Grantee Research Project Results
Final Report: Looking Inside the Black Box: Microlevel Empirical Analyses of the Impact of State and Federal Policy Instruments on Hazardous Waste Generation and Management
EPA Grant Number: R828952Title: Looking Inside the Black Box: Microlevel Empirical Analyses of the Impact of State and Federal Policy Instruments on Hazardous Waste Generation and Management
Investigators: Good, David , Richards, Kenneth
Institution: Indiana University - Bloomington
EPA Project Officer: Aja, Hayley
Project Period: April 1, 2001 through March 31, 2003 (Extended to May 20, 2005)
Project Amount: $180,917
RFA: Market Mechanisms and Incentives for Environmental Management (2000) RFA Text | Recipients Lists
Research Category: Environmental Justice
Objective:
Although many have developed analytic ideas about how Federal and state regulations will affect the production and treatment of hazardous wastes, there is little empirical evidence about how these regulations work in practice. The primary objective of this research project was to empirically evaluate the effects of the several alternative policy instruments that states have at their disposal.
The specific objectives of this research project were to: (1) develop a useful database of the hazardous waste policy instruments, particularly taxes, that states used from 1985 to 2002; and (2) start a series of papers analyzing the rationality of states’ use of taxes as a policy instrument.
Summary/Accomplishments (Outputs/Outcomes):
Inventory of State-Level Hazardous Waste Policy Instruments
Preliminary Review of Instruments. The first step in building the policy instrument database was to characterize the instruments used by each state, differentiating each instrument according to the entity that the state had targeted. The exercise involved, for each state, an initial review of the Bureau of National Affairs Environment and Safety Laboratory Digests of State Programs, which lists states’ individual hazardous waste laws that exceed the Federal Resource Conversation and Recovery Act (RCRA) requirements. The review covered reporting requirements, command and control regulations, liability laws under the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA), programs and initiatives, taxes, and fees. The instruments were differentiated according to whether they were aimed at: (1) generators, (2) transporters, or (3) treatment, storage and disposal (TSD) facilities. For each state a summary table was prepared detailing the types and characteristics of instruments that were employed.
The result of this general review of states’ use of policy instruments was informative. It showed that states largely have focused on a small subset of the many environmental policy instruments available. Four instruments particularly stand out. First, many states have adopted voluntary programs involving technical assistance and public recognition. Second, the vast majority of states have developed tracking and reporting requirements that exceed RCRA requirements in scope of materials covered, detail of reports, or frequency of filing. Third, most states have adopted one of many forms of fees for generators; transporters; and treatment, storage and disposal facilities. Finally, virtually all of the states have adopted some form of liability law related to the generation and management of hazardous waste.
Liability Laws. The Environmental Law Institute (ELI) has maintained a periodic examination of state liability laws regarding hazardous materials. These have the limitation that they are student projects and consequently judgments can change from year to year primarily because the people making the judgments change with no real change in the rules. For our analysis, which is based on panel data, it is more essential that our data be based on consistently measured information over time instead of on judgments that might seem appropriate to make at one point in time. As a result, we spent considerable time in reevaluating the ELI studies and have made approximately 60 changes to their 250 observations. This also has given us the opportunity to interpolate the data between years and extrapolating results through 2001. The end product is that our revised data more consistently reflects the laws on the state books and is less reliant on changing judgments on the part of students compiling the data or on changing judgments of state regulators answering the survey regarding the nature of implementation.
Audit Privilege Laws. Firms that generate hazardous wastes face mixed incentives about how much information they gather about that generation. Knowing what waste is generated and how it is treated should be important information so that it can be managed efficiently. Alternatively, having this information could potentially be used against companies in legal proceedings. Some states have responded to this quandary by adopting audit privilege laws. These laws allow firms to perform environmental audits and keep the information of those audits confidential.
We have completed gathering information about the status of state audit laws each year between 1989 and 2001. The largest changes to state laws in this area occurred in 1995-1997, making that time period ideal for evaluation. Approximately one-half of the states have adopted such a procedure.
Pollution Prevention Programs—Voluntary and Mandatory. Since the 1980s, many states have adopted some form of pollution prevention program. As of 1996, 38 states had passed pollution prevention legislation establishing mandatory or voluntary programs. Voluntary programs generally combine a legislative finding in favor of waste reduction over treatment or disposal with incentives and education programs designed to encourage firms to reduce their level of hazardous waste generation. Mandatory programs generally require various forms of participation in state-sponsored efforts. One of the primary tools in mandatory programs is a requirement that firms draw up pollution prevention plans to assess their opportunities to reduce hazardous waste. Even then, implementation of the pollution prevention program is generally voluntary ¾ only development of the plan is mandatory. Some states also include reporting requirements among the elements of their mandatory programs.
States Taxes and Fees. This study has identified three basic forms for taxes: one-time fees, annual fees, and quantity-based fees. One-time fees are generally applied when a new hazardous waste activity commences or is significantly modified. For example, in the year 2000, Indiana charged a fee of $21,700 for a permit to establish either an onsite or offsite hazardous waste incinerator and $40,600 to permit a new landfill. Like the one-time fee, the annual fee is a facility fee that can be assessed on any of the ongoing hazardous waste activities. For example, in the year 2000, Iowa assessed a $250 annual fee on hazardous waste generators, Illinois assessed a $9,000 annual fee for onsite land application operations and a $35,000 annual fee for offsite land application operations, and Michigan charged hazardous waste transporters $216.67 per year.
In contrast to the one-time and annual fees, the quantity-based taxes are applied to activities rather than to facilities. Most quantity-based taxes are expressed as a fee per weight or a fee per volume. For example, in 2000, Kansas charged $10 per ton of hazardous waste treated at an offsite incinerator, whereas Illinois charged a fee of $0.09 per gallon for land application and landfill disposal. Regardless of the form of the fee, the important point is that unlike the one-time and annual fees, which are fixed fees, the quantity-based fees are proportional to the amount of hazardous waste activity. Thus, the quantity-based fees have the potential to send price signals to hazardous waste managers regarding the states’ preferences and values.
From our review and cataloging of fees, we were able to develop a taxonomy of fees. We identified four structures of fees and 14 points of applications for those fees. The result was a total of 56 distinct fee types. Interestingly, in the completed database of the 50 states from 1985 to 2002, we can observe almost all of these 56 fee types in use by one state or another at one time or another.
As we looked forward to employing the database of fees as an explanatory variable in the econometric analysis, we also realized that we could not use the fees in their fully disaggregated form. This required that we develop quantitative variables that capture most of the variation among the states’ fees.
Rationality of States’ Use of Hazardous Waste Taxes
This portion of the project examines if state tax (charge, fee) regimes for hazardous waste management are rationally designed to promote plausible policy goals. To examine the issue of rational design, we drew on the extensive database of state laws, including detailed structures for taxes and fees at various stages of the life cycle. We also have developed a framework for examining the apparent policies embedded in state hazardous waste tax structures. We look for three kinds of rationality. First, we examine if states comply with our informal a priori hypotheses of expected preferences as revealed through hazardous waste tax structures. Generally, we would expect to see states adopting policies that discourage the generation and offsite transportation of wastes. Second, we explore if states exhibit a consistent set of policy preferences among states, regardless of our a priori expectations. Third, we consider if states at least exhibit an internal consistency in their tax structures.
In our preliminary examination of 12 Midwest states we found little consistency between our expectations of policy-driven state tax structures and those that have been adopted by the states. We also found little consistency among the states with respect to revealed policy preferences. Finally, we concluded that 4 of the 12 state programs ( Illinois, Kansas, Missouri, and Ohio) are internally inconsistent.
In the case of hazardous waste, state governments may have several different environmental policy goals. These goals could include: (1) reducing the amount of hazardous waste generated; (2) reducing the amount of hazardous waste transported offsite; (3) changing the mix between treatment and disposal; (4) reducing the amount of one treatment practice relative to the others; (5) reducing the amount of one disposal practice relative to the others; and (6) reducing the amount of waste brought in from out of state for storage, treatment, or disposal. In this discussion we will focus on three primary policy goals. First, states may care if firms generate hazardous waste and how much. Second, they may care where the hazardous waste goes, onsite or offsite. And third, state governments may care what firms do with the hazardous waste and how it is treated, stored, and disposed.
To pursue their three primary policy goals, states can use any of the three types of taxes ¾ one-time facility fees, annual facility fees, or fees based on the level of an activity (generally dollars per ton). Of the three types of taxes, the quantity-based fees are the ones favored by policy analysts as providing an appropriate price signal to decision makers. Thus, in this analysis, we will examine how states can use quantity-based hazardous waste fees to effect all three types of management decisions ¾ whether, where, and what ¾ that private parties make with respect to hazardous waste.
Given that states can design taxes to target whether, where, and what firms do in terms of hazardous waste management decisions, it follows that there are actually a limited number of tax regimes. Table 1 provides an enumeration of the eight combinatorial possibilities that follow from the three issues that states can target. In Regime 1, for example, the state structures hazardous waste taxes to address all three issues. In Regime 5, the state addresses where firms transport their waste and what is done with it but not if or how much waste is generated. Regime 8 represents a complete lack of incentive-based fees, suggesting that the states adopting this approach are not trying to provide incentives to change hazardous waste management decisions.
Table 1. State Tax Regimes for Hazardous Waste Policy, Year 2000
Regime | Whether | Where | What | Midwest States |
---|---|---|---|---|
1 | √ | √ | √ | Iowa, Missouri |
2 | √ | |||
3 | √ | |||
4 | √ | Michigan, South Dakota | ||
5 | √ | √ | Illinois, Kansas, Ohio | |
6 | √ | √ | Minnesota, Nebraska, Wisconsin | |
7 | √ | √ | ||
8 | Indiana, North Dakota |
From the data on state hazardous waste fees, it is possible to categorize the 12 Midwest states according to the hazardous waste tax regimes that they have adopted. Of the eight possible regimes, five are represented in the Midwest region (Table 2). Only Iowa and Missouri have adopted tax structures that provide incentives to decrease generation, keep hazardous waste onsite, and adopt particular treatment and disposal options ¾ whether, where and what . Two states, Indiana and North Dakota, have tax regimes, or a lack thereof, that provide no direct incentives for firms to alter their management decisions.
To the extent that states adopt fees to further environmental policy goals, the relative level of fees within a state can indicate states’ preferences across possible management options. Table 2 provides a rank ordering of state preferences across treatment and disposal options as revealed in the level of taxes. As mentioned above, Indiana and North Dakota have no quantity-based fees, indicating that they have not used this policy tool to promote their hazardous waste policy. Michigan taxes only the landfill option, treating deep well injection and land application equally with the treatment options.
Looking at the onsite preferences of the remaining nine states, two observations stand out. First, all prefer incineration and other treatment options to any of the disposal options and none distinguish between incineration and the other treatment options. Second, the treatment of the disposal options is mixed. Five states treat land application, landfill, and deep well injection equally, two states prefer land application and landfill ( Minnesota and Missouri), and two states ( Ohio and Kansas) prefer deep well injection. Only two of the states ( Michigan and Kansas) distinguish between land application and landfill, both providing preferential treatment to land application. Thus, we see little consistency regarding revealed policy preferences among the states.
Table 2. Comparison of Apparent State Preferences for Treatment and Disposal 1
State |
Onsite Preferences |
Offsite Preferences |
||||||
1st Choice |
2nd Choice |
3rd Choice |
4th Choice |
1st Choice |
2nd Choice |
3rd Choice |
4th Choice |
|
IN, ND |
Incin., OT, LA, LF, DWI |
|
|
|
Incin., OT, LA, LF, DWI |
|
|
|
MI |
Incin., OT, DWI, LA |
LF |
|
|
Incin., OT, DWI, LA |
LF |
|
|
IA, Neb, SD, WI |
Incin., OT |
LA, LF, DWI |
|
|
Incin., OT |
LA, LF, DWI |
|
|
IL |
Incin., OT, DWI |
LA, LF |
|
|
DWI |
Incin., OT |
LA, LF |
|
MN, MO |
Incin., OT |
LA, LF |
DWI |
|
Incin., OT |
LA, LF |
DWI |
|
OH |
Incin., OT |
DWI |
LA, LF |
|
Incin., OT |
DWI |
LA, LF |
|
KS |
Incin., OT |
DWI |
LA |
LF |
DWI |
LA |
LF |
Incin., OT |
1 Notation for Table 2: Incin = incineration, OT = other treatment, LA = land application, LF = landfill, and DWI = deep well injection.
Intrastate consistency is evaluated on the basis of two relatively simple assertions. The first principle of rationality states that, for policy purposes, the incentive related to where a waste is treated or disposed should be independent of what the ultimate treatment or disposal is. For example, if it is riskier to transport waste offsite for land application than to leave it onsite, then it is also riskier to transport it offsite for incineration than to incinerate onsite. The second rationality principle is the converse of the first: the risk of one type of treatment or disposal method relative to another is independent of the location at which they take place. So, for example, if onsite landfill disposal imposes additional risk to society relative to onsite incineration, then it imposes the same additional risk offsite and should be subject to the same tax premium. These two principles can be applied to our analysis of the internal rationality of states’ hazardous waste fee structures.
The transportation fee in Missouri appears to apply only to the waste headed to treatment or deep well injection (i.e., there is no disincentive to transport waste offsite for land application or landfill). The three states that have adopted Regime 5 fees are all internally inconsistent. Ohio, the least inconsistent, charges an offsite premium for all forms of treatment and disposal, but that premium ranges from $2/ton for treatment to $2.5/ton for deep well injection up to $5/ton for land application and landfill. Both Illinois and Kansas charge an offsite premium for treatment but not for disposal.
Thus, four states ( Ohio, Illinois, Kansas, and Missouri) violate the first rationality principle by applying offsite fees, either directly or indirectly, that vary according to the type of treatment or disposal methods adopted.
To examine if states violate the second rationality principle, we can turn to the offsite preference summary in Table 2. We see that only two states, Illinois and Kansas, show a different ordering among offsite preferences than for onsite preference ranking. Whereas the Kansas state fees for disposal are greater for onsite waste, when the waste is moved offsite, the fees for treatment are greater. Interestingly, for both states, deep well injection receives the most preferential treatment offsite whereas incineration and other treatment options drop in preference. Although the practice is not internally inconsistent, it merits mention that the fees for either onsite or offsite deep well injection are 0.0091 dollars per ton, less than a penny per ton. One wonders why Kansas bothers.
Issues in Modeling Firm Response to Environmental Instruments
As a conceptual framework, all of the models we analyze examine how different hazardous waste policies in a state affect firm decisions on location, generation, and treatment. Our discussion focuses on three main issues that are intertwined. (1) What industries should be studied? (2) What models describe firm behavior? (3) What econometric difficulties need to be either avoided by answering Issues 1 and 2 in clever ways or confronted head on through the use of econometric technique?
Our first econometric difficulty was to avoid simultaneity. A number of states with relatively high levels of waste generated also tend to have fairly strict environmental taxes and programs. As it is just as plausible that states set laws in response to wastes generated as it is that firms adjust wastes generated in response to laws, the simultaneity issue must be confronted first. Although some might chose to address the problem through the use of econometric technique, such as instrumental variables estimation, the usefulness of the results depend heavily on the plausibility of the instruments. Motivation for a set of instruments in this setting is painfully weak, as we would be forced to argue that we could identify variables that affect the creation of laws directly, but they only affect the waste generation decision indirectly through those laws. We chose a different track in our analysis by relying on the microeconometric argument against simultaneity. If we do disaggregate analysis, at the firm- or industry-level, then state laws are less likely to be adopted as a result of these decisions. Choices of industries should include industries that do generate wastes but that are not so large that single firm decisions are likely to motivate public policy. Consequently, we should avoid industries that others have looked at in the past (e.g., petrochemicals, extraction, and pulp mills).
For several reasons, we chose six industries to examine in our preliminary analysis: commercial printing (Standard Industrial Classification System [SIC] 2752/North American Industry Classification System [NAICS] 323110 and 323114), paint manufacturing (SIC 2851/NAICS 325510), electroplating and metal finishing (SIC 3471/NAICS 332813), printed circuit board manufacturing (SIC 3672/NAICS 334412), auto parts (SIC 3714/ NAICS several) and wooden unupholstered furniture (SIC 2511/NAICS 337122 and 337215). First, there were a large number of facilities in the Biennial Reporting System (BRS) that report generating these wastes; second, these appear to be industries that have been targeted by many states for pollution prevention activities; third, there was a clean SIC/NAICS bridge, allowing us to use information prior to adoption of NAICS and after without substantial redefinition of the industry (which would confound any determination of effects over time); fourth, there was an expectation of reasonable homogeneity within industry group (i.e., all of the facilities were performing similar kinds of activities); and, finally, these facilities are widely distributed across several states so that state level variation in laws could be used to identify impacts. Ultimately, we dropped the wooden furniture industry because there was an insufficient distribution of facilities across states and the SIC/NAICS bridge was substantially weaker than with our other industries. We dropped auto parts because there was too much heterogeneity in the products produced/processes used.
Our analysis requires that we integrate two other types of information with our database of state policies. These include economic data that describe other decisions by the firm (industry with which it identifies, levels of employment, and output). The second type of information describes their hazardous waste generation and treatment/disposal decisions. Our initial plan was to use the Longitudinal Research Database (LRD) from the Census Bureau to identify the economic decisions. Unfortunately, given our decisions on the industries we examined, the typical firm size was quite small (80 or so employees), and Census stratifies the sampling severely at this firm size. Additionally, there is only one year (1997) where a Census is completed that would be usable for our problem. Instead, we used data from the Dun & Bradstreet Million Dollar Directory (electronic version) from 1991 through 2002. This in itself was a major task as Dun & Bradstreet no longer keep historical information. Our initial plan was to use BRS to describe firm decisions. This data runs every other (odd) year from 1989 through 2001 with some 2003 data now becoming available. We chose not to include the 1989 data in our analysis because changes in reporting laws would confound any analysis that focused on changes in behavior over time. We chose to redefine the waste streams so that they were consistent with which national report definition was most stringent. As a result, when a firm’s wastes change over time, it is a result of their behavior changing, not a change in the definition of what constitutes waste. Our having access to the data in an environment that necessitated less confidentiality allowed us to more aggressively pursue facility identification and matching with the waste generation data, as well as lead to a substantially larger number of facilities for analysis. The effects we are seeking are likely to be small. The costs of compliance with state policy or the cost of hazardous waste fees are likely to be a very small part of the factors influencing the decision. This emphasizes our need to be econometrically efficient and the need for a large sample. Ultimately, we were able to match data for 8,343 observations (facility years) from our 4 industries.
The second major issue to confront is sample truncation. Sample truncation occurs when observations are systematically excluded from the analysis based on the value of the dependent variable. In our study, if we excluded facilities that were not large quantity generators (as BRS does) but based our analysis entirely on that data, a truncation bias would occur that would reduce the magnitude of estimated coefficients (bias is toward no evident effect). This truncation bias is the most severe when there is a large amount of unexplained heterogeneity because waste generation decisions are idiosyncratic. Instead, we avoided the problem and based our sample on facilities that were part of the Dun & Bradstreet data where inclusion is based on the amount of sales and the number of employees at the local facility. No truncation bias occurs when sample selection is based on the value of independent variables. When it comes to decisions about hazardous waste generation, we can only say that they are not large quantity generators (the conditions for inclusion in BRS). Thus, we trade a censored data problem in return for the more difficult sample truncation problem.
Measurement issues with independent variables pose yet another difficulty. The mismeasurement of an independent variable causes additional econometric difficulty. The effect of this mismeasurement is a bias in the estimated coefficient toward zero. In our analysis, this is most likely to occur with the measurement of hazardous waste taxes. In our original proposal, we developed models that had nonlinear responses on the part of facilities in response to linear prices. In our initial investigation into state level taxes, it is clear that the pricing is very nonlinear. It is not uncommon to have very high taxes per ton for small quantity generators and the tax rate to decline or even have a cap on total taxes paid. This means that the facilities that are most likely to be responsive to environmental taxes are the facilities that are least likely to be included in LRD and BRS. It also suggests that analyses that focus on state-level generation decisions will be heavily influenced by facilities in which the effective marginal tax rate is zero. This suggests that analysis at that level is less likely to identify a significant relationship because of econometric problems but also because they are looking for a relationship precisely where it is unlikely to exist. This reinforces our decision to examine the relationship between hazardous waste policy and decisions at small facilities.
The nonlinearity of state hazardous waste taxes also forces us to confront the measurement problem of characterizing state taxes in ways that are the most relevant for our sample of facilities. We confront this problem by the construction of several index numbers for state tax policies for generation, treatment, and disposal activities. In effect, these indices measure what the fee per ton would be if each facility in an industry from across the country was located in the particular state. This necessitated the development and calculation of the nonlinear tax functions on a facility-by-facility basis using as much information about the context of generation (e.g., some states adopt special tax rates dependent on industry or other firm characteristics) as was available. These are then totaled with an average tax per ton assigned. We calculated tax rates for several types of activities: generation, transportation, recycling, other treatment, and disposal. Because some state policies with respect to recycling retroactively change other tax rates, some of our taxes can be negative (i.e., if there is a generation tax, but it is eliminated by recycling, the generator tax is positive but the tax on recycled hazardous waste is negative, offsetting the generator tax). To get a tax index for all activities (generation, transportation, treatment, and or/disposal), the individual tax indices can be summed. The negative value of the recycling index sometimes restricts the functional form that can be examined.
The final econometric challenge for our models is unexplained persistent heterogeneity. Variables that effect firm decisions have been excluded from our model for several reasons. Inevitably, some of these variables are related to state-level policy variables, especially as there are many variables that might be related to both firm decisions and state policies. Fortunately, because of the panel nature of our data we can use either firms or states as their own controls. Because we are more concerned with the effects that misspecification can have on biasing the parameter estimates than we are in potential inefficiencies in estimation, we construct fixed effects panel data estimators wherever possible. There are some instances where these fixed effects estimators are not well-developed (the censored panel data tobit model) and in those cases, we attempt to estimate the importance of the bias of using random effects models through approximate Hausman tests. One can also plausibly argue that a policy-by-policy string of models is inappropriate as states that adopt a strict set of hazardous waste policies along one dimension (e.g., liability) are also likely to adopt strict policies along other dimensions (e.g., mandatory pollution prevention programs). Failure to include the entire suite of policy instruments makes interpretation of the coefficients on the policy variables tenuous. To the extent we can argue that the variables have a collective effect, we can test such hypotheses with generalized F tests in the linear case or Wald tests for maximum likelihood estimates.
Econometric Models of Policy Instrument Impacts
In our summary of the results of our econometric investigation, we focus on four of the several model variations we estimated. These tend to be the models that were estimated with robust methods (we are trying to measure what are likely small effects) rather than the models that happen to have statistically significant coefficients.
Facility Location. To estimate our model of facility location, we conduct an aggregate analysis of the number of facilities in each state for our four industries (for the years 1987, 1992, 1997, and 2002) collected from the Census of Manufacturers. We controlled for state population (an exposure variable in which the coefficient was restricted to be one) and the wage rate in the industry. To account for both persistent selectivity and other persistently excluded variables, we estimate a negative binomial model including fixed effects with conditional maximum likelihood. Our suite of policy variables include the generator tax index, transportation and disposal tax index, a dummy variable for strict liability, dummies for audit privilege and immunity, dummies for mandatory and voluntary pollution prevention programs, and dummies for whether those pollution prevention programs involved technical assistance or information clearinghouses. Pooling the data across industries so that the impacts of the policy variables could be estimated over a larger sample greatly improved the precision of the estimates (industry technical coefficients were not restricted to be the same). The model generated estimates of the effect of taxes that are very precise (the elasticity had a standard error of .001 or less). Moreover, the estimate of the elasticity was positive (rather than negative as economic theory would suggest) and were very close to zero, .001 to .003. To the extent that a measurement bias exists, it would suggest that the real effect is even more positive, though still substantively irrelevant. Because of the precision for estimation, we can conclude that firms really do not respond to hazardous waste taxes when making location decisions. Thus, when examining policy issues like the pollution haven hypothesis, states might set hazardous waste taxes to attract potentially polluting manufacturing jobs, but firms do not respond to them.
State-Level Aggregate Industry Generation. We estimate a model in which the decisions for all of the generators in a state (large quantity generators in BRS and smaller quantity generators from Resource Conservation and Recovery Information System [RCRIS]) are aggregated. This has the desirable effect of averaging away firm level heterogeneity and allowing us to focus on systematic effects. Using the assumption that the industries we selected are still a small fraction of the manufacturing base in a state, simultaneity is avoided (though not as convincingly as with a facility level analysis). We control for simple production related variables, such as the number of employees and the level of sales in the industry, allowing the hazardous waste generating function to be different across industries. State-level fixed effects are included to limit the consequences of persistent selectivity and correlated excluded variables. Stacking the industries to estimate the same proportionate effects across the industries yields no policy variables that are statistically significant. An increase in $1 per ton (on average about a 5% increase) in hazardous waste taxes leads to a 0.6 percent reduction in wastes generated. Although this is fairly precisely estimated with a standard error of 0.5 percent, it is both not statistically significant and not substantively important for policy purposes. None of the other policy variables showed results that were both precisely estimated and substantively important.
Facility Level Waste Generation Decisions. In our third group of models, we examine the quantity of hazardous wastes generated at the firm level. Our preferred model, similar to the previous two, would use fixed individual effects, allowing the heterogeneity across firms to be controlled by examining only within firm variation. This cannot be cleanly done for this situation. Because we exchanged censoring for sample truncation, we must estimate a panel data tobit model (keeping the truncated sample approach would have been far more difficult). We estimate this model two ways, a random effects tobit model in which firm effects are not correlated with either policy variables or firm level control variables and a fixed effects model in which the quantity of wastes generated is estimated by the facilities RCRIS status (small quantity or conditionally exempt). Both models yield similar results. The effect of increasing the tax index by $1 (or about 5%) is only a 0.1 to 0.2 percent reduction in the quantity of wastes generated. This effect is precisely estimated, though not statistically significant (or at least sensitive to model specification). This suggests again that hazardous waste taxes at the levels at which they are set are not a substantively useful policy instrument. Other policy parameters appear to have useful impacts. Strict liability appears to reduce the quantity of wastes generated by about 5 percent. The existence of mandatory pollution prevention programs reduces wastes generated by approximately 10 percent. As a group, we can reject hypotheses that the policy parameters play no role in the determination of wastes generated. Individually, because of the high degree of multicollinearity among these variables, we cannot identify their individual effects precisely.
Summary. Our models have made very reasonable econometric attempts to avoid several pitfalls that have plagued other studies. We use states or individual facilities as their own control and examine how variation in the policy over time leads to variation in response. This has the advantage of reducing bias and improving precision of our estimates. Our stacked models that assume that the effects are proportionate across industries leads to further efficiencies in estimation. Our attempts have lead to models that produce sensible, though not sensational, results. The impacts of individual policies are hard to identify because there are so many of them and there is a tendency for them to all be adopted around the same time (1994-1996). If hazardous waste policies work, their effects are quite small.
Journal Articles:
No journal articles submitted with this report: View all 7 publications for this projectSupplemental Keywords:
hazardous waste, environmental policy, policy implementation, taxes, fees, charges, policy instruments, empirical analysis, econometric, RCRA, audit privilege, liability, pollution prevention,, RFA, Economic, Social, & Behavioral Science Research Program, Scientific Discipline, Waste, Hazardous Waste, Corporate Performance, decision-making, Hazardous, Market mechanisms, Environmental History, Economics & Decision Making, Social Science, Environmental Law, hazardous waste management, environmental quality, market incentives, market-based mechanisms, policy instruments, effects of policy instruments, impact of state policy instruments, impact of federal policy instruments, policy making, policy incentives, technology-based regulation, incentives, decision making, environmental decision making, risk management, hazardous waste generation, environmental policy, pollution reduction, public policy, environmental behavior, behavior change, pollution preventionRelevant Websites:
http://www.spea.indiana.edu/kenricha/research.htm Exit
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.