Environmental protection in the federalist system: the political economy of NPDES inspections.

Helland, Eric

I. INTRODUCTION

Federal regulatory programs, from occupational safety to protection of the environment, are implemented jointly by national and state agencies. More broadly, almost all regulatory programs involve some delegation to branch offices (meat inspections) or regional districts (tax collection). Given the pervasiveness of regulatory federalism, surprisingly little research has been done on the political consequences of delegating national policy to local actors. In particular, the small but growing literature on the politics of regulation has almost completely ignored the role of federalism in implementing regulatory policy.(1)

Perhaps the most widely accepted model of bureaucratic decision making, Peltzman's [1976] model of support maximization implies that delegation can have a profound impact on the implementation of regulatory policy. Peltzman's model posits that regulators maximize support while minimizing opposition. In the context of environmental protection, a rational regulator would calculate the marginal support from pollution abatement generated by inspecting a given plant and equate that to the marginal opposition, arising from job loss, generated by that action. Pashigian [1985] points out that in environmental legislation local and national interests are unlikely to be identical because industry, the primary target of pollution control measures, has an uneven geographic distribution. Thus a local regulator, performing Peltzman's calculation, is quite likely to reach a very different decision than a national regulator.

This paper examines the extent to which local regulators are able to respond to local interests when enforcement of a national policy has been delegated to them. There are three possibilities. One is that delegation of enforcement to local government represents a complete abdication of control by the federal government. At the other extreme is the possibility the federal government can delegate with no loss of control because of its ability to enforce its goals through the use of threats (loss of authority over the program) and incentives (additional grant money). The third possibility is that both local and national interests can influence a delegated program. The federal government will lose some control when it delegates, but the loss will not be total.

The Clean Water Act of 1972 is one of the largest federal regulatory programs ever delegated to the states. Under the Act, the federal government took over the previously state level responsibility for regulating water pollution.(2) Under the Clean Water Act, industrial polluters are issued permits, which specify types and amounts of pollutants allowed in effluent discharged during a given time period. These limits are set by state or regional authorities but must conform to minimum national standards. Permit writers may consider the desires of the local community if those desires demand more stringent standards than the federal minimums, but may not respond to community demands for lower standards. Local communities with considerable employment in the industry, for example, might want a lower standard than the federal mandate. Although a politically-minded regulator does not have the ability to actually grant a permit with a lower standard, a second source of discretion is available in the form of selective enforcement.

While permits are difficult enough to issue, they are meaningless without some method of verifying compliance.(3) The state or regional Environmental Protection Agency (EPA) conducts regular on-site inspection of polluters. Although the EPA requests that delegated states inspect polluters at least once a year, it is not required. In practice, states have considerable discretion on how and when to inspect polluters. This should not come as a surprise. There are 7,500 NPDES facilities in the country. Variation in state performance could be due to location alone?

This study, using data from the EPA's Permit Compliance Database, which tracks compliance with the Clean Water Act, analyses plant level inspection, violation and effluent discharge data. The analysis focuses on the pulp and paper industry, which is the largest industrial polluter of the nation's waterways and therefore most affected by the provisions of the Clean Water Act. The results of the study support the theory that interest groups do have an impact on state-level regulatory bureaucracies. Agencies make marginal adjustments in order to satisfy changing interests of both national and local actors.

II. SPECIFICATION

The three possible outcomes of delegation offer differing predictions on the sources of agency discretion. To address this question, three empirical specifications of the determinants of inspections, violations and effluent discharge are estimated. The importance of inspections has been noted above. The determinants of the probability a firm is in violation of its NPDES permit and the average pollution discharge at the mill are estimated to address the question of whether or not inspections are important in deterring violations and reducing pollution. It is possible that a decline in pollution or a compliant mill causes a decline in the probability of inspection, not political concerns; that is, the level of pollution and the probability of a violation at the mill also determine the probability of inspection. If this is the case, then estimating these three equations independently will yield inconsistent estimates of the parameters. For this reason the three equations are estimated via a two-stage instrumental variables approach.

The analysis is further complicated by the fact that the data on violations do not represent a random sample. The fundamental problem is that there is no verification of compliance without an inspection. Mills listed as "in compliance" may actually be "in violation" because they were fortunate enough to not have been detected.(5) This problem is compounded by the fact that mills can report a violation on their quarterly noncompliance reports even if there is no inspection.

In the case of violations, an assumption that mill operators are rational can reduce the partial observability problem. A rational owner or manager will not voluntarily report a violation unless one actually exists. Thus only one subcase suffers from partial observability: the uninspected firms who do not report a violation. A log-likelihood function that corrects for this partial observability is given by,

[Mathematical Expression Omitted]

where [Phi] denotes the normal cdf, [y.sub.1i] denotes the violation variable, [x.sub.1] are the independent variables in the violation equation, [[Beta].sub.1] are the coefficients for the violation equation, [y.sub.2i] denotes the inspection variable, [x.sub.2] denotes the inspection equation variables, and [[Beta].sub.2] denotes the coefficients for the inspection equation. The problem is less (or more) acute for effluent discharges because the data set contains no information on which effluent observations have been verified and which have not. For this reason it is impossible to estimate a sample selection model of effluent discharge and the OLS results are reported.

The three equations are estimated as

[Y.sub.it] = f ([[Beta].sub.i1] [x.sub.i1], [[Beta].sub.i2] [x.sub.i2], [[Gamma].sub.i] [z.sub.i] + [[Epsilon].sub.i]

where [Y.sub.it] is one of the three dependent variables: i = inspection, violation, or effluent discharge; [x.sub.i1] are the federal variables; [x.sub.i2] are the local variables; [z.sub.i] are the inspection, violation or effluent equation control variables; and [[Beta].sub.i1], [[Beta].sub.i2], [[Gamma].sub.i] are coefficients on the respective model components.

A. State and Local Political Variables

A key variable relating to the cost of environmental protection is the economic stability of the mill. The EPA [1993] argues that the probability of a mill closing is captured by the difference between the plant's shutdown point and the current price of its product. While the actual shutdown point is not available, the EPA has produced an econometric model for the industry which can be used to estimate just how far the plant's cost can be raised before it exits the industry. As the distance between price and the shutdown point grows, the probability of an inspection should rise because the plant is less likely to close if a violation is detected. A decrease in the distance between the shutdown point and the market price should increase the probability of a violation and the level of effluent discharge for similar reasons.

The impact of environmental groups, whose importance to local policymakers is well documented, is measured by the percentage of the state's residents who are members of the Sierra Club. The percentage of environmentalists is predicted to have a positive effect because the size of the environmental constituency determines the importance of their support or opposition to the regulator. For similar reasons it is expected that the probability of a violation and the level of effluent discharge will decrease as the size of environmental groups increases.

Because a relatively affluent community is more likely to react to undetected violations the per capita income for the average household in the mill's county is also included. The more affluent the county, the greater is the probability of an inspection, the lower the probability of a violation and the lower the level of effluent discharge. The unemployment rate in the county in which the mill is located is included because an increase in the unemployment increases the implicit price of environmental protection (in terms of jobs). This increase causes the probability of an inspection to fall, the probability of a violation to rise and the level of effluent discharge to increase. In addition, the percentage of the county labor force employed at the plant is included to capture the importance of the plant to the local economy. The larger the level of employment, the greater the impact if the plant closes. For this reason, the probability of inspection should decline, the probability of a violation increase, and the level of effluent discharged increase as the employment at the plant increases. An interaction term between local employment and the unemployment rate is also included to capture the opportunity cost of job loss in the county. If the unemployment rate is high, it will be more difficult for the local labor market to absorb a large number of unemployed mill workers. Thus, it is expected that as unemployment and the size of the mill's employment increase, inspections will become even less likely.

The variable Governor is designed to capture the governor's preference for environmental protection and is equal to one if the state's governor is a member of the Democratic party. Traditionally, Democratic governors have been more pro-environment than their Republican counterparts. Therefore, Governor is expected to be positively related to probability of inspection. The majority party in the state Senate and House are included (1 = Democrat). Again, it is expected that Democratic majorities will be relatively more pro-environment.

B. Federal Political Variables

The congressional oversight variables are the median League of Conservation Voters (LCV) scores for the House Energy and Commerce Committee, House Oversight, and the Senate Environment and Public Works Committee, Senate Oversight, respectively. The more pro-environment the committee (higher LCV scores), the greater is the probability of inspection, the lower the probability of a violation and the lower the level of effluent discharge. These two variables proxy the ideological location of two of committee members.(6) Because national politicians may intervene differently in their home states, i.e. lobby for more stringent regulation for their home state or district, the LCV score for district or state's member on the House Energy and Commerce Committee and the Environment and Public Works Committee are also included. The predicted effects on the independent variables are similar to those of the committee's median member. In addition, because some states do not have a representative on the respective committees, two dummy variables controlling for membership are included. These dummy variables equal one if the state or district has a member on the respective committee.

The national unemployment rate is also included. It is expected that when the unemployment rate rises, the probability of an inspection fall, the probability of a violation and effluent discharges rise. As the opportunity cost of pollution abatement rises, the federal government, if it controls local policy, will reduce abatement efforts.

C. Inspection Control Variables

To control for seasonal effects the quarter of the year is included.(7) Because the national government still retains control over several state enforcement efforts, a dummy variable is used to differentiate between states in which the federal government, at least legally, is in charge of inspections. The real dollar amount the state spends on water quality divided by the number of manufacturers in the state is also included.(8) The state's water pollution control budget is not available on an annual basis and the 1988 budget is used for 1989 while the 1991 level is used for 1990-1993.(9)

The EPA suggests that local authorities inspect plants at least once a year. Therefore the number of quarters since the plant has been inspected in the last year is included and is expected to have a positive impact on the probability of inspection. The EPA also suggests that plants found to be in significant noncompliance, i.e. far in excess of their permit, be inspected again within one year, therefore the number of significant violations in the last year is included.(10) Finally, the EPA requires plants to supply monthly reports to local regulators. If plants fail to report a violation, which is then detected, this should trigger future inspections because it is a visible sign that a firm is not complying with EPA directives. A series of variables, equal to one if the plant failed to file a report one to four quarters ago, are included. Given the EPA's guideline we would expect these variables to have a positive impact on the probability of inspection. To control for industry specific effects, such as the difficulty in inspecting certain types of mills, the SIC code of the mill is used to construct a series of four variables: pulp mills, paper mills, paperboard mills, paperboard containers and boxes and converted paper products.

In addition, two structural parameters are included in all three specifications: (1) if there is a predicted violation at the mill in the current quarter and (2) the predicted effluent emission. If a violation is expected at the mill and/or the predicted effluent level rises, inspections should become more likely.(11) Two different versions of the full model are estimated. The first version contains only an intercept term. The second version substitutes the EPA region of the country in which the mill is located to control for any regional fixed effects. Finally, all inspections are not equally stringent; an inspection that involves only a check of paperwork is not as stringent as an inspection that involves sampling. For this reason the full model is also estimated using only inspections which involve sampling as the dependent variable.

D. Violation and Effluent Discharge Control Variables

The state-local and federal political variables are also included in the violation and effluent equations. Because local regulators, in conjunction with the national EPA, write the permits that govern pollution discharges, it is possible that political forces will also impact violation rates and discharge rates. The control variables for violations and effluent discharge follow Magat-Viscusi [1990]. Because both violations and effluent discharges may have a seasonal component, variables for the first through the third quarters are included. To capture economies of scale in pollution abatement the daily output, in tons, of the mill is included. Given the nature of the permitting process, larger plants need not be in violation more often than smaller plants. However, the EPA has explicitly included the disproportionate impact of pollution abatement regulations on small plants as a factor in its cost-benefit studies. This suggests that the EPA believes there are some economies of scale in pollution control. Plant size should be a key component in the amount of effluent discharged from the mill because the greater the output of the mill the more water used in the production process.

A third set of factors is the plant's history of past inspection. Inspections are expected to have a deterrent effect on future violation by raising the probability of detection. To capture this effect, four variables were created, numbering one if the plant was inspected one to four quarters ago. These variables are expected to have a negative effect on the probability of a violation and reduce effluent discharges. Another factor contributing to noncompliance and effluent discharge is the mill's history of past violations. There are two possible impacts of past violations. One possibility is that, because violations can only be corrected with the installation of new abatement technology which takes time, past violations will increase the probability of current period violations and increase effluent discharges. Alternatively, it may be that violations have a deterrent effect. Once the mill's employees or regulators discover the violation, efforts are made to correct it, reducing the probability of future violations and future effluent discharges. To capture these effects, a series of variables was created, numbering one if the plant was in violation one to four quarters ago.

The product type that the mill produces may also contribute to the probability of a violation and the level of effluent discharge. This effect is again captured by the SIC code(s) of the mill. The EPA alleges that pulp, paper and paper board mills utilize more water and have a production process that is subject to more spill overs, the usual cause of a violation.(12) The predicted probability of an inspection is included in both equations. An increase in the probability of an inspection this quarter is expected to decrease the probability of violation and the level of pollution discharged at the mill.(13)

E. The Data Base

The data used in this study is derived from the EPA's Permit Compliance System database. The study focuses on the pulp and paper industry because: (1) it is a large industry in terms of pollution discharge, (GAO [1987]) and can be an issue in local politics, and (2) there are many sites so the inspection issue is important. There are 232 plants in the sample covering 30 states. The time period considered for this study is from 1989 to 1993. The analysis is conducted by quarter because the same mill is only rarely inspected more than once a quarter. Table I compares the sample to the population of mills in the United States. The sample has good coverage, with 80% of the regulated industry included in the data set.

The sample characteristics can be found in Table II. The inspection rate is similar to previous studies, at about 34%. This is slightly higher than what one would expect if local authorities followed the EPA's one inspection per year guide. The non-compliance rates in this sample are actually lower than in either Magat-Viscusi [1990], the EPA [1986] or the more general GAO [1987] study. Magat and Viscusi found a 25% violation rate, the EPA found 16% noncompliance, while the GAO found 21% noncompliance. This study finds a noncompliance rate of 11% for the full sample, closer to the EPA estimates than either Magat-Viscusi or the GAO.(14)

III. RESULTS

The results for the inspection equations are presented in Table III. The results for the violation and effluent equations are presented in Table IV.

TABLE I

Number of Plants in the Population and Sample by State

State Population Sample

Alaska 2 0
Alabama 16 13
Arkansas 8 7
California 2 1
Connecticut 2 0
Florida 6 6
Georgia 10 10
Idaho 1 0
Iowa 2 1
Illinois 3 3
Indiana 4 4
Kentucky 3 1
Louisiana 13 9
Maryland 1 0
Massachusetts 16 11
Maine 8 8
Michigan 15 15
Minnesota 5 3
Mississippi 7 7
Missouri 3 0
North Carolina 7 7
New Hampshire 11 9
New Jersey 10 4
New York 24 24
Ohio 9 9
Oklahoma 3 1
Oregon 14 9
Pennsylvania 10 3
South Carolina 6 6
Tennessee 4 4
Texas 9 6
Virginia 4 2
Washington 18 16
Wisconsin 22 21

A. State and Local Political Variables

The natural logarithm of the difference between the current market price of the plant's product and its shutdown point is positive and significant in all specifications. A one standard deviation change in the difference produces a 2% change in the probability of an inspection. The importance of this effect indicates that local inspectors consider the possibility that a mill will be forced to shut down due to detected violations.(15)

The size of the environmental movement in the state is significant in all of the inspection specifications. A one standard deviation change in the percentage of the state's residents in the Sierra Club produces a 15% increase in the probability of inspection. An increase [TABULAR DATA FOR TABLE II OMITTED] in the size of environmental membership in the state also causes a statistically significant reduction in the probability of a violation (a one standard deviation increases in the environmental membership reduces the probability of a violation by 4.2%) and reduces the rate of effluent discharge at the mill by a statistically significant amount in all specifications which do not contain regional intercepts.

The logged household Per Capita Income (PCI) of the surrounding community is significant and positive in all of the inspection specifications without regional intercepts. A one standard deviation increase in the natural log of PCI increases the probability of an inspection by 8.2%. The importance of PCI suggests that higher income communities value environmental protection more and are more likely to complain when inspections are not conducted. In addition, an increase in PCI reduces effluent discharges by a statistically significant but small amount; a one standard deviation increase in PCI reduces discharges by half a percent.

The level of employment at the mill is negative and significant in all inspection specifications. The marginal effect indicates that a one standard deviation change in the value of the natural log of mill employment reduces the probability of an inspection by 2.7%. This is consistent with the hypothesis that the local agencies charged with enforcement maximize political support by considering the economic interest of the surrounding communities. In particular, the sign and significance of the coefficient suggest that local EPA's are concerned with the possible employment effects of detected violations. Further confirmation of this hypothesis can be found in the effect of the interaction term between the level of employment at the mill and the unemployment rate in the county in which the mill is located. The interaction term is negative and significant in all specifications indicating that as the unemployment rate rises, larger employers are inspected less frequently. The marginal effect indicates that a one standard deviation change in the interaction term reduces the probability of an inspection by 3.4%. The interaction term is also statistically significant and the marginal effect indicates that a one standard deviation increase in the interaction term increases the probability of a violation by 6.8%.

B. Federal Political Variables

Local representation on the national oversight committees has an important impact on the implementation of pollution control laws. The LCV score of the district's representative on the House oversight committee has a significant and positive effect. Merely having a member on the House Energy and Commerce Committee reduces the probability of inspection; in the 6% of the observations that have a member on the oversight committee, 14 total mills, the inspection rate is 30% lower; the largest effect of any of the independent variables. When the member's ideology is included, the effect of membership on the oversight committee is still negative. Evaluated at the mean LCV score of local representatives, an LCV score of 34, the presence of a representative still reduces the probability of inspection by 13%. For the most pro-environment representatives, those with an LCV score of 100, the probability of an inspection in a given quarter increases by 34%, a 100% increase. The results are similar for the level of effluent discharged at the mill. Mills located in committee member's districts pollute more; however, the more liberal the member the lower the pollution levels. The results are similar for the Senate oversight committee. The magnitude of the marginal effects is, however, smaller. A Senator with committee membership reduces the probability of an inspection by only 7% and when the mean LCV score for those states with a committee member is included the marginal effect drops to only -3.9%. The effect on effluent discharges is similar to those of the House.

In all specification without regional intercepts the median member of the House Energy and Commerce Committee is significant and positive. A one standard deviation increase in the median LCV score (the oversight committee becomes more liberal) raises the average probability of inspection by five points; a 15% change in the average inspection rate. The effect of a change in the Environment and Public Works Committees median member's LCV score is also statistically significant and positive. A one standard deviation increase in the median LCV score increases the probability of inspection by 3.5%.

The results from estimating the models using sampling inspections are substantively similar. The coefficient estimates and standard errors are reported in column 5 of Table III.(16) In the specification containing regional intercepts, several of the local variables are no longer significant. However, employment effects and the effects of local representation on the oversight committees remain significant.

[TABULAR DATA FOR TABLE III OMITTED]

C. Control Variables

The control variables used in this study are quite similar to other studies of enforcement and are not discussed at length (see Magat-Viscusi [1990]). There are, however, a few interesting features in the control variables' results. First, the effect an increase in the pollution abatement budget is not significant. This is surprising because the level of spending is likely determined by the same political factors that determine inspections. However, even when political variables are omitted from the model spending levels, though always positive, do not have a statistically significant impact on the probability of an inspection. An alternative explanation for this lack of significance is the biennial nature of the budget variable, which adds noise to the data.

[TABULAR DATA FOR TABLE IV OMITTED]

Also interesting is the effect of the structural parameters. An increase in the predicted effluent discharge at the mill increases the probability of an inspection. Violations are deterred by an increased probability of an inspection in the contemporaneous quarter and from inspections three and four quarters before. While neither an increased probability of inspection nor past inspections appear to reduce pollution levels, violations three and four quarters ago reduce pollution levels in the current period. One way to interpret these findings is that violations that involve excess pollution discharge are difficult to correct and take nine months to a year before pollution levels at the mill begin to fall.

IV. CONCLUSION

My central objective in this paper is to characterize the source of discretion found in local bureaucratic agencies and to ascertain the consequences of delegation by national policymakers. The three models which were estimated are characterized as the state/local, federal, and full specifications. The theory predicts that local regulators will alter the stringency with which pollution control regulations were enforced in order to maximize the level of political support from external actors. The federal-specification predicts that local regulators respond only to external signals from a national coalition. By contrast, the state/local-specification predicts that local regulators respond only to local interests.

The results suggest that state agencies are responsive to local and national interests groups and that delegation allows different political forces to have an impact in different states. One remaining question is why, given the loss of policy control, does the national government choose to delegate? One possible motivation is suggested by the impact of having local members on the oversight committees. By delegating, the members of oversight committees, who have been charged with "watching the watchers," are able to pass a sizable portion of the costs of enforcement on to state governments. The GAO estimates that by delegating, the federal government passed $154 million on to the states in 1995 alone. Furthermore, the results indicate that members of the oversight committees, particularly in the House, are quite successful in protecting the interest of their constituents, at least to the extent that these interests are captured by LCV scores. The cost-control tradeoff is typical of all regulation; however, the impact of committee membership shows that this loss of control is by no means complete. There are alternative explanations for delegation that cannot be ruled out by the results. One possibility is that oversight committees delegate when local authorities are going to do what Congress wants in any case. An alternative possibility is that Congress delegates in those areas where the marginal opposition to enforcement is particularly high.

The author is grateful to William Olbrich, Jr., and his staff. Without their assistance this paper would not have been possible. The author would also like to acknowledge helpful comments from Art Denzau, Marilyn Flowers, James Hamilton, John Lott, Bill Lowry, Gary Miller, Mary Olson, Gary Santoni, Brian Staihr, Mike Sykuta, Mark Zupan, the anonymous referees, and seminar participants from Ball State University and Washington University. All errors are my own responsibility.

1. A notable exception is the work of Scholz [1986].

2. Enforcement of the Act was initially a federal responsibility, with inspections conducted either jointly with the EPA or by independent contractors hired by the EPA. The current program, delegating oversight authority to some states, was developed in 1984.

3. Becker's [1968] analysis of crime indicates that as the probability of an inspection increases, the frequency of violations would decrease. This decrease is caused by the increase in the expected cost to the violator from engaging in criminal activity. In the context of industrial dischargers, Becker's model suggests that as the probability of an inspection increases, the firm's costs increase because it must undertake costly pollution abatement or risk an increased probability of detection.

4. The role of fines in NPDES inspections is more ambiguous. Unlike OSHA inspections, a reliable record of fines assessed does not exist. Resources for the Future conducted a survey of states to ascertain the average penalty per violation and found the average state penalty per Notice of Violation (NOV) was 393 dollars between 1978 and 1983. Fifteen percent of the inspections resulted in an NOV. The small dollar amount of state fines has led Russell et al. [1986] to conclude that fines do not represent a serious deterrent for polluters. Such a conclusion may be premature. First the EPA and delegated states recently gained the right to fine without legal action on the part of the Justice Department. In addition, Notices of Violation are often used in court cases when individuals bring legal action against polluters for damages. The EPA recently fined Louisiana-Pacific and Simon Paper Company 5.8 million dollars as a settlement to a law suit. The plant's history of non-compliance was considered key evidence in the San Francisco Chronicle [1991]. While the dollar value of a given penalty may be low, a sequence of violations, kept on file by the EPA or state agencies, can cost a company far more than a one time penalty. Typically, mills in repeated violation are pressured into installing new and more costly abatement technology. Such investments would not be necessary without some legal record of a mill's compliance.

5. For an excellent treatment of this problem implicit in all empirical tests of compliance, see Feinstein [1990].

6. The LCV scores for the various committees represent a difficult choice for a researcher. Some 35 committees and subcommittees oversee the Environmental Protection Agency (Lazarus [1991]). A potential solution is to regress the LCV scores on the number of inspections. The problem with such a test is that the relatively short time span produces extreme multicollinearity between the committees and their subcommittees. For this reason the House Energy and Commerce Commission, which according to Lazarus had the largest number of EPA testimonies in the House, and the Senate Environment and Public Works Committee, which occupies a similar position in the Senate, were selected. The motivation for this selection is theoretical. The usual method used by committees to oversee executive agencies is testimony. The number of times EPA officials where called to testify before Congressional committees can be found in Lazarus [1991].

7. To avoid perfect collinearity with the constant term or the regional intercepts quarter four is suppressed.

8. The number of manufacturers is the total number manufacturing establishments in the state listed in the Dunn and Bradstreet directory. Although not all of these firms are regulated, the Council of State Governments argues that this is a good measure by which to compare state's regulatory mandate.

9. The Council of State Governments, which collects the data, does so only every two years. In addition, given the differing ways in which states present their budgets, the Council adjusts the budget numbers into similar categories making comparisons possible.

10. Ideally I would like to know the magnitude of the violation. However the EPA's data does not contain information of the amount by which a mill exceeded its permit.

11. Effluent is measured as the pounds per day of biological oxygen demand, the amount of oxygen, which the mill's effluent removes from the stream. It is the conventional measure of pollution.

12. A spillover is untreated water over flowing from the treatment pond.

13. Because divergent magnitudes in the independent variables can cause problems for maximum likelihood estimation, the natural log of a number of the variables is used.

14. The principal reason for these differences may be the period in which the studies where conducted.

15. The economic impact of the independent variables is evaluated using the marginal effects from the full model unless otherwise noted.

16. The results for the violation equation and the effluent equation using the probability of sampling inspections as a structural parameter are not reported because they are almost identical to the results reported above.

REFERENCES

Becker, G. "Crime and Punishment: An Economic Approach." Journal of Political Economy, March/April, 1968, 169-217.

Feinstein, J. "Detection Controlled Estimation." Journal of Lava and Economics, April 1990, 233-77.

General Accounting Office. Water Pollution: Application of National Cleanup Standard to Pulp and Paper Industry, 1987.

Lazarus, R. "The Neglected Question of Congressional Oversight of EPA: Quis Custodiet Ipsos Custodes (Who shall watch the watchers themselves)." Law and Contemporary Problems, Autumn 1991,205-48.

Magat, W., and K. Viscusi. "Effectiveness of the EPA's Regulatory Enforcement: The Case of Industrial Effluent Standards." Journal of Law and Economics, 33, October 1990, 361-30.

Pashigian, P. "Environmental Regulation: Whose Self-Interests are Being Protected?" Economic Inquiry, October 1985, 551-84.

Peltzman, S. "Toward a More General Theory of Regulation." Journal of Law and Economics, August 1976, 211-40.

Russell, C., W. Harrington, and W. Vaughan. Enforcing Pollution Control Laws. Washington, D.C.: Resources for the Future, 1986.

San Francisco Chronicle. "Pulp Mills Pollution Settlement." 10, September 1991, Sec. A, p. 18:1.

Scholz, J., and F. Wei. "Regulatory Enforcement in a Federalist System." American Political Science Review, December 1986, 1,249-70.

U.S. Environmental Protection Agency: Water Management Division, Facilities Performance Branch. Study of the Pulp and Paper Industry in Region IV, 1986.

-----. Economic Impact and Regulatory Flexibility Analysis of Proposed Effluent Guidelines NESHAP for the Pulp, Paper, and Paperboard Industry, 1993.

Helland: Assistant Professor, Ball State University, Muncie, Ind., Phone 1-764-285-5378, Fax 1-765-285-8024 E-mail 00eahelland@bsubc.bsu.edu