corporate environmentalism and environmental statutory ... - CiteSeerX

CORPORATE ENVIRONMENTALISM AND ENVIRONMENTAL STATUTORY PERMITTING* CHRISTOPHER S. DECKER University of Nebraska at Omaha

Abstract Studies have shown that despite infrequent inspections and low penalties for statutory violations, a large fraction of firms comply with environmental restrictions. What then motivates compliance? I investigate this question by focusing on the length of time it takes environmental agencies to process and issue new source construction permits pursuant to Clean Air Act regulations and new industrial discharge permits pursuant to Clean Water Act regulations. I find that plants (or firms) with fewer instances of noncompliance receive permits for major projects more quickly. In addition, I find that permit delays are sensitive to economic conditions as well, such as local area unemployment. As far as voluntary pollution control behavior is concerned, I find that regulators that issue permits for plant modifications focus primarily on statutory compliance, but when permitting new plant construction, where there is no plant compliance history to go on, voluntary pollutant releases do matter.

I.

Introduction

The motivations for compliance with environmental law continue to puzzle researchers. Nearly all reliable data on environmental compliance indicate that inspections are far too infrequent and penalties for noncompliance too small to be considered a reasonable deterrent. For instance, Clifford Russell describes a Resources for the Future survey that found that large air pollution sources were visited by state agencies’ enforcement personnel an average of once every 8 months, and large water pollution sources were visited only once every 5 months.1 Since operating permits generally set discharge limits on a per-hour or per-day basis, such inspection activity seems rather infrequent. Moreover, the same survey revealed that penalties per discovered * I would like to thank Tom Lyon, John Maxwell, Sam Peltzman, an anonymous referee, and participants at Indiana University’s Business Economics and Public Policy seminar series for their helpful comments. Any errors remaining herein are my own. 1 Clifford S. Russell, Monitoring and Enforcement, in Public Policies for Environmental Protection 243 (Paul R. Portney ed. 1990). [Journal of Law and Economics, vol. XLVI (April 2003)] 䉷 2003 by The University of Chicago. All rights reserved. 0022-2186/2003/4601-0005$01.50

103

104

the journal of law and economics

violation ranged anywhere from zero to $2,500.2 More recently, Kelly Lear found that in 1995, the average penalty for noncompliance with Clean Air Act regulations was $10,181, with fewer than 200 violating firms actually being fined.3 Yet despite lax enforcement, compliance among regulated firms remains high. As Wesley Magat and W. Kip Viscusi report, the average compliance rate in the U.S. pulp and paper industry between 1982 and 1985 was on the order of 75 percent.4 Winston Harrington cites data from the U.S. Environmental Protection Agency (EPA) that put compliance rates for some industrial sectors at 90 percent or higher.5 However, with such weak enforcement, one must question what motivates compliance with environmental regulation. Often referred to as the “Harrington paradox,” a growing literature has emerged that tries to explain this phenomenon.6 Harrington and other researchers adapt models from the income tax literature to explain high compliance rates.7 In these models, firms are placed in groups on the basis of their previous compliance records. Firms with poor compliance records are placed in a group subject to some combination of frequent inspection, tough regulatory and permitting standards, and maximal fines for discovered violations. Firms with good compliance records, by contrast, are placed in a second group subject to less intensive scrutiny and lower penalties for noncompliance. These models then show that under such a scenario, firms can be induced to comply even though penalties are rarely levied. Other studies have looked for costs other than statutory penalties as a motivation for firms’ developing good environmental records. For example, in an investigation of the relationship between hazardous-waste lawsuits and stockholder returns, Michael Muoghalu, H. David Robison, and John L. Glascock use event study techniques to determine that a firm’s stockholders 2 Winston Harrington, Enforcement Leverage When Penalties Are Restricted, 37 J. Pub. Econ. 29 (1988), provides similar evidence of weak enforcement of environmental law. For instance, over the period 1978–83, the state of Connecticut issued 800 notices of violations to sources found to be in violation of their water discharge permits. Only 21 of those cases involved a penalty, and the average fine was only $363. 3 Kelly K. Lear, Environmental Regulation: A Theoretical and Empirical Analysis of Enforcement and Compliance (unpublished Ph.D. diss., Indiana Univ., Kelley Sch. Bus. 1998). 4 Wesley A. Magat & W. Kip Viscusi, Effectiveness of the EPA’s Regulatory Enforcement: The Case of Industrial Effluent Standards, 33 J. Law & Econ. 331 (1990). 5 Harrington, supra note 2. 6 See Mark A. Cohen, Monitoring and Enforcement of Environmental Policy, in 4 Int’l Y.B. Envtl. & Resource Econ. (Henk Folmer & Thomas H. Tietenberg eds. 1999), for a comprehensive survey of this literature; Anthony Heyes, Implementing Environmental Regulation: Enforcement and Compliance, 7 J. Reg. Econ. 107 (2000). 7 Harrington, supra note 2; Clifford S. Russell, Winston Harrington, & William J. Vaughan, Enforcing Pollution Control Laws (1986); see Russell supra note 1; Jon D. Harford & Winston Harrington, A Reconsideration of Enforcement Leverage When Penalties Are Restricted, 45 J. Pub. Econ. 391 (1991); Jon D. Harford, Measurement Error and State-Dependent Pollution Control Enforcement, 21 J. Envtl. Econ. & Mgmt. 67 (1991).

environmental permits

105

suffer a 1.2 percent loss in market value (about $33 million) when a suit is filed against a firm for violation of solid-waste-management laws.8 Returning to regulatory behavior, there is reason to believe that other benefits to maintaining a good statutory compliance record exist. For instance, companies that wish to build new plants or modify existing plants that emit pollutants into the air or water are required to obtain special permits from the relevant state environmental protection agency. In fact, permitting is an integral part of the major environmental statutes in the United States, and most of the time enforcement actions are brought against plants for failure to meet the guidelines specifically stipulated in their permits. Since permits are largely issued on a plant-by-plant basis, regulators are in a position to utilize a plant’s compliance history when determining whether a permit should be granted. Conversely then, firms may be in a position to influence the behavior of these permitting authorities. Indeed, the transaction costs associated with permitting delays, complications, repeated information requests, and appearances before zoning boards can be quite high. James Boyd, Alan Krupnick, and Janice Mazurek assert that the costs associated with production delays can be exacerbated by longer permitting times.9 Focusing particular attention on the high-technology, microprocessor industry, they state that “permit modification and review processes can impose delays for weeks, months, or even years. In the presence of competition, delays threaten to erode slim technological and marketing leads.”10 Moreover, “some managers posit that production delays cost Intel a million dollars in lost revenue each day.”11 Kelly Robinson makes a similar point, stating that “production delays due to extended permit preparation and review . . . can interrupt revenue streams and extend financing costs and may cause the firm to incur performance penalties or miss short-lived strategic opportunities.”12 Finally, Geoffrey Keogh and Alan Evans cite evidence that tying up land in the development process costs the U.K. building industry as much as £35.5 million (about $50–$55 million) per week.13 Some case study evidence even suggests that a good environmental record can reduce these costs. Marie Christel Cothran cites several case studies in which corporations were able to gain building permits from local government authorities in record time, and receive variances on existing regulations more 8 Michael I. Muoghalu, H. David Robison, & John L. Glascock, Hazardous Waste Lawsuits, Stockholder Returns, and Deterrence, 57 S. Econ. J. 357 (1990). 9 James Boyd, Alan J. Krupnick, & Janice Mazurek, Intel’s XL Permit: A Framework for Evaluation (Discussion Paper No. 98-11, Resources for the Future 1998). 10 Id. at 3. 11 Id. at 1. 12 Kelly Robinson, One-Stop Permitting? A Critical Examination of State Environmental Permit Assistance Programs, 13 Econ. Dev. Q. 245, 246 (1999). 13 Geoffrey Keogh & Alan W. Evans, The Private and Social Costs of Planning Delay, 29 Urb. Stud. 687 (1992).

106


easily, by promoting environmental activities, undertaking voluntary environmental investments, and achieving high compliance rates.14 For instance, in 1991, Vulcan Materials, Inc., sought a construction permit to open several additional quarries in Virginia. Quarries are required by Virginia state law to maintain a buffer zone—or wildlife preserve—around each quarry. The company voluntarily extended its buffer zones, created water holes, and planted shrubs and other vegetation to further support area wildlife. In response, it found faster permitting times and easier acquisition of construction permits. As Cothran writes, “Permitting boards view corporations with a history of environmental consciousness more favorably.”15 From a firm’s perspective, however, whether this regulatory response is beneficial depends on the cost of developing a good environmental record versus the cost savings associated with quicker permitting. For firms building new facilities, for instance, these cost savings will depend on three factors: the number of new facilities the firm expects to build, the costs of regulatory delay, and the amount of time saved (measured here in days) from avoiding environmental noncompliance at the firm’s existing plants. Specifically, if (number of new plants # cost of regulatory delay # days saved)

(1)

1 compliance costs,

then the firm would opt to maintain compliance in order to avoid costly delays. The question is whether (1) holds. Clearly, permitting plays a prominent role in environmental law and may have significant implications for environmental compliance. To my knowledge, however, no formal studies exist that investigate the environmental permitting process. I attempt to fill this void by addressing the following questions. First, does a history of good environmental stewardship hasten permit issuance?16 Second, if so, by how much? Finally, is there reason to believe that the incentives to avoid permit delays offer a possible solution to the Harrington paradox? My results indicate that plants (or firms) with fewer instances of noncompliance indeed receive their permits more quickly, particularly for larger projects. Turning attention to the other control variables, I also find that 14 Marie Christel Cothran, Proactive Environmental Activity Eases Permitting Process, 2 J. Envtl. Permitting 293 (1993). 15 Id. at 293. Other evidence suggests that positive environmental profiles facilitate permitting. The New England Office of the Environmental Protection Agency’s StarTrack Program offers an example. This purely voluntary program asks participating firms to improve their internal compliance auditing practices. In return, participants can expect certain regulatory benefits. Prominent among these benefits is that participants can expect “express lane” service for permits and other regulatory actions (U.S. Environmental Protection Agency, New England StarTrack Program Draft: 1997/98). 16 According to state environmental authorities, permits are rarely rejected, and if they are, the permit record is simply discarded.


107

permits are issued more quickly to those plants that are located in counties in which the unemployment rate is relatively high and in those states that tend to be more politically conservative. As far as voluntary pollution control behavior is concerned, my results indicate that regulators that issue permits for plant modifications focus primarily on that plant’s historical statutory compliance. However, when permitting for new facilities, where there is no compliance history to go on, voluntary environmental behavior also matters. The paper is organized as follows. In Section II, I will highlight certain elements of environmental statutory permitting. Section III includes the basic model to be estimated and the data used in the analysis. In Section IV, I discuss the econometric methodology used in this paper, and in Section V, I present the results. In Section VI, I implement the cost-benefit analysis implied by equation (1) using some of my estimates, and in Section VII, I conclude by suggesting some useful research extensions. II.

A Brief Discussion of Environmental Permitting

In this study, I have chosen to concentrate principally on New Source Review (NSR) permits, which fall under the New Source Control Program pursuant to the Clean Air Act (CAA), and National Pollutant Discharge Elimination System (NPDES) permits, which are required under the Clean Water Act (CWA). Under these statutes, it is primarily the responsibility of the states to assess and issue these permits.17 Moreover, states also have primary authority over the monitoring and enforcement of granted permits. In this section, I will discuss certain elements of both the NSR and NPDES permitting processes. A.

New Source Review Permitting

The 1970 and 1977 amendments to the CAA established and developed an implementation strategy for new and modified sources of air pollution called New Source Performance Standards (NSPS). States implement these standards through a program called New Source Review, under which new and modified air pollution sources are subject to preconstruction review and permitting. A project is tagged as either “major” or “minor” depending on certain characteristics such as each source’s potential for harming the environment (see Table 1 for details). Permit applications are submitted directly to the appropriate state environmental management agency and include a detailed description of the new or modified facility’s design and construction schedule as well as a description 17

Lynn M. Gallagher, Clean Water Act, in Environmental Law Handbook 109 (Thomas F. P. Sullivan & R. Craig Anderson eds., 14th ed. 1997); F. William Brownell, Clean Air Act, in Environmental Law Handbook (Thomas F. P. Sullivan & R. Craig Anderson eds., 13th ed. 1995). Both discuss in detail permitting and enforcement under the Clean Water Act and Clean Air Act, respectively.

108

the journal of law and economics TABLE 1 Permit Characteristics

Permit Coverage

Deemed major if

Public notification period Other characteristics

New Source Review Permits

Industrial Discharge Permits

Sulfur dioxide, nitrogen dioxide, ozone, volatile organic compounds, and particulate matter

Many substances such as heavy metals, dredged soils, solid waste, incinerator residue, sewage, garbage, munitions, chemical wastes, biological materials, radioactive materials, heat, discarded equipment, and other industrial, municipal and agricultural wastes a Attainment: (1) source has the potential “Large” dischargers (determined to emit over 250 tons per year of a within the individual permit) for regulated pollutant or (2) source is one facilities that discharge any of the U.S. EPA’s 28 listed source amount of heavy metals into categories and has the potential to bodies of water (such as steel emit over 100 tons per year of a mills, metal finishing and b regulated pollutant electroplating plants, and oil Nonattainment: source has the potential refineries) to emit between 10 and 100 tons per year of a regulated pollutant 30 day minimum 30 day minimum Issued on a project-by-project basis Issues on a plant-by-plant basis for 5 years

a A region (primarily county delineated) is classified as in “attainment” by the U.S. Environmental Protection Agency (EPA) if the region is meeting nationally set ambient air quality standards. Otherwise, the region is deemed to be in “nonattainment.” b These 28 source categories include chemical plants, coal-cleaning plants, kraft pulp mills, Portland Cement plants, steel mills, copper smelters, and petroleum refineries.

of the pollution control technology to be implemented. In addition, primarily for major projects, the application must provide information on the new source’s potential impact on air visibility, soils, and vegetation in the area. Note that NSR permits are issued on a project-by-project basis. Hence, when a firm wishes to expand or modify an existing plant, an NSR permit is required. Depending on the frequency of modifications, a single plant can have several different NSR permits over the course of its operating life. Once the application is received, the state environmental agency has 30 days to request from the applicant any additional information deemed necessary to make an informed decision regarding permit issuance. There is no stated limit on the number of information requests an agency can make, and particularly for major projects, repeated information requests are often made. After the agency is satisfied with the application’s information content, the agency must make a preliminary determination on permit issuance, give public notice of its decision, and provide an opportunity for comment and public hearings before the permit is issued.

environmental permits B.

109


The NPDES is the principal means through which CWA provisions are implemented. It is a permitting program that requires any source that discharges a designated CWA pollutant into the nation’s lakes, rivers, and other navigable waterways to obtain a discharge permit from the relevant state environmental agency. All state permits are subject to EPA review. Unlike the NSR permits, each NPDES permit is issued on a plant-by-plant basis and is active for 5 years. These permits cover a wide variety of plants, from publicly owned waste treatment plants to privately owned industrial plants. The focus of this study is on permits granted to industrial entities often called Industrial Discharge Permits (IDPs). Given the broad definition of “pollutants” subject to CWA regulation, the permit application itself requires an extensive amount of information about the facility, the nature of the pollution discharges, and the proposed pollution control technology to be adopted. As with NSR permits, IDP-permitted facilities are classified as either major or minor dischargers (see Table 1). Once an application has been submitted, discussions with the applicant are initiated by the state agency to address detailed features of the individual permit. This can be the most time-consuming element of the permitting process since the information demands for each permit are immense. Once a permit document is completed, the agency must post a “notice of issuance” to allow for public response before a final permit can be issued. III.

The Model and Permit Data

For NSR permits and IDPs issued between 1990 and 1998, I have collected data on the proposed location of the new or existing plant, the Standard Industrial Classification, the dates when the application was received and issued, and whether the project is major or minor. To these data I have linked plant-level information on toxic chemical releases, noncompliance history, county demographic and economic information, and some state-level data, discussed in detail below. Since state environmental agencies are responsible for permit issuance (as well as enforcement of permit restrictions), permit data are often available only by directly contacting each state’s environmental protection office. I have chosen to focus my study on six midwestern states comprising the EPA’s Region V: Michigan, Illinois, Indiana, Ohio, Minnesota, and Wisconsin. Moreover, I have chosen to focus only on permits for industrial sources because compliance and toxic chemical release data are reliably available on a plant-by-plant basis for these sectors.


110

TABLE 2 Variable Definitions and Their Expected Effect of Permit Delays

Variable DURATION MAJOR ENFLAST3 TRI1 CHTRI PART3350 RELUNEMP FIRMREV GREENRAT TOURINC POPDEN

EAIRQUAL EWATERQUAL CONSERV

Definition

Expected Effect

Length of time (days) from when the application is received to when the permit is issued Dummy variable that equals one if the project (plant) is categorized as “major” Number of enforcement actions taken against a plant in the 3 years prior to the application date Level of TRI releases (in 1,000 of tons) at a plant 1 year prior to the application date (that is, if the application was made in period t, then the TRI releases are as of t ⫺ 1) Change in TRI releases (in 1,000s of tons) 1 year prior to the application date Dummy variable that equals one if the plant is owned by a company that participated in EPA’s 33/50 Program County percentage unemployed where the plant is located minus the state percentage unemployed where the plant is located Total revenues ($ millions) for the firm seeking an NSR permit or IDP for its plant as of the year when the permit was issued The number of Sierra Club and Natural Resource Defense Council members per every 10,000 state residents Tourist revenue generated at a state’s parks and other state-funded recreation areas by tourist activity per $100,000 of state personal income (1996) Population (measured in 1,000s) within a 3-mile radius of the plant requesting a permit (for the IDPs, POPDEN is the population (measured in 1,000s) divided by county land area where the new plant will be located) State expenditures on air quality in fiscal year 1994 per $1,000 of state personal income (1996) State expenditures on water quality in fiscal year 1994 per $1,000 of state personal income (1996) Percentage of state voters who voted Republican in the most recent presidential election preceding the permit request date (includes Reform party voters)

⫹ ⫹ ⫹ ⫹ ⫺ ⫺ ⫺ ⫹ ⫹ ⫹

⫺ ⫺ ⫺

Note.—TRI: Toxic Release Inventory; EPA: Environmental Protection Agency; NSR: New Source Review; IDP: Industrial Discharge Permit.

The general model is DUR p f (MAJOR, ENFLAST3, TRI1, CHTRI, PART3350, RELUNEMP, GREENRAT, TOURINC, FIRMREV,

(2)

POPDEN, ExQUAL, MAJOR # GREENRAT), where DUR measures the time (in days) between NSR (IDP) permit application and permit approval. Table 2 provides a list of variable definitions as well as my prior expectation of each variable’s expected impact on the duration of the permitting process. Table 3 provides summary statistics.

TABLE 3 Descriptive Statistics New Source Review Permits Total N DURATION: Mean Median Standard deviation ENFLAST3: Mean Median Standard deviation TRI1: Mean Median Standard deviation CHTRI: Mean Median Standard deviation PART3350: Mean Median Standard deviation RELUNEMP: Mean Median Standard deviation GREENRAT: Mean Median Standard deviation TOURINC: Mean Median Standard deviation POPDEN: Mean Median Standard deviation ExQUAL (x p AIR, WATER): Mean Median Standard deviation CONSERV: Mean Median Standard deviation FIRMREV: Mean Median Standard deviation

Major

Minor

Total

174.09 106 208.77

305.34 227 269.83

94.82 73 96.73

648.35 275.5 754.01

914.04 546 875.77

428.02 190 550.45

.97 0 2.24

1.32 0 2.81

.76 0 1.79

3.55 0 5.57

5.86 4.5 6.56

1.64 0 3.66

.82 .06 4.99

1.34 .08 7.88

.5 .06 1.53

6.67 .70 14.94

13.25 3.79 20.07

1.21 0 3.26

.54 ⫺.00003 8.36

.12 .00022 .19

68

Minor

221

.07 ⫺.00002 .06

150

Major

587

.34 ⫺.00003 6.60

366


.14 .00838 .21

82

.1 .00007 .17

.75 1 .43

.77 1 .42

.74 1 .44

.81 1 .39

.75 1 .44

.87 1 .34

⫺.88 ⫺.85 1.48

⫺1.04 ⫺1.03 1.46

⫺.78 ⫺.74 1.49

⫺.71 ⫺.64 1.71

⫺1.06 ⫺.78 1.71

⫺.42 ⫺.17 1.67

18.15 17.96 4.68

18.18 17.96 4.92

18.13 17.96 4.53

17.76 16.66 5.73

18.91 16.66 4.86

16.81 16.66 5.62

9.2 7.95 5.73

9.39 7.95 5.74

9.10 7.95 5.74

6.81 7.54 3.18

6.07 3.22 3.3

6.27 7.54 3.09

36.56 14.65 53.09

37.69 13.24 61.7

35.88 15.35 47.26

47.66 .17 197.69

65.12 .55 234.41

33.18 .05 16.12

8.83 7.59 3.7

9.11 8.23 3.82

8.67 7.42 3.62

1.4 .85 .93

1.81 .85 1.17

1.07 .85 .47

56.3 55.7 4.54

56.66 55.7 4.55

56.03 55.7 4.52

55.74 55.7 4.93

55.53 55.70 4.69

55.91 55.71 5.14

19.3 2.5 45.8

19.26 3.05 44.98

19.32 2.2 46.35

9.95 3.3 22.88

14.70 4.7 32.55

6.01 2.4 7.20

112

the journal of law and economics A.

Explanatory Variables

The variable MAJOR is a dummy variable equal to one if the construction project is major and zero if minor. The data presented in Table 3 strongly suggest that major projects take much longer to permit than minor projects. Major projects are likely to have a greater impact on the environment, and therefore permit authorization for such projects may take longer. For the NSR data, POPDEN measures the number of people living within a 3-mile radius of the permit-seeking plant. These data were provided by the EPA’s Office of Environmental Compliance and Assurance (OECA). However, this variable was not available for IDPs. As an alternative, for the IDP data set, POPDEN is the population of the county in which the new facility is to be located divided by the land area of that county.18 I expect permitting authorities to view projects in more densely populated regions to be a greater overall health risk and hence to direct more attention to those permit requests. The variables ENFLAST3, TRI1, CHTRI, and PART3350 relate specifically to issues concerning voluntary compliance. As a measure of the plant’s (firm’s) compliance history, ENFLAST3, again available from OECA, indicates the number of enforcement actions taken against the plant (firm) for CAA, CWA, and Resource Conservation and Recovery Act statutory violations over the 3-year period prior to the permit issue date (see Section IIIB for further details regarding this variable). This is the key variable for testing the effect of compliance on permitting. If compliance facilitates permitting, then fewer historical enforcement actions should reduce permitting times. The variable TRI1, available from the EPA, measures the level (in thousands of tons) of chemicals listed in the Toxic Release Inventory (TRI) released by a plant (or group of plants owned by a firm) roughly 1 year prior to a firm’s permit application date.19 The TRI data are often used as a measure of voluntary environmental control.20 By law, facilities are required to report annually the total amount of certain chemicals (over 600 currently) both released on-site and transferred off-site. The data are self-reported and are not an indicator of a plant’s compliance with environmental law. Most of the chemicals reported under the TRI are unregulated, meaning that the actual 18 County population data come from the U.S. Department of Commerce, Bureau of Economic Information, Bureau of Economic Analysis, Regional Regional Accounts Data: Local Area Personal Income (http://www.bea.doc.gov/bea/regional/reis/). County land area data were obtained from the U.S. Bureau of the Census, County and City Data Book: 1994 (1994). 19 The TRI program was established as part of the U.S. EPA’s Emergency Planning and Community Right to Know Act. The reason for the 1-year lag is that the TRI data are released with roughly a 1-year lag. 20 John W. Maxwell, Thomas P. Lyon, & Steven C. Hackett, Self-Regulation and Social Welfare: The Political Economy of Corporate Environmentalism, 43 J. Law & Econ. 583 (2000); Madhu Khanna & Lisa A. Damon, EPA’s Voluntary 33/50 Program: Impact on Toxic Releases and Economic Performance of Firms, 37 J. Envtl. Econ. & Mgmt. 1 (1999); Shameek Konar & Mark A. Cohen, Information as Regulation: The Effect of Community Right to Know Laws on Toxic Emissions, 32 J. Envtl. Econ. & Mgmt. 109 (1997).


113

level of release is not subject to any federal or state regulatory statute. According to Cothran, we should expect shorter permitting times to be associated with lower TRI releases. In addition to looking at release levels, regulators may respond favorably to TRI reductions (CHTRI). A final indicator of environmental stewardship is whether the firm participated in the EPA’s 33/50 Program. Briefly, this was a purely voluntary toxic pollution control program in which participants were charged with reducing their levels of 17 particularly toxic TRI chemicals by 33 percent by 1992 and 50 percent by 1995, relative to their 1988 levels.21 The variable PART3350 is a dummy variable that equals one if the plant seeking an NSR permit or IDP is owned by a company that was a 33/50 Program participant and zero otherwise. The variables RELUNEMP and FIRMREV are other control variables that may influence permitting. The variable RELUNEMP measures the percentage not employed in the county where the permit-seeking plant is located, minus the state unemployment rate during the year when the permit was applied for. Permitting authorities may be under pressure from state legislators, state economic development councils, large companies operating in the state, and even the general voting public to use permitting to facilitate regional economic development. Hence, the estimated effect on permit duration should be negative. The variable FIRMREV measures, in millions of dollars, the total revenues for the firm seeking an NSR permit or IDP for its plant as of the year in which the permit was issued.22 Regulators may favor high-revenuegenerating companies since these companies tend to build larger operations and employ more of the local labor force. I also control for interest-group effects on permitting by including the following variables. The variable GREENRAT measures membership rates (per 10,000 state residents) in two prominent environmental groups, the Sierra Club and the Natural Resource Defense Council.23 Higher environmental, or “green group,” membership rates mean that permitting authorities may be pressured to scrutinize permits more closely. The variable TOURINC measures the revenue generated at a state’s parks and other state-funded recreation 21

Khanna & Damon, supra note 20, provides a detailed discussion of this program. These data were taken from various issues of Dun and Bradstreet, Million Dollar Directory (published annually). 23 Membership rates were supplied by the Sierra Club and Natural Resource Defense Council. The latest year made available was 1992. State population data come from U.S. Department of Commerce, supra note 18. Since the membership data are at the state level, we cannot determine whether some residents are members of both environmental groups. Regardless, it is unclear whether this double counting represents a problem. If a resident is willing to pay membership dues to two different environmental groups, then it seems reasonable to assume that such individuals place more value on the environment than members of a single group and are likely to support additional pressure on regulatory authorities. 22


114

areas by tourist activity per $1,000 of state personal income.24 States that have a high level of tourism may desire fewer polluting plants. Therefore, more pressure would likely be placed on permitting authorities. The variable CONSERV represents the percentage of state voters that voted Republican in the most recent presidential election preceding a firm’s permit request date.25 A higher percentage is taken to indicate a more “pro-business” political climate. If regulators are responsive to majority opinion, then we should see a quicker turnaround in permitting the higher the percentage of votes for the Republican candidate. The next variable included in the model is ExQUAL, where x p AIR for the NSR equation and WATER for the IDP equation. These data come from the Council for State Governments and capture state expenditures on air and water quality (see Table 2). I expect states with larger budgets for air (water) quality to issue permits more quickly. Finally, I include an interaction variable, MAJOR # GREENRAT, to test whether green groups pressure regulators to scrutinize larger projects more thoroughly than smaller ones. B.

Characteristics of the Permit and Enforcement Data

There are some salient features of the NSR permit and IDP data that are worth discussing before proceeding with the econometric analysis. As stated in Table 1, NSR construction permits are issued on a project-by-project basis. Since most of the permits listed in the data are for construction modification rather than new facility construction, I will focus only on those NSR permits issued to existing sources for plant construction modification. Therefore, for NSRs, the variable ENFLAST3 measures the number of enforcement actions logged against the plant seeking an NSR permit. Industrial Discharge Permits are issued on a plant-by-plant basis and are active for 5 years, at the end of which time the plant must reapply to be reissued a new IDP. Since reapplication is largely a matter of procedure, of primary concern to firms is obtaining operating permits for new facilities.26 Therefore, initial permit delays are likely to be of significant interest. Maintaining a good compliance record at the permit-seeking company’s other plants may be beneficial in facilitating future permitting. 24 Tourist data come from the National Association of State Park Directors and are published in the U.S. Bureau of the Census, Statistical Abstract of the United States: 1993 (113th ed. 1993); the state personal income data come from U.S. Department of Commerce, supra note 18. 25 U.S. Bureau of the Census, supra note 24; U.S. Bureau of the Census, Statistical Abstract of the United States: 1997 (117th ed. 1997). Note that CONSERV includes Reform party voters in the 1992 and 1996 elections. 26 Failure to reapply within 5 years usually results in immediate revocation of the permit and possibly other enforcement actions. However, if reapplication is made within the 5-year horizon and the regulatory authority fails to reissue a new permit before the existing permit expires, then the plant is granted an administrative extension.


115

Since these IDP observations are for new facilities, plant-level historical chemical release and compliance data are not available. Therefore, for IDPs, ENFLAST3 measures the compliance record of all the other plants located in the same state and owned by the same company.27 Similarly, the toxic chemical release levels for these permits are the release levels of all the other plants located in the same state and owned by the same company. As Table 3 illustrates, IDPs take much longer to be approved than NSR permits. There are at least two likely reasons for this. First, IDPs are permits issued to new facilities, whereas the NSR permits examined here are plant modification permits. Having existing information on the plant may help regulators issue NSR permits more quickly. Second, the sheer number of items considered pollutants under the CWA can make IDPs quite complex. Therefore, more time may be needed to ensure that all relevant aspects of the proposed plant’s operation are covered. IV.

Econometric Methodology

This paper examines the conditional probability that a plant will be issued either an NSR permit or an IDP in t days given that it was not issued in t ⫺ 1 days. There are essentially two states of the world in this analysis: the initial, or “permit not issued,” state and the subsequent “permit issued” state. If I find that, say, fewer historical enforcement actions taken against a plant increase the probability of exiting the permit-not-issued state, I can say that the plant’s good historical compliance behavior is effective at speeding up, or facilitating, permit issuance. The econometric methodology employed here closely follows that of William Greene and Nicholas Kiefer.28 The function of interest in duration analysis is the hazard function, defined as l(t) p

f (t) , 1 ⫺ F(t)

(3)

where F(t) p Pr (T ! t) is the cumulative density function and f (t) p dF(t)/dt is the corresponding distribution function. Equation (3) is interpreted as a conditional probability. Depending on the characteristics of the density 27 Differences in ENFLAST3 between the two permit types afford an opportunity to see how important other firm and geographic characteristics are to regulators when facility compliance information is not available. As discussed in detail below, my results suggest that plant compliance is of primary interest to regulators when granting authorization to construct. Other firm-level information seems to be of secondary concern. However, at the suggestion of a referee, I did estimate the NSR equation using an ENFLAST3 measure consistent with the IDP equations. As with the original variable definition, this variable was statistically significant, although the effect on permit duration was lower and the precision of the estimated effect was smaller (that is, the resulting p-values were slightly higher). 28 William H. Greene, Econometric Analysis 725, 726 (3d ed. 1997); Nicholas M. Kiefer, Economic Duration Data and Hazard Functions, 26 J. Econ. Literature 646 (1988).

116


function, t can affect the hazard function in various ways. If at some point t ∗, dl(t)/dt 1 0, then the hazard function exhibits positive duration dependence. For instance, if this were true for my permit data, then the longer the permit application remains outstanding, the greater the probability that the permit will be issued shortly. Conversely, if at t ∗, dl(t)/dt ! 0, then the hazard function exhibits negative duration dependence. If this were the case for my permitting data, then the longer the permit application remains outstanding, the less likely the permit will be issued shortly. In order to proceed with full parametric maximum-likelihood estimation, a distribution is required.29 It is often possible to construct a sample estimate ˆ , particularly in the absence of data censoring.30 for the hazard function, l(t) ˆ Since my NSR and IDP samples are not censored, I can construct l(t) in the following way. For both the NSR and IDP samples, I first order the duration events (of sample size n) from smallest to largest, t1 ! t 2 ! t 3 ! … ! tk. The number of completed durations k is usually less than n because often two or more observations have the same duration.31 Define hj to be the number of completed spells of duration tj, for j p 1, … , k. Finally, define nj to be the number of spells not completed before tj:

冘 k

nj p

i≥ j

hi .

(4)

ˆ is Therefore, a convenient sample estimator for l(t) ˆ p hj . l(t) nj

(5)

Thus the estimated sample hazard is the number of event completions at duration tj divided by the number of “survivors” at tj. Figures 1 and 2 plot ˆ against duration length for both the NSR and IDP data. By inspection, l(t) one can see that in both cases, the estimated sample hazard functions appear to be monotonically increasing, a characteristic of a Weibull distribution (commonly used in duration studies) that exhibits a positive duration de-

29 Greene, supra note 28. It is possible to conduct a semiparametric duration analysis in which no distributional assumption is necessary, but interpreting estimated coefficients can be extremely difficult. 30 Censoring of the dependent variable is pervasive in duration studies simply because, over the sample period of interest, some spells may have not been completed yet. Within my data sets, censoring has not proved to be an issue for two reasons. First, the average length of a permitting spell was roughly 1 to 2 years, and there were very few permit applications for industrial sources in 1999 and 2000. Of these, however, none was a TRI reporting facility. Since TRI is an integral part of the analysis, I was left with a data set of completed spells. 31 For instance, in my NSR data set, n p 587 but k p 301.


117

Figure 1.—Sample hazard, New Source Review

pendence.32 The corresponding hazard function for the Weibull distribution can be expressed as l(t) p lr t 1⫺rr,

(6) 33

where t is the duration of a spell and r 1 0 is a shape parameter. The estimation procedure is conducted as follows. The distribution function for the Weibull distribution is f (t) p lr t 1⫺rr exp [⫺(lt) r].

(7)

The explanatory variables enter f (t) through l. Given its nonnegativity requirement, this can be defined for each observation i as l i p exp (xib),

(8)

where, for each permit observation, xi is a 1 # z row vector of the z explanatory variables and b is a z # 1 column vector of coefficients. We can represent the log-likelihood function for the Weibull distribution as

冘 n

L(xi , b, r, t) p

ip1

冘 n

ln f (wi ) p

[ln (ti ) ⫹ ln (r) ⫹ wi ⫺ exp (wi )],

(9)

ip1

32 While I use the Weibull distribution in what follows, I also estimated the model using the log logistic distribution (which exhibits both positive and negative duration dependence) and Cox’s proportional hazard specification. While quantitative comparisons are difficult, the results using these models are qualitatively similar to those obtained using the Weibull distribution. 33 Note that r is a parameter to be estimated. Since my data suggest positive duration dependence, my estimated r should be greater than one.

118


Figure 2.—Sample hazard, Industrial Discharge Permits

where wi p r[xib ⫹ ln (ti )]. Equation (9) is maximized with respect to b and r. The resulting sign of the estimated coefficient indicates the effect it has on the hazard function for each variable. For instance, a positive estimated coefficient indicates that an increase in a particular variable increases the probability of the permit being issued shortly, given that it has not been issued previously, ceteris paribus. Determining the magnitude a particular variable, xi, has on the hazard function involves both bl and r. Given (6) and (8), it can easily be shown that 1 ⭸l(ti ) ⭸l(ti )/l(ti ) p p bl r. l(ti ) ⭸xl ⭸xl

(10)

The term bl r is thus a semielasticity indicating that a unit change in variable xl will generate a percent change in the hazard function of bl r. The expected duration of a spell (of more interest to this study) is simply equal to the inverse of the hazard rate.34 Thus, the magnitude of a change in a variable xl on the expected duration, E(ti ), is a semielasticity indicating that a unit change in xl will generate a percentage change in E(ti ): ⭸E(ti )/E(ti ) p ⫺bl r. ⭸xl

(11)

If the estimated bl r is positive, an increase in variable xl will shorten the 34 Paul D. Allison, Event History Analysis: Regression for Longitudinal Event Data (1984); Louis W. Nadeau, EPA Effectiveness at Reducing the Duration of Plant-Level Noncompliance, 34 J. Envtl. Econ. & Mgmt. 54 (1997).


119

expected length of the spell. Conversely, if bl ! 0, then an increase in xl will lengthen the spell. While the nature of data used in this study implies the use of duration analysis, it is nonetheless possible to estimate effects using standard multivariate regression analysis.35 Therefore, as part of my results, I also include ordinary least squares (OLS) estimation results where I model ln (DUR) as a function of the explanatory variables in equation (2). While it should be noted that the two procedures are fundamentally different, the majority of those variables that prove significant in the Weibull maximum-likelihood estimation also prove significant using OLS. V.

Estimation Results

Tables 4 and 5 present the estimation results for NSR permits and IDPs, respectively. For each permit type, I estimate two different models. The first model, labeled Weibull (1) and OLS (1), includes those variables listed in equation (2). The second, Weibull (2) and OLS (2), includes the interaction term, ENFLAST3 # MAJOR, to address whether historical noncompliance matters more for major projects seeking authorization. Moreover, since MAJOR appears to have a substantial effect on the length of the permitting process, Tables 6 and 7 present separate estimation results for major and minor projects. Most of what follows addresses those results presented in Tables 4 and 5 since the likelihood ratio statistics for the full-sample Weibull models (particularly Weibull (2)) suggest that they provide the best overall “explanation” of these duration events. The first column underneath Weibull (1) and Weibull (2) of each table shows the estimated bl coefficient for the hazard function, and the second column shows the semielasticity, ⫺bl r, that measures the impact each variable has on the expected length of permitting times. Note that the value of the shape parameter, r, is shown to be greater than one for both NSR permit and IDP duration equations, conforming to expectations that the data exhibit positive duration dependence. For NSR permits, whether a construction modification is a major project (MAJOR) has considerable influence on permitting times. It can take two to three times as long to permit a major construction modification project than a minor one. This conforms to expectation. Major projects are likely to have a greater impact on the environment, and therefore such permit applications receive greater scrutiny. Moreover, the evidence suggests that for major projects, a plant’s historical noncompliance proves to have positive and significant effect on the duration of the permitting process. While ENFLAST3 is significant in Weibull (1) 35 Paul Joskow, Contract Duration and Relationship-Specific Investments: Empirical Evidence from Coal Markets, 77 Am. Econ. Rev. 168 (1987); Mary K. Olson, Firm Characteristics and the Speed of FDA Approval, 6 J. Econ. & Mgmt. Strategy 377 (1997).


120

TABLE 4 Estimation Results: New Source Review Permits Weibull (1) bl CONSTANT MAJOR ENFLAST3

⫺11.4** (0) ⫺1.097** (.003) ⫺.045** (.007)

OLS (1)

OLS (2)

⫺bl r

bl

Weibull (2) ⫺bl r

bl

bl

22.518

⫺11.272** (0) ⫺.852* (.025) .021 (.373) ⫺.116** (.002) ⫺.002 (.803) .001 (.903) .076 (.5) .084* (.012) ⫺.031 (.112) ⫺.051** (.001) ⫺.0018* (.032) ⫺.001 (.977) .038** (.005) .0005 (.619) ⫺.038⫹ (.061) 2 332.97

22.526

4.917** (0) .844** (0) .022** (.007)

4.902** (0) .774** (0) ⫺.0002 (.989) .094* (.019) .003 (.476) .001 (.784) .011 (.828) ⫺.031* (.046) .019⫹ (.071) .023** (.002) .0005 (.197) ⫺.005 (.637) ⫺.01⫹ (.063) ⫺.00001 (.979) ⫺.003 (.736)

2.166 .089

MAJOR#ENFLAST3 TRI1 CHTRI PART3350 RELUNEMP GREENRAT TOURINC POPDEN EAIRQUAL CONSERV FIRMREV MAJOR#GREENRAT r Likelihood ratio statistic Adjusted R2

⫺.009 (.417) .0000004 (.946) .078 (.488) .081* (.015) ⫺.034⫹ (.083) ⫺.05** (.001) ⫺.0019* (.023) ⫺.001 (.956) .044** (.001) .0006 (.541) ⫺.029 (.146) 2 323.54

.018 ⫺.001 ⫺.154 ⫺.16 .068 .099 .0038 .002 ⫺.087 ⫺.0012 .057

1.702 ⫺.042 .232 .004 ⫺.002 ⫺.151 ⫺.167 .063 .102 .0036 .001 ⫺.076 ⫺.001 .075

.004 (.371) .001 (.752) .007 (.895) ⫺.033* (.039) .02⫹ (.053) .024** (.002) .0005 (.210) ⫺.005 (.64) ⫺.011⫹ (.063) ⫺.00005 (.918) ⫺.004 (.639) .41

.42

Note.—Values in parentheses are p-values. ⫹ Significant at the 10% level. * Significant at the 5% level. ** Significant at the 1% level.

and OLS (1), the results for Weibull (2), OLS (2), and those presented in Table 6 strongly imply that noncompliance matters only when considering permits for major projects. According to Weibull (2) in Table 4, one additional violation increases the time it takes to receive a permit by roughly 23 percent. For the average (median) major NSR permit, this translates into an increase of about 52 days. Several other socioeconomic variables prove significant as well. Referring to Weibull (1) and (2) in Table 4, construction projects proposed in more densely populated regions of a state (POPDEN) tend to be permitted less


121

TABLE 5 Estimation Results: Industrial Discharge Permits Weibull (1)

CONSTANT MAJOR ENFLAST3

OLS (1)

OLS (2)

bl

⫺bl r

bl

⫺bl r

bl

bl

⫺14.619** (0) ⫺1.494* (.041) ⫺.035* (.044)

29.509

⫺14.28** (0) ⫺1.447* (.049) .024 (.469) ⫺.074* (.049) ⫺.023** (.003) .139 (.805) .417⫹ (.082) .221** (0) ⫺.034 (.211) ⫺.014 (.744) .0002 (.594) .005 (.973) .050⫹ (.084) ⫺.0024 (.586) .024 (.554) 2.1 159.6

29.149

7.154** (0) .741* (.04) .015⫹ (.074)

6.774** (0) .659⫹ (.066) ⫺.016 (.344) .039* (.037) .015 (.001) .193 (.499) ⫺.225⫹ (.07) ⫺.084** (.007) .023 (.122) .018 (.402) .00001 (.878) ⫺.141 (.11) ⫺.026⫹ (.063) .0032 (.149) ⫺.006 (.737)

3.017 .071

MAJOR#ENFLAST3 TRI1 CHTRI PART3350 RELUNEMP GREENRAT TOURINC POPDEN EWATERQUAL CONSERV FIRMREV MAJOR#GREENRAT r Likelihood ratio statistic Adjusted R2

⫺.024** (.002) .166 (.757) .443⫹ (.065) .215** (0) ⫺.042 (.126) ⫺.0004 (.991) .0003 (.538) .041 (.796) .059* (.037) ⫺.0024 (.583) .018 (.663) 2 156.03

.048 ⫺.335 ⫺.894 ⫺.434 .084 .001 ⫺.0006 ⫺.083 ⫺.12 .0049 ⫺.036

Weibull (2)

2.954 ⫺.049 .151 .047 ⫺.283 ⫺.851 ⫺.451 .07 .028 ⫺.0005 ⫺.011 ⫺.102 .0048 ⫺.049

.016** (.001) .195 (.501) ⫺.226⫹ (.73) ⫺.08⫹ (.1) .026⫹ (.76) .007 (.717) .00001 (.855) ⫺.171⫹ (.052) ⫺.033* (.016) .0032 (.156) ⫺.005 (.784) .55

.56


quickly than projects proposed in less densely populated regions.36 The variable RELUNEMP also proves to have a statistically significant negative effect on permit length times. Projects proposed in counties where the unemployment rate is higher than the state’s overall unemployment rate tend to get permitted more quickly, irrespective of project size. While the effect is small, this result suggests that environmental regulatory authorities are 36 Although insignificant when using ordinary least squares, the estimated effect remains positive.

122

the journal of law and economics TABLE 6

Estimation Results: Major versus Minor Projects, New Source Review Permits Major Weibull

CONSTANT ENFLAST3 TRI1 CHTRI PART3350 RELUNEMP GREENRAT TOURINC POPDEN EAIRQUAL CONSERV FIRMREV r Likelihood ratio statistic Adjusted R2

Minor OLS

Weibull

OLS

bl

⫺bl r

bl

bl

⫺bl r

bl

⫺17.747** (0) ⫺.057** (.005) ⫺.004 (.671) 1.77 (.158) ⫺.196 (.292) .15** (.009) .056⫹ (.086) .039 (.141) ⫺.002 (.153) ⫺.145** (0) .103** (0) .0005 (.764) 2.1 78.46

36.768

7.373** (0) .036 (.001) .003 (.48) ⫺.672 (.319) .096 (.306) ⫺.073* (.014) ⫺.033⫹ (.053) ⫺.015 (.312) .0009 (.188) .044** (.01) ⫺.028 (.018) ⫺.0006 (.461)

⫺9.259** (0) .034 (.22) ⫺.015 (.703) .003 (.708) .069 (.632) .02 (.636) ⫺.107** (.0001) ⫺.12** (.0002) .0004 (.764) .097** (.0001) .006 (.73) .0001 (.985) 2.2 45.3

20.031

3.788** (0) ⫺.007 (.627) .003 (.873) .00001 (.976) ⫺.01 (.869) ⫺.005 (.759) .049** (0) .046** (0) ⫺.0002 (.713) ⫺.035** (.002) .002 (.761) .0004 (.452)

.118 .008 ⫺3.667 .406 ⫺.31 ⫺.116 ⫺.081 .0041 .3 ⫺.213 ⫺.0011

.19

⫺.074 .032 ⫺.005 ⫺.149 ⫺.043 .232 .26 ⫺.0008 ⫺.21 ⫺.012 ⫺.0001

.1

Note.—Values in parentheses are p-values. ⫹ Significant at the 10% level. ** Significant at the 5% level. ** Significant at the 1% level.

sensitive to local area economic conditions and may be tempted (or perhaps pressured) to foster economic growth in relatively depressed regions by facilitating permitting. Both GREENRAT and TOURINC have a positive and statistically significant effect on permitting times as well. Thus, both the interests of green groups and those economic sectors dependent on the natural environment do influence permitting decisions. Finally, the results generally support the notion that environmental permits are approved more quickly in more politically conservative states. Toxic Release Inventory chemical releases (TRI1) and the change in TRI releases (CHTRI) are insignificant. Moreover, whether the plant’s parent company was a member of the EPA’s 33/50 Program (PART3350) is insig-


123

TABLE 7 Estimation Results: Major versus Minor Projects, Industrial Discharge Permits Major Weibull

CONSTANT ENFLAST3 TRI1 CHTRI PART3350 RELUNEMP GREENRAT TOURINC POPDEN EWATERQUAL CONSERV FIRMREV r Likelihood ratio statistic Adjusted R2

Minor OLS

Weibull

OLS

bl

⫺bl r

bl

bl

⫺bl r

bl

⫺18.353** (0) ⫺.084** (0) ⫺.025** (.003) 1.565⫹ (.066) .349 (.31) .304** (.001) ⫺.036 (.378) ⫺.001 (.984) .0007 (.204) .302 (.187) .07 (.167) ⫺.0038 (.399) 2.3 66.47

41.784

9.072** (0) .035** (.001) .013** (.007) ⫺.49 (.205) ⫺.179 (.301) ⫺.127** (.006) .01 (.586) .009 (.794) ⫺.0003 (.388) ⫺.2 (.119) ⫺.05⫹ (.07) .0025 (.266)

⫺13.076 (0) .053 (.181) ⫺.12* (.016) ⫺.527 (.57) 1.105** (.005) .143⫹ (.064) ⫺.001 (.978) ⫺.027 (.741) .0002 (.782) ⫺.651 (.144) .02 (.219) ⫺.0223 (.219) 2.1 31.5

27.575

6.651 (0) ⫺.027 (.128) .06** (.002) 1.007* (.024) ⫺.397* (.035) ⫺.051 (.206) .029 (.165) ⫺.013 (.634) .000001 (.985) ⫺.062 (.769) ⫺.024 (.155) .013 (.166)

.19 .056 ⫺3.563 ⫺.795 ⫺.692 .082 .003 ⫺.0016 ⫺.687 ⫺.159 .0086

.49

⫺.111 .252 1.112 ⫺2.331 ⫺.301 .003 .058 ⫺.0005 1.373 ⫺.042 .047

.23


nificant as well. Hence, there seems to be little evidence in the data that voluntary activities by the plant (or firm) have any impact on environmental regulators when it comes to construction modification permit issuance. Permitting authorities seem more inclined to condition their decisions on a plant’s historical compliance record rather than nonstatutory indicators of environmental stewardship. The results for IDPs are similar to those for NSR permits, but there are some differences as well. If the proposed new facility is a major discharger, then receiving an IDP will take about three times longer than it would for a minor discharger. Consistent with the NSR results, for major dischargers, ENFLAST3 is shown to have a (statistically significant) positive effect on the length permitting times. Using results from Weibull (2) in Table 5, one


124

additional violation at any one of the firm’s plants will increase subsequent permitting times by 15 percent. For the average (median) major IDP, this translates into an increase of about 81 days. Again, consistent with both intuition and the NSR results, RELUNEMP’s effect on the permitting times for both major and minor dischargers is negative and significant and the effect is much larger. Moreover, in all four models, CONSERV has a negative and significant effect on permit times. Contrary to expectation and the NSR results, TOURINC and GREENRAT are insignificant, as are CHTRI, EWATERQUAL, and FIRMREV.37 Unlike the NSR equation, TRI1 and PART3350 are significant for IDPs. Greater TRI releases lengthen permitting times, while participation in the 33/50 Program substantially reduces permitting times. These inconsistencies pose an interesting question. Why should a firm’s voluntary environmental actions facilitate new plant permitting and not existing plant permitting? The answer seems to be that a plant’s historical compliance is the best indicator of its willingness and ability to meet any new permit requirements. In the absence of such information (since new plants have no historical compliance data), regulators may need to broaden their scope and consider other firmlevel environmental information when assessing the likely compliance behavior of a new plant. Perhaps the most important finding here is that in both the NSR and IDP cases, when it comes to major construction projects, historical enforcement activity is shown to have a crucial impact on regulators. If we think of voluntary compliance as compliance efforts undertaken by firms in order to reap benefits other than penalty avoidance, then the results presented here may offer at least a partial solution to the Harrington paradox discussed in Section I. Firms may be recognizing that good compliance records can reduce the regulatory red tape associated with environmental permitting. VI.

Benefits and Costs of Increased Compliance

If firms comply with environmental regulations to reap benefits other than penalty avoidance, the question remains whether it is worth it to firms to comply in order to avoid permit delays of the magnitudes estimated here. The answer ultimately depends on the firm’s marginal costs of compliance relative to the cost savings associated with a shorter permit delay. While a detailed benefit-cost analysis is beyond the scope of this study, it is possible to obtain an admittedly rough estimate of the cost savings associated with reduced permitting times from increased compliance, relative to compliance costs, by utilizing equation (1). As an illustrative example, I focus on new plant construction—the IDP results. This calculation is performed over a 3-year period since my regression 37

Other measures of firm size, such as firm-level employment, prove insignificant as well.


125

results show that the number of enforcement actions over a 3-year period has a significant effect on permit times. As a measure of compliance costs, I will utilize the U.S. Census Bureau’s Pollution Abatement Cost and Expenditure data, last published for the year 1994.38 In that year, the U.S. census reported that U.S. manufacturers spent $18.772 billion in pollution abatement operating expenditures. According to the U.S. Census Bureau’s 1992 Census of Manufactures, the total number of manufacturing establishments was 370,912.39 Dividing this figure by expenditures provides a rough compliance cost estimate of about $50,476 per plant. I estimate the number of new plants a firm is likely to build per year in the following manner. Utilizing data from the U.S. Small Business Administration, for 1999 (the latest year of available data), the ratio of new to existing manufacturing establishments was .10.40 Using the TRI database, I found that in 1995, the average number of plants owned by companies that produce in the six states in my sample was approximately 3.6. By multiplying .10 by 3.6, we obtain a rough measure of the average number of new plants per year a firm will likely build.41 Recalling that one fewer environmental violation in a 3-year period can reduce IDP issuance by 81 days, by multiplying .36 # 81 # (cost of regulatory delay) # 3, we can arrive at an estimate of the cost savings realized from avoiding one environmental violation. Unfortunately, the cost of regulatory delay is not easily estimated. However, utilizing the above estimates and solving for the cost of regulatory delay, equation (1) implies that if the daily cost of regulatory delay is greater than $577, then it is beneficial for the average firm to incur compliance costs to avoid permit delays.42 To my knowledge, no systematic study has been undertaken to measure regulatory delay costs. However, as pointed out by Boyd, Krupnick, and Mazurek, permit delays cost Intel an estimated $1 million per day in lost sales revenue.43 Moreover, Keogh and Evans estimated that delay costs can run between $10 and $15 million per day for the U.K. building sector.44 According to the U.K.’s Department of Trade and Industry’s Construction Statistics Annual, 2001 Edition, there were 7,043 38 U.S. Bureau of the Census, Pollution Abatement Cost and Expenditures, 1992 (Current Industrial Reports MA200(94)-1, 1996). 39 U.S. Bureau of the Census, Economics and Statistics Administration, 1992 Census of Manufactures (MC92-S-1, October 1996). 40 These data are available directly from the Small Business Administration and can be queried at http://www.sba.gov/ADVO/stats. 41 Hence, the average firm that operates plants in the Midwest will build a new plant every 2–3 years. Given the rate of economic growth experienced over the latter part of the 1990s, this number seems reasonable. 42 That is, 577 p $50,476/(.36 # 81 # 3). 43 Boyd, Krupnick, & Mazurek, supra note 9. 44 Keogh & Evans, supra note 13.

126


building and civil engineering contractors operating in 1995.45 This implies that the per-contractor cost of delay is roughly between $1,419.85 and $2,129.77. Note, however, that this figure represents only the delay cost of stalled development and does not include the cost of delayed revenue generation from operating a new plant. If this cost were included, total delay costs would be quite high.46 While the above calculation is admittedly rough, these results suggest not only that firms are in a position to directly influence regulatory behavior through their environmental compliance activities but also that their incentives to comply with existing statutes can be substantial. Hence, my findings may offer at least one additional answer to the Harrington paradox. Perhaps part of the reason is to avoid costly regulatory delays in permitting. VII.

Conclusion

In this paper, I have investigated the relationship between environmental compliance, voluntary corporate environmental activities, and the environmental permitting process. I find that, controlling for other variables, better compliance records can shorten the length of time it takes for a firm to receive construction or pollution discharge permits for major projects. This result may partially resolve the counterintuitive result that compliance rates tend to be high even when inspections and penalties are low. Firms may be investing in compliance not simply to avoid regulatory sanctions, but rather to receive positive regulatory benefits such as less bureaucratic red tape during the permitting process. In addition, I find that projects in regions with higher unemployment rates tend to get permitted more quickly and that states that are more politically conservative permit construction projects more quickly. Finally, as far as voluntary environmental activity is concerned, there does seem to be some evidence that TRI releases and participation in voluntary environmental programs facilitate permitting, but only when plant-level compliance information is not available. Ultimately then, permit authorities seem to be principally concerned with compliance rather than with nonstatutory signals of environmental stewardship. Yet when no compliance data are available, they are indeed likely to consider voluntary environmental behavior when deliberating over environmental permits. This paper offers several opportunities for research extensions. Certainly, a broader geographic focus would be beneficial in verifying the results presented here. Moreover, investigating permitting data other than environmental 45 U.K. Department of Trade and Industry, Construction Statistics Annual, 2001 Edition (2001). In this publication, the number of “general contractors” is substantially higher. However, most contractors work primarily on residential construction projects. I elected to focus on the building contractors subcategory in an effort to capture those contractors who are mostly engaged in industrial and commercial construction projects. 46 I am indebted to the referee for clarification of this point.


127

permits, such as building permits issued by municipal governments, might prove illuminating. More important, however, this paper highlights the need for additional research into the cost of regulatory delay. More refined numerical estimates of these costs to both firms and regulatory agencies are necessary to better quantify the relative benefits of increased compliance and reductions in regulatory delay. Obtaining better cost estimates would best be achieved through an extensive survey of manufacturing facilities. However, designing an appropriate survey would require a great deal of time and care. I leave these suggestions for future research. Bibliography Allison, Paul D. Event History Analysis: Regression for Longitudinal Event Data. Beverly Hills, Cal.: Sage Publications, 1984. Boyd, James; Krupnick, Alan J.; and Mazurek, Janice. “Intel’s XL Permit: A Framework for Evaluation.” Discussion Paper No. 98-11. Washington, D.C.: Resources for the Future, 1998. Brownell, F. William. “Clean Air Act.” In Environmental Law Handbook, 13th ed., edited by Thomas F. P. Sullivan and R. Craig Anderson. Rockville, Md.: Government Institutes, Inc., 1995. Cohen, Mark A. “Monitoring and Enforcement of Environmental Policy.” In Vol. 3 of International Yearbook of Environmental and Resource Economics, edited by Henk Folmer and Thomas H. Tietenberg. Lyme, N.H.: Edward Elgar, 1998. Cothran, Marie Christel. “Pro-active Environmental Activity Eases Permitting Process.” Journal of Environmental Permitting 2 (1993): 293–300. Deily, Mary E., and Gray, Wayne B. “Enforcement of Pollution Regulations in a Declining Industry.” Journal of Environmental Economics and Management 21 (1991): 260–74. Downing, Paul B., and Kimball, James N. “Enforcing Pollution Control Laws in the United States.” Policy Studies Journal 11 (1982): 55–65. Dun and Bradstreet. Million Dollar Directory. New York: Dun and Bradstreet, annual issues. Environmental Protection Agency. Office of Enforcement and Compliance Assurance. Audit Policy: Incentives for Self-Policing. Washington, D.C.: Environmental Protection Agency, March 5, 1998. Environmental Protection Agency. Office of Environmental Stewardship, Region 1, New England. StarTrack Fact Sheet: Better Environmental Performance through Environmental Management Systems and Third Party Certification. Washington, D.C.: Environmental Protection Agency, December 2, 1998. Gallagher, Lynn M. “Clean Water Act.” In Environmental Law Handbook, 14th ed., edited by Thomas F. P. Sullivan and R. Craig Anderson, pp. 109–60. Rockville, Md.: Government Institutes, Inc., 1997.

128


Gray, Wayne B., and Deily, Mary E. “Compliance and Enforcement: Air Pollution Regulation in the U.S. Steel Industry.” Journal of Environmental Economics and Management 31 (1996): 96–111. Greene, William H. Econometric Analysis. 3d ed. New York: Prentice-Hall, 1997. Harford, Jon D. “Measurement Error and State-Dependent Pollution Control Enforcement.” Journal of Environmental Economics and Management 21 (1991): 133–38. Harford, Jon D., and Harrington, Winston. “A Reconsideration of Enforcement Leverage When Penalties Are Restricted.” Journal of Public Economics 45 (1991): 391–95. Harrington, Winston. “Enforcement Leverage When Penalties Are Restricted.” Journal of Public Economics 37 (1988): 29–53. Helland, Eric. “The Enforcement of Pollution Control Laws: Inspections, Violations, and Self-Reporting.” Review of Economics and Statistics 80 (1998): 141–53. Hemphill, Thomas A. “Corporate Environmentalism and Self-Regulation: Keeping Enforcement Agencies at Bay.” Journal of Environmental Regulation 3 (Winter 1993/94): 145–54. Heyes, Anthony. “Implementing Environmental Regulation: Enforcement and Compliance.” Journal of Regulatory Economics 17 (2000): 107–29. Joskow, Paul L. “Contract Duration and Relationship-Specific Investments: Empirical Evidence from Coal Markets.” American Economic Review 77 (1987): 168–85. Keogh, Geoffrey, and Evans, Alan W. “The Private and Social Costs of Planning Delay.” Urban Studies 29 (1992): 687–99. Khanna, Madhu, and Damon, Lisa A. “EPA’s Voluntary 33/50 Program: Impact on Toxic Releases and Economic Performance of Firms.” Journal of Environmental Economics and Management 37 (1999): 1–25. Kiefer, Nicholas M. “Economic Duration Data and Hazard Functions.” Journal of Economic Literature 26 (1988): 646–79. Konar, Shameek, and Cohen, Mark A. “Information as Regulation: The Effect of Community Right to Know Laws on Toxic Emissions.” Journal of Environmental Economics and Management 32 (1997): 109–24. Lancaster, Tony. The Econometric Analysis of Transition Data. Cambridge: Cambridge University Press, 1990. Lear, Kelly K. “Environmental Regulation: A Theoretical and Empirical Analysis of Enforcement and Compliance.” Unpublished Ph.D. dissertation. Bloomington: Indiana University, Kelley School of Business, 1998. Magat, Wesley A., and Viscusi, W. Kip. “Effectiveness of the EPA’s Regulatory Enforcement: The Case of Industrial Effluent Standards.” Journal of Law and Economics 33 (1990): 331–60. Maxwell, John W.; Lyon, Thomas P.; and Hackett, Steven C. “Self-Regulation


129

and Social Welfare: The Political Economy of Corporate Environmentalism.” Journal of Law and Economics 43 (2000): 583–619. Muoghalu, Michael I.; Robison, H. David; and Glascock, John L. “Hazardous Waste Lawsuits, Stockholder Returns, and Deterrence.” Southern Economic Journal 57 (1990): 357–70. Nadeau, Louis W. “EPA Effectiveness at Reducing the Duration of PlantLevel Noncompliance.” Journal of Environmental Economics and Management 34 (1997): 54–78. Olson, Mary K. “Firm Characteristics and the Speed of FDA Approval.” Journal of Economics and Management Strategy 6 (1997): 377–401. Robinson, Kelly. “One-Stop Permitting? A Critical Examination of State Environmental Permit Assistance Programs.” Economic Development Quarterly 13 (1999): 245–68. Russell, Clifford S. “Monitoring and Enforcement.” In Public Policies for Environmental Protection, edited by Paul R. Portney, pp. 243–74. Washington, D.C.: Resources for the Future, 1990. Russell, Clifford S.; Harrington, Winston; and Vaughan, William J. Enforcing Pollution Control Laws. Washington, D.C.: Resources for the Future, 1986. U.K. Department of Trade and Industry. Construction Statistics Annual, 2001 Edition. London: Queen’s Printer and Controller of Her Majesty’s Stationary Office, 2001. U.S. Bureau of the Census. County and City Data Book: 1994. Washington, D.C.: U.S. Government Printing Office, 1994. U.S. Bureau of the Census. “Pollution Abatement Cost and Expenditures, 1992.” Current Industrial Reports MA200(94)-1. Washington, D.C.: U.S. Government Printing Office, 1996. U.S. Bureau of the Census. Economics and Statistics Administration. 1992 Census of Manufactures. Washington, D.C.: U.S. Government Printing Office, October 1996. U.S. Department of Commerce. Bureau of the Census. Statistical Abstract of the United States: 1993. 113th ed. Washington, D.C.: U.S. Government Printing Office, 1993. U.S. Department of Commerce. Bureau of the Census. Statistical Abstract of the United States: 1997. 117th ed. Washington, D.C.: U.S. Government Printing Office, 1997. Weil, David. “If OSHA Is So Bad, Why Is Compliance So Good?” Rand Journal of Economics 27 (1996): 618–40.