the clean air act and volatile organic compounds: did

34
The Clean Air Act and volatile organic compounds: Did plants reduce their health-indexed air emissions or shift their emissions into other media? Shanti Gamper-Rabindran 1 Assistant Professor GSPIA University of Pittsburgh [email protected] Presented at the Maxwell School, Syracuse University April 2009 Abstract Do plants respond to medium-specific regulation of pollutants by reducing their emissions into that medium or by shifting their emissions into other media? I examine the impact of the U.S. Clean Air Act’s ground-level ozone regulations (CAA) on the chemical manufacturing sector’s health-indexed emissions of volatile organic compounds (VOC) that are reportable to the Toxic Release Inventory. I construct a database of 1,964 plants across the United States between 1988 and 2002. I estimate the CAA’s regulatory effects by comparing the changes in health-indexed emissions of plants in counties that are in non-attainment status for ground-level ozone, where the CAA regulation is stricter, with their counterparts in attainment counties. I also control for changes in plant-level output, using changes in employment as a proxy. The results indicate that the CAA reduced health-indexed VOC emissions into air and that the CAA did not increase emissions into water, onto land or to offsite transfers. JEL codes: Q53, Q58, I18, L65 Keywords: Clean Air Act, cross-media substitution, health-indexed emissions, volatile organic compounds 1 I thank Michael Greenstone for the VOC list, Patrick Conway, James Hamilton, John Mendeloff, David Popp and colleagues at the Environmental Protection Agency and ChemAlliance for helpful comments. Funding from the National Science Foundation BCS 0351058 is gratefully acknowledged. All errors are mine. 1

Upload: others

Post on 03-Jan-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The Clean Air Act and volatile organic compounds: Did

The Clean Air Act and volatile organic compounds: Did plants reduce their health-indexed air emissions

or shift their emissions into other media?

Shanti Gamper-Rabindran1 Assistant Professor

GSPIA University of Pittsburgh [email protected]

Presented at the Maxwell School, Syracuse University

April 2009

Abstract

Do plants respond to medium-specific regulation of pollutants by reducing their emissions into that medium or by shifting their emissions into other media? I examine the impact of the U.S. Clean Air Act’s ground-level ozone regulations (CAA) on the chemical manufacturing sector’s health-indexed emissions of volatile organic compounds (VOC) that are reportable to the Toxic Release Inventory. I construct a database of 1,964 plants across the United States between 1988 and 2002. I estimate the CAA’s regulatory effects by comparing the changes in health-indexed emissions of plants in counties that are in non-attainment status for ground-level ozone, where the CAA regulation is stricter, with their counterparts in attainment counties. I also control for changes in plant-level output, using changes in employment as a proxy. The results indicate that the CAA reduced health-indexed VOC emissions into air and that the CAA did not increase emissions into water, onto land or to offsite transfers.

JEL codes: Q53, Q58, I18, L65 Keywords: Clean Air Act, cross-media substitution, health-indexed emissions, volatile organic compounds

1 I thank Michael Greenstone for the VOC list, Patrick Conway, James Hamilton, John Mendeloff, David Popp and colleagues at the Environmental Protection Agency and ChemAlliance for helpful comments. Funding from the National Science Foundation BCS 0351058 is gratefully acknowledged. All errors are mine.

1

Page 2: The Clean Air Act and volatile organic compounds: Did

1 Introduction

The Environmental Protection Agency (EPA) has historically taken a single-medium

approach to regulating pollution emitted by the manufacturing sector. For example, the

Clean Air Act regulates emissions into air, while the Clean Water Act regulates emissions

into water. Despite its adoption of a few cross-media programs, 2 the EPA’s regulatory

and administrative framework constrains it to a traditional single-media approach to

regulating emissions from the manufacturing sector (Hahn and Males, 1990; Funke,

1993; Barrette, 1995; Ma and Crawford-Brown, 1997; McMahon, 2006). This approach

may lead to unintentional outcomes, i.e. plants may shift the release of their pollution

from the regulated medium into other media categories (Hendrickson and McMichael,

1985; Hahn and Males, 1990; Borys, Skarzinskas, and Green, 1996; Ma and Crawford-

Brown, 1997).

One stark example of pollutants shifting across media is the methyl tertiary-butyl ether

(MTBE) crisis in the 1990s. In response to air regulations, the petroleum industry

switched from tetra-ethyl lead to MBTE. The use of MTBE reduced emissions of lead

into the air and other smog-producing air pollutants. Unfortunately, the switch to MBTE,

alongside the problem of leaking underground storage tanks, caused groundwater

contamination and polluted the drinking water in several cities (Mcgarity, 2004). The

EPA failed to consider cross-media contamination issues in that case (Mcgarity, 2004).

Nevertheless, there is little systematic evidence on whether medium-specific regulation

has cause cross-media substitution, with Sigman (1996), Dombrowski (2000) and

Greenstone (2003) reporting contrasting results.

I study the Clean Air Act Amendments’ ground-level ozone regulation (CAA) for the

control of air emissions of non-methane volatile organic compounds (VOC).3 In

particular, I examine the impact of the CAA on the chemical manufacturing sector, which

produces VOC as a by-product of their manufacturing processes. VOC are of interest

2The EPA has taken cross-media approaches in its Persistent, Bio-accumulative and Toxic (PBT) Chemical Initiative (EPA, 2000) and the Mercury Initiative (EPA, 2006). 3 These ‘ground-level ozone’ regulations are different from those regulations governing emissions of substances that may damage the stratospheric ozone layer.

2

Page 3: The Clean Air Act and volatile organic compounds: Did

because they affect human health; indirectly, as precursors to ground-level ozone, and

directly, as toxins (WHO, 2006; Delfino et al, 2003; Delfino et al, 2003b). Plants can

either adopt an overall pollution reduction strategy that reduces their emissions to all

media, or adopt end-of-pipe abatement technologies that may result in the shifting of

emissions to other media. Small shifts in toxins across media can create sizable effects

(Hahn and Males, 1990). Operating alongside the CAA, the Toxic Release Inventory

program publicizes the ranking of plants according to reported pounds of emissions. This

focus on medium-specific pounds of emissions could potentially distract plants from the

goal of reducing the health impact of these pollutants. Therefore, the CAA raises two

policy questions. First, did the CAA reduce health-indexed VOC emissions into air?

Second, did the CAA cause plants to shift their health-indexed emissions of VOC from

air, which is targeted by the CAA, to other media categories?

I construct a database of plants in the chemical manufacturing sector, covering plants that

report both their emissions to the TRI and their employment to Dun & Bradstreet. I

extend two important previous studies (Greenstone, 2003; Sigman, 1996) in four ways.

First, I study the chemical manufacturing sector, which released 57% of the health-

indexed VOC emissions from the manufacturing industry in 1991.4 Second, I examine

the impact of the CAA on toxicity-weighted or health-indexed emissions. Health-index

emissions provide a better indication of the health-impact of emissions because chemicals

vary in their health-impacts

ed

5 (Bouwes and Hassur, 1997; Wright, 2007). Third, I control

for changes in output, which is an important determinant of emissions, using plant-level

changes in employment as a proxy, albeit imperfect. Fourth, in considering emissions

into non-air media, I extend my analysis beyond onsite releases, as in Greenstone (2003),

and consider offsite transfers.

Focusing on plants that report VOC emissions in at least 9 out of the 15 years between

1988 and 2001, I find that the CAA reduced health-indexed emissions into air by 1.4%

annually. Over the study period of 15 years, this figure translates to a sizable reduction of

4 My study is limited to VOC that reportable to the TRI since 1988.

3

Page 4: The Clean Air Act and volatile organic compounds: Did

21% of health-indexed emissions. I do not find evidence that the CAA increased health-

indexed VOC emissions into water or land or to offsite transfers. These results suggest

that the CAA ozone regulation resulted in genuine reductions in health-indexed emissions

into air.

2. The Clean Air Act and the chemical manufacturing sector

2.1 The Clean Air Act regulation

Under the Clean Air Act (CAA), the EPA sets national ambient standards for priority air

pollutants, such as ground-level ozone. The control of VOC is relevant for the attainment

of the national ozone ambient standards as VOC (along with oxides of nitrogen) form the

precursors to ozone (Henderson, 1996; Becker and Henderson, 2000). The EPA

designates counties whose ambient concentration exceeds a pollutant-specific national-

ambient-standard as non-attainment counties for that pollutant.

The CAA requires counties “that are in non-attainment to bring themselves into

attainment, or else, face federal penalties. The primary way of achieving attainment is

through the regulation of VOC-emitting … sources within one’s jurisdiction –

particularly manufacturing plants in certain industries. As a result, these plants in non-

attainment areas face much stricter environmental regulation than their counterparts in

attainment areas. For example, in non-attainment areas, plants with the potential to

pollute are subject to more stringent and more costly technological requirements on their

capital equipment” (Becker and Henderson, 1999). Table 1 highlights various aspects of

the stricter and costlier regulations in the non-attainment counties. Consistent with their

documentation that plants in non-attainment areas face stricter environmental regulations

than those in attainment areas, Becker and Henderson (1999) find that productions costs

are higher for plants in non-attainment areas.

“Both the states and the federal EPA are given substantial enforcement powers to ensure

that the CAAs’ statutes are met” (Chay and Greenstone, 2005). “Plants in non-attainment

5 A complete accounting of the health impacts would require an ecological modeling effort that is outside the scope of this paper.

4

Page 5: The Clean Air Act and volatile organic compounds: Did

areas face a greater likelihood of being inspected and fined than their counterparts in

attainment areas” (Becker and Henderson, 1999)

2.2 The Clean Air Act and cross-media considerations

The 1990 Clean Air Act for criteria air pollutants requires that the choice of the best

available control technology take into account “non-air quality health and environmental

impacts” (Hahn and Males, 1990). However, in practice, the technologies that have been

chosen are those that achieve the greatest reduction in air emissions without causing

major economic dislocation (Hahn and Males, 1990). In the late 1990s and in the 2000s,

the EPA began implementing Maximum Available Control Technologies for hazardous

air pollutants, which includes some VOC. In its consideration of these technologies, the

EPA has taken into account several cross-media issues (EPA, 2001). For example, the

2005 Final Rule for the Organic Chemical Industry specifies requirement for effluent

from air control devices (EPA, 2005).

2.3 Cross media substitution of VOC6

Plants in the chemical manufacturing sector emit VOC into the air as a by-product of

their manufacturing processes (Ackermann et al, 1998; Moretti, 2002; EPA, various

years). For example, in the organic chemical industry (SIC-286), VOC air emissions

include direct emissions from stacks or vents, fugitive emissions from tanks, valves and

mechanical seals, and emissions from cooling towers and wastewater treatment units. In

the plastics, resins and man-made fiber industry (SIC-282), VOC air emissions are

produced during the dry spinning process. In the pharmaceutical industry (SIC-283),

VOC emissions into air occur during chemical synthesis and natural product extraction.

In theory, plants may respond in two ways to regulations that limit VOC emissions.

Plants may adopt pollution reduction strategies, thereby, reducing overall emissions to all

categories. These strategies include product design changes, process modification, input

or material changes, in-process recycling and reuse and good operating practices, such as,

6 This section draws heavily from EPA’s sector notebooks (EPA, various years), Ackermann et al (1998) and Moretti (2002).

5

Page 6: The Clean Air Act and volatile organic compounds: Did

waste segregation and preventative maintenance (Moretti, 2002). Under these

circumstances, reduction in air emissions would be complementary with reductions in

emissions to other media categories.

Alternatively, plants may adopt end-of-pipe abatement technologies that reduce air

emissions, but these technologies may increase emissions to water, land or offsite

transfers for treatment, recycling or disposal. Under these circumstances, substitution

from air emissions into other media may occur. Several abatement technologies exist for

VOC control (Moretti, 2002). The input and by-products of the abatement process, such

as scrubbers, solvents, coolants and wastewater, need to be treated, disposed or recycled

(Moretti, 2002). First, the thermal oxidization process oxidizes the VOC, and after the

oxidization process, water or caustic7 scrubber is used to remove highly corrosive gases

e.g. hydrochloric acid. Second, absorption or scrubbing, employing solvents such as

water, mineral oils, or other non-volatile petroleum, is used to separate gaseous streams

using solvents such as water, mineral oils, or other non-volatile petroleum. The stripping

process is then used to re-generate the solvents, and VOC are desorbed from the solvents,

and recovered as a liquid in a condenser. The stripping process may create water disposal

problems (Moretti, 2002).

Two factors limit plants’ incentives to adopt air abatement technologies that shift

emissions into other medium such as water. First, the low solubility and the high vapor

pressure of some VOC limit the extent of substitution from a technical standpoint.

Second, several VOC that face the CAA regulations are also subject to water regulations,

limiting the incentive for plants’ to shift their emissions into water. Out of the 147 VOC

in my study, only 87 are not listed as chemicals subject to water regulations, i.e., they are

not listed in the Priority Pollutant List (PPL) and in the Safe Drinking Water Act

(SDWA).

In my analysis across media, I examine emissions into water and onto land. I also

examine the category of offsite transfers. While proper recycling serves as a key waste

7 These are hydroxides of alkali and alkaline earth metals, such as sodium, potassium, calcium, and barium.

6

Page 7: The Clean Air Act and volatile organic compounds: Did

management tool, improper recycling and disposal activities have raised concerns of a

‘toxic shell game’ (Office of Technology Assessment, 1988; Orum, 1991).

2.4 Limitations in the detection of cross-media substitution

My study can detect cross-media substitution if plants shift from TRI-reportable VOC air

emissions to TRI-reportable VOC emissions into other media. My study can also detect,

albeit with much less power, if plants switch from TRI-reportable VOC emissions to

TRI-reportable non-VOC emissions into other media. The abatement process for air VOC

may employ non-VOC scrubbers or solvents or may produce non-VOC by-products

(Moretti, 2002). If shifts across media occur, I expect in emissions of VOC into air but

an increase in emissions into other media of chemicals including non-VOC. As an

imperfect way to examine the possibility of increases in emissions of other chemicals, I

examine the impact of CAA on all TRI-reportable chemicals. Unfortunately, this analysis

is biased against finding substitution, as the non-VOC scrubbers, solvents or by-products

are likely to be a subset of the TRI chemicals.

My study cannot detect if plants substitute from VOC air emissions to releasing non-TRI-

reportable chemicals into other media. Indeed, TRI-reportable chemicals and non-TRI

reportable chemicals can serve as substitutes (Crumpler, 1996). For example, in the wood

products industry, TRI-reportable paint solvents are used (Crumpler, 1996). In addition,

“non-toxic solvents such as mineral spirits [that] are not reported on the TRI are used in

coating and gluing operations. [The latter] are significant as volatile organic compounds

(VOC) that are photo-reactive and contribute to ground-level ozone pollution”

(Crumpler, 1996).

3. Estimation strategy

As in Greenstone (2003), I compare changes in emissions from plants in counties that are

in non-attainment status for ozone ('treatment' regions) relative to changes in emissions

from plants in counties that are in attainment status for ozone ('control' regions). Previous

studies, summarized in Table 2, have documented that CAA regulations are more

stringent in non-attainment counties than attainment counties, and have implemented this

7

Page 8: The Clean Air Act and volatile organic compounds: Did

strategy of comparing non-attainment and attainment counties in order to estimate the

regulatory impact of the CAA.

The strength of this “difference-in-difference” (DID) approach is that the estimate is not

biased by events that affect both plants in the attainment region and the non-attainment

region, for example, a recession that affected the entire US. However, the limitation in

this approach is that it cannot control for confounding factors that may have had

differential impacts on plants in non-attainment and attainment counties.8

3.1 Estimation strategy - reviewing the assumption

In my estimation strategy, I have assumed that if any impacts are detected, the direction

of causality is from the non-attainment status, through the regulatory effect described

above, to the reductions in emissions at the plant-level in the chemical manufacturing

sector. This assumption would be defensible if the direction of causality were not the

reverse, i.e. the changes in plant-level emissions in the chemical manufacturing sector did

not influence the county-level non-attainment or attainment status. As described below,

the chemical manufacturing sector contributes only a small fraction of the VOC

emissions at the county-level, and therefore, it is unlikely that the changes in the

emissions at the plant-level influence the county-level attainment or non-attainment

status.

VOC is produced by manufacturing plants, road emissions and non-road emissions, such

as agricultural activity (EPA, 1999). The National Emissions Inventory (NEI) provides

data on air toxins from all three sources (EPA, 1999). Out of the 146 VOC chemicals in

my study, the NEI has compiled data for 124 chemicals. The NEI database provides

useful information on the contribution of the chemical plants in my study to the county-

level VOC emissions. The sample of plants in the NEI is likely to be comparable to the

sample of plants in my database. As the NEI contains plants that are major sources under

the Clean Air Act regulation, the NEI is likely to contain the larger plants in the chemical

8

Page 9: The Clean Air Act and volatile organic compounds: Did

manufacturing industry. Similarly, as larger plants are more likely to report to the D&B,

my database, which contains the TRI-reporting plants from the chemical manufacturing

industry that are successfully linked to the D&B sample of plants, contains the larger

plants in that industry. It is the larger plants are more likely to report to the D&B.

Table 3 tabulates the contribution of the chemical manufacturing plants to the county-

level health-indexed VOC emissions, as calculated from the 1999 NEI.9 The sample of

counties, described below, is restricted to those counties with at least one chemical

manufacturing plant. Table 3 shows that the chemical manufacturing plants contribute

only a small fraction of the VOC emissions at the county-level, even for those counties

which are at the 90th percentile in their share of VOC emissions from the chemical

manufacturing plant. For the counties at the median and the 75th percentile, chemical

manufacturing plants contribute only 0.05% and 0.8% of the county-level health-indexed

VOC emissions, respectively. Even for the counties at the 90th and 95th percentile, the

contributions of the chemical manufacturing plants are only 6% and 16% respectively.

Only in twelve counties at the extreme upper tail of the distribution, i.e., at and above the

99th percentile, do these contributions rise to 45%.

3.2 Estimation model

My estimation model is based on Greenstone (2003), with the following modification. I

include two additional control variables to the model: the percentage change in

employment (to proxy for output) and the demographic characteristics of the census tract

where each plant is located. These demographic factors are correlated to plant-level

emissions (Arora and Cason, 1999).

Observations are for plant k in county c at time t. The program impact is measured by the

difference in the percentage changes in emissions between plants located in non-

attainment counties and plants located in attainment counties.

8E.g., the DID approach cannot address recessions that hit the non-attainment counties but that spare the attainment counties. The fairly wide geographical variation in the location of plants may reduce this potential estimation issue. 9 The 1993 and 1996 NEI compiled data on only 33 chemicals (EPA, 2004).

9

Page 10: The Clean Air Act and volatile organic compounds: Did

% ∆Y k c t = [ (Yk c t – Y k c t-1 ) ÷ ( Y k c t + Y k c t-1) /2 ] --- Equation 1

= β1 Z c t-1 + β2 % ∆ L k c t + β3 T t + β4 S k + β5 X k + e k c t

where ∆ Y k c t is a measure of the percentage change in emissions of pollutant i for plant

k in county c between time t and t-1, ∆ L k c t is the percentage change in employment for

plant k in county c between time t and t-1, Z is a binary variable that takes the value 1 for

county c that is in non-attainment status in time t-1, and takes the value 0 if that county is

in attainment status at that time, T are time dummies, S are dummies at the 3-digit SIC

code, and X is a vector of the census tract characteristics where the plant is located. The

coefficient of interest, β1, measures the annual percentage change in plant-level

emissions in non-attainment counties relative to those in attainment counties. I allow the

error to be correlated for the same plant across time. The time period t is between 1989

and 2002.10 The year dummies controls for year specific variations in emissions. For

example, the year dummies can take into account the fact that offsite transfers to

recycling, treatment and energy recovery became reportable only in 1991.

The ideal analysis would control for fluctuations in emissions that are driven by changes

in output. Unfortunately, output data is not publicly available, forcing numerous studies

(e.g. Khanna and Damon, 1999; Greenstone, 2003) to study reductions in emissions

without controlling for output. I use plant-level employment to control for output. This

variable is an imperfect control for output. As an illustration, consider the following case.

For simplicity, assume that the ratio of health-indexed emissions to output is constant.

Consider the case when: (a) plant-level output in non-attainment counties are

systematically declining relative to that in attainment counties, and (b) the ratio of

employees to output is declining. Because the percentage decline in plant-level

employment does not fully account for the actual percentage decline in plant-level output

in non-attainment counties relative to attainment counties, my study would overstate the

CAA’s impact on reducing plant-level emissions. Nevertheless, some types of

manufacturing industries, such as the iron and steel industry, are more susceptible to this

10

Page 11: The Clean Air Act and volatile organic compounds: Did

potential estimation bias. The iron and steel industry has experienced a systematic decline

in the Rust Belt counties, which are also the counties that had been in non-attainment for

criteria air pollutants (Kahn, 1999). In contrast, the geographical variation in the location

of plants in the chemical manufacturing sector serves to mitigate this estimation issue.

4. Data

My sample is restricted to plants in the chemical manufacturing sector that report their

emissions to the TRI and whose plant-level employment is reported in the Dun and

Bradstreet (D&B) database.11 Specifically, I focus on those plants that report VOC

emissions for at least nine years out of the possible fifteen years between 1988 and 2002.

The sample size is about 20,000 observations, representing 1,964 plants, between 1988

and 2002. I omit the first year of the TRI, 1987, as plants’ lack of familiarity with TRI

make that year’s data less reliable (Levinson, 2001).

I use Greenstone’s (2003) list of VOC with two modifications. A Ph.D. chemist had

produced Greenstone’s (2003) list by determining which of the chemicals reportable to

the TRI since 1988 are VOC. I shortened the VOC list in two ways. First, I omit three

chemicals that are considered non-photo-reactive by the EPA and therefore these

chemicals are not subject to the CAA regulations. Second, out of these 171 VOC

chemicals, I focus on the 146 chemicals that have been assigned oral or inhalation

toxicity-weights, in order to create the measure of health-indexed emissions.12 13

10 The first year of the analysis is for changes between 1988 and 1989. 11 As the D&B database contains employment data for larger plants, my sample is skewed towards the larger TRI plants. 12 The 18 VOC chemicals that have not been assigned oral or inhalation toxicity weights comprise a tiny fraction of the VOC non-health-indexed emissions. The efforts to assign toxicity weights have been concentrated on chemicals perceived as more harmful. These 17 chemicals comprise 1% of non-health-indexed VOC emissions in the US manufacturing sector. 13: Greenstone’s (2003) shortened list contains 171 chemicals. Among these 171 chemicals, 146 have toxicity weights. [25 do not have both oral and inhalation toxicity]. Out of the 146 VOC with tox weights, 124 of them are in the NEI. Out of the 171 VOC, 135 are in the NEI.

11

Page 12: The Clean Air Act and volatile organic compounds: Did

The EPA’s TRI data provides emissions information by chemical, by media, by plant and

by year.14 Plants that emit chemicals beyond a threshold self-report their emissions

(Hamilton, 2005). Most plants estimate their emissions using emission factors related to

their production processes while few plants monitor their emissions directly (Hamilton,

2005). The TRI is the only source for plant-level emissions of numerous toxic chemicals

(Hamilton, 2005). Several studies (Khanna and Damon, 1999; Greenstone, 2003;

Gamper-Rabindran, 2006; Morello-Frosch and Jesdale, 2006; Bui, 2005) have relied on

the TRI plant-level emissions. The EPA has taken several steps to improve the accuracy

of the TRI data, though limitations still exist.15 The EPA provides technical assistance to

plants on estimating their emissions.16 The Emergency Planning and Community Right-

to-Know Act, the legislation which established the TRI, authorizes the EPA to penalize

plants that fail to report their emissions and that fail to report their correct level of

emissions (EPA, 2001).17 The EPA has taken enforcement actions against some of these

plants.18

The chemical-specific toxicity weights are from the Risk Screening Environmental

Indicator (RSEI) (EPA, 2004). Emissions to air are weighted using inhalation toxicity

weights. Emissions to other media are weighted using oral toxicity weights. Using

Chemical Abstracts Service (CAS) numbers, I link the list of VOC with the chemical-

level emissions data in the TRI. Filtering for TRI chemicals that are VOC, I then

calculate health-indexed VOC emissions. I use analogous filtering techniques to calculate

health-indexed emissions for the subset of VOC that are not on the Priority Pollutant List

(PPL) and the Safe Drinking Water Act (SDWA).

14 I use the TRI data assembled by the Risk Screening Environmental Indicators (RSEI) project. Among advantages of this database is that the EPA has geo-coded the latitude and longitude of plants, ensuring more accurate linkage of the plants to their census tract locations (EPA, 2004). 15 Marchi and Hamilton (2006) report that large drops in plant-level TRI reporting for lead and nitric acid are not matched with corresponding drops in measured concentrations from EPA monitors. 16 The EPA’s sector guidance documents contain emission factors for typical processes for that particular industry. Plants can use these emission factors to estimate chemical releases (Marchi and Hamilton, 2006). The EPA has also undertaken specific projects to improve the TRI data quality, such as the National Nitrate Compliance Initiative, in which plants undertook voluntary audits of their TRI reporting (EPA, 2002). 17 The EPA is authorized to issue Civil Administrative Complaints against “data quality errors such as the failure to provide reasonable estimates of releases and off-site transfers” (EPA, 2001). 18 Data quality errors, such the failure to report or the failure to provide reasonable estimates, can result in penalties as high as $27,500 per day (EPA, 2001).

12

Page 13: The Clean Air Act and volatile organic compounds: Did

The D&B database provides annual plant-level employment data. The Code of Federal

Regulations provides the lists of counties that are in non-attainment for ozone by year.19

The 1990 Decennial Census provides demographic information of the census tract

surrounding each plant. I use Geographical Information System overlay tools to link

plants, using their latitude and longitude, to their census tracts.

5. Summary statistics

Table 4 shows the number and the percentage of counties that are in non-attainment or

attainment status for ground-level ozone in the period 1989-2002. Both the number and

the percentage of counties that are in non-attainment status for ground-level ozone have

declined, despite a temporary increase from 1992-1994. In the earlier period my study,

about 36% of counties were in non-attainment, but by the later period of my study, only

25-26% of the counties were in non-attainment. Table 5 tabulates the summary statistics.

Graph 1 shows the trends of health-indexed VOC emissions into air, onsite emissions and

offsite transfers. While health-indexed VOC emissions into air have gradually declined,

health-indexed VOC emissions onsite (other than air) and offsite transfers have not

declined. While onsite emissions show significant fluctuations, they have increased over

time.

Graph 2 provides a closer look at the various components of onsite health-indexed

emissions. Again, while the air component of onsite health-indexed emissions have

shown a gradual decline, emissions into underground injection and onto land has

fluctuated and not declined. The health-indexed emissions into water, however, show a

decline beginning in 1996. Graph 3 provides a closer look at the various components of

the health-indexed emissions to offsite media categories. Offsite transfers for recycling,

treatment and energy recovery became reportable in 1991. Despite a dip in 2000, offsite

transfers have increased between 1991 and 2002. While most categories for offsite

13

Page 14: The Clean Air Act and volatile organic compounds: Did

transfer do not show a clear pattern, offsite transfers for recycling have grown between

1991 and 2002, with a dip in 2000.

6. Regression results

6.1 Health-indexed VOC emissions into air

To recall, in my regression analysis, the coefficient of interest, β1, measures the

regulatory effect of the CAA. Specifically, the coefficient measures the annual

percentage change in plant-level emissions in non-attainment counties relative to those in

attainment counties. The regression results for health-indexed air emissions for the

chemical manufacturing sector are presented in Table 6. Results from the sparser

regression model, with the non-attainment variable, employment and SIC and year

dummies, are presented in column 1. These results suggest that the non-attainment status

reduces plant-level health-indexed emissions by 1.5% annually. Results from the full

regression model, which adds the census tract characteristics where the plants are located,

presented in column 2, are similar. These results suggest that CAA reduced health-

indexed VOC emissions into air by 1.4% annually. Over the study period of 15 years, this

translates to a sizable reduction of 21% of health-indexed emissions. In both models, a

1% change in plant-level employment is positively associated with a 6.5% change in

plant-level health-indexed emissions.

6.2 Health-indexed VOC emissions into other media

The regression results for health-indexed emissions into other media are presented in

Table 7. The results for emissions into water are presented in column 1. The results for

emissions offsite (which include offsite transfers for disposal and treatment) are

presented in column 2. The results for emissions into onsite categories except air (i.e. to

water, land, and underground injection) are presented in column 3. The results for

emissions into all media categories other than air are presented in column 4. In all of

these regression models, the coefficient on the non-attainment variable is not statistically

different from zero. Therefore, I conclude that there is no evidence that the CAA caused

19 As in Greenstone (2003) I consider a county to be in non-attainment status for ozone if it is listed as nonattainment for either ozone or nitrogen dioxide. Chemical processes that involves nitrogen dioxide and VOC create ozone.

14

Page 15: The Clean Air Act and volatile organic compounds: Did

plants to increase their health-indexed emissions into other media. In most of these

regression models, the changes in plant-level employment are positively correlated with

changes in health-indexed emissions. For two of the regressions models (column 1 and

2), a higher proportion the population being white in the census tract where the plants are

located is associated with a decline in health-indexed emissions.

6.3 VOC unregulated under SDWA and PPL

Plants are more likely to shift their emissions of VOC from air to water for those VOC

that are unregulated by water regulations. Therefore, I examine the impact of the CAA on

health-indexed emissions of VOC that are not listed under SDWA and PPL. The results

are presented in Table 8. The coefficient for the percent change in employment is positive

and statistically significant as expected. Unfortunately, the impact of the non-attainment

status on the health-indexed emissions of VOC to air and water are not precisely

estimated for this subset of VOC.

6.4 TRI chemicals

Plants may shift from emitting VOC into the air to emitting other TRI-reportable

chemicals. For example, the solvents used to adsorb the VOC during the abatement

process may be non-VOC TRI-reportable chemicals.20 Therefore, I examine the impact

of CAA on health-indexed emissions of TRI chemicals that are reportable since 1987.

The limitation in this analysis is that the relevant solvents make up only a fraction of the

TRI emissions and thus the analysis is biased against finding evidence for cross-me

substitution. The results are presented in Table 9. The results in column 1 indicate that

the CAA reduced health-indexed TRI air emissions. Based on the results in column 4, I

conclude that there is no evidence to indicate that the CAA increased health-indexed TRI

emissions to the non-air media categories. In both regression models, I find that changes

in employment are related to changes in emissions.

dia

6.5 Scrubbers or solvents and by-products of VOC abatement – in progress

20 This study cannot detect switches to non-TRI-reportable solvents.

15

Page 16: The Clean Air Act and volatile organic compounds: Did

I plan to examine the subset in the TRI list of chemicals that (1) are typical scrubbers or

solvents for VOC or (2) are typical chemical by-products during the abatement processes

for air VOC emissions. Discussions are ongoing with chemical engineers in EPA and

ChemAlliance to identify a relevant list of chemicals. Hydroxides of alkali and alkaline

earth metals such as sodium, potassium, calcium and barium have been used as scrubbers

in the abatement process for VOC air emissions. Hydrochloric acid is as one by-product

of the VOC abatement process, particularly in the oxidation of VOC (Moretti, 2002).

7. Conclusion and policy implication

This study on the chemical manufacturing sector asks whether the CAA ozone

regulations reduced plant-level health-indexed air emissions, or whether the CAA caused

plants to shift their emissions to other media. The results indicate that these regulations

successfully reduced plant-level health-indexed emissions into air by 1.4% annually.

Over the study period of 15 years, this translates to a sizable reduction of 21% of health-

indexed emissions. I also do not find evidence that the CAA ozone regulations caused an

increase in the release of health-indexed VOC into other media categories. My study

provides complementary results to Greenstone’s (2003) finding that the CAA reduced

plant-level pounds of emissions into air, but did not increase those emissions to water or

land in the iron and steel industry between 1987 and 1997.

These results have two policy implications. First, the TRI, which operates alongside the

CAA and publicizes the ranking of plants based on their pounds of emissions, may create

an incentive for plants to focus on reducing their pounds of emissions instead of on

reducing their health-indexed emissions. These results indicate that, in practice, in

response to the CAA, plants have indeed reduced their health-indexed air emissions.

Second, in theory, in response to medium-specific regulations, plants may substitute their

emission from the regulated media to other media categories. These results indicate that,

in practice, for the CAA ozone regulations, plants have truly reduced their health-indexed

emissions and not simply shifted their emissions to non-air media categories.

16

Page 17: The Clean Air Act and volatile organic compounds: Did

17

This study has two limitations. My study finds that there is no evidence to support cross-

media substitution among TRI reportable chemicals. The study would have provided

stronger evidence for or against cross-media substitution if the results for health-indexed

emissions into other media had been statistically significant from zero. For example, if

the estimated coefficients for regressions on health-indexed emission had been negative

and statistically different from zero, one could proceed the following way. One could

apply a stronger test against cross-substitution and one could then draw the conclusion

against cross-media substitution if the CAA had reduced health-indexed emissions into

other media by a greater percentage than it had reduced health-indexed emissions.

Second, in reality, plants may shift from emitting TRI reportable VOC into the air to

releasing non-TRI reportable chemicals to other media. However, my study detects cross-

media substitution only if plants switch from TRI-reportable air VOC emissions to

emitting TRI-reportable VOC or TRI-reportable chemicals into other media.

In the future, I plan to discuss in greater detail the end-of-pipe pollution abatement

processes in the chemical industry with chemical engineers at the EPA and in the

chemical industry and at ChemAlliance. I plan to improve the study by examining

specifically TRI-reportable chemicals that serve as typical solvents for VOC or that are

emitted as typical by-products of the VOC abatement processes.

Page 18: The Clean Air Act and volatile organic compounds: Did

Bibliography R. Ackermann, G. Hughes, D. Hanrahan, A. Somani, S. Aggarwal, A. FitzGerald, Pollution Prevention and Abatement Handbook 1998, The World Bank, Washington, D.C, 1999. S. Arora, T. Cason, Do Community Characteristics Influence Environmental Outcomes? Evidence from the Toxics Release Inventory, Southern Economic Journal, 65 (1999) 691-716. M. D. Barrette, Design and Implementation of Multimedia and Sector Strategies at EPA, Journal of Environmental Regulation, 5 (1) (1995) 1-15. R.A. Becker, J.V. Henderson, Effects of Air Quality Regulations on Polluting Industries, Journal of Political Economy, 108 (2000) 379-421. R.A. Becker, J.V. Henderson, Costs of Air Quality Regulation, National Bureau of Economic Research Working Paper No. W7308, 1999. R.A. Becker, J.V. Henderson, Costs of Air Quality Regulation, in C. Carraro and G.E. Metcalf (Eds.), Behavioral and Distributional Effects of Environmental Policy, Chicago, The University of Chicago Press, 2001, 159-82. D. S. Borys, R. D. Skarzinskas, and L. D. Green, Assessing Cross-Media Impacts: A Comparative Risk Approach, Risk Analysis, 16 (5) (1996) 683 – 692. N.W. Bouwes, S.M. Hassur, Toxic Release Inventory: Relative Risk-Based Environmental Indicators: Interim Toxicity Weighting Summary Document, Economics Exposure and Technology Division, US Environmental Protection Agency, 1997. L.T.M. Bui, Public Disclosure of Private Information as a Tool for Regulating Environmental Emissions: Firm-Level Responses by Petroleum Refineries to the Toxic Release Inventory, US Census Bureau Center for Economic Studies CES 05-13, 2005. K.Y. Chay, M. Greenstone, Air Quality, Infant Mortality, and the Clean Air Act of 1970, National Bureau of Economic Research, October 2003. K.Y Chay, M. Greenstone, Does Air Quality Matter? Evidence from the Housing Market, Journal of Political Economy, 113 (2) (2005) 376-424. P. Crumpler, Sector Assessments: An Analysis of Pollution Prevention Opportunities and Impediments in the Wood Products Manufacturing Sector in Georgia, Pollution Prevention Engineer, Pollution Prevention Assistance Division, Georgia Dept of Natural Resources, 1996.

Page 19: The Clean Air Act and volatile organic compounds: Did

J.M. Dombrowski, Cross-Media Transfers of Pollution and Risk, Ph.D diss., University of Massachusetts Amherst, 2000. S. de Marchi, J.T. Hamilton, Assessing the Accuracy of Self-Reported Data: An Evaluation of the Toxics Release Inventory, Journal of Risk and Uncertainty, 32 (2006) 57-76. R.J. Delfino, et al., Asthma Symptoms in Hispanic Children and Daily Ambient Exposures to Toxic and Criteria Air Pollutants, Environmental Health Perspectives 111 (2003) 647-656 R.J. Delfino, et al., Respiratory Symptoms and Peak Expiatory Flow in Children with Asthma in Relation to Volatile Organic Compounds in Exhaled Breath and Ambient Air, Journal of Exposure Analysis and Environmental Epidemiology 13 (2003) 348-363. L. Fernandez, The Case For Cross-Media Environmental Policy, Contemporary Economic Policy, 23(4) (2005) 555-567. O. Funke, Struggling with Integrated Environmental Policy: The EPA Experience, Policy Studies Review, 12 (3/4) (1993) 137-161. General Accounting Office, Clean Air Act: EPA Should Improve the Management of its Air Toxics Program, GAO-06-669, 2006. M. Greenstone, Estimating Regulation-Induced Substitution: The Effect of the Clean Air Act on Water and Ground Pollution, American Economic Review Papers and Proceedings, 93 (2) (2003) 442-448. M. Greenstone, The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufactures, The Journal of Political Economy, 110 (6) (2002) 1175-1219. R.W. Hahn, E.H Males, Can Regulatory Institutions Cope with Cross Media Pollution? Journal of the Air & Waste Management Association, 40 (1) (1990) 24-31. J.T. Hamilton, Regulation Through Revelation: The Origin and Impacts of the Toxics Release Inventory Program, Cambridge University Press, New York, 2005. J.V. Henderson, Effects of Air Quality Regulation, American Economic Review, 86 (1996) 789-813. C.T. Hendrickson, F.C. McMichael, Controlling Contradictions among Regulations, American Economic Review, 75 (4) (1985) 876-877. M.E. Kahn, The Silver Lining of Rust Belt Manufacturing Decline, Journal of Urban Economics, Elsevier, vol. 46(3) (1999) 360-376.

Page 20: The Clean Air Act and volatile organic compounds: Did

M. Khanna, L.A. Damon, EPA’s Voluntary 33/50 Program: Impact on Toxic Releases and Economic Performance of Firms, Journal of Environmental Economics and Management, 37 (1) (1999) 1–25. A. Levinson, An Industry-Adjusted Index of State Environmental Compliance Costs, in: G. Metcalf, C. Carrero (Eds.), Behavioral and Distributional Effects of Environmental Policy, University of Chicago Press, Chicago, 2001. H. Ma and D.J. Crawford-Brown, Comparison of Systems and Single-Medium Approaches in Environmental Risk-Based Decision Making: A Case Study of a Sludge Management Problem, Journal of Environmental Systems, 26 (3) (1997-1998) 215-247. T.O. McGarity, MTBE: A Precautionary Tale, Harvard Environmental Law Review, 28 (2) (2004) 281-342. R. McMahon, The Environmental Protection Agency: Structuring Motivation in a Green Bureaucracy: The Conflict Between Regulatory Style and Cultural Identity, Sussex Academic Press, Portland, 2006. R. Morello-Frosch, B. Jesdale, Separate and Unequal: Residential Segregation and Air Quality in the Metropolitan U.S., Environmental Health Perspectives, 113 (2006) 386-393. E.C. Moretti, Practical Solutions for Reducing Volatile Organic Compounds and Hazardous Air Pollutants, American Institute of Chemical Engineers, 2001. E.C. Moretti, Reduce VOC and HAP Emissions, CEP Magazine (June 2002) 30-40. National Emission Standards for Hazardous Air Pollutants: Miscellaneous Organic Chemical Manufacturing; Final Rule and Proposed Rule, 40 C.F.R § 63 (2007). Office of Technology Assessment, Are We Cleaning Up? 10 Superfund Case Studies, Special Report (OTA-ITE-362), US Government Printing Office, Washington, DC, 1988. P. Orum, Recycling Loophole in the Toxics Release Inventory: Out of Site, Out of Mind: Working Notes on Community Right to Know: A Working Paper on Our Right-to-Know about Toxic Pollution, The Newsletter of the Working Group on Community Right-to-Know, February/March 1991. D. Popp, Pollution Control Innovations and the Clean Air Act of 1990, National Bureau of Economic Research, Working Paper 8593, 2001. H. Sigman, Cross-Media Pollution: Responses to Restrictions on Chlorinated Solvent Releases, Land Economics, 72 (1996) 298–312. U.S. Clean Air Act of 1990, 42 U.S.C §7412 (2006).

Page 21: The Clean Air Act and volatile organic compounds: Did

US Environmental Protection Agency, National Emission Inventory Documentation and Data - Final Version 3.0, 1999. US Environmental Protection Agency, Enforcement Response Policy for Section 313 of The Emergency Planning Community Right-to-Know Act (1986) and Section 6607 of the Pollution Prevention Act (1990), Office of Compliance Monitoring and Office of Prevention, Pesticides, and Toxic Substances, 2001. US Environmental Protection Agency, Hazardous Air Pollutant Emissions for Miscellaneous Organic Chemical Manufacturing: Supplementary Information Document for Proposed Standards, Emission Standards Division, 2001. US Environmental Protection Agency, PBT Program Accomplishments: 2000, 2001. US Environmental Protection Agency, The National Nitrate Compliance Initiative, Office of Enforcement and Compliance Assurance, 2002. US Environmental Protection Agency, National Monitoring Strategy Air Toxics Component, Office of Air and Radiation, 2004. US Environmental Protection Agency, User’s Manual for RSEI Version 2.1.2 (1988-2002 TRI Data) Economics, Exposure, and Technology Division, Office of Pollution Prevention and Toxics, 2004. US Environmental Protection Agency, Changes to the TRI List of Toxic Chemicals, 2005. US Environmental Protection Agency, EPA’s Roadmap for Mercury, 2006. US Environmental Protection Agency, Sector Notebook: Profile of the Inorganic Chemical Industry, 1995. US Environmental Protection Agency, Sector Notebook: Profile of the Plastic Resins and Man-made Fibers Industry, 1997. US Environmental Protection Agency, Sector Notebook: Profile of the Pharmaceutical Industry, 1997. US Environmental Protection Agency, Sector Notebook: Profile of the Organic Chemical Industry 2nd Edition, 2002. US Environmental Protection Agency, Sector Notebook: Profile of the Rubber and Plastics Industry, 2nd Edition, 2005.

Page 22: The Clean Air Act and volatile organic compounds: Did

M. Walls, K. Palmer, Upstream Pollution, Downstream Waste Disposal, and the Design of Comprehensive Environmental Policies, Journal of Environmental Economics and Management, 41 (1) (2001) 94-108. World Health Organization, WHO Air Quality Guidelines for Particulate Matter, Ozone, Nitrogen Dioxide and Sulfur Dioxide: Global Update 2005: Summary of Risk Assessment, World Health Organization, Geneva, 2006. D.W. Wright, Toxicity-Weighting: A Prioritization Tool for Quality Assurance of Air Toxics Inventories, Paper presented at the US Environmental Protection Agency’s 16th Annual International Emission Inventory Conference, May 2007.

Page 23: The Clean Air Act and volatile organic compounds: Did

Table 1. Comparison of the more stringent regulations in non-attainment counties and the less stringent regulations in attainment countiesStatus of county In Attainment In Non-attainmentOzone concentrationa Below the National Ambient Air Quality Standards Below the National Ambient Air Quality StandardsState Implementation SIPS have to specify details on how states will bring theirPlans (SIP)a submittted by violating counties into attainment.e The SIP outlines the the State to the Fed. EPA detailed regulatory requirements for each plant.a

Technical requirements imposed on plants' capital equipment for new plants and existing plants that are undertaking major expansion* plants emit < 100 Plants face no specific technological requirementsa Plants are subject to Lowest Achievable Emission Rates tonnes of criteria and are "essentially unregulated".d (LAER) standards.b

air pollutants. New plants are required to purchase pollution offsets fromexisting plants.e

* plants emit > 100 Plants are required to install the Best Available Control Plants are subject to Lowest Achievable Emission Rates tonnes of criteria Technology (BACT), which is negotiated on a case-by-case (LAER) standards.b

air pollutants. basis and which is sensitive to cost considerations. a New plants are required to purchase pollution offsets fromexisting plants.eLAER standards required the installation of the “cleanest” available technology that is in use in any state.The costs of such technology can be considered onlyif the costs were so high that they would prohibit the new source from operating at all.b

Technical requirements imposed on plants' capital equipment for existing plants Plants face no specific technological requirementsa Plants that are grandfathered from the LAER standards,

and are "essentially unregulated".d are required to install the Reasonably Available Control Technology (RACT), which requires retrofitting, and which takes into account the economic burden imposed

on the plants.a

State impose emission limits on plants d

EPA inspections Rates of inspection and penalties are lower Rates of inspection and penalties imposed are higher(a) Becker and Henderson (1999), Costs of Air Quality Regulations NBER; (b) Popp (2001); (c) EPA’s On-Line SIP Processing Manual, “The On-line State Implementation Plan Processing Manual”; (d) Greenstone and Chay (2000) JPE; (e) Becker and Henderson (2000) JPE

Page 24: The Clean Air Act and volatile organic compounds: Did

Table 2. Studies that use the county-level nonattainment/attainment status to estimate the impacts of the more stringent Clean Air Act regulations in the non-attainment counties relative to attainment counties.(1) Henderson (1996) Counties that are in non-attainment for ground-level ozone experience greater reductions

in ground-level ozone than those counties that are in attainment.(2) Becker and Henderson In a study of industries that are highly polluting in VOCs, they find counties that are

(2000) in non-attainment for ground-level ozone experience plant births that are lower by 26-45%.(3) Becker and Henderson In a study of industries that are highly polluting in VOCs, they find counties that are

(1999) in non-attainment for ground-level ozone experience greater production costs.(4) Greenstone Plants in the iron and steel industries located in counties that are in non-attainment

(2003) for ozone experienced greater reductions in their pounds of VOC emitted into air than did comparable plants in attainment counties.

(5) Chay and Greenstone Non-attainment counties experience greater reductions in the level of total suspended (2003) particulate matter and infant mortality rates than attainment counties.

(6) Chay and Greenstone Non-attainment counties experience greater reductions in the level of total suspended (2005) particulate matter and greater increases in county-level housing prices than attainment counties.

Page 25: The Clean Air Act and volatile organic compounds: Did

Table 3 The contribution of chemical manufacturing sector to the county-level health-indexed VOC emissions.Panel A: The contribution of the chemical manufacturing sector to the total health-indexed VOC emissions at the county-level.In counties with values at the: 50th percentile 0.05% 75th percentile 0.8% 90th percentile 6% 95th percentile 16% 99th percentile 45%Panel B: The contribution of the chemical manufacturing sector to the health-indexed emissions from point sources at the county-levelIn counties with values at the: 50th percentile 0.02% 75th percentile 1.7% 90th percentile 12% 95th percentile 32% 99th percentile 65%Notes: This analysis is restricted to counties with at least one chemical manufacturing plant.

Page 26: The Clean Air Act and volatile organic compounds: Did

Table 4: Percentage counties that are in attainment and non-attainment for ground-level ozone between 1989 and 2002Year 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002# attainment counties 312 327 340 319 338 348 357 389 422 439 447 448 437 442# non-attainment counties 176 182 186 235 233 233 226 204 168 160 151 159 155 151Total counties 488 509 526 554 571 581 583 593 590 599 598 607 592 593% counties in attainment 64 64 65 58 59 60 61 66 72 73 75 74 74 75% counties in non-attainment 36 36 35 42 41 40 39 34 28 27 25 26 26 25 Notes: During the latter period of our study, we have more plants in the sample and these plants are spread out in a larger number of counties.

Page 27: The Clean Air Act and volatile organic compounds: Did
Page 28: The Clean Air Act and volatile organic compounds: Did
Page 29: The Clean Air Act and volatile organic compounds: Did
Page 30: The Clean Air Act and volatile organic compounds: Did

Table 5. Summary statisticsMean Std. Deviation

% individuals that are white 0.76 0.29% individuals above 25 0.33 0.17 with less than high school education% households that are below the poverty rate 0.16 0.15 Year=1991% change in Employment 0.02 0.27Non-attainment 0.60 0.49% change in health-indexed emission Air -0.02 0.79 Water -0.05 0.62 Onsite except air -0.08 0.72 Offsite 0.46 1.16 All except air 0.42 1.19 Year=1997% change in Employment 0.02 0.23Non-attainment 0.51 0.50 % change in Health-indexed emissions Air -0.01 0.68 Water -0.03 0.46 Onsite except air -0.04 0.51 Offsite 0.06 0.89 All except air 0.06 0.88 Year=2001% change in Employment 0.00 0.25Non-attainment 0.48 0.50 % change in Health-indexed emissions Air -0.09 0.69 Water -0.01 0.48 Onsite except air -0.01 0.53 Offsite -0.05 0.86 All except air -0.05 0.85Note: The construction of the % change variable for employment and emissionsand emissions allows figures to range from +2 (for entrants) to -2 (for exiters).

%∆ = (Yt - Yt-1) / [(Yt + Y t-1)/2], where Y denotes emissions or employmentand t indexes time.

Page 31: The Clean Air Act and volatile organic compounds: Did

Table 6: Impact of the CAA ozone regulations on plant-level health-indexed VOC emissions into air in the chemical manufacturing sector (1) (2) Media Air Air Non-attainment -0.015* -0.014* (0.008) (0.008) White -0.003 (0.016) Low education -0.024 (0.027) Poor 0.040 (0.033) Percentage change 0.065** 0.065** in employment (0.021) (0.021) Year dummies Y Y SIC dummies Y Y Observations 20000 20000 No. of plants 1964 1964

Robust standard errors are in parenthesis. * statistically significant at the 5%** and 10% * respectively

Page 32: The Clean Air Act and volatile organic compounds: Did

Table 7: Impact of the CAA ozone regulations on plant-level health-indexed VOC emissions into non-air media in the chemical manufacturing sector

(1) (2) (3) (4)Media Water Offsite Onsite All

except air except airNon-attainment -0.005 0.008 -0.004 0.005

(0.005) (0.009) (0.006) (0.009)White -0.020** 0.012 -0.020* 0.009

(0.009) (0.019) (0.010) (0.019)Low education -0.008 0.008 -0.004 0.010

(0.016) (0.031) (0.019) (0.031)Poor -0.020 0.028 -0.018 0.015

(0.016) (0.037) (0.020) (0.036)Percentage change 0.029** 0.060** 0.011 0.049* in employment (0.014) (0.028) (0.017) (0.028)SIC dummies Y Y Y YYear dummies Y Y Y YObservations 20000 20000 20000 20000No. of plants 1964 1964 1964 1964 Robust standard errors are in parenthesis. * statistically significant at the 5%** and 10% * respectively

Page 33: The Clean Air Act and volatile organic compounds: Did

Table 8: Impact of the CAA ozone regulations on plant-level health-indexed VOC emissions: Analysis is limited to VOC that is not listed in the Safe Drinking Water Act and the Priority Pollutant List

(1) (2)Media Air WaterNon-attainment -0.010 -0.004

(0.009) (0.005)White -0.007 -0.020**

(0.019) (0.009)Low education -0.060** -0.013

(0.030) (0.017)Poverty 0.077** -0.007

(0.035) (0.015)Percent change 0.048* 0.012in employment (0.025) (0.014)No. obs. 16906 16906No. facilities 1670 1670Robust standard errors are in parenthesis. * statistically significant at the 5%** and 10% * respectively

Page 34: The Clean Air Act and volatile organic compounds: Did

Table 9: The impact of the CAA ozone regulations on plant-levelhealth-indexed TRI emissions in the chemical manufacturing sector

(1) (2) (3) (4)Media Air Water Onsite All except

except air air Non-attainment -0.021** 0.001 0.005 0.004

(0.008) (0.006) (0.007) (0.010)White 0.009 -0.014 -0.012 0.006

(0.017) (0.011) (0.012) (0.020)Low education -0.007 -0.022 0.005 0.000

(0.028) (0.021) (0.023) (0.031)Poor 0.027 -0.000 0.002 0.038

(0.033) (0.019) (0.021) (0.038)Percentage change 0.047** 0.026 0.011 0.051*in employment (0.022) (0.017) (0.019) (0.030)SIC dummies Y Y Y YYear dummies Y Y Y YObservations 20000 20000 20000 20000No. of plants 1964 1964 1964 1964 Robust standard errors are in parenthesis. Significant at the 5% ** and 10% * respectively.