S0022381613000170 506..519


Compliance Bias and Environmental (In)Justice

David M. Konisky Georgetown University
Christopher Reenock Florida State University

Scholarship on race- and class-based disparities in regulatory outcomes has failed to provide a theoretically
grounded account of this bias’ origin. We address this shortcoming by providing a microlevel explanation of how
demographics influence compliance bias or the failure to detect noncompliant firms. We argue that regulatory
compliance is best understood as a dual-agent—firm and regulatory officer—production function, and that
community mobilization and agency decision-making authority shape bureaucrats’ incentives to report non-
compliance. We test our argument with an original dataset on community mobilization and agency structure that
delineates the political costs and benefits of state regulatory officers implementing the U.S. Clean Air Act. Using
detection-controlled estimation, we find that while certain communities are vulnerable to compliance bias, such
bias is mitigated in the presence of either politically mobilized communities or decentralized enforcement authority
within the implementing agency.

E
qual protection under the law is a fundamental
principle of democratic public policy. When
governments fail to treat citizens equally, the

very legitimacy of democracy is threatened.1 Accord-
ingly, the roots of equal protection extend deeply and
broadly in the discipline. Scholars recognize its impor-
tance as a key component of the rule of law and the civil
liberties protected in society (Maravall and Przeworski
2003). Bureaucracy scholars refer to doctrines of admin-
istrative fairness and representation (Meier 1993) and
judicial scholars to impartiality and neutrality (Raz
1977). The core questions on equal protection turn on
whether government is treating citizens differently, and,
if so, what remedies might reduce such inequities?

In approaching this question, scholars have inves-
tigated policy domains that are ripe for violations of
equal protection such as housing, pay, environment,
education, and employment. Of these areas, environ-
mental protection is particularly salient because of the
potential for adverse public health effects. Advocates
for ‘‘environmental justice’’ often allege that minority
and low-income communities experience dispropor-
tionate environmental hazards, in part, as a result
of unequal enforcement of laws (e.g., Bryant 1995;
Bullard, 1993; Bullard & Johnson 2000). Some recent

research investigating these claims has found that
governments conduct fewer inspections and impose
fewer punitive sanctions when firms are located in
poor and minority areas, but other work has identified
few such disparities (Dion, Lanoie & Laplante 1998;
Gray and Shadbegian 2004; Helland 1998b; Konisky
2009; Konisky and Schario 2010; Scholz and Wang
2006).

Notwithstanding the mixed results, these studies
share a common approach of establishing correlations
between community demographics and regulatory
enforcement outputs. However, while demonstrating
these correlations helps to diagnose the presence of
enforcement bias, prior research has not developed a
strong theoretical account of the sources of this bias.
Enforcement actions rely upon initial determinations
of firm compliance, which are generally ascertained
through government detection efforts. As a result, ob-
served correlations between community demographics
and enforcement actions may be either evidence of bias
originating in an agency’s decision to take that action
or evidence of bias originating in the initial compliance
determination. Moreover, bias in regulatory outputs is
generated from a process that involves the strategic
interaction of both firms and regulatory officers.

The Journal of Politics, Vol. 75, No. 2, April 2013, Pp. 506–519 doi:10.1017/S0022381613000170

� Southern Political Science Association, 2013 ISSN 0022-3816

1
An online appendix for this article is available at http://journals.cambridge.org/jop containing supplemental analyses. Data and

supporting materials necessary to reproduce the numerical results in the article will be made available at http://mailer.fsu.edu/
;creenock/ by January 2013.

506



Observed correlations between community demo-
graphics and regulatory outputs may reflect a firm’s
decision to be compliant, an officer’s decision to
accurately identify noncompliant firms, or both. When
studying regulatory outputs alone, these sources of
bias are observationally equivalent.

We believe a more productive path for revealing
the sources of bias is to focus on compliance. Com-
pliance is the foundation upon which other regulatory
outputs are based, and for this reason, studying firm
compliance is not vulnerable to the first problem
noted above. In this sense, exploring environmental
justice issues within a compliance framework high-
lights the core process within which we believe bias is
likely to originally arise.2 Second, while compliance
does entail dual-agent (firm and officer) production,
we can explicitly model each data-generating process,
allowing us to determine if the bias originates with
firms and/or regulators.

In this article, we develop a novel theoretical
account of compliance bias, or the systematic non-
detection of violations, in the context of environmental
justice. Building on past work (e.g., Feinstein 1990;
Helland 1998a, 1998b; Scholz and Wang 2006), we
argue that compliance bias is best characterized as a
dual-agent production function of firms and regulatory
officers. While pursuing different goals, both firms and
bureaucrats attempt to minimize their costs. We argue
that the costs associated with noncompliance (firms)
and failure to detect noncompliance (bureaucrats) are
lower in poor and minority communities because these
communities have fewer resources with which to
document and protest noncompliance. For this reason,
firms in these areas are more likely to be noncompliant
and less likely to be detected by regulatory officers.

This dual-agent approach reveals two potential
remedies for such bias. First, politically mobilized
communities are better equipped to generate compli-
ance costs for firms and political costs for bureaucrats,
and, as a result, bias will diminish in poor and minority
communities that have overcome the collective action
problems necessary to exert political pressure. Second,
we argue that high-level agency managers have a strong
incentive to overreport compliance rates in order to
craft positive impressions of agency performance. For
this reason, they are more likely to engage in what we
refer to as motivated, nondetection of compliance when a
firm is located in a poor or minority area, since these

communities have fewer political resources and are
therefore less likely to discover the bureaucrat’s behavior.

The empirical setting for our analysis is firm-level
compliance with the federal Clean Air Act (CAA).
Using an original dataset on both local mobilization
and the decision-making authority of the state regu-
latory officers largely responsible for implementing
the CAA, we employ detection-controlled estimation
(DCE) to model the effects of demographic character-
istics on both individual firm and regulatory officer
compliance decisions. We find that compliance bias is
more likely in Hispanic (but, not in African American)
communities, but that such bias is mitigated in the
presence of either politically mobilized communities or
decentralized enforcement authority. Our results sug-
gest that compliance bias in policy implementation can
be curbed not just by investing in the political capacity
of communities but also by modifying agency decision-
making structures.

The article proceeds as follows. We begin by
developing our theoretical expectations about the link-
age between compliance bias and community demo-
graphics. We then discuss our research design and
explain the usefulness of using a DCE approach to test
these expectations. Next we discuss our empirical
analysis of compliance bias in the case of enforcement
of the CAA and then conclude with a discussion of the
implications.

Theory and Hypotheses

Models of regulatory compliance recognize two distinct
processes: Firms make decisions over compliance, and
regulatory officers make decisions over compliance
determination. Strategic-deterrence models provide ba-
sic expectations that are useful for our purposes. In such
models, firms and bureaucrats base their behavior on an
expected utility calculation in which each attempts to
maximize their expected payoffs given their beliefs
about the other’s action (Braithwaite and Makkai
1991; Scholz 1991; Winter and May 2001). Of the many
outcomes of this strategic interaction, we are interested
in features of the policy environment that may increase
the potential for ‘‘opportunistic’’ behavior—that is,
when firms pursue noncompliance based on their lower
expectations of being caught and/or punished by
regulatory officers. The end result of this outcome is
undetected noncompliance—the object of our study.

We assume that firms choose whether to remain
in compliance or to violate the law. In making this
choice, firms are interested in maximizing their
individual profit at the lowest cost of compliance.

2
A few studies look at disparities in firm compliance with mixed

findings (Earnhart 2004a, 2004b; Mennis 2005; Scholz and Wang
2006).

compliance bias and environmental (in)justice 507



The expected costs of compliance are related to various
features of the market and to policy and political
factors that shape the consequences of noncompliant
behavior. Expectations about the costs of compliance
are also related to the detection efforts pursued by
regulatory officers, and specifically, the likelihood that
noncompliant behavior will not only be discovered but
treated as violations. It is for this reason that one needs
to conceptualize compliance outcomes as a dual-agent
production function. When firms expect agency offi-
cials to pursue less rigorous enforcement strategies,
they have greater incentive to avoid full compliance.
Alternatively, when firms believe that agency officials are
more likely to pursue a maximal deterrence strategy,
they will have greater incentive to stay in compliance.

We assume that bureaucrats choose whether to
determine a given firm’s compliance status. This deci-
sion is perhaps the most fundamental enforcement
activity for a regulatory officer, since it is not only the
first action in a long line of potential actions directed at
an individual firm but also represents an important
performance indicator for an agency at the aggregate
level. In making this choice, officers are interested in
maximizing the political benefits of successful detection
(i.e., getting credit for making correct compliance
determinations) while minimizing the political costs
of errors (i.e., getting blamed for making incorrect com-
pliance determinations). Political benefits stem from
being responsive to policy demands such as managing
the aggregate consequences of noncompliance (e.g.,
pollution levels, accident rates) and political demands
including the preferences of political principals and
other stakeholders. As we argue below, an important
element in being responsive is whether the agency is
viewed as being ‘‘effective.’’ In addition to the various
types of transaction costs involved for any given case
(i.e., search and information costs, bargaining and
decision costs, policing and enforcement costs), regu-
latory officers must also consider the political costs of
their decisions. Specifically they must weigh the con-
sequences of making wrong decisions, and we argue that
they will allocate their effort to minimize the net costs of
different incorrect compliance determinations.

Incorrect compliance decisions come in two
forms. First, the officer could wrongly attempt detec-
tion of a compliant firm (so called ‘‘harassment’’;
Scholz 1991). Second, the officer could fail to detect
a noncompliant firm. Each of these errors generates
different political costs from unique clientele. For the
first error, business interests are more likely to gen-
erate costs for officers, while environmental advocacy
groups are more likely to generate costs for the second
error. In the aggregate, at the level of a region or a

community, officers attempting to minimize one set of
errors necessarily increase the probability of the other
(this is of course not the case at the individual firm
level). On the whole, clientele generate cross-cutting
incentives for officers, who attempt to minimize costs
associated with making errors, offering officers a trade-
off. A rational regulatory officer, seeking to be respon-
sive to stakeholders, will attempt to minimize costs
generated from relevant clientele. We are interested in
how community demographics affect the second type
of errors, or failures to detect noncompliance. As we
will argue below, factors such as demographics alter the
relative political costs of these errors, making one error
more attractive than the other, ceteris paribus.

In the context of firm compliance, officers are
unable to fully and regularly assess the behavior of all
regulated entities. Even if bureaucrats are motivated
primarily by functional or intrinsic preferences (Brehm
and Gates 1997), they are still constrained by practical
limits. For this reason, bureaucrats may engage in some
satisficing behavior as a decision-making shortcut
(Simon 1976) or, alternatively, choose to strategically
allocate more effort to accurately determine the com-
pliance status of some firms more than others. This
could take the form of directing agency efforts to easier
cases (Wilson 1989) or directing resources to cases
depending on the estimation of the political costs of a
wrong decision. We assume that firms are aware that
officers face such limitations but are uncertain of
precisely which officer type they are dealing with—one
preferring less or more rigorous detection effort.

Given that firms and regulatory officers are un-
certain about each other’s preferences, they develop
beliefs using signals in the policy environment. These
signals inform a firm’s (agency’s) expectations about an
agency’s (firm’s) strategy, and they originate from
political and policy task factors alike (Potoski 1999;
Ringquist 1993; Scholz and Wei 1986; Scholz, Twombly
and Headrick 1991; Wood 1992). When signals from
the policy environment are sufficiently suggestive that
regulatory officers may pursue less rigorous detection,
firms will be more likely to risk noncompliance.

Community Demographics, Mobilization,
and Enforcement Authority

We posit that community characteristics shape firms’
and regulatory officers’ behavior by providing signals
about each others’ preferences over compliance and en-
forcement, respectively. For firms, demographics pro-
vide informative signals about the expected costs of
compliance. For reasons we specify below, bureaucrats
are less likely to devote their limited resources to

508 david m. konisky and christopher reenock



correctly detecting noncompliance when the firms are
located in less politically active areas. Because large
minority and poor communities tend to have fewer
political resources with which to engage in advocacy,
these communities as well as the regulatory officers who
serve them will be less likely to employ aggressive
detection efforts. Strategic firms respond accordingly
and should be more inclined to risk noncompliance,
leading to our first hypothesis:

H1: Firms in minority (or poor) neighborhoods are
more likely to be in noncompliance.

Community demographics may generate costs for
regulatory officers in several ways. First, failure to detect
noncompliance, if uncovered, can lead to objections
from interested stakeholders. Rational bureaucrats seek-
ing to minimize costs generated from a specific clientele
over incorrect decisions will devote fewer resources to
carefully determining compliance for cases where the
costs associated with wrongful detections are lower.
Conversely, they will dedicate more attention to those
cases where the costs of making incorrect decisions are
higher. To be clear, officers cannot simply ignore these
resource-intense cases to avoid the costs associated with
them. Doing so will generate either undetected non-
compliance or detection against compliant firms, both
of which generate costs, albeit from different clientele.
In this sense, the costs that a regulator faces for failing
to detect a noncompliant firm vary, depending on the
probability that an incorrect determination will be re-
vealed. Community characteristics enter this decision-
making equation because some communities have better
capacity to identify and dispute an incorrect compliance
decision. In particular, poor and minority communities
tend to have fewer political resources to demand and
secure accurate compliance decisions from bureaucrats.
This raises the relative attractiveness of compliance bias
in these communities.

A second reason regards the identification of
potential violations. In policy contexts where third-
party monitoring assists bureaucrats in detecting
problems, communities with reserve time, resources,
and social capital are better equipped to serve in this
role. Poor and minority communities are less likely to
possess this capacity. As a result, regulatory officers
are more likely to mischaracterize noncompliant
firms as compliant in these areas. Whichever reason
holds, the end result is the same—poor and minority
neighborhoods are less likely to be able to generate
political costs for regulatory officers. Therefore,
bureaucrats have less incentive to actively pursue
noncompliant firms in these communities, leading to
the following hypothesis:

H2: Bureaucrats are less likely to code violating firms
as noncompliant when they are located in minority
(or poor) communities.

The effect of community demographics on firm and
regulatory officer behavior, however, should not be
constant across all contexts. Rather, we argue that the
effect should be mediated by two contextual features—
the ability of a community to generate political costs
through effective mobilization and the vulnerability
of a regulatory officer to these costs. We discuss each
in turn.

Community Mobilization. Past work demonstrates
that communities with higher political capacity
are better able to influence firm decision making
(Hamilton 1993, 1995; Hamilton and Viscusi 1999),
and we argue that a similar dynamic exists with firm
compliance. While community demographics provide
signals to bureaucrats (and firms) about potential
political mobilization, they do not account for the
effects of actual mobilization. Politically mobilized
communities should alter the incentives of firms and
bureaucrats in decisions regarding compliance. Regard-
ing firms, communities that overcome collective action
problems and exercise their political voice can modify
the costs of compliance through direct pressure, ‘‘sham-
ing’’ poor performers, or litigation. Well-organized
communities should also increase the political costs to
a bureaucrat of making an incorrect decision, since they
can protest and publicize what they believe are any
erroneous decisions. In turn, these higher costs should
reduce the likelihood of officers wrongly classifying a
firm violating the law as compliant and of firms being
noncompliant. Two specific political-mobilization hy-
potheses follow from this logic:

H3a: Firms in minority (or poor) communities are
less likely to be noncompliant when they are located
in politically mobilized communities.

H3b: Bureaucrats are more likely to code violating
firms in minority (or poor) communities as noncom-
pliant when these communities are politically mobilized.

Enforcement Authority. To this point, we have implic-
itly assumed that all bureaucrats share similar incentive
structures when it comes to determining firm compli-
ance. Yet, within an agency, different bureaucrats have
varied tasks. Front-line agents focus on the day-to-day
activities of policy delivery, while high-level managers
focus on achieving overall organizational goals and
managing relationships with external stakeholders
(Hammond 1986). As a consequence, the incentives
that influence bureaucrats’ compliance decisions may
vary with their position in the agency. Critically
important for our purposes, not all states assign the

compliance bias and environmental (in)justice 509



same personnel final decisionmaking authority over
compliance decisions. Some states allow field officers
wide discretion over compliance decisions, while others
require final authorization from higher-level agency
managers. We argue that there is strong reason to
believe that, depending upon their position and task
responsibilities, bureaucrats are not equally vulnerable
to the political costs that a community may generate.

Specifically, we posit that compared to lower-level
field officers (e.g., career civil servant front-line com-
pliance officers), higher-level bureaucrats serving in
agency management roles (e.g., Department Secretaries,
Deputy Secretaries, Division Managers, or Regional
Directors) have additional incentives to inaccurately
characterize noncompliant firms as compliant. Com-
pliance rates are a key indicator by which stakeholders
evaluate the performance of regulatory agencies, and
they are often included in annual reports to state leg-
islatures and to federal overseers. In many areas of
policy, where it is difficult to precisely measure bureau-
cratic performance, for reasons of both asymmetric
information and causal ambiguity between agency
action and policy outcomes, stakeholders rely on these
types of metrics instead. The use of such performance
indicators often results in goal displacement where
agencies focus on producing metrics to satisfy external
stakeholders, rather than achieving more relevant policy
outcomes (Blau and Meyer 1971; Bohte and Meier
2000; Downs 1967). Knowing this, bureaucrats serving
in high-level management positions responsible for
crafting impressions of agency performance have a
stronger incentive to deliver high compliance rates.

High-level managers, faced with these incentives,
may be tempted to overlook noncompliant firms in
some circumstances. We refer to this behavior as
motivated, nondetection of compliance, and it is analo-
gous to Bohte and Meier’s (2000) idea of organizational
cheating in public agencies. Although bureaucrats enjoy
informational advantages over many stakeholders, there
are potentially adverse consequences to engaging in this
behavior, and strategic bureaucrats are more likely to
pursue motivated, nondetection of compliance when
the risks of being caught are smaller.3 It is here again
that we return to the important role of community
characteristics. Since there is a higher probability of the
deception not being detected, bureaucrats will be more

likely to deliberately mischaracterize a firm violating
the law as compliant when the community in which the
firm is located has fewer political resources (i.e., poor
and minority communities). To be clear, this incentive
is above and beyond the reasons stated previously
regarding the political mobilization capacity of these
communities, leading us to our final hypothesis:

H4: Bureaucrats who are more likely to be held
accountable for agency performance are more likely to
mischaracterize violating firms as compliant when these
firms are located in minority (or poor) communities.

Research Design

To test our expectations we employ DCE, a statistical
technique that enables us to jointly model the dual
production of compliance. Originally developed by
Feinstein (1990), DCE statistically controls for the
possibility that some portion of noncompliant firms
may remain undiscovered by regulatory officers. As a
result, facilities may be entered into a database as
compliant when they are actually violating the law.
DCE estimates the likelihood of a firm being non-
compliant, as well as the likelihood that a given entity
was correctly coded as noncompliant by a regulatory
officer. Failure to account for the two reasons for
observing compliance (actual compliance and the failure
to detect noncompliance), which is implicitly done when
compliance is modeled with a single-equation probit or
logistic regression specification, can bias inferences.

DCE techniques have been utilized to correct for
compliance bias in a variety of settings, including
taxpayer compliance (Feinstein 1999) and firm com-
pliance with environmental (Brehm and Hamilton
1996; Helland 1998a, 1998b; Scholz and Wang 2006),
occupational health and safety, (Feinstein 1990), and
food and drug (Olson 1995) regulation. The DCE
approach has two specific benefits for our purposes.
First, it corrects for bias in estimating demographic
effects in firm-compliance models. Second, it enables
us to explicitly account for the dual-agent production
problem. With DCE we can consider the sources of
bias by separately testing whether demographics
affect firm decisions on compliance and regulators’
determinations of facility compliance.

Specifically, we utilize Feinstein’s (1990) DCE
estimator, which consists of two binary choice models:
one that models the likelihood of a ‘‘true’’ violation
and a second that models the likelihood of detection of
‘‘true’’ compliance. Because the likelihood of a viola-
tion and the likelihood of detection are separately
unobservable, these likelihood functions are estimated

3To be clear, the mechanisms by which motivated nondetection
occur are not necessarily nefarious. We do not necessarily envision
managers explicitly manipulating compliance documents or turn-
ing the other way to blatant violations—although we do not rule
this out. Rather, such motivations may translate into directing
subordinates’ detection efforts either to or away from specific
geographical areas of concern, based upon political demands.

510 david m. konisky and christopher reenock



jointly via maximum-likelihood estimation. The DCE
estimator is equivalent to a bivariate probit model
with partial observability (Abowd and Farber 1982;
Poirier 1980). More formally, we model the observed
dichotomous compliance variable, zi, as the product of
two unobserved latent dichotomous variables, yi1 and
yi2, such that zi is only observed, where zi 5 yi1 X yi2. In
our application, we assign the interpretations of firm
noncompliance and agency detection to yi1 and yi2,
respectively, such that:

Zi ¼
1; yi1 and yi2 ¼1
0; otherwise

n

Thus, we expect to observe a firm being listed as
noncompliant (zi 5 1) when both the firm is non-
compliant and the agency has detected this noncom-
pliance. In all other combinations of firm and agency
behavior, we observe a case of compliance (zi 5 0),
which highlights the fact that we cannot distinguish
between true compliance and undetected violators.4 In
this model the probabilities are jointly and simulta-
neously determined with a correlated error term, r,

Pr z ¼ 1ð Þ¼ F2 x1b1; x2b2; rð Þ
Pr z ¼ 0ð Þ¼ 1 � F2 x1b1; x2b2; rð Þ

yielding the following likelihood to be estimated:

lnL ¼+N
i¼1fzilnF2ðx1b1; x2b2; rÞ
þ ð1 � ziÞ ln½1 � F2ðx1b1; x2b2; rÞ�g

Identification of this model requires an exclusion of
at least one exogenous variable to ensure that the
parameters being estimated in each model are not
identical. Moreover, identification is enhanced when
the exogenous variable exhibits sufficient variation
over the sample. This condition is likely to be met
when the exogenous variable is continuous (Poirier
1980). In our model, Signature Authority (which as we
describe below is our measure of agency structure) is a
continuous exogenous variable which we restrict to the
detection equation based on our theoretical argument
that the location of decision making in a state
administrative agency creates different incentives for
bureaucrats. We do not include Signature Authority in
the firm model given what we think is a reasonable
assumption that firms are largely unaware of this
feature of administrative agencies, and, even if they

were aware, it is unlikely to affect their compliance
costs. The other variables not included in both
equations are the firm-level indicators of past enforce-
ment. Because firms’ compliance costs are affected by
the temptation to cheat, which is in turn a function of
the likelihood of getting caught, we include measures
of past government inspections in the firm equation.
For their part, we argue that regulators will devote
their time and effort to ‘‘bad actors,’’ so we include
measures of whether recent violations resulted in a
punitive action in the detection equation.

The empirical setting for our analysis is the
regulatory compliance of individual firms within
the context of air pollution control across the U.S.
states. We use an original dataset that combines firm-
level compliance with the federal CAA with data on
community characteristics and contextual variables
across county, state administrative region, and U.S.
states. The federal government sets most standards
under the CAA, but the states have responsibility for
enforcing many of its provisions. To best capture the
policy and political demands on state regulatory
officers, we use novel data on how each state agency
divides enforcement responsibilities among bureau-
crats internal to the agency. We discuss these data in
greater detail below. Our attention to the details of
administrative program structure does, however,
generate a trade-off. Given the demands of gathering
these data, we can only examine a limited time frame
(2001–2004) for which we have compiled administra-
tive data. However, we do not believe that this trade-
off compromises our ability to draw valid inferences
because our hypotheses only require cross-sectional
variance. Moreover, by limiting our analysis to a single
Presidential term, we can hold constant variations in
national factors that might influence state enforcement
patterns. Of course, we must be cautious extending
our inferences beyond this temporal domain.

Measuring Compliance

The dependent variable, HPV Status, reflects whether
a regulated facility was designated as a High Priority
Violator (HPV) of the CAA. An HPV facility is one that
is failing to meet core CAA obligations, usually pollu-
tion performance standards. Dissimilar to minor paper-
work violations, noncompliance of this sort can trigger
significant punitive sanctions including substantial
monetary penalties. For this reason, designating a firm
as HPV is of high political relevance for regulatory
officers. HPV Status is a dichotomous variable that takes
a value of one for a facility designated as a HPV at any
time during the year, and zero otherwise. The variable is

4It is possible that there could be ‘‘false’’ noncompliance, but this
would be exceedingly rare in our context. HPV determinations
represent cases of severe noncompliance and typically include
major infractions of emissions standards, minimizing errors, even
by overly zealous regulators.

compliance bias and environmental (in)justice 511



measured on annual basis from 2001 to 2004, and we
examine all federally reportable facilities under the CAA.
The result is a panel dataset that includes a total of
160,896 cases (40,224 firms per year),5 although the
number of cases analyzed is slightly fewer due to missing
data.6 Table A.1 in the online appendix includes sources
and descriptive statistics for all data.

Measuring Community Characteristics

Environmental justice concerns typically emerge with
respect to race and class. We use two standard
measures in the literature to assess the racial compo-
sition of a community: Percentage of African-American
Population and Percentage of Hispanic Population. The
literature is less consistent when accounting for class,
with most using some combination of income, pov-
erty, and education measures. Rather than emphasize
any one measure, we constructed a scale based on four
standardized variables, median household income,
percent below poverty line, percent college educated,
and percent high school educated, each measured at
the zip-code level. The resulting scale, Class, obtained a
Cronbach’s alpha of .86, and an investigation of the
dimensionality of the scale revealed a single factor with
an eigenvalue of 2.73, accounting for 79% of the total
variance.

Measuring Community Political
Mobilization

To assess the degree of community political mobiliza-
tion, we require a measure that captures whether citizens
have overcome collective action costs and organized
around environmental justice concerns. Standard meas-
ures of political mobilization, such as voter turnout and
campaign donations, are too broad to capture our
concept of interest. Moreover, it is likely that it is
precisely in communities where traditional political
mobilization is low that the pressures for organization
around equity issues are highest. As a result, we use a
more direct measure: the presence of locally oriented
environmental justice advocacy organizations.

We assembled data from the People of Color
Environmental Groups Directory (Environmental

Justice Resource Center 2000), which publishes a list
of organizations whose activities include advocating
for environmental justice. Specifically, we coded the
presence of groups listed as ‘‘environmental justice
resource groups’’ at the zip-code level, as well as
information about the constituency served by each
group and its geographic focus. Using these data, we
first aggregated this group information up to the
three-digit zip-code level to take into account the
positive spillovers from the presence of a group, under
what we believe is a reasonable assumption that the
reach of these groups extends beyond the specific
mailing zip code of their main office.7 We then
constructed a scale based on five standardized varia-
bles, the total number of groups, the total number
of groups focused on African Americans, the total
number of groups focused on Hispanic groups, the
total number of groups with a neighborhood focus,
and the total number of groups with a local focus.
The resulting scale, Mobilization, obtained a Cronbach’s
alpha of .80, and an investigation of the dimensionality
of the scale revealed a single factor with an eigenvalue
of 2.54, which accounts for 92% of the total variance.
Section 3 of the online appendix discusses the meas-
urement of this variable in more detail.

Measuring Regulatory Officer Incentives

To test our hypothesis regarding motivated, nondetec-
tion of compliance, we must know whether the author-
ity to issue enforcement-related actions rests with field
officers, is centralized in the hands of high-level agency
officials, or lies somewhere in between. We believe that
a reasonable proxy is the location within the agency
where (or more precisely within whose hands) final
signature authority to issue enforcement actions lies.
We measure the locational authority over three sets of
enforcement actions (Reenock and Gerber 2008). For
each state air-pollution control agency, this measure
divides (Final Authority—1) by (Vertical Depth—1),
where Final Authority represents the location of final
signature authority for a given action within the chain
of command, and Vertical Depth represents the number
of entities in the direct chain of command from the
field officer up to and including the individual or
committee at the top of the chain of command.
Tennessee, for example, has a very centralized authority
structure. Typically, enforcement actions are pushed up
the agency’s chain of command all the way to the top,

5We correct for the panel structure of the data by including time
dummy variables as well as estimating robust standard errors that
are clustered in the firm. Additional analysis is reported in the
online appendix, Table A.6.

6About 4.5% of our cases are missing demographic data at the
zip-code level from the U.S. Census (likely due to shifting zip
codes). The missing zip codes have slightly lower HPV rates and
likely possess higher than average incomes. Given this pattern, we
likely underestimate demographic effects on compliance bias.

7
A detailed description of this measure and a robustness check,

using five-digit zip codes (the results were substantively similar)
is presented in the online appendix (Section 3, Table A.3).

512 david m. konisky and christopher reenock



eventually being signed off by the agency head. Com-
pare this to a state such as South Carolina, where these
actions are signed off by civil servants located five levels
below the agency head. The resulting variable ranges
between (0) and (1), where zero represents perfectly
centralized decision-making authority, and one repre-
sents authority decentralized to the field-officer level.
This equation yields a measure of locational authority
and is available for three levels of enforcement action
across each state—Level I actions (informal and formal
notices that typically are reserved for the first step in a
case of noncompliance), Level II actions (formal
administrative actions, which may include penalties),
and Level III actions (civil and criminal cases filed
against a noncompliant entity). We standardized each
of these measures and added them to construct an
equally weighted scale of sign-off authority. The final
measure, Signature Authority, has a mean of approx-
imately zero and a standard deviation of 2.5.

Control Variables

Firms and regulatory officers rely on a host of signals
to inform their decisions. Given that these signals
may be correlated with their compliance decisions as
well as with the demographic composition of the
relevant community, we include several controls to
avoid drawing incorrect inferences. To account for
relevant policy task factors, we include measures of
problem severity and complexity in the local policy
arena. We measure problem severity with Nonattain-
ment, a count of the number of CAA ambient air-
quality standards a county fails to meet on an annual
basis. We also include Policy Entropy, a diversity index
of state air-emissions sources in a county, where
higher values represent a more complex implementa-
tion environment (Potoski 1999). Last, many states
have decentralized their CAA compliance-monitoring
activities to regional offices. To control for each office’s
workload, we include Regional Scale, which is the total
number of regulated firms in the region.

Firms’ and bureaucrats’ compliance decisions may
also be influenced by economic and political condi-
tions. To account for economic conditions, we use
Unemployment Rate, measured at the county-level. In
addition, we include a measure of industry salience
as the percentage of a given county’s total nonfarm
income that derives from Air Polluting Industries
(Ringquist 1993). Political factors at both the state
and local level may also affect compliance decisions. A
stronger Democratic presence in state government has
been associated with greater regulatory activity in
general (Scholz, Twombly, and Headrick 1991; Scholz

and Wei 1986), and in enforcement of the CAA in
particular (Konisky 2007; Wood 1992). To account for
state-level political influence, we include Democratic
Governor, a dummy variable reflecting gubernatorial
partisan control, and Percentage Democrats in State
Legislature, which is the total percentage of Democrats
in both state houses. Past research has also shown that
regulators are responsive to local political conditions
(Scholz, Twombly, and Headrick 1991; Scholz and
Wei 1986; but see Huber 2007). Given that minorities
tend to live in more Democratic areas, and state
governments led by Democrats tend to pursue more
enforcement, it is important to control for any local
political effects, which we do with Percent Democratic
Vote, which is the county-level percentage of the 2000
presidential vote for the Democratic candidate.

Last, we include a set of firm-level controls to
account for heterogeneity in compliance costs across
firm types and for variation in the political costs that
officers face in handling cases for different types of
firms. First, firms with a recent agency-inspection
history are more likely to assign a higher cost to future
noncompliance. Therefore, we include two dichoto-
mous inspection variables in the firm model that
indicate whether it has been inspected in the prior
year, State Inspection and Federal Inspection. Moreover,
regulatory officers will be more likely to incur greater
political costs for failing to detect a prior noncompliant
firm. Therefore, in the regulatory detection model, we
include two dichotomous enforcement actions variables
that indicate whether the firm has been punished with
at least one enforcement action in the prior year, State
Enforcement Action and Federal Enforcement Action.

We also include three firm-level dummy variables
reflecting a facility’s industrial classification code:
Manufacturing, Utilities, and Transportation. Last, we
also included a firm-level variable reflecting the level of
pollution generated by the regulated entity. This
variable, Major Source, is a dummy variable that is
coded one if the firm is classified as a major source of
air pollution (generally greater than 100 tons per year)
and zero if it is not.

Empirical Results

Before describing the results from estimating the
DCE model, it is important to recall that the model
consists of two sets of estimated parameters. For the
first set, the firm model, positive coefficients reflect an
increase in the probability of a firm being an HPV. For
the second set, the detection model, positive coeffi-
cients reflect an increase in the probability of a firm

compliance bias and environmental (in)justice 513



being detected as an HPV. In addition, it is important
to note that the mean of the group-mobilization and
signature authority variables are essentially zero. Ac-
cordingly, the parameters on the constituent terms,
Percentage of African American Population and Percent-
age of Hispanic Population, represent the effect of either
community demographic when both mobilization and
signature authority are at their respective means.

The results are displayed in Table 1. Given space
constraints, we focus our discussion on the variables
of central theoretical interest. Across both the firm
and detection models, the results are generally con-
sistent with our expectations. The estimates from the
firm model suggest that community demographics
have statistically significant associations with a firm’s
HPV classification. Specifically, firms located in His-
panic communities and lower socioeconomic status
communities are more likely to be noncompliant with
the CAA. Firms located in African American commun-
ities, however, are not more likely to be major
violators—a finding that cuts against our expectations
and claims made by many environmental justice
advocates. It is also worth noting that the coefficients
on the Hispanic and class variables are of greater
magnitude than those derived from a single-equation
probit model, which highlights the problem of model-
ing compliance without taking into account the com-
pliance bias inherent in the data.8 (We present estimates
from a probit model in Section 2 of the online appendix.)

The estimates in the detection model suggest that
community demographics are also systematically as-
sociated with the likelihood of regulatory officers
accurately detecting noncompliant firms. The percent-
age of Hispanics (but not African Americans) in a
community and a community’s socioeconomic class
each have a statistically significant effect on regulatory
officers’ reliably detecting noncompliance. The results
suggest that, on average, regulators are less likely to
detect HPVs in these communities. Taking the results
of both models together, we now have a clearer picture
of the process behind disparities in environmental
regulatory outputs. Relative to those located in more
upper-class, non-Hispanic communities, firms are both
more likely to be significant violators of the CAA and
less likely to be characterized as such as by bureaucrats.
This supports our expectations in both Hypothesis 1
and Hypothesis 2.

What of the ability to mitigate this pattern of bias
via political mobilization? Recall that our expectation
is that, in more mobilized communities, firms will face

greater costs associated with noncompliance, and
bureaucrats will face greater costs associated with
failing to detect noncompliance. To the extent that
discriminatory patterns exist across community dem-
ographics, these patterns ought to be attenuated in the
presence of actual political mobilization. The analysis
is consistent with our expectations in Hypothesis 3.
The interaction terms in both the firm and detection
models suggest a diminishment of bias—that is, in the
presence of mobilized, Hispanic communities, firms
are less likely to be noncompliant, and bureaucrats are
more likely to detect noncompliance.9 (We would
have expected an analogous effect in African American
communities, but as noted above, the coefficient on
the initial percentage African American variable does
not suggest any baseline bias to attenuate.) This
finding lends additional support to a widely held belief
among scholars that political mobilization can effec-
tively counter environmental inequities. The novelty of
our findings, however, lies in demonstrating that
political mobilization can alter not only firms’ com-
pliance decisions, but also regulatory officers’ detection
decisions.10

It also appears that the ability of political mobi-
lization to mitigate this bias does not extend to other
standard indicators of mobilization. We gathered
data on both voter turnout from the 2000 Presiden-
tial election at the county level and the number of
campaign donations made to candidates annually
(2001–2004) at the zip-code level. When we estimate
the effect of these indicators of political mobilization,
neither successfully attenuates the compliance bias
present in Hispanic neighborhoods—the coefficient
on the interaction term is null in both cases. We
suspect that voting and campaign donations repre-
sent diffuse mobilization across policy issues—too
diffuse to alter the incentives of firms or regulatory
officers in Hispanic communities. (These results are pro-
vided in Section 4, Table A.5 in the online appendix.)

Finally, the parameter estimate on the Signature
Authority interaction term is consistent with our last
hypothesis that institutional structures alter regula-
tory officers’ incentives to detect noncompliance.

8Moreover, with a single equation, the analyst must decide whether
to interpret an effect as influencing compliance or detection.

9Our mobilization measure primarily reflects groups representing
persons of color and may be a less valid measure of mobilization
around class issues. Including a multiplicative interaction term
between mobilization and class yields a null coefficient.

10
Our results are also robust to alternative specifications of group

mobilization. Specifically, we used a simple count of environ-
mental justice groups, and differentiated between groups focused
on African American and Hispanic issues when creating our
interaction terms. These results are reported in the online
appendix (Section 3, Table A.4).

514 david m. konisky and christopher reenock



TA B L E 1 High Priority Violator Status for Individual Regulated Firms

HPV Status

Bivariate Probit, with Partial Observability

Firm Model
Pr(Observing HPV)

Detection Model
Pr(Detection)

b s.e. b b s.e. b

Environmental Justice Indicators
% African American 0.0002 0.0010 0.0003 0.0013
% Hispanic 0.0069*** 0.0011 -0.0069*** 0.0014
Class -0.0632** 0.0254 0.0635** 0.0320
Mobilization 0.1011** 0.0402 -0.0962* 0.0494
Mobilization X African American -0.0011 0.0008 0.0008 0.0009
Mobilization X Hispanic -0.0024*** 0.0007 0.0022** 0.0009
Signature Authority — — -0.0216*** 0.0054
Sig. Authority X African American — — 0.0000 0.0002
Sig. Authority X Hispanic — — 0.0008*** 0.0002

Policy Task Factor Indicators
Regional Scale 0.0003*** 0.0001 -0.0001 0.0001
Policy Entropy -0.0474 0.0282 0.0860** 0.0351
Nonattainment 0.0994*** 0.0431 -0.0634 0.0535

Economic Context Indicators
Unemployment 0.0850 0.0117 -0.0761*** 0.0143
% Income Air Pollution Industry 0.0076 0.0041 -0.0060 0.0049

Political Context Indicators
Democratic Governor -0.1617*** 0.0302 0.1584*** 0.0396
% Democratic in State Legislature -0.0012 0.0013 0.0006 0.0017
% Democratic Vote -0.0095*** 0.0015 0.0080*** 0.0018

Firm-Level Factors
State Inspection (1 Year Lag) 0.1759*** 0.0258 — —
Feb Inspection (1 Year Lag) -0.0341 0.0219 — —
State Enforcement Action (1 Year Lag) — — 5.4527*** 0.5632
Feb Enforcement Action (1 Year Lag) — — 1.2510*** 0.2173
Manufacturing Firm 0.0148 0.0402 0.0913* 0.0551
Utility Firm 0.0073 0.0476 -0.0621 0.0570
Transportation Firm 0.0234 0.1167 -0.0583 0.1368
Major Source 0.6429*** 0.0344 -0.2262** 0.0984

Year Dummy (2002) -0.1832*** 0.0373 0.2145*** 0.0503
Year Dummy (2003) -0.1932*** 0.0392 0.1815*** 0.0503
Year Dummy (2004) -0.2502*** 0.0388 0.2185*** 0.0499
Intercept -1.1353*** 0.1032 0.5389* 0.2785

rho -0.867**
Log-Likelihood -27858.16
x2 (49) 1241.56***
Number of Cases 153580

Note: *p , .10, **p , .05, ***p , .01, two-tailed test. Standard errors clustered on firm.

compliance bias and environmental (in)justice 515



The results suggest a statistically significant interaction
between percentage of Hispanic residents in a com-
munity and the location of enforcement authority.
When this authority rests with field officers, while a
certain degree of failed detection still exists, field officers
are less sensitive to whether a community is Hispanic.
In other words, the detection of noncompliance is
relatively flat across levels of Hispanic communities.
However, when enforcement authority is centralized
near the top of the agency’s chain of command, a
different type of pressure arises. For these high-level
managers, the detection of noncompliance wanes in the
presence of an increasingly more Hispanic community.
This relationship is consistent with our argument that
bureaucrats near the top of the agency’s structure feel
greater pressure for their agency to perform well
(i.e., produce high compliance) and as such are more
sensitive to registering noncompliance in heavily
Hispanic communities. It is worth noting that this
relationship is not conditional on party control of the
governor’s office. In models not reported, we included
a three-way interaction term—Democratic Governor
X Signature Authority X % Hispanic (and similarly
for % African-American)—but the coefficient was not
statistically significant, suggesting that high-level bu-
reaucrats’ compliance determinations are not condi-
tioned by executive political control. This is consistent
with the idea that high-level managers, across political
contexts, worry about maintaining a positive image of
their agency.

Figure 1 better demonstrates the substantive
effects of these findings. Each of the panels display
the marginal effect of a 10% increase in the percent-
age of the population that is Hispanic in a zip code
on the joint probability of observing a noncompliant,
undetected firm.11 The top panel displays the mar-
ginal effect of Hispanic across the mobilization scale,
while the lower panel displays the effect across the
signature authority scale.12

The top panel of Figure 1 shows that regulatory
officers respond to Hispanic communities differently,
conditioned on whether those communities are more
mobilized around environmental justice concerns. In
fact, the number of undetected noncompliant firms
(nationwide) can decrease from a high of approx-
imately 150 in the absence of mobilized communities
to nearly 50 in the presence of highly mobilized ones.
A decrease of 100 HPVs nationally may appear to
some to be a diminutive effect, but the presence of a
single HPV can be quite impactful for the surround-
ing community. Direct mobilization of at-risk com-
munities is not the only incentive mechanism that
affects regulator decision making. The lower panel
suggests that the location of signature authority
within an agency is a powerful incentive device as
well. While all agency bureaucrats engage in some
amount of failed detection (i.e., the expected number

FI GU R E 1 Effect of Percent Hispanic on
Undetected HPVs across Mobilization
and Signature Authority

11
The substantive effect that we report is the change in the

unconditional joint probability of observing a noncompliant
firm, that is likely to have gone undetected, or Pr y1 ¼ 1;ð
y2 ¼ 0Þ¼ F x1b1ð Þ� F2 x1b1; x2b2; rð Þ multiplies by the total
number of firms for whom the condition is reasonable (see next
note for explanation).

12The figures were created by estimating the mean expected
probability and associated standard errors from 10,000 draws off
of the variance-covariance matrix. To display a substantively
interesting outcome, we multiplied the marginal probability for a
given condition by the nationwide subset of the firm population
for whom an additional 10% Hispanic gain is reasonable
(approximately 18%, or 6,913 firms for our data). This calcu-
lation produced the expected change in the number of unde-
tected noncompliant firms across the United States.

516 david m. konisky and christopher reenock



of undetected noncompliant firms is positive over the
entire range of signature authority), those officers
nearer the top of an agency’s chain of command
succumb to an additional disincentive to code non-
compliance. At the extreme, in the presence of a more
Hispanic community, when decision making over
enforcement rests with high-level bureaucrats, nearly
140 noncompliant HPVs will go undetected compared
to 80 when such authority rests with field officers. We
believe that this is due to high-level bureaucrats being
more likely to engage in organizational cheating in the
presence of a community that is less politically mobi-
lized and therefore less likely or able to engage in fire-
alarm oversight.13

The relative impact of the mobilization and
signature authority variables on the joint probability
of observing an undetected violating firm compares
reasonably well with other variables in the model. For
example, having a Democratic governor decreases the
number of undetected violators by an estimated 290
firms, nationally. This suggests that the greatest impact
of mobilization or signature authority (;150) is nearly
half the magnitude of changing the party of the
governor’s office. Another factor, local-level partisan
representation, also motivates firms and regulators.
An increase of a standard deviation (about14 points)
above the average percent of the Democratic county
vote in the 2000 presidential election (around 46%),
decreases the number of undetected violators by an
estimated 200 firms, nationally. These results are
consistent with long-held findings in the literature that
central and local partisan actors shape policy delivery
(Scholz and Wei 1986). But, our analysis reveals two
alternate and meaningful pathways to influence com-
munity compliance bias—one emphasizing commun-
ity empowerment and political mobilization and the
other emphasizing regulatory officers’ incentives struc-
tures via creative-agency design features.

Discussion and Conclusion

Democratic governments continue to wrestle with
shortfalls in equitable policy delivery. Examining the
occurrence of such shortfalls has been a central focus
of scholars and practitioners alike. In this article, we

focused on one area, environmental policy, for which
research has failed to arrive at a consensus on the
causal process that generates bias in regulatory out-
comes. We have attempted to move this discussion in
a different direction—one that considers the incen-
tives of both firms and regulatory officers making
decisions about compliance.

Our argument suggests that disparities in U.S.
environmental policy outcomes are rooted in the
incentives facing both firms and the bureaucrats
tasked with implementation. Although past research
has recognized the dual production of compliance
bias, we offered a novel theoretical account focusing
on the incentives of bureaucrats to code firm com-
pliance status. This approach enabled us to then show
how decisions change in the presence of variation in
these incentives. We demonstrated that when com-
munities overcome collective action problems, bu-
reaucrats will dedicate more resources to the firms in
their areas. We also showed that agency design can
accentuate bias in compliance decisions. High-ranking
bureaucrats have an incentive to overreport compli-
ance rates, since these rates are used as indicators of
agency performance by external stakeholders. Because
these bureaucrats are less likely to get caught engaging
in this behavior when communities have fewer polit-
ical resources to detect it, the outcome is likely to
further disadvantage poor and minority communities
since they tend to lack these resources.

An unanticipated finding is that the bias revealed
from our analysis pertains to poor and Hispanic
communities, but not African American ones. It is
worth noting that this result would have been masked
had we just considered minorities collectively, which
is often done in the literature. What might explain this
pattern, and particularly the lack of bias in firm and
regulatory compliance decisions toward firms located
in African American communities? The simplest
interpretation, of course, is that there is no systematic
bias. Although this runs counter to claims made in the
literature (e.g., Bryant 1995; Bullard 1993), it would be
welcome news for these communities which past work
has found to be subject to disproportionate environ-
mental burdens (Ringquist 2005). It may also reflect
the time period analyzed. Our analysis was limited
to 2001–2004, and by this time, the environmental
justice movement was quite mature in many African
American communities, and firms and regulators alike
may have factored in the costs of committing major
violations and failing to detect noncompliance, re-
spectively. In other words, advocacy in these com-
munities may have paid considerable dividends by the
time of our study. For Hispanics, however, the results

13We also considered the possibility that mobilized groups might
change the decision calculus for upper-level bureaucrats. We
examined whether a three-way interaction was present between
Hispanic community, mobilization, and signature authority, but
the coefficient was not statistically significant.

compliance bias and environmental (in)justice 517



suggest a more traditional story of environmental
disparities. Firms in these communities are not only
more likely to be significant violators of their CAA
obligations, but they are less likely to be designated as
such by regulatory officers.

There are several additional implications of our
study. While our findings suggest that substantial
disparities in compliance outcomes exist, they also
suggest the possibility that patterns of bias stem from
the decision-making structures of regulatory officers,
not necessarily from intentional discrimination. This
by no means discounts the importance of the disparate
outcomes, but it does suggest that the source of bias
need not rest with deliberate prejudice, but with
bureaucrats’ rational responses to their resource con-
straints. In other words, biased outcomes are possible,
indeed probable, even in the absence of explicit
discrimination.

Our results also offer a cautionary note on goal
displacement. Political principals can use various
performance measures to determine agency ‘‘suc-
cess.’’ However, in certain contexts, indicators such
as compliance rates may present perverse incentives
to regulatory officers. Our analysis suggests that when
bureaucrats face such perverse incentives, they may
substitute the appearance of successful policy delivery
for equitable policy delivery. Assessing the way in
which such indicators incentivize regulatory officers
is essential for any effort to reducing compliance bias.
Given the increasing prevalence of performance
measurement at all levels of government, the poten-
tial implications of this type of goal displacement on
equitable policy implementation likely extends into
many different areas of public policy.

Last, our study reveals evidence of multiple ways
that class- and race-based disparities could be re-
dressed. Communities could be empowered through
capacity-building efforts. Such efforts designed to
help poor and minority areas organize around issues
of environmental protection will produce a demand
for more equitable treatment from government and
put direct pressure on firms to improve their perform-
ance. Second, our results suggest that changes to the
institutional decision-making structure of the agencies
responsible for implementing policy would also be
effective. Specifically, our analysis suggests that decen-
tralizing authority to regulatory officers in agencies
could result in fewer cases of deliberate nondetection
of compliance. That is, moving the location of
compliance determinations away from high-ranking
officials with incentives to overreport compliance
may result in fairer—although not necessarily fair—
treatment of communities hosting regulated firms.

Acknowledgments

We would like to thank Don Haider-Markel, Eliz-
abeth Rigby, and Evan Ringquist for helpful com-
ments and discussion, as well as John Mortara and
Kristen Holder for their excellent research assistance.

References

Abowd, John M., and Henry S. Farber. 1982. ‘‘Job Queues and
The Union Status of Workers.’’ Industrial and Labor Relations
Review 35 (3): 354–67.

Blau, Peter M., and Marshall W. Meyer. 1971. Bureaucracy in
Modern Society. 2nd ed. New York: Random House.

Bohte, John, and Kenneth J. Meier. 2000. ‘‘Goal Displacement:
Assessing the Motivation for Organizational Cheating.’’ Public
Administration Review 60 (2): 173–82.

Braithwaite, John, and Toni Makkai. 1991. ‘‘Testing an Expected
Utility Model of Corporate Deterrence.’’ Law & Society Review
25 (1): 7–40.

Brehm, John, and Scott Gates. 1997. Working, Shirking, and
Sabotage: Bureaucratic Response to a Democratic Public. Ann
Arbor: The University of Michigan Press.

Brehm, John, and James T. Hamilton. 1996. ‘‘Noncompliance in
Environmental Reporting: Are Violators Ignorant, or Evasive,
of the Law?’’ American Journal of Political Science 40 (2):
444–77.

Bryant, Bunyan. 1995. Environmental Justice: Issues, Policies, and
Solutions. Washington, DC: Island Press.

Bullard, Robert D. 1993. ‘‘Anatomy of Environmental Racism
and the Environmental Justice Movement.’’ In Confronting
Environmental Racism: Voices from the Grassroots, ed. R. D.
Bullard. Boston: South End Press, 15–39.

Bullard, Robert D. and Glenn S. Johnson. 2000. ‘‘Environmental
Justice: Grassroots Activism and Its Impact on Public Policy
Decision Making.’’ Journal of Social Issues 56 (3): 555–78.

Dion, Catherine, Paul Lanoie, and Benoit Laplante. 1998.
‘‘Monitoring of Pollution Regulation: Do Local Conditions
Matter?’’ Journal of Regulatory Economics 13 (1): 5–18.

Downs, Anthony. 1967. Inside Bureaucracy. Boston: Little, Brown.

Earnhart, Dietrich. 2004a. ‘‘Regulatory Factors Shaping Environ-
mental Performance at Publicly-Owned Treatment Plants.’’
Journal of Environmental Economics and Management 48 (1):
655–81.

Earnhart, Dietrich. 2004b. ‘‘The Effects of Community Charac-
teristics on Polluter Compliance Levels.’’ Land Economics
80 (3): 408–32.

Environmental Justice Resource Center. 2000. People of Color
Environmental Groups Directory. Atlanta: Environmental Jus-
tice Resource Center.

Feinstein, Jonathan S. 1990. ‘‘Detection Controlled Estimation.’’
Journal of Law & Economics 53 (1): 233–76.

Feinstein, Jonathan S. 1999. ‘‘Approaches to Estimating Non-
compliance: Examples from Federal Taxation in the United
States.’’ The Economic Journal 109 (456): 360–69.

Gray, Wayne B., and Ronald J. Shadbegian. 2004. ‘‘‘Optimal’
Pollution Abatement: Whose Benefits Matter, and How
Much?’’ Journal of Environmental Economics and Management
47 (3): 510–34.

518 david m. konisky and christopher reenock



Hamilton, James T. 1993. Politics and Social Costs: Estimating
the Impact of Collective Action on Hazardous Waste Facility.
The RAND Journal of Economics 24 (1): 101–25.

Hamilton, James T. 1995. Testing for Environmental Racism:
Prejudice, Profits, Political Power? Journal of Policy Analysis
and Management 14 (1): 107–32.

Hamilton, James T., and W. Kip Viscusi. 1999. Calculating Risks?
The Spatial and Political Dimensions of Hazardous Waste
Policy. Cambridge, MA: MIT Press.

Hammond, Thomas H. 1986. ‘‘Agenda Control Organizational
structure and Bureaucratic Politics.’’ American Journal of
Political Science 30 (2): 379–420.

Helland, Eric, 1998a, ‘‘Environmental Protection in the Federalist
System: The Political Economy of NPDES Inspections.’’
Economic Inquiry 36 (2): 305–19.

Helland, Eric, 1998b, ‘‘The Enforcement of Pollution Control
Laws: Inspections, Violations, and Self-Reporting.’’ The Re-
view of Economics and Statistics 80 (1): 141–53.

Huber, Gregory A. 2007. The Craft of Bureaucratic Neutrality:
Interests and Influence in Governmental Regulation of Occupa-
tional Safety. Cambridge: Cambridge University Press.

Konisky, David M. 2009. ‘‘Inequities in Enforcement? Environ-
mental Justice and Government Performance.’’ Journal of
Policy Analysis and Management 28 (1): 102–21.

Konisky, David M., and Tyler S. Schario. 2010. ‘‘Examining
Environmental Justice in Facility-Level Regulatory Enforce-
ment.’’ Social Science Quarterly 91 (3): 835–55.

Maravall, José Marı́a, and Adam Przeworski. 2003. Democracy
and the Rule of Law. Cambridge: Cambridge University Press.

Meier, Kenneth J. 1993. ‘‘Representative Bureaucracy: A Theo-
retical and Empirical Exposition.’’ In James Perry, ed.,
Research in Public Administration. Greenwich, CT: JAI Press.

Mennis, Jeremy L. 2005. ‘‘The Distribution and Enforcement of
Air Polluting Facilities in New Jersey.’’ The Professional
Geographer 57 (3): 411–22.

Olson, Mary K. 1995. ‘‘Regulatory Agency Discretion among
Competing Industries: Inside the FDA.’’ Journal of Law,
Economics, & Organization 11 (2): 379–407.

Poirier, Dale J. 1980. ‘‘Partial Observability in Bivariate Probit
Models. Journal of Econometrics 12: 209–17.

Potoski, Matthew. 1999. ‘‘Managing Uncertainty through Bu-
reaucratic Design: Administrative Procedures and State Air
Pollution Control Agencies.’’ Journal of Public Administration
Research and Theory 9 (4): 623–40.

Raz, Joseph. 1977. ‘‘The Rule of Law and Its Virtue.’’ The Law
Quarterly Review 93: 195–211.

Reenock, Christopher M., and Brian J. Gerber. 2008. ‘‘Political
Insulation, Information Exchange, and Interest Group Access
to the Bureaucracy.’’ Journal of Public Administration Research
and Theory 18 (3): 415–40.

Ringquist, Evan J. 1993. Environmental Protection at the State
Level. Armonk, NY: M.E. Sharpe.

Ringquist, Evan J. 2005. ‘‘Assessing Evidence of Environmental
Inequities: A Meta-Analysis.’’ Journal of Policy Analysis and
Management 24 (2): 223–47.

Scholz, John T. 1991. ‘‘Cooperative Regulatory Enforcement and
the Politics of Administrative Effectiveness.’’ American Polit-
ical Science Review 85 (1): 115–36.

Scholz, John T., Jim Twombly, and Barbara Headrick.
1991. ‘‘Street-Level Political Controls over Federal Bu-
reaucracy.’’ American Political Science Review 85 (3): 829–
50.

Scholz, John T., and Wang Cheng-Lung. 2006. ‘‘Cooptation or
Transformation? Local Policy Networks and Federal Regula-
tory Enforcement.’’ American Journal of Political Science
50 (1): 81–97.

Scholz, John T., and Wei Feng-Heng. 1986. ‘‘Regulatory Enforce-
ment in a Federalist System.’’ American Political Science
Review 80 (4): 1249–70.

Simon, Herbert. 1976. Administrative Behavior, 3rd Edition.
New York: The Free Press.

Wilson, James Q. 1989. Bureaucracy: What Government Agencies
Do and Why They Do It. New York: Basic Books.

Winter, Soren, and Peter May. 2001. ‘‘Motivation for Compli-
ance with Environmental Regulation.’’ Journal of Policy
Analysis and Management 20 (4): 675–98.

Wood, B. Dan. 1992. ‘‘Modeling Federal Implementation as a
System: The Clean Air Case.’’ American Journal of Political
Science 36 (1): 40–67.

David M. Konisky is Assistant Professor at
Georgetown University, Washington, DC 20057.

Christopher Reenock is Associate Professor at
Florida State University, Tallahassee, FL 32306.

compliance bias and environmental (in)justice 519