. Reza. 2015. Agency problems of corporate philanthropy. Review of Financial Studies
28:592–636.
Mian, A., and A. Sufi. 2011. House prices, home equity–based borrowing, and the us household leverage crisis.
American Economic Review 101:2132–56.
Saiz, A. 2010. The geographic determinants of housing supply. Quarterly Journal of Economics 125:1253–96.
Shapira, R., and L. Zingales. 2017. Is pollution value-maximizing? The Dupont case. Working Paper, Law School,
IDC.
Shapiro, J. S., and R. Walker. 2018. Why is pollution from us manufacturing declining? the roles of environmental
regulation, productivity, and trade. American Economic Review 108:3814–54.
Starks, L. T., P. Venkat, and Q. Zhu. 2017. Corporate esg profiles and investor horizons. Working Paper, University
of Texas at Austin.
634
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 634
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
Hasbrouck, J. 2009. Trading costs and returns for us equities: Estimating effective costs from daily data. Journal
of Finance 64:1445–77.
Financial Constraints and Corporate Environmental Policies
Walker, W. R. 2013. The transitional costs of sectoral reallocation: Evidence from the clean air act and the
workforce. Quarterly Journal of Economics 128:1787–835.
Wardlaw, M. 2020. Measuring mutual fund flow pressure as shock to stock returns. Journal of Finance
75:3221–43.
[07:44 18/12/2021 RFS-OP-REVF210059.tex]
Page: 635
576–635
Downloaded from https://academic.oup.com/rfs/article/35/2/576/6265483 by National University of Singapore user on 19 January 2024
635