• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Enforcement Principles and Environmental Agencies: Principal-Agent Relationships in a Delegated Environmental Program

Posted on: Sunday, 23 December 2007, 06:00 CST

By Atlas, Mark

This article examines whether states shirked enforcement responsibilities in their principal-agent relationship with the federal government when implementing a delegated environmental program. It evaluates determinants of environmental enforcement stringency, particularly whether penalties were less when imposed by states than by the federal government. It analyzes 6,048 hazardous waste administrative penalties across 32 states and all U.S. Environmental Protection Agency Regions over 14 years. It finds that state penalties are typically substantially lower than federal penalties, and that penalty amounts are typically also related to the partisan composition of elected officials and the characteristics of enforcement actions, such as the type, seriousness, and number of violations. Factors such as the influence of organized interest groups, agency sensitivity to economic conditions or the economic importance of regulated industries, and environmentalist preferences of elected officials and the public are typically unrelated to enforcement stringency. A dominant pattern in implementing U.S. public policy in recent decades is the federal government delegating to states the legal authority to implement and enforce certain federally designed policies, often termed "partial preemption" (Scheberle 1997; Hedge et al. 1991; Crotty 1987). Partial preemption has been used for major environmental, social welfare, occupational safety, and other programs (Scicchitano & Hedge 1993; Conlan 1988). The process of delegating authority to states ordinarily begins with the federal government creating a detailed regulatory program that states can seek authorization, and partial federal funding, to implement if they enact into state law their own programs at least as stringent as the federal program and adequately fund and enforce them (Zimmerman 2005; Hedge et al. 1991; Crotty 1987). States authorized to conduct a program are monitored by the federal government to ensure their compliance with federal requirements, while the federal government directly implements the program itself in unauthorized states (Scheberle 1997; Conlan 1988; Crotty 1987).

Among the purported benefits of such a delegation of authority is that it allows states to lead the implementation of important and often controversial public policies in their jurisdictions, with the federal government as their partner, rather than having federal authority imposed directly (Zimmerman 2005; Scicchitano & Hedge 1993; Welborn 1988). Both regulated entities and the general public may be more tolerant of increased government regulation if it is perceived to originate from their state, rather than federal, government (Dwyer 1997; Scicchitano & Hedge 1993). States also have some flexibility in customizing policies to address their specific problems, reflect their priorities, and satisfy their particular political and economic constituencies, as long as federal minimum standards are met (Zimmerman 2005; Dwyer 1997; Scicchitano & Hedge 1993). States also can bring to bear their presumably more intimate knowledge of the regulated entities and problems to implement more effective and efficient programs (Welborn 1988). Furthermore, delegation reduces the need for federal funds and personnel to implement programs in states (Dwyer 1997; Welborn 1988; Crotty 1987).

Along with these potential benefits of delegated programs, however, is the concern over whether states can be trusted to enforce laws as rigorously as the federal government. This concern can be viewed through the perspective of the principal-agent model of bureaucratic control frequently used to examine variations in the development and implementation of regulatory policy (Davis & Davis 1999; Hedge et al. 1991; McCubbins et al. 1987). The principal- agent model assumes that regulatory actors interact in a hierarchically ordered system of relationships. Principals at the top try to control decisions and actions of bureaucratic subordinates lower in the political hierarchy, whose primary responsibility should be faithful implementation of the law. Relationships between principals and agents can be characterized by a divergence of interests and agents' tendency to shirk responsibilities imposed by principals. This can occur as bureaucratic agents develop separate interests through institutionalization, external politics, and "capture" by external groups. Principals try to control agents' behavior through oversight activities, sanctions, and incentives, which agents try to resist through means such as grants of discretion from and information advantages over principals. Principals can try to monitor agents' activities to offset information imbalances, but this can be costly and time-consuming. Whether agents' shirking is prevented is affected by principals' access to and willingness to use control mechanisms, but some slack in the principal-agent relationship is expected. A key issue under principalagent theory is how principals can overcome agents' shirking tendency.

The delegation to states of federal environmental programs can be viewed as a three-tiered principal-agent process. The first tier involves Congress, as the principal, delegating to the U.S. Environmental Protection Agency (EPA), as its agent, responsibility for developing and implementing an environmental program. The second tier involves EPA, now as the principal, delegating to its Regional Offices, as its agents, responsibility for implementing the environmental program in the states in their respective regions. The third tier involves EPA Regional Offices, now as the principals, overseeing authorized states, as their agents, as they implement the environmental program.

Empirical studies have attempted to assess how state environmental standards changed when the federal government supposedly gave states more responsibility for their environmental programs (Millimet 2003; Potoski 2001; List & Gerking 2000, 1996) or how performance measures differed when environmental or occupational safety and health programs were managed by state or federal governments, respectively (Scholz & Wang 2006; Helland 1998; Hedge et al. 1991; Scholz & Wei 1986; Thompson & Scicchitano 1985b; Marvel 1982). These studies reached conflicting conclusions, with states sometimes performing better, worse, or the same as the federal government. An inherent problem in using environmental standards as the basis for testing environmental stringency is that, as with other delegated programs, the federal government sets minimum standards (e.g., pollutant limits or pollution abatement technologies) for states in major areas of environmental regulation. Thus it should be impossible for an authorized state to have laws that are explicidy inadequate.

Consequently, inadequate implementation should only be able to occur through states' enforcement of environmental laws. Inadequate enforcement, however, should be more difficult for the federal government to notice than inadequate laws and could undermine a program even more. Most states have been delegated the authority to implement most major federal environmental programs, so states do the overwhelming share of enforcing those laws (Brown & Green 2001). Concerns have often been expressed by scholars (Mintz 2001; Rabe 2000; Steinzor 2000), federal agencies (U.S. EPA 2001, 1999a, 1999b, 1998e, 1997a; U.S. General Accounting Office [GAO] 2000, 1988), and the news media (Johnson 2002; Polakovic 2001; Edwards 2000; Braile 1999) that states are not enforcing environmental laws as stringently as would the federal government-in effect, state agents are shirking their enforcement responsibilities. Thus evaluators of environmental standards may assume that a state was environmentally conscientious after reviewing its laws, not realizing it was pursuing its own interests, divergent from those of EPA, by lagging in enforcing those laws. The result could be that violators are treated more leniently if prosecuted by states than by the federal government. Due to this concern, some scholars have advocated against delegated programs or for closer federal supervision of such programs (Markell 2000; Steinzor 2000; Kuehn 1996).

Some scholars have argued that the interests of state environmental enforcers have diverged from those of EPA because they are more vulnerable to pressures from elected officials or interest groups, pleas of economic hardship from violators, enforcement budget constraints, and too-close relationships between regulators and regulated entities (Mintz 2001; Kuehn 1996). For example, a state government has an inherent economic interest in creating a hospitable business climate compared to other states. Thus a state might use weaker environmental enforcement to make itself more attractive to industry. Indeed, a primary theoretical rationale for federal environmental regulation is that, without it, states might "race-to-the-bottom" in environmental standards (Engel 1997; Revesz 1997, 1992; Kuehn 1996). "Race-to-the-bottom" refers to a progressive relaxation of state environmental standards, perhaps motivated by interstate competition to attract industry, that reduces social welfare below levels that would exist in the absence of such competition (Saleska & Engel 1998; Engel 1997; Revesz 1997, 1992). One study found that substantial percentages of small survey samples of various government and interest group officials claimed that states imposed, or advocated imposing, smaller penalties on environmental law violators to encourage industries to locate in their states (Engel 1997). An EPA audit (1991c) of some states' environmental enforcement efforts concluded that states were hesitant to take strong enforcement actions against violators for fear of losing business. In addition, EPA stated: [A]bsent Federal enforcement capability, some State and local governments would be more susceptible to economic and other pressures from industry that could actually make State and local enforcement less effective than it currendy is. . . . [A]bsent a nationwide, Federal enforcement presence, industry would be inclined to build, or move, [pollution] sources to States with the least effective enforcement efforts. Such a possibility would give businesses more leverage over the State governments and could foster a competition among the States to actually relax enforcement efforts (1989b:27277).

Furthermore, examining state enforcement of federally delegated environmental programs offers an especially useful opportunity to assess if shirking occurred in a principal-agent regulatory relationship because, in this situation, the principal was in a particularly disadvantageous position to prevent it. For example, although states must periodically report information to EPA on their enforcement programs (U.S. GAO 1998)-such as numbers of inspections, violations identified, and penalties assessed-tins does not eliminate the information imbalance that facilitates state agents' shirking. The thoroughness of inspections matters at least as much as their mere numbers in assessing the quality of enforcement efforts. In addition, the quantities of violations and penalties do not necessarily indicate how many should have resulted from adequate enforcement, as low numbers could reflect either lackluster enforcement or stringent enforcement that deterred violators or serious offenses.

EPA also has very limited resources to monitor possible shirking by state enforcers. Under EPA guidelines, only a few percent of state inspections of regulated facilities are followed up annually by EPA inspections of the same facilities to check the accuracy of the state's findings (U.S. GAO 1993, 1987; Cannon 1983:22). In addition, EPA Regional Office enforcement resources always have been scarce (U.S. GAO 1993, 1990, 1988, 1987, 1986, 1985a), so it is unlikely that EPA can dioroughly examine state enforcement actions to evaluate their adequacy.

Even if EPA believed a state agent was shirking its enforcement obligations, EPA would not necessarily have much credibility in complaining. Although many government reports have identified shortfalls in state enforcement efforts, at least as many have criticized EPA for the same problem (U.S. GAO 1996, 1991a, 1991b, 1990, 1988, 1987, 1985b). Consequently, in this situation, the agent could be justifiably uncertain whether the principal's professed instructions actually were to be followed. Thus it is unclear if EPA would be motivated to "call the kettle black" and if a miscreant state agent would assign much weight to the criticism.

Even if EPA wanted to express its dissatisfaction about a state agent's sluggish enforcement efforts, it has few available sanctions. Rather than depending solely upon the state, EPA could initiate enforcement actions itself against violators, even possibly against those who were already penalized by the state but in a way EPA believed was inadequate. Other than perhaps affronting the independence and diligence of state enforcers, however, this inflicts no substantive punishment on the state agent, but rather only on the violators. Furthermore, given the scarcity of EPA enforcement resources, this might be an empty threat. EPA could withdraw the state agent's authorization and implement the environmental program itself, but this would be a drastic step and, given EPA's limited resources, might penalize EPA more than the state and not necessarily result in more aggressive enforcement efforts (Zinn 2002; U.S. GAO 1995). In addition, if EPA tried to punish the state, this principal-agent relationship is atypical in that the agent (the state) could complain about the principal (EPA) to someone (a member of Congress) who is a principal in its own relationship with EPA as the agent. Therefore, the principal might fear repercussions from trying to discipline the agent.

For these reasons, examining if bureaucratic shirking exists in an environmental enforcement context could be especially fruitful. Thus the purpose of this article is to examine a possible weakness of partial preemption by empirically assessing if there is a difference in the stringency of state and federal enforcement in a delegated environmental program. It does so in the context of administrative penalties for hazardous waste law violations. Specifically, it evaluates if federal and state penalties differ, controlling for the characteristics of an enforcement action and of the political environment in which it occurs. Prior research has focused on other environmental or occupational safety and health enforcement indicators, such as inspections or notices of violations (NOVs) (Helland 1998; Hedge et al. 1991; Thompson & Scicchitano 1985b; Marvel 1982), and only one has also examined penalties (Scholz 8c Wei 1986). This article examines 6,048 environmental penalties in 32 states and all EPA Regions over a period of 14 years to determine whether and why federal and state governments differ in environmental enforcement stringency.

Determinants of Enforcement Stringency

Aside from an inherent tendency for states to shirk enforcement obligations, other theories have been proposed and empirically tested to explain variations in state or federal enforcement stringency. Naturally, enforcement stringency should vary depending on the characteristics of an enforcement action, such as the seriousness and number of violations and past noncompliance. Aside from these characteristics, however, some theorize that enforcement stringency varies due to the political environment of the regulatory regime. These theories, concisely summarized in Kagan (1989), assume that agency behavior might be affected by interest groups and political leaders who seek to impose their preferences and priorities on regulators, and regulators who seek to avoid political trouble by acceding to those pressures.

One theory is that enforcement stringency is related to partisan control of government. Although the executive branch is responsible for law enforcement, the legislative branch could, through control over appropriations for an agency or other tactics, also influence enforcement. This is essentially another aspect of the principal- agent model of government behavior tested in prior studies (McKinney 1998; Hedge et al. 1991; Scholz & Wei 1986). Rather than focusing on the principal-agent relationship between the EPA Region and the state, it focuses on the relationship between elected officials and the environmental agency. Essentially, this theory asserts that agency bureaucrats implement policy in a manner responsive to the preferences of elected officials who control the policy, financial, and oversight levers of government power.

Some prior research concluded that partisan control of government is related to federal or state environmental enforcement activity (Lynch et al. 2004b; Ringquist & Emmert 1999; Ringquist 1998; McKinney 1998; Wood 1992). Democratic Party office holders are perceived to be more aggressive in environmental protection than Republicans. For example, League of Conservation Voters (LCV) ratings of Democratic federal and state legislators' environmental protection voting records typically are much higher than for Republicans.1 Consequendy, partisan control of the executive and legislative branches can be used as a proxy for environmental policy preferences of elected officials. Thus one hypothesis is that enforcement stringency varies depending on which political party controls the executive branch2 and legislature, respectively, with more stringent enforcement the more the Democratic Party is in control.

Another theory is that pressure from organized interest groups influences agency enforcement. Some prior research concluded that the presence of environmental groups affects environmental enforcement activity (Davis & Davis 1999; Helland 1998; Hamilton 1996). For example, an agency could be motivated to impose larger penalties if it were concerned about adverse public reaction, stoked by environmental groups, to smaller penalties. Although there is no evidence the general public is aware of environmental enforcement actions, because such information is publicly accessible, agencies might be concerned that environmental groups would obtain and publicize it. Thus another hypothesis is that the more environmental group members in an area, the more stringent the enforcement.

Another theory is that regardless of the presence of organized interest group members, the more sentiment in an area toward a particular interest-such as environmentalism-the more actual or potential political pressure agencies might feel for more stringent enforcement to assuage important political interests. Thus the environmentalism that could pressure environmental enforcers could be among the general public or among political office holders, or both. If such pressure comes from political office holders, this might somewhat overlap with partisan control of government, as the latter assumes Democratic office holders are more environmentalist than Republicans. Some prior research concluded that environmentalist sentiment of office holders affects environmental enforcement (Scholz & Wang 2006; Davis & Davis 1999; Ringquist & Emmert 1999; Ringquist 1998) or that labor sentiment of office holders or union strength in a state affects occupational safety enforcement (Scholz & Wei 1986; Thompson & Scicchitano 1985a; Marvel 1982). Thus another hypothesis is that the more environmentalist sentiment in an area, the more stringent the enforcement. Possibly counteracting the influence of group, public, or office holder environmentalism is the theory that the more important to the economy the industries being regulated, the less stringent an agency's enforcement of their operations. The more important such industries are, the more actual or potential political pressure governments might feel to impose less-stringent enforcement to appease important economic interests. Some prior research concluded that the economic importance of an industry is related to environmental enforcement actions against that industry (Decker 2005; Lynch et al. 2004b; Davis & Davis 1999; Ringquist & Emmert 1999; McKinney 1998; Ringquist 1998). In addition, as noted earlier, a study found substantial percentages of samples of government and interest group officials claiming that states impose, or advocate imposing, smaller penalties on environmental law violators to encourage industries to locate in their states (Engel 1997).

Another theorized source of political pressure on enforcers might be economic conditions. Some prior research concluded that the unemployment rate or business climate in an area affects environmental enforcement (Decker 2005; McKinney 1998; Gray & Deily 1996; Wood 1992; Deily & Gray 1991). The more depressed economic conditions are, the more pressure environmental enforcers might feel to lower penalties so as not to further undermine the financial viability of defendants and risk job losses. Thus the hypothesis is that the worse the economic conditions in an area, the less stringent the enforcement.

Finally, another theory is that enforcement is less stringent against "public" (i.e., government or quasi-government) defendants than against businesses (Adas 2001; Kagan 1989). It is reasonable to assume that enforcers would receive less internal gratification, fewer public relations kudos, and more political pressure from severely penalizing fellow government entities (or even quasi- public defendants such as private hospitals and educational institutions) (Kuehn 1996). For example, EPA favors municipalities under the federal Superfund clean-up program and EPA's Clean Water Act penalty policy, making their penalties less than those for businesses (U.S. EPA 1998d, 1995b), and there have been policies to forgo penalties against small municipalities (U.S. EPA 2004, 1995c). If states act consistent with EPA policies, this relationship also should exist for state enforcement actions. In addition, some studies of EPA enforcement concluded that public entities are penalized less than business defendants (Adas 2001; Ringquist & Emmert 1999; McKinney 1998; Ringquist 1998).

Hazardous Waste Enforcement

This article evaluates state versus federal enforcement stringency in the context of hazardous waste administrative enforcement actions under the federal Resource Conservation and Recovery Act (RCRA) (1976) and its state law counterparts. Hazardous wastes are industrial process wastes the government designates as dangerous if they are not managed and disposed of with special precautions. A comprehensive regulatory program exists under RCRA to ensure that hazardous waste is properly identified, stored, managed, transported, and disposed of. This includes requirements for generators and transporters of hazardous waste, and treatment, storage, or disposal facilities (TSDFs) managing it. After Congress and EPA create part of the RCRA program, states can be authorized to implement and enforce it, and receive supporting federal grants, if they enact laws consistent with and at least as stringent as federal law (42 U.S.C. [section]6926). Almost all states are authorized to manage some of the RCRA program, but none yet manages it all.3

A state seeking RCRA authorization signs a Memorandum of Agreement (MOA) with the Regional Administrator of whichever of the 10 EPA Regions it is in (40 C.F.R. [section]271.8 [2005]). MOAs describe the rights and responsibilities of the state and EPA under the authorized program (U.S. EPA 1984c). EPA's Regional Offices oversee authorized states in their Regions and directly enforce RCRA in unauthorized states. Although EPA claims it can pursue RCRA enforcement actions even in authorized states, as a matter of policy it typically does not (U.S. EPA 1996a, 1987). Because of this policy and because almost all states are delegated some of the RCRA program, states conduct the overwhelming majority of RCRA enforcement actions (U.S. EPA 2000b). Indeed, federal appellate courts have split (Harmon Industries, Inc. v. Browner 1999; U.S. v. Power Engineering Company 2002) over whether RCRA allows EPA to pursue an enforcement action against an alleged violator when a state with an authorized RCRA program already initiated its own enforcement action (a practice known as "overfilling").

After identifying a violation, the government might pursue it through a criminal action or a civil administrative or judicial process, although some states lack legal authority to impose administrative penalties (Brown & Green 2001; U.S. GAO 2000, 1995). In an administrative enforcement action, the case is processed and the penalty is imposed by the agency (almost always through a setdement negotiated with the defendant), rather than through a judicial process. More than 90 percent of all "formal" (i.e., serious) state and federal RCRA enforcement actions are handled administratively (U.S. EPA 2000b).

The RCRA statute and regulations provide almost no framework for how RCRA penalties should be determined. Thus EPA developed guidance policies for its staff on how to assign RCRA penalties (U.S. EPA 1990a, 1984b). The enforcement staff is first supposed to select a penalty in a specified range applicable to a violation, based on its gravity (i.e., potential for harm and extent of deviation from RCRA). Each day of noncompliance, up to a limit, constitutes a separate violation that should be penalized. In addition, ordinarily the economic benefit of noncompliance to the defendant must be recovered in EPA cases (U.S. EPA 1993a, 1984b). The penalty can be adjusted, within limits, to reflect the defendant's good faith efforts to comply, degree of willfulness, history of noncompliance, and ability to pay. In addition, a defendant can reduce its penalty by receiving credit for a Supplemental Environmental Project (SEP), which is activity benefiting the environment that is not legally required, such as pollution prevention (U.S. EPA 1998b, 1995a, 1991a, 1990a).

Internal and external reviews of EPA RCRA administrative enforcement actions during the 1980s and 1990s, however, always demonstrated EPA staff's widespread noncompliance with its penalty guidance policies (U.S. EPA 1997a, 1989a; U.S. GAO 1992, 1991a, 1988). As one EPA report concluded, "Regional offices have not closely adhered to the national policies and procedures. . . . Regions had: (i) not considered nor negated the economic benefits of violator noncompliance; (ii) proposed insufficient gravity-based penalties and excessively mitigated them during negotiations . . ." (1989a:3). Another EPA report concluded, "[a] common perception is that the RCRA program does not seek penalties in appropriate cases, or that the penalties proposed or assessed in the RCRA program are low, both when compared to the other major EPA media programs (air and water) and in terms of creating an effective deterrent" (1990b:61).

These reviews also always concluded that state RCRA administrative enforcement programs were even less stringent than federal programs. EPA's position is that "state enforcement penalties must be at least equivalent to penalties provided for in RCRA; and state enforcement activities must be equivalent to those performed by EPA" (1998g:1-5). States are not required, however, to have RCRA penalty policies identical to those of EPA (U.S. EPA 1997a, 1993a, 1986).4 Some states had no written RCRA penalty policies, some adopted at least parts of EPA's policies (though they did not necessarily implement some of those parts), and some developed their own, with varying statutory or regulatory limits on penalties inconsistent with those of EPA (U.S. EPA 2001, 1997a). Just as with RCRA, an outside observer could review a state's hazardous waste statutes and regulations and not know if the state even had a hazardous waste penalty policy, much less its provisions.

Some of these characteristics of the RCRA administrative penalty process would be expected to facilitate shirking by state enforcers. Despite EPA's instructions that state penalties should be at least as stringent as what EPA would impose, there are no legally binding requirements on how to determine penalties. Furthermore, EPA does not even require states to have written policies on how to determine penalties, much less have policies similar to those of EPA. Thus state agents can take advantage of this grant of discretion and by EPA's lack of information on the state's penalty procedures. In addition, because almost all administrative penalties are settled through negotiations between state enforcers and violators, there typically is no public record on how penalty amounts are decided. In addition, EPA's unimpressive experience in adhering to its own RCRA penalty policies and imposing stringent penalties makes it a less credible source of sanctions against state agents for their lackluster enforcement.

RCRA enforcement actions comprised about 10 percent of formal state environmental enforcement actions from the mid-1980s through the 1990s (EPA 2000b) and about 10 and 25 percent of EPA administrative penalty cases and fines, respectively, during the 1990s (U.S. EPA 1991b, 1992, 1993b, 1994, 1995d, 1996b, 1997b, 1998f, 1999d, 2000b). Although it cannot be assured that any differences between federal and state governments in RCRA enforcement are representative of their other environmental enforcement efforts, it is reasonable to assume that any such differences reflect basic, broadly applicable approaches. Penalties are particularly useful in evaluating the consistency and stringency of enforcement efforts. First, penalties are measured in units- dollars -that have the same meaning everywhere. In contrast, inspections vary in quality (e.g., thoroughness) and NOVs vary in seriousness, making comparisons uncertain. Second, agencies must annually conduct certain numbers of environmental inspections. These federally imposed minima make it more difficult to assess if inspections simply meet required levels or exceed them. In contrast, there are no federal minima for penalties. Third, the penalty imposed should reflect the severity of an enforcement action, whereas the number of inspections or NOVs might at least partly reflect the severity of environmental concerns in an area. Indeed, a low number of enforcement actions might indicate not an agency's laxness, but rather a high level of compliance due to its past stringent enforcement efforts. Consequendy, although imperfect, penalty amounts have advantages as a measure of enforcement stringency. Data

Selection of Enforcement Actions

Data from various sources were used to test these enforcement stringency theories. The RCRA enforcement actions were from EPA's Resource Conservation and Recovery Information System (RCRIS) National Oversight Database (U.S. EPA 2000a). RCRIS is supposed to contain information about every facility that has notified the government that it generates or manages hazardous waste, or both.5 "Core" data from EPA Regional databases were extracted to create the RCRIS National Oversight Database. Core data include data necessary (as negotiated by states, EPA Regions, and EPA headquarters) for national oversight of the RCRA program (U.S. EPA 1998a). Some of these data are publicly available, including details on RCRA enforcement actions.

Pursuant to EPA procedures and MOAs with states, not all RCRA enforcement actions are in RCRIS. Although reportable enforcement actions might vary, ordinarily they involve the most serious violations. Of my analyses' enforcement actions, 84 percent were categorized by EPA as "Class I" violations, the most serious category. These are

[d]eviations from regulations, or provisions of compliance orders, consent agreements, consent decrees or permit conditions which could result in a failure to: (a) Assure that hazardous waste is destined for and delivered to authorized treatment, storage, or disposal facilities (TSDFs); or (b) Prevent releases of hazardous waste or constituents, both during the active and any applicable post-closure periods of the facility operation where appropriate; or (c) Assure early detection of such releases; or (d) Perform emergency clean-up operation or other corrective action for releases (U.S. EPA 1998a: 12.2.4).

Thus my analyses focus on the more serious enforcement actions. Although this makes these data unrepresentative of all RCRA enforcement actions, those included should be of greatest interest to policy makers and the public. In addition, using these enforcement actions inherently controls for some variation in the severity of violations.

Although RCRIS contains various types of enforcement actions, my analyses included only final compliance orders under RCRA [section]3008(a) (42 U.S.C. [section]6928[a]) or its state law equivalent, through which the government can issue an order imposing a civil penalty on, and require immediate compliance by, a facility violating RCRA. These final [section]3008(a) orders comprised 67 percent of enforcement actions in RCRIS. Most other enforcement actions were initial [section]3008(a) orders (15 percent), which were largely duplicates of their accompanying final [section]3008(a) orders; judicial cases (10 percent), the penalties for which were thus not decided by agencies; or informal verbal or written administrative actions (7 percent), most with penalties under $1,000. Consequently, [section]3008(a) orders encompass the overwhelming majority of all RCRIS enforcement actions and essentially all relevant ones.

Of 6,759 listed [section]3008(a) enforcement actions, I deleted 367 that ended before 1986 or after 1999. Pre-1986 cases were excluded because EPA issued its formal guidance policy in 1984 requiring that its penalties recover any economic benefit of noncompliance to the defendant (U.S. EPA 1984a). Thus penalties in such cases might be systematically lower than later ones due to this required recapture of economic benefits.

I also excluded 81 enforcement actions whose data were suspect (e.g., likely duplicate records or enforcement actions in which one penalty was listed for violations at multiple facilities). I also excluded enforcement actions that had no penalties (presumably these [section]3008[a] orders required other actions by defendants), occurred outside of the 50 U.S. states, or were from 16 states in which less than 20 enforcement actions each were listed or from two states whose numbers of enforcement actions combined with the numbers of their applicable EPA Region enforcement actions were less than 100. This left 4,978 state and 1,070 federal enforcement actions for my analyses. Although these comprised only one type of enforcement action, there was wide variation in the typical penalties imposed. The median penalties in states ranged from $2,816 to $43,244, with half the states having median penalties below $10,000, and one-quarter each with median penalties between $10,000 and $20,000 and more than $20,000, respectively. The median penalties in EPA Regions ranged from $13,714 to $54,247.

Dependent Variable

The dependent variable in my analyses is the dollar amount of the penalty imposed in the enforcement action, as reported in RCRIS. This penalty comprises the money paid by the defendant to the government plus, where applicable, the cost to the defendant of an SEP. To measure the severity of each penalty in constant dollars, I transformed the penalties into 1999 dollars by adjusting them consistent with the annual Consumer Price Index (from the U.S. Bureau of Labor Statistics Web site at http://www.bls.gov/cpi). Because the distribution of penalties was significantly skewed, due to relatively few very large penalties, a log transformation was used to produce more normally distributed transformed penalties.

Independent Variables: Political Environment Characteristics

Partisan Control of Government

Dummy variables were used to indicate partisan control of the government (i.e., federal or state) imposing the penalty. Control of the executive branch was indicated by the political party of the governor or president for state or federal actions, respectively. Control of the legislative branch was determined by how many legislative houses were controlled6 by each party. Thus an EPA case would reflect partisan control of the U.S. presidency, House, and Senate at the time the case ended. A state case would reflect partisan control of the governorship and state legislative houses at the time the case ended.7

These dummy variables reflected the following scenarios: (a) Democratic Party controlled executive branch and both legislative houses (Dem. executive/legislature), (b) Democratic Party controlled executive branch and one legislative house (Dem. executive/ divided legislature), (c) Democratic Party controlled executive branch and neither legislative house (Dem. executive/Rep. legislature), (d) Republican Party controlled executive branch and neither legislative house (Rep. executive/Dem. legislature), (e) Republican Party controlled executive branch and one legislative house (Rep. executive/ divided legislature), and (f) Republican Party controlled executive branch and both legislative houses (Rep. executive/ legislature).

The scenario excluded as the dummy variable reference group in my regression analyses was Dem. executive/legislature. Whichever of the remaining dummy variables matched the situation when the penalty was imposed was coded as a one, and the others as zeroes. Thus the sign of the coefficient for each partisan control dummy variable indicates if penalties imposed by that type of government were more or less stringent than penalties from a government completely controlled by Democrats.

Environmental Group Presence

Pressure from organized public interest groups was represented by the percentage of the population that belonged to the Sierra Club in the state in the year the penalty was imposed (Environmental group members).8 The state population estimates for each year from 1986 to 1999 were from the U.S. Census Bureau (http://www.census.gov/popest/ estimates.php). Although the Sierra Club provided its membership totals by state, only 1989-1999 were available. To fill the gap from 1986 to 1988, I used time-series analyses to estimate those membership totals.

There was considerable fluctuation in the state membership data from 1989 to 1999, but the 1989-1991 data in each state had a consistent linear pattern. Thus each state's Sierra Club membership from 1986 to 1988 was estimated through a time-series analysis of its 1989-1991 membership data. Though only three data points were in these analyses, each state's time-series equation produced an excellent fit to its 1989-1991 membership data, and thus hopefully to its 1986-1988 data.9 Therefore, this combination of actual and estimated Sierra Club membership was used and, because this variable's relationship with penalties might be nonlinear, a quadratic form of it was included. As noted later, the importance of this variable did not meaningfully change if the analyses excluded the 1986-1988 estimated data. Environmentalism

To represent generalized pressure for more-stringent enforcement from Environmentalism sentiment, I used the mean LCV rating of a state's U.S. House of Representatives delegation in the year the penalty was imposed and, because its relationship with penalties might be nonlinear, a quadratic form of it. Although some prior research concluded that congressional LCV ratings are associated with other measures of state environmentalism (Mazur & Welch 1999), this still is an uncertain measure of the latter. LCV ratings should, however, be a meaningful measure of political office holders' environmentalism.10 Consequendy, these officials' environmentalism-rather than or in addition to that of the public- might be influencing environmental enforcers' penalty decisions.

Economic Importance of Hazardous Waste Industries

To represent pressure from the regulated industry, I first determined the industries most involved with hazardous waste. Although the hazardous waste management industry (i.e., hazardous waste transporters and TSDF operators) obviously is involved with hazardous waste, no reliable economic data on it were available until separate categories for it were established in the U.S. Census Bureau's 1997 Economic Census. Therefore, the absence of data alone makes it difficult to consider this industry for this analysis. However, the 1997 Economic Census also indicates that this industry ordinarily should be of little economic importance to a state. In the entire nation in 1997, only 414 establishments collected hazardous waste, and only 512 treated or disposed of it (U.S. Census Bureau 2001)-a combined average of fewer than 20 per state (and some establishments might have done both). These establishments accounted for an average of only about 500 employees and $80 million in receipts per state.

Consequently, if political influence flows from economic importance, the hazardous waste management industry should add little to the political influence of other hazardous waste facilities-those that generate (and perhaps treat or dispose of their own, but not others') hazardous waste. Although some facilities in probably almost every industry generate hazardous waste, most hazardous waste generation is concentrated in a few industries. To identify these industries, EPA's Biennial Reporting System (BRS) data were used. Beginning in 1983, large-quantity generators (LQGs) of hazardous waste (i.e., those generating at least 1,000 kilograms of hazardous waste in at least one month a year) have submitted information about their hazardous waste generation in odd-numbered years (40 C.F.R. [section]262.41 [2005]), which is made publicly available by EPA in BRS. Although LQGs comprise a small percentage of hazardous waste generators, they account for more than 90 percent of hazardous waste generated annually in the United States (Adas 2002). Among the information reported by LQGs for each of more than 1.3 million hazardous waste streams from 1991 to 1999 is the Standard Industrial Classification (SIC) code of the industrial process from which the waste came, which indicates an LQG's industry.

Thus for each BRS year from 1991 to 1999, I compiled the unique two-digit SIC code(s) that applied to the hazardous waste stream(s) generated by each LQG. For each BRS year, I determined which two- digit SIC codes appeared most often. Although the legal definition and volume of hazardous waste generated, and BRS reporting requirements, changed substantially over the years, the same four industries were reported most frequently in each BRS year. These industries-Chemicals and Allied Products (SIC code 28), Fabricated Metal Products (SIC code 34), Electronic and Other Electric Equipment (SIC code 36), and Transportation Equipment (SIC code 37)- accounted for 58.4 to 61.2 percent of all SIC codes reported in a BRS year. The percentage accounted for by any one of these SIC codes varied litde across the BRS years, and no other SIC code accounted for more dian a few percent of SIC codes. Therefore, these data were very stable over time.

I then determined the total value added by manufacturing by these four industries for each state for each year from 1986 to 1999 from U.S. Census Bureau Annual Surveys of Manufactures and Economic Censuses.11 Finally, I determined the percentages these amounts were of the gross state products, as reported for each year in each state in the U.S. Bureau of Economic Analysis Survey of Current Business reports. This percentage was used to represent the economic importance of hazardous waste industries in the state in the year the penalty was imposed (Industry importance) and, because its relationship with penalties might be nonlinear, a quadratic form of it was included.

Economic Conditions

To represent pressure for less stringent enforcement when economic conditions are adverse, I used the statewide Unemployment rate three months before the penalty was imposed (from U.S. Bureau of Labor Statistics data at http://www.bls.gov/lau/ home.htm) and, because its relationship with penalties might be nonlinear, a quadratic form of it. It is uncertain how long before a penalty was formally imposed that the amount was decided upon, but the three- month time lag seems reasonable. I also tried five-and seven-month lagged unemployment, but they were almost perfectly correlated with the three-month figures and thus it makes no meaningful difference which was used.12

Business vs. Public Defendant

To represent the bias for less-stringent enforcement against public entities, I reviewed how defendants were categorized in RCRIS (which has a code indicating the owner/operator type) and examined their names to verify the RCRIS categorization. I corrected or, when missing, supplied a few percent of the defendants' categorizations. Public defendants included government agencies, medical facilities, and educational institutions (the latter two even if they were private institutions), and were coded as zeroes for this Business defendant dummy variable.

Independent Variables: Enforcement Action Characteristics

Naturally, the characteristics of an enforcement action also should be related to its resulting penalty. Thus in addition to the above-described independent variables representing political environment factors possibly affecting enforcement stringency, variables were included controlling for aspects of the enforcement action that should determine penalty severity.

Seriousness of Violation

I hypothesized that penalties are higher for more serious violations. Some past research concluded that the seriousness of environmental violations affects penalty amounts (McKinney 1998; Oljaca et al. 1998; Cohen 1992, 1987). In addition, pursuant to RCRA [section]3008(a)(3), the government must take into account "the seriousness of the violation" in determining a penalty. EPA's RCRA penalty policies require the seriousness of the violation to be based upon the "potential for harm; and extent of deviation from a statutory or regulatory requirement" and explicitly state that more serious violations should be penalized more severely (U.S. EPA 1990a, 1984b). If states act consistent with EPA policies, this relationship also should exist for state enforcement actions.

The only RCRIS information specifically, though only generally, indicating the overall seriousness of violations is if they were Class I violations, which were defined earlier. Thus a dummy variable was included indicating if an enforcement action was so designated.

Number and Type of Violations of Each Law

A violation's seriousness also might be reflected in the specific RCRA requirement violated, as some might inherently pose greater potential harm than others (e.g., EPA's penalty policies recommend considering the risk of exposure to hazardous waste posed by a violation [U.S. EPA 1990a, 1984b]). Some prior research concluded that a key determinant of a penalty is the type of environmental violation (Atlas 2001; Helland 2001; Ringquist & Emmert 1999; Kleitetal. 1998; McKinney 1998; Ringquist 1998; Hamilton 1996).

In addition, I hypothesized that the more violations in an enforcement action, the higher the penalty because the government can penalize each violation. Therefore, as indicated by prior research, more environmental violations should lead to higher penalties (Kleit et al. 1998; Ringquist 1998; Hamilton 1996). In addition, the number of violations might reflect the seriousness of the matter, as more violations likely pose greater potential harm and show more deviation from legal requirements. Consequently, more violations might indicate more-serious offenses, thereby justifying higher penalties.

Thus independent variables were included for the number of violations of each of certain categories of violations in an enforcement action. RCRIS assigns violations to categories, all of which I included if there were any such violations in my selected cases, although I combined similar categories with few violations. The following are the categories included:

1. Compliance agreement violations: not fulfilling a prior legal agreement with the government (three RCRIS categories were combined into this category),

2. TSDF combustion violations: thermal treatment of hazardous waste (seven RCRIS categories were combined into this category),

3. TSDF closure violations: actions required to close and, if necessary, clean up a TSDF,

4. TSDF financial assurance violations: assuring funds are available to pay for spills or closure activities,

5. TSDF groundwater monitoring violations: groundwater monitoring required for land disposal facilities,

6. TSDF land ban violations: requirements that hazardous waste be treated before disposal on land,

7. Other TSDF violations,

8. Generator land ban violations: requirements that hazardous waste be treated before disposal on land,

9. Other generator violations,

10. Transporter violations: hazardous waste transportation requirements, and 11. Other violations.

Thus independent variables were included for the number of violations in each of these categories in an enforcement action. I hypothesized a positive relationship between the number of violations in a category and the penalty.

SEPs

I hypothesized that the penalty is higher if a SEP credit is part of it. This is because both the government and defendants have incentives to exaggerate the expected SEP cost. For a defendant, the higher the claimed SEP cost, the lower (to a limit) its monetary penalty due to the resulting SEP credit. For the government, the higher the claimed SEP cost, the more severe the publicized penalty, which might have some public relations and deterrence value (Dana 1998). Thus a dummy variable was included indicating if the penalty included a Supplemental Environmental Project.

Prior Violations

I hypothesized that the more prior violations by a defendant, the higher the penalty in the current enforcement action. Under EPA's RCRA penalty policies, a defendant's prior noncompliance with hazardous waste laws justifies a higher penalty, as it demonstrates that the defendant was undeterred by prior enforcement (U.S. EPA 1990a, 1984b). If states act consistent with EPA policies, this relationship also should exist for state enforcement actions. Some research concluded that a history of prior environmental violations affects subsequent penalties (Lynch et al. 2004a, b; Ringquist & Emmert 1999; McKinney 1998; Oljaca et al. 1998; Ringquist 1998).

An independent variable was included for the number of Prior violations by a defendant and, because its relationship with penalties might be nonlinear, also a quadratic form of it. A defendant's violations prior to a particular enforcement action were determined through the "violation control number(s)" assigned by EPA to the violation(s) in the enforcement action.13 Part of that number indicates that violation's sequence within all RCRA violations for that defendant. Thus the number of prior violations was determined by subtracting one from the lowest such sequence number among the violations in the enforcement action in question.14

Time

I hypothesized a relationship between penalties and the timing of the enforcement action. EPA has asserted, as supported by its overall enforcement results, that it has been more aggressive in enforcement since the mid-1980s. It is possible states might claim the same. In particular, the number of EPA and state RCRA administrative actions generally increased from the mid-1980s to the early 1990s, then decreased for a few years before increasing again (Zinn 2002; U.S. EPA 2000b). In addition, according to EPA data, its median RCRA administrative penalty increased almost linearly from the early 1980s to at least the early 1990s, unlike penalties in its other programs (U.S. EPA 1993b). Some prior research controlling for other factors concluded that EPA RCRA administrative penalties and EPA civil judicial penalties increased over this time (Atlas 2001; Ringquist & Emmert 1999; Hamilton 1996). Of course, if states are pursuing a race-to-the-bottom in enforcement stringency, one would expect their penalties to decrease over time.

I represented time as the years from 1986 to the date of the enforcement action in question (i.e., number of days divided by 365). Because the relationship between it and penalties might be nonlinear, a quadratic form of it also was included. Because the trend over time might be different for federal and state governments, two sets of time variables were included for federal and state actions, respectively. Thus, EPA time equaled zero if the enforcement action in question was a state action, and equaled the years from 1986 to the date of the enforcement action if it was a federal action. The State time variable was calculated in a similar manner.

Table 1 displays the means and standard deviations of the variables in my analyses, separated into the groups of federal and state enforcement actions, respectively.

Analyses

Analytic Approach

To test federal versus state environmental enforcement stringency, as well as the other above-described hypotheses, I began with a separate ordinary least squares regression model for each of the 32 states that had more than 20 state enforcement actions and more than 100 combined state and EPA Region actions. It was necessary to compare each state's penalties against federal penalties of the EPA Region it was in because EPA Regional Offices have a substantial flexibility in pursuing enforcement actions according to their own priorities and policies, though within constraints imposed by EPA Headquarters (U.S. EPA 1996a, 1987, 1984c). Thus there could be variation between EPA Regions based on the types of RCRA enforcement actions they pursue, the severity of penalties they impose, and the degree of oversight and pressure they bring to bear upon the RCRA-authorized states in their respective Regions. This inter-Region variation, both among EPA Regions and the states within them, has been noted elsewhere (U.S. GAO 2000, 1988; Hamilton 1996; Hunter & Waterman 1992).

Thus to determine if state penalties differ from EPA penalties, it was most appropriate to compare the penalties for each state within an EPA Region against that Regional Office's penalties. That Regional Office is the principal to the state's agent. For example, a state's penalty could be higher than the typical national EPA penalty in a comparable enforcement action only because the state was in an EPA Region that pressured states to be more severe in their punishments. Because ordinarily the alternative to a state enforcement action is one by the EPA Regional Office, not EPA headquarters, the former's actions should be used as the benchmark. This prevents differences between EPA Regions' enforcement approaches from biasing the results.

Consequently, each state regression model also included a dummy variable (State action) indicating if the state or the EPA Region prosecuted the enforcement action in question. Therefore, the statistical significance and direction of this variable's coefficient indicated if state enforcement actions were more (positive coefficient) or less (negative coefficient) stringent than EPA enforcement actions in the EPA Region in question. The hypothesis was that these State action variables would not be statistically significant, as this would indicate no difference in the outcomes of EPA versus state enforcement actions, and thus no shirking by the state agents. Because an EPA Region's enforcement stringency should not vary by state within that Region (other than due to some of the political environment characteristics described earlier), the regression analysis for each state included all of that state's enforcement actions and all of the EPA enforcement actions of the EPA Region it was in, including those EPA pursued in other states in the Region.

To lessen multicollinearity concerns,15 any quadratic term that was not statistically significant after the initial run of each regression model was deleted. This eliminated all but 10 of the initial 224 quadratic terms. In addition, any variable that was not statistically significant in subsequent model runs and produced a variance inflation factor score exceeding 10 was deleted, and the model was rerun. This occurred in only 18 of the 32 regression models, with typically only one variable being deleted. Thus these infrequent removals only of variables that were not statistically significant helped lessen multicollinearity concerns without losing any meaningful information.16 Each model also was tested for heteroskedasticity, and the results used robust standard errors for five states' models in which it was statistically significant.

Table 2 displays a summary of the results of the final regression analysis for each state. Although the full results are available from the author, for purposes of brevity-given the large number of regression models-and to make comparisons across states easier for readers, only the numbers and coefficient signs of the statistically significant variables are reported. The state results are also grouped by EPA Region to facilitate comparisons across Regions.

State versus Federal Enforcement Stringency

The mean r^sup 2^ of the 32 models was 0.376, which places them in the upper range of r^sup 2^'s reported by other studies of environmental penalties (Adas 2001; Ringquist & Emmert 1999; McKinney 1998; Oljaca et al. 1998; Ringquist 1998; Hamilton 1996; Cohen 1992, 1987). Six of the models produced r^sup 2^'s over 0.5, which are higher than any prior study of environmental penalties, and nine produced r^sup 2^'s between 0.4 and 0.5. There were substantial differences in r2's across EPA Regions, with the state models in Regions 2 and 5 having consistently lower r^sup 2^'s, as well as the one-state models in Regions 7 and 9. Despite these relatively high r^sup 2^'s, some of the determinants of penalties clearly were not captured by the models. This is not surprising, as much of the information that should be used in setting penalties (e.g., days of noncompliance, cooperation of defendants, and seriousness of violation) were entirely unavailable or insufficiently specific in the data. The lack of adequately detailed data makes all penalty studies inherently difficult.

Of the 32 state models, 23 produced statistically significant differences in penalties between state and EPA enforcement actions. Thus this indicates that differences exist in state and federal enforcement stringency. Indeed, this variable was statistically significant more often than any other variable. Furthermore, in 14 of the 23 states in which State action was statistically significant, the absolute value of its coefficient was the largest of the statistically significant dummy variables, and its coefficient was among the three largest in the other nine states, indicating the importance of its relative effect on penalties. In addition, in 18 of the 23 states in which State action was statistically significant, the absolute value of the change in penalty due to the state, instead of EPA, taking the enforcement action was larger than a two standard deviation change in any of the statistically significant continuous variables. Of the 23 states with statistically significant differences in penalties, 21 had lower penalties than their EPA Regions. Thus this supports concerns about state agents shirking their enforcement obligations. To assess the extent of the difference in federal and state enforcement stringency, I used the final regression model for each state to estimate what the penalty would have been for each state enforcement action if it instead had been prosecuted by the applicable EPA Region. This was done by simply changing the value of the State action variable in these cases from one to zero and running the values of it and the other independent variables through the final model equation to predict the estimated EPA penalty in each case. The mean estimated EPA penalty was $80,829, versus the mean actual state penalty of $30,718. So on average, state penalties were 62 percent lower than what the EPA penalty would have been. The difference between median penalties was even larger. The median estimated EPA penalty was $27,399, versus the median actual state penalty of $7,868, a reduction of 71 percent. Consequently, the disparity in environmental stringency is substantial.

Of 11 states that did not have statistically significant lower penalties than EPA, four were in EPA Region 5. Therefore, it appears that something in this EPA Region drew together the states' and EPA's enforcement outcomes. One possibility is that EPA Region 5 imposed more pressure on its states to be consistent with its enforcement policies. As a GAO report noted:

EPA's [Region 5] Chicago office, for example, has long held a reputation for having an aggressive enforcement program in which the region would act quickly and forcefully if it determined that the state was not performing its responsibilities adequately. A Chicago office official told us that the region believes that it is important to maintain an "enforcement presence" in states as a deterrent to the regulated community, in contrast to other regions that believe having to take an enforcement action is a sign of failure (2000:38).

Thus, EPA Region 5 might have been more aggressive in using monitoring and sanctions to police its principal-agent relationship with states.

It also is notable that EPA's Regional Offices were located in four of the seven states outside Region 5 that did not have statistically significant lower penalties than EPA. Perhaps geographical closeness between state officials and the federal officials monitoring them makes oversight easier by reducing the information imbalance between the principal and agent, thereby promoting more consistency between their enforcement policies. In addition, during at least some of the time period studied, some of the seven states had the explicit statutory authority to include in their environmental penalties the economic benefits defendants gained due to noncompliance. Consequendy, this was more consistent with EPA's policy and thus was less likely to lead to lower penalties than from EPA.

Effects of Political Environment Characteristics

For the political environment characteristics, the results indicate that partisan control of government is often related to penalties. At least one partisan control scenario was statistically significant in 20 of 32 state models. Furthermore, 80 percent of the time they were statistically significant, it was in the expected direction-penalties were lower when the Democratic Party did not completely control the executive and legislative branches.17 I tested another set of models in which the partisan control scenarios were replaced with just a dummy variable for the party controlling the executive branch, which should be most responsible for environmental enforcement. Although a less refined reflection of partisan control, this variable was statistically significant in 13 of the 32 models, all but one of which indicated that Democratic executive branches were associated with higher penalties.

The other independent variables reflecting potential political pressures on environmental enforcers, however, were typically not statistically significant in more than five states each. The one exception was that Environmental group members in a state had a statistically significant relationship with penalties in seven states, five of which were in the hypothesized positive direction. Four of those five states are in EPA Region 5, indicating that the effect of environmental group members was typically important in- and only in -Midwestern states. Excluding 1986 to 1988 enforcement actions and their accompanying estimated, rather than actual, Sierra Club membership data, would not change the overall results for this variable.

Effects of Enforcement Action Characteristics

All types of variables representing enforcement action characteristics frequently had statistically significant relationships with penalties, almost always in the hypothesized directions. First, the most serious violations received significantly higher penalties than other violations. A Class I violation was statistically significant in 18 of the 32 state models, always with a positive relationship. Second, all but one of the state models produced a statistically significant relationship with one or more types of RCRA violations. There were 81 statistically significant relationships between penalties and violation types across the models, 89 percent of which were in the hypothesized positive direction. Thus the more violations of almost any particular RCRA requirement, the higher the penalties.

Third, there was a st

Source: Law & Society Review

More News in this Category