1. Introduction
What factors explain variations in environmental performance across political jurisdictions? A typical list would include political considerations such as government ideology and interest group pressures, economic influences such as wealth and unemployment rate, and factors related to the severity of environmental problems. In addition to these and other longstanding considerations, however, recent research suggests an additional possibility: that the bureaucratic structure of the implementing agency itself may impact its performance in systematic and predictable ways.
Agencies that combine environmental protection responsibilities with other functions like public health or natural resource management, this line of argument goes, regulate pollution emissions less stringently than those that focus strictly on environmental protection. Broadly speaking, this is due to competing constituencies, cultures, and priorities in agencies with multiple functions, which may lead them to de-emphasize environmental protection relative to other agency goals.
This idea, which may be termed the agency structure hypothesis, has largely been developed and tested within the context of the American states. However, it is potentially relevant in many other contexts. In order to avoid redundancy and foster synergies across complementary policy areas, many countries have combined environmental functions with others, such as public health, energy, or natural resource conservation [
1]. Examples include the German Federal Ministry for Environment, Nature Conservation, Nuclear Safety, and Consumer Protection (BMUV), and the French Ministry of Ecological Transition, which is responsible for implementing not only environmental protection policies but also those related to national parks, transportation, civil aviation, energy, and housing.
Nonetheless, the American states form an ideal arena with which to test the bureaucratic structure hypothesis for several reasons, including the fact that while there is sufficient variation in bureaucratic, political, and environmental factors to allow analysts to examine their effects, there is not so much variation as to make meaningful comparison difficult [
1]. Moreover, the core environmental statutes implemented by state agencies are derived from overarching federal legislation that applies to every state. This uniformity in the laws themselves makes it much easier to isolate the role that bureaucratic factors play in generating different regulatory outcomes. These regulatory differences may significantly impact pollution levels and thus overall environmental sustainability.
Using a set of panel data models, this study assesses the bureaucratic structure hypothesis (1) across states and over time, (2) across multiple indicators of regulatory stringency, and (3) across multiple environmental programs, including those regulating air pollution, water pollution, and the disposal of hazardous waste. The results should provide a much broader and more comprehensive picture of the effects of agency design on environmental regulation than currently exists.
2. The Organizational Design of State Environmental Agencies and Why It Matters
States play crucial role in the implementation of most major U.S. environmental policies. Seminal environmental protection laws such as the Clean Air Act, Clean Water Act, and Resource Conservation and Recovery Act (which governs hazardous solid waste) contain provisions that allow state agencies to implement and enforce them. If states choose to assume primary implementation authority, or primacy, they are the principal actors in charge of the implementation and enforcement of these federal laws within their borders [
2]. By now, the vast majority of states have chosen to assume this authority under major federal environmental statutes [
3,
4].
Today, the state agencies in charge of implementing these policies are housed in one of three basic structures: (1) agencies with a specific focus on environmental protection (state EPA), (2) agencies that combine environmental protection with natural resource management functions, such as parks, forests, fish and game, or agriculture (NREPA), or (3) agencies that combine environmental protection with public health functions, such as vaccine administration and health care finance (PHEPA) [
5]. This distribution of agency types is not static; rather, it reflects an ongoing pattern of institutional evolution. Over time, environmental agency structures have evolved from being primarily PHEPs and “decentralized aggregations of boards or commissions” [
6] (p. 38) to more consolidated agencies such as state EPAs or NREPAs [
7]. At present, the latter two are the dominant forms of agency structure; only five states continue to employ PHEPAs, and none employ a decentralized approach [
5].
For the bureaucratic structure hypothesis, the critical distinction is that environmental agencies have an exclusive focus on environmental protection, while NREPs and PHEPs house their environmental protection apparatus within agencies that contain other functions. Agency design theorists contend that this initial design choice has several interrelated implications that lead state EPAs to protect the environment more aggressively than multi-function agencies. First, the agencies should have different constituencies. Agencies with more unified constancies are likely to behave differently than agencies with competing constituencies [
8,
9]. State EPAs are expected to be responsive, first and foremost, to environmental interest groups and the mass environmental movement, and they may often have adversarial relationships with the industries that they regulate. In other types of agencies, however, environmental constituencies would face greater competition from other stakeholders. This effect may be particularly pronounced in NREPs since their constituencies may often have financial interests that sometimes lie directly in opposition to environmental protection. For instance, DNRs often have mandates that include developing and utilizing natural resources such as timber and fisheries and often have important external constituents among developers and extraction industries [
10]. Thus, as Ringquist notes: “While {NREPs} can enhance comprehensive environmental management, it is equally likely that environmental protection may be de-emphasized in favor of an agency’s natural resource development responsibilities” [
6] (p. 38).
Second, these agencies should have different cultures. Agencies with combined functions are more likely to have agency personnel with different professional backgrounds and cultural norms, and when cultures conflict within an organization, certain tasks are likely to be devalued [
11]. While EPAs often have strong regulatory cultures that emphasize compliance by affected industries, natural resources and public health officials are more likely to come from backgrounds emphasizing negotiation and accommodation with regulated interests [
1]. Natural resources management plans often involve extraction industries as participants in the planning process, and successful conservation efforts often require collaboration among a number of stakeholders and industries [
10]. Public health officials likewise tend to express a preference to include stakeholders in agency decision-making processes directly [
1]. Thus, in multi-function agencies, there may be an internal conflict between cultures that emphasize strict regulatory compliance and those that emphasize participation, negotiation, and accommodation.
Finally, because of differences in statutory mandates, interest group pressures, and organizational cultures, agencies should have different priorities. Goal ambiguity may lead to competition among internal priorities in multitask agencies [
12,
13], which, in turn, may reduce their level of environmental aggressiveness relative to state EPAs. At times, these priorities may directly oppose environmental protection goals, as with the economic development concerns of NREPs. Even in cases where they do not conflict, other priorities may push environmental enforcement down the list. For instance, less than 3% of the budget of Kansas’ PHEP agency goes toward environmental protection activities, which are of low priority for many public health officials who are typically more concerned with administering the state’s health care finance system, which consumes early 90% of the agency budget [
5]. Moreover, “a…divergence of culture may exist between the enforcement divisions within state environmental agencies, which view protection of the environment as their mission, and enforcement divisions in Departments of Natural Resources, which view their primary mission as making sure deer season goes off without a hitch” [
14] (p. 88).
Such concerns are not new. In the early stages of state environmental agency reorganization, observers detailed how differences in agency structure, culture, and priorities may result in uneven policy implementation across states [
15]. In subsequent years, a relatively small number of studies empirically examined the effect of agency structure on policy adoption, expenditures, or enforcement, with mixed results [
14,
16,
17,
18,
19].
The most in-depth, detailed, and convincing treatment of the bureaucratic structure hypothesis, however, comes in a series of recent works by Hopper [
1,
5,
10]. Taken together, these works marshal a variety of evidence, including interviews of agency personnel, text analysis of press statements and other public documents, and multivariate analyses of enforcement actions, to convincingly argue that competing constituencies, cultures, and priorities lead to a weakened emphasis on environmental regulation in state agencies that combine environmental protection with other functions.
Though persuasive, Hopper’s work does raise questions regarding the generalizability of these results. In particular, her multivariate analyses of regulatory stringency examine a single policy area (clean air regulation), using specific indicators of regulatory stringency (number of enforcement actions and dollar value of penalties assessed) during a particular window of time (2010 to 2014). Do these results hold across policy areas, regulatory activities, and time? The latter issue may be especially pertinent because there was little or no change in the bureaucratic structure of state environmental agencies during the time period she examines, and her dataset contains no examples of regulation by decentralized environmental boards or commissions. The latter might be expected to regulate less aggressively than state EPAs due to their fragmented nature [
19], but they don’t fit neatly into the single issue vs. multitask agency dichotomy at the heart of the bureaucratic structure hypothesis, and little theorizing has been done about them. This study addresses these questions in order to assess the scope and durability of the effects of agency design more comprehensively.
3. Materials and Methods
3.1. Dependent Variables: Regulatory Activities
Regulation of pollution emissions constitutes the most traditional approach to achieving environmental sustainability [
20,
21]. The analyses reported below assess two types of state agency regulatory activities: agency monitoring actions (inspections) and agency enforcement actions (an unweighted sum of total punitive actions). Each variable is measured as an inspection or enforcement rate per facility for the federal Clean Air Act, Clean Water Act, and Resource Conservation and Recovery Act (which governs hazardous and solid waste). These are standard measures of environmental regulatory activities and have been employed in a wide variety of studies [
22,
23,
24,
25]. One additional alternative indicator of regulatory stringency that is sometimes employed is the dollar value of penalties imposed for noncompliance. This indicator is utilized in addition to enforcement actions in Hopper’s work [
1,
5,
10] but is not included here due to data limitations.
3.2. Dependent Variables: Regulatory Compliance Costs
In addition to these program-specific variables, an additional model is run using the private costs of pollution abatement as the dependent variable. This is the broadest indicator of regulatory stringency because (1) it captures the effects of a variety of environmental programs including, but not limited to, those listed above, and (2) it incorporates not only the aggressiveness of environmental monitoring and enforcement actions but also the stringency of environmental regulations themselves. These data come from Levinson’s industry-adjusted index of state environmental compliance costs, which is constructed from state-level data on private pollution abatement and control expenditures [
26]. The original data, which were compiled annually by the U.S. Census Bureau until 1994, were adjusted by Levinson to account for differences in state industrial composition so that they indicate the average pollution abatement costs per dollar of state output if each firm in the state conformed to the national average for its industry. The resulting measure has been used as a proxy for regulatory stringency in a variety of other studies [
4,
27,
28,
29].
3.3. Independent Variables
The primary independent variables are a set of dummy variables representing the types of agencies discussed above: NREPs, PHEPs, decentralized agencies, and state EPAs. The data on state agency structures over time were originally compiled by Emily Bedwell through an exhaustive study of original state documents [
7]. The agency structure hypothesis suggests that NREPs and PHEPs should regulate environmental protection less stringently than state EPAs, so state EPAs form the omitted reference category.
A variety of control variables are also employed in order to capture the effects of internal state factors that may influence regulatory stringency. These variables are similar to those employed in other studies of environmental enforcement in the U.S. states [
1,
10,
22,
23,
24,
25] and, broadly speaking, capture factures related to politics, economics, the severity of environmental problems, and the degree of state agency regulatory authority. Unless otherwise noted in the text below, these control variables come from the Correlates of State Policy database housed at Michigan State University [
30].
A political factor that is often considered to be an important driver of regulatory behavior is the ideological preferences of elected political officials. An updated version of Berry et al.’s dynamic measure of government ideology is employed to capture these preferences, with the expectation that more liberal political officials will press for more aggressive regulatory enforcement [
31].
Another important aspect of a state’s political environment is its interest group context. Manufacturing employment per capita is included in the models as a proxy measure of the manufacturing industry’s economic and political clout, which may lead to less stringent environmental regulation. Similarly, annual per capita Sierra Club Membership is included as an indicator of organized environmental group influence, which may lead to more stringent environmental regulation. The Sierra Club provided the latter data.
Wealthier states may be more likely to prefer and be able to afford more expansive environmental protections. State wealth is measured as gross state product per capita. State economic conditions may also affect state officials’ political pressures, with states in poorer economic climates facing greater political pressure to reduce regulatory stringency. The state’s lagged unemployment rate is therefore included to assess the impact of the state’s economic conditions.
Environmental regulation may be more stringent in situations where pollutant emissions may have more serious human consequences, as where a greater number of people will be exposed. At the same time, effective regulation may be more challenging in larger states. The models therefore include variables measuring area and population density. States with more serious environmental problems may also be more likely to produce more stringent environmental standards and more aggressive enforcement. To capture the severity of the problem, this study includes the logged toxic emissions in the state, which come from the U.S. Environmental Protection Agency’s Toxic Release Inventory website.
Finally, although the large majority of states had primacy over the relevant federal programs over this time period, in a small handful of states, the U.S. EPA retained primary implementation authority. The analyses of these programs thus include a dummy variable coded 1, if the state does not have primacy over the relevant base program (New Source Performance Standards for the CAA, National Pollutant Discharge Elimination System for the CWA, and the base RCRA program) in a given year. These data were provided by the U.S. Environmental Protection Agency. For the analysis of environmental compliance costs, the three separate variables are summed to create a composite measure ranging from 0–3. Further descriptions and summary statistics for the independent and dependent variables are presented in
Table 1.
3.4. Model Estimation
These models are estimated using a set of panel data models. Care must be taken in handling potential violations of ordinary least squares assumptions, as pooled models are likely to exhibit the estimation problems characteristic of both time serial and cross-sectional design. Specifically, OLS regression is often inappropriate since two assumptions regarding the error term of the model—homoskedasticity and lack of autocorrelation—are likely to be violated [
32].
Because the core independent variables are weakly time-invariant, a fixed effects specification is not appropriate. The models therefore employ a Prais–Winston generalized least squares procedure that corrects for autocorrelation. The models also employ panel-corrected standard errors to deal with heteroskedasticity across panels.
The dataset contains observations for the 48 contiguous American states. The models are estimated over the time period 1988–1994, a period chosen due to data constraints. While this period was some years ago, it has several advantages for this study. First, it gives us the advantage of assessing the bureaucratic structure hypothesis at a substantially different time period than the more recent period covered in Hopper’s work discussed above. Second, some states had decentralized environmental regulatory structures over most of this time period (though they had disappeared by the end of it), allowing one to estimate the effects of bureaucratic structure across a type of agency that has not previously received much scholarly attention. Finally, several states made changes in their agency structure during this period, providing some degree of temporal variation in the independent variables of interest.
Figure 1 presents the distribution of environmental agency structures in the American states during the period covered by the analyses. As indicated in the figure, nine states converted from one type of bureaucratic structure to another during this period. Perhaps most strikingly, decentralized agency structures, which existed in five states in 1988, had disappeared from the American environmental regulatory landscape by 1994. The last three states to have such structures, Oklahoma, Texas, and Virginia, ended them in 1993, with Oklahoma shifting to an EPA structure while Texas and Virginia opted for a NREP.
4. Results
The results of the analyses are presented in
Table 2. The first two columns present the results for Clean Air Act inspections and enforcement activities, respectively. The results of enforcement actions are striking. Both NREPs and PHEPs produce significantly lower enforcement rates, ceteris paribus, than state EPAs, a result fully consistent with the bureaucratic structure hypothesis. States that utilize a decentralized system of boards or commissions evidence lower enforcement rates as well, and this is the only structural variable that is likewise significant for inspections.
The enforcement results are consistent with Hopper’s [
1,
5,
10] results supportive of the bureaucratic structure hypothesis, which are similarly based on analyses of enforcement activity within the clean air program. Thus, they provide empirical validation of the bureaucratic structure hypothesis for that program over a substantially different period. Moreover, the results for decentralized agency structures suggest that they too, may lead to weaker environmental protection activity.
The next two columns present the results of inspections and enforcement actions under the Clean Water Act. Here, PHEP’s also evidencelower enforcement rates than state EPAs. In this policy arena, however, NREPs do not appear to regulate in a systematically different manner. Thus, the bureaucratic structure hypothesis receives some support for water pollution enforcement actions, though it is not as robust as for air pollution enforcement. As with the Clean Air Act results, states with decentralized systems show significantly weaker inspection and enforcement activity than state EPAs. For inspections, decentralized system is the only structural variable that is statistically significant, though it is worth noting that the coefficient on NREPs is negatively signed and narrowly misses significance at conventional levels (p = 0.12).
The third set of columns provides the results for the RCRA program, which are notably different than those for the previous programs. In this case, the models indicate that both NREPs and PHEPs have higher inspection rates than mini-EPAs, a result that is the opposite of what the regulatory structure hypothesis would suggest. Moreover, PHEP agencies have significantly higher enforcement rates as well. In this program area, the regulatory structure hypothesis does not fare nearly as well. Nor do decentralized program structures appear to make an appreciable difference in either RCRA inspection or enforcement rates.
The final column presents the results for environmental compliance costs, a broad measure that captures all three of the above programs along with other environmental protection initiatives. In this case, too, the results fail to comport neatly with the expectations generated by the bureaucratic structure hypothesis. Controlling for the other variables in the model, the results indicate that, on average, PHEP agencies impose higher environmental compliance costs on industry than state EPAs, the opposite of what the bureaucratic structure hypothesis would suggest. Decentralized agencies, on the other hand impose significantly lower compliance costs, while NREPs show no significant effect.
The results for the control variables indicate that wealthier states tend to average higher levels of environmental regulatory stringency, ceteris paribus, while in larger states, stringency is often lower. These results are consistent with expectations. States with greater problem severity, as measured by TRI emissions, often also have higher levels of stringency, as expected, though for clean air enforcement rates, the effect is reversed. Somewhat puzzlingly, states with a higher percentage of Sierra Club members tend to evidence lower inspection and enforcement rates.
Lack of state primacy is a significant predictor almost across the board, and it is consistently associated with lower inspection and enforcement rates but higher compliance costs. This pattern is not unexpected, as states that have not been delegated primary enforcement authority leave most enforcement to the U.S. EPA; in some cases, the state agency makes no inspections and/or takes no enforcement actions in a given year. Unlike state-level inspections and enforcement actions, however, compliance costs are affected by both state and federal actions, and higher costs in states where the U.S. EPA takes the primary enforcement role are consistent with the common perception that the EPA is generally a more aggressive regulator than state agencies [
4].
5. Conclusions
Do agencies that combine environmental protection with other functions like public health or natural resource management regulate pollution emissions less stringently than those that focus exclusively on environmental protection? The results of these analyses suggest that the answer is complicated. On the one hand, the findings of CAA and CWA enforcement actions are broadly consistent with that hypothesis. On the other hand, the findings for inspections, RCRA enforcement, and environmental compliance costs are generally inconsistent with that hypothesis, and in some cases the results of these analyses produce significant effects in the opposite direction from that hypothesized. While this produces a somewhat murky overall picture, these analyses suggest at least four broad takeaway points.
First, the organizational design of a state environmental agency does appear to impact its regulatory behavior. Across a variety of indicators of regulatory stringency, variables representing agency design features were significant in these analyses. In fact, at least one of these three variables was significant in all seven of the empirical models. In this broad sense, agency design theorists do appear to have isolated an important determinant of agency regulatory activity, though the effects of structure may be more nuanced and context-dependent than articulations of this theory commonly suggest.
Second, the degree of centralization of environmental regulatory functions appears to impact regulatory activity in addition to the choice of a single purpose or multitask agency. Although little theory exists about the impact of agency decentralization on the outputs of environmental agencies, the variable representing decentralized environmental agency structures produced some of the most consistently significant effects across the different models in this study. One needs to be careful in interpreting this result, as it is based on a fairly small number of observations in this sample. Nonetheless, these findings indicate that the impact of centralization of environmental functions is a topic worthy of systematic theorizing and further empirical investigation.
Third, these results provide no indication that the bureaucratic structure hypothesis holds for regulatory inspections. Neither of the variables for multitask agency structure (NREP or PHEP) were significant in any of the inspection models, with the exception of one case in which the effect was opposite what the hypothesis would lead us to expect. This is curious because inspections are a similar regulatory activity to enforcement actions and are generally responsive to the same set of factors.
Finally, the impact of organizational design features does not appear to be consistent across programmatic areas. Though the effect of the structure appears to hold fairly strongly for both air and water pollution enforcement actions in a way that generally supports the bureaucratic structure hypotheses, analyses of both RCRA enforcement and a general measure of environmental costs produce effects that are not supportive of the hypothesis at all.
These results are generally in accordance with Hopper’s [
1,
10] findings supportive of the bureaucratic structure hypothesis, at least with respect to the specific program (CAA) and regulatory activity (enforcement actions) that she studies, though over a substantially earlier time period. This bolsters the argument that agency design produces stable, enduring impacts on regulatory outputs. Moreover, the results are, broadly speaking, similar to those of prior studies of the CWA, with scholars finding that bureaucratic structure has significant effects on both point-source enforcement [
18] and nonpoint-source water pollution policy activism [
17] but not on inspections [
14]. However, the dramatically different results for other activities and programs illustrate the importance of theory testing across multiple contexts, and raise questions about the scope of the bureaucratic structure hypothesis. Though the theory, as it has so far been articulated, has general applicability, these results suggest the possibility that the impact of agency structure on regulatory activity may be mitigated either by substantive aspects of particular policy areas or features of programmatic design.
These possibilities provide fertile ground for future research. Such research could address current gaps in our understanding of this phenomenon by examining additional regulatory activities (such as standards setting or permitting), testing the bureaucratic structure hypothesis in non-U.S. contexts (or cross-nationally), or assessing the implications of agency design for environmental outcomes (such as pollution emissions), rather than the bureaucratic outputs that researchers have thus far investigated. The latter would help answer a crucial remaining question that has yet to be addressed: What difference does the design of implementing agencies ultimately have for pollution levels? Pollution control is a critical component of environmental sustainability [
33,
34], and research into this question may make significant contributions to our understanding of how sustainability may be achieved.