University of Southern California Dissertations and Theses

Essays in environmental policy

(USC Thesis Other)

Essays in environmental policy

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ESSAYS IN ENVIRONMENTAL POLICY
by
Paul Jeremy Hughes
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment o f the
Requirements for the Degree
Doctor of Philosophy
(POLITICAL ECONOMY AND PUBLIC POLICY)
May 2005
Copyright 2005 Paul Jeremy Hughes
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
UMI Number: 3180356
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
®
UMI
UMI Microform 3180356
Copyright 2005 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
DEDICATION
to
Professor Jeff Nugent
Whose tireless enthusiasm and dedication are an
inspiration to all around him
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
Table of Contents
Dedication ii
List of Tables iv
Abstract v
Introduction 1
Chapter I: Air Quality and Ethnic Divisions
Introduction 6
Literature Review 7
Background and Framework 15
Data and Empirical Model 19
Summary and Suggestions for Further Research 53
Chapter II: Political Effects in Environmental Litigation
Introduction 55
Literature Review 57
Institutions and Policy 64
Data and Empirical Model 69
Summary and Suggestions for Further Research 80
Conclusion 82
Bibliography 86
Appendix 1 92
Appendix 2 104
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
List o f Tables
Chapter I: Air Quality and Ethnic Divisions
1. Division of Air Pollutants by Point and Non-Point Sources 21
2. Air Pollutant Residence Times 22
3. Sources of Sulfur Dioxide Air Pollution 24
4. Descriptive Statistics 27
5. Matrix of Correlation Coefficients 35
6. Changes in Variables over Time 36
7. OLS Regressions- Sulfur Dioxide Air Pollution- Observed 41
8. OLS Regressions- Sulfur Dioxide Air Pollution-
Emissions from All Sources Divided by Land Area 42
9. Fixed Effect Panel Data Regressions Including 1970 Data-
Sulfur Dioxide Air Pollution- Observed 43
10. Fixed Effect Panel Data Regressions Excluding 1970 Data-
Sulfur Dioxide Air Pollution- Observed 44
11. OLS Regressions- Carbon Monoxide Air Pollution- Observed 45
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
List o f Tables- Continued
12. OLS Regressions- Carbon Monoxide Air Pollution-
Emissions from All Sources Divided by Land Area 46
13. Fixed Effect Panel Data Regressions-
Carbon Monoxide Air Pollution- Observed 47
14. OLS Regressions- PM10 Air Pollution- Observed 48
15. OLS Regressions- PM 10 Air Pollution-
Emissions from All Sources Divided by Land Area 49
16. OLS Regressions- Nitrogen Air Pollution- Observed 50
17. OLS Regressions- Nitrogen Air Pollution-
Emissions from All Sources Divided by Land Area 51
18. Source and Mobility of Air Pollutants 53
Chapter II: Political Effects in Environmental Litigation
19. Abnormal Daily Returns to Companies with EPA Litigation
Outstanding during the 2000 Election 73
20. Abnormal Daily Returns to Matched Companies
during the 2000 Election 79
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
List o f Tables- Continued
Appendix
21. OLS Regressions- Sulfur Dioxide Air Pollution- Observed 96
22. OLS Regressions- Sulfur Dioxide Air Pollution-
Emissions from All Sources Divided by Land Area 98
23. OLS Regressions- Carbon Monoxide Air Pollution- Observed 100
24. OLS Regressions- Carbon Monoxide Air Pollution-
Emissions from All Sources Divided by Land Area 102
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
vii
ABSTRACT
The theme of this dissertation is the impact that political effects have on
environmental policy.
The first chapter concerns itself with the relationship between air quality and ethnic
diversity in American metropolitan areas. Recent research has demonstrated that
ethnic divisions may be inversely related to investment in public goods. This paper
applies this finding to the most pure public good, that is, air quality, presenting
empirical results for air quality in U.S. metropolitan statistical areas. The results are
mixed. For carbon monoxide, the results are consistent with the literature, in that
ethnic diversity is associated with more air pollution. For sulfur dioxide, the opposite
results are found. It is suggested that these different results are caused by differences
in the levels of government at which point as opposed to non-point air pollution is
regulated. The implications of this finding for the literature are discussed.
The second chapter uses a new method for determining whether or not the stringency
of environmental regulation is dependent upon political changes. In particular, it uses
the change in the share price of companies with environmental litigation outstanding
at the time of an election. The election of Congressional Republicans in 2000 was
associated with increases of approximately five percent in the share prices of firms
that were being sued by the Environmental Protection Agency (EPA). No similar
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
effects were found to be associated with the election of George W. Bush as
President. These findings are robust to a number of different specifications, and are
not found in an otherwise similar sample of companies without litigation
outstanding. It is suggested that the enforcement of environmental law is dependent
upon political pressure.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
1
INTRODUCTION
The theme of this dissertation is the impact that political effects have on
environmental policy. The first essay concerns itself with the relationship between
air quality and ethnic diversity in American metropolitan areas. The second attempts
to quantify the relationship between changes in political leadership and
environmental policy. This introduction will place these ideas into their broader
context. The motivation of the papers is that economic analysis of environmental
issues may, by itself, be insufficient to lead to meaningful and satisfying policy
implications. The political environment in which decisions are made is also
important.
In the case of the first paper, there are many examples in the broader context of the
literature of the importance of ethnic diversity to political economy. Ethnic diversity
is defined in this literature as diversity across ethnic, racial, religious and tribal
groups. By far the most harrowing example is the desperate poverty in Africa,
resulting from the failure to achieve economic growth. According to the literature
(Easterly and Levine, QJE, 1995) Africa’s adoption of growth retarding policies
owes much to its ethnic diversity. The idea is that because African countries contain
so many different tribal, ethnic and religious groups, they spend all of their energy
fighting among themselves, leaving none left to foster economic growth. Fourteen of
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
2
the fifteen most ethnically diverse countries in the World are in Africa. Partly this
was because the 19th century European imperial powers carved up the continent with
no regard to ethnic divisions (p. 1213), but even before the arrival of the Whites,
Africa had exceptionally high levels of ethnic and linguistic diversity. There are
numerous examples of how conflict between Africa’s ethnic groups retarded growth.
For example, cocoa production in Ghana is concentrated in one ethnic group, and it
was often subject to punitive policies. The motivation for these policies was not to
maximize the tax revenue from cocoa for the benefit of all of Ghana’s ethnic groups,
but rather to dissipate the rents that would otherwise have accrued to one ethnic
group. The result was to kill the goose that laid the golden egg. The contribution of
cocoa to the Ghanaian economy collapsed; cocoa exports went from 19% of GDP in
1955 to 3% in 1983 (p. 1218). Notably, one of the most economically successful
African countries, Botswana, is also one of the most ethnically homogenous. Its
Tswana tribes have a history of co-operation with each other (Alesina et al, NBER,
p.20, 2002). The East Asian ‘economic miracle’ is at least partly attributable to
ethnic homogeneity; Japan, South Korea, Singapore and Hong Kong are among the
most ethnically homogenous countries in the World. They stand in sharp contrast to
Africa.
The finding of a relationship between the adoption of economic growth retarding
policies and ethnic diversity is not unique to studies across countries. There is
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
3
voluminous research documenting the difficulties of local governments in the United
States in investing in public goods, such as education, when they are riven by ethnic
divisions. In the case of education, it can usefully be seen as a ‘loan’ from one
generation to another (Goldin and Katz, NBER, 1998). When the demographic
composition of the neighborhood changes to include more old people, especially
when the old are of a different ethnic group from the young, they are increasingly
unwilling to make this loan to pay for education as a public good. The same
calculations of group benefit and cost can be seen in investments in other public
goods, such as roads and waste disposal, investment in which suffers when their tax
base is ethnically diverse. People seem to be unwilling to see their taxes go to benefit
people in another group.
The present paper applies these findings to the most pure public good, that is, air
quality. The results are mixed. For carbon monoxide, the results are consistent with
the literature, in that ethnic diversity is associated with more air pollution. For sulfur
dioxide, the opposite results are found. These findings are explained further in the
paper itself. The most interesting implication of all of this literature is to show that
an assumption of much of the social sciences, that people behave as individuals, is an
oversimplification. In many economic models this assumption can still usefully be
made. However, much of the time it is more accurate to imagine people as members
of groups, not as individuals.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
4
The second paper tries to measure whether election results in the U.S. Presidency
and Congress have any effects on the prosecution of environmental offenders. It is a
common complaint among voters in democracies that ‘all politicians are the same’.
This sentiment sometimes leads to low voter turnout, especially in the U.S. In some
respects, the claim that all politicians are the same is clearly untrue; tax and fiscal
policies do differ across administrations. In respect of environmental policies,
however, this claim is largely untested. The second paper uses a recent innovation,
the Iowa Electronic Market (IEM) to see which candidates and environmental
policies the stock market was ‘voting’ for. Some of the academic literature supports
the idea that ‘all politicians are the same’ in that political and regulatory changes
often have no quantifiable impacts on stock market valuations (Binder, RJE, 1985).
The reason for this is because the prices of stocks react instantly to new information
that is publicly released. Because of the fact that political and regulatory changes
usually happen slowly over a period of months and years, often with uncertainty and
reversals along the way, it is difficult to decide how much to attribute any change in
price to political change, and how much to other things that may be happening at the
same time. A good recent example of the effects of political change on stock market
prices was the attempt of the first Clinton administration to overhaul healthcare in
America (Ellison and Mullin, JLE, 2001). This would have had a severe impact on
the profitability of the pharmaceuticals industry, had it been successful. In fact, after
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
5
several years of stop-start efforts, the plan was declared dead by Republican
Congressional leaders in July 1994. However, most of the 50% decline in
pharmaceutical stock prices had already happened before the plan was even made
public, suggesting that there was considerable leakage of information prior to
publication. Even after the plan’s defeat, the stocks did not regain their lost value.
The uncertainty over when the market knew of the plan confuses any quantification
as to its impact. This is complicated further by other bad news for pharmaceutical
stocks that came out at around the same time.
In truth, political and regulatory changes do have impacts on stock market
valuations. It is important, however, that any test for such effects is specified
carefully, so as to be able to detect subtle changes. That is the contribution of the
present paper. It uses the IEM and abnormal returns to the stock prices of companies
being sued by the Environmental Protection Agency as a test of which politicians
were deemed by the stock market to be the most likely to ‘crack down’ on
environmental polluters.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
6
CHAPTER 1- AIR QUALITY AND ETHNIC DIVISIONS
INTRODUCTION
Considerable recent economic research indicates that the provision of public goods is
negatively related to ethnic diversity. That is, the greater the likelihood that the
taxpayer and the beneficiary of a public good would be of different ethnicities or
races, the lower is the share of spending on productive public goods (Alesina, Baqir
& Easterly, 1999). This has been alleged to be true of education, roads, sewers, waste
disposal etc. However, it should be noted that none of these is a pure public good in
the sense that it has the qualities of non-excludability and non-rivalry. They are at
most local public goods, or club goods1 , in the sense that they can be of much better
quality in one area of a city than another (Los Angeles Times, October 21st 2001,
p.l). When such local public goods can be provided privately as well as collectively,
there could be an incentive for ethnically divided communities to substitute private
for public provision. An example of this in the American context is the proliferation
of'gated communities', private clubs and schools, or disaggregated and decentralized
school districts or centralized school districts but with rules against school bussing.
The topic of this paper's analysis is air quality, which comes much closer to the
abstract qualities of a public good. It is much harder to imagine how air quality could
1 These two terms, ‘local public goods’ and ‘club goods’, are used interchangeably throughout this
paper. They are distinct from purer public goods such as air quality.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
7
be privatized. There is evidence that air quality is capitalized into housing prices
(Chay & Greenstone, 1998), although this is clearly established only from one city to
another, and not for different areas within the same city, so all residents of a city can
be assumed to enjoy a relatively standard level of air quality. This paper casts doubt
on whether the findings of the literature can be applied to a pure public good such as
air quality. Ethnically fragmented areas actually have higher spending on local
public goods, although that is paid for by intergovernmental transfers, not local taxes
(Alesina, Baqir & Easterly, 1999, p. 1274). Although air quality legislation in the
U.S. is formulated at the federal level, it is implemented at the state and local level.
There is evidence that it is enforced much more stringently in some local areas than
in others. This is helpful for this paper because, if air quality regulation were
enforced federally, then there would be no locally-enforced variation in air quality
standards which would vary with ethnic divisions. The strongest results of this paper
are for those air pollutants which generally come from stationary rather than mobile
sources. This also is also important for this paper since stationary source regulation
is much more subject to local regulation than are mobile sources.
LITERATURE REVIEW
Several excellent recent papers discuss the effect of ethnic and racial heterogeneity
on public policies, especially on the provision of public goods. I will first examine
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
8
the contribution of this research, and then outline how the present paper will
contribute to this field.
The central insight of this literature is that people do not value the welfare of people
outside their own group as equal to their own. This seemingly obvious statement has
far-reaching implications for political economy and public finance.
In most of this literature, the measure of ethnic fragmentation used was the
probability that two randomly drawn people will belong to different ethnic groups
(Alesina, Baqir and Easterly, 1999, p. 1255);
ETHNIC = 1-XiCRacei)2
where Race i denotes the share of the population self-identified as of race i and i =
|White, Black, Asian and Pacific Islander, American Indian, Other|. ETHNIC varies
from a minimum of zero (complete ethnic homogeneity) to a maximum of one minus
the reciprocal of the number of different ethnic groups there are (i.e; if there are five
different ethnic groups, ETHNIC could be a maximum of 0.8, if each of the five
ethnic groups accounted for exactly 20% of the population (Alesina, Baqir and
Easterly, 1999, p.1257).
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
9
The most fundamental effect of ethnic diversity in the literature is on trust. Alesina
and LaFerrara QJE, 2000; JPE, 2002) find that participation in social activities such
as churches, labor unions and clubs is lower in more ethnically diverse localities, and
survey respondents admit to trusting others less. This is especially true of
organizations that it is difficult to exclude people from, and for people who express
views against racial mixing.
This absence of trust affects redistributive public policies. For example, Luttmer
(1998, unpublished paper) analyzes survey data to find two effects. The negative
exposure effect, that is people decrease their support for welfare the more welfare
recipients there are in their local area, and the racial group loyalty effect, that is that
they increase their support for welfare if more recipients belong to their racial group.
Interestingly, in view of my own empirical results, he finds that his results hold true
when areas are defined as census tracts, metropolitan areas or states. This is offered
as an explanation for why welfare payments are lower in racially heterogeneous
states such as Mississipi. Similar findings, that centralized transfer payments are
lower when ethnic fragmentation is higher are reported by McCarty (1993). Using
actual rather than survey data, Alesina, Baqir and Easterly (NBER, 1998) find
evidence to corroborate Luttmer’s findings. They find that city employment in U.S.
cities is higher when racial heterogeneity and income inequality is higher. They
explain this as a ‘disguised’ redistributive policy. They hypothesize that as direct
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
10
redistribution is infeasible, indirect redistribution is mandated instead. Their findings
are not unique to the U.S; they report that in Italy, there are more forest rangers in a
small region in the poor South than in the whole Italian Alps (p.l). In addition to the
evidence on greater public employment in ethnically diverse areas, there is more
evidence that such areas receive greater transfers per capita from higher levels of
government (Alesina et al, QJE, 1999, p. 1266). Why this is so is an unresolved
question. Alesina et al. guess that it could be to help the supply of public goods to
areas that have difficulty providing them. More cynically, they also suggest that it
could be because such areas have more effective pressure groups.
Ethnic diversity also affects the geographic size of political jurisdictions. Alesina,
Baqir and Hoxby (NBER, 2000) consider heterogeneity in income, ethnicity, race
and religion in American school districts and municipalities. Their strongest
evidence is of a tradeoff between economies of scale and racial heterogeneity. That
is to say, large school districts, for example, can have economies of scale because
they can provide libraries, sports facilities and administration to a greater population.
This results in gains in efficiency and in lower per unit costs. The tradeoff between
income inequality and economies of scale is not as strong. In other words, people are
more willing to lose the benefits of economies of scale in order to avoid mixing of
the races than they are to avoid mixing of income groups. A theoretical treatment of
similar issues is given in Alesina and Spolaore (QJE, 1997).
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
11
There are two substantive differences between the existing literature and the present
paper. Those are, that it deals with a pure public good, and that that good is not
allocated directly by voting.
The existing literature, to the extent that it deals with tangible effects of public
policy, as opposed to intangibles such as trust, purports to examine the effects of
ethnic diversity on the provision of public goods. In fact, these goods are better
characterized as local public goods or club goods rather than as pure public goods. A
good example of these is education spending. Alesina Baqir and Easterly (1999,
QJE) found that ethnic diversity is associated with a lower share of expenditure on
productive public goods, which supposedly includes education, roads and waste
disposal. Obviously, if ethnic diversity is associated with lesser shares of public
spending on some goods, then it must be associated with greater shares on spending
on other goods. Alesina et al. (1999, QJE, p. 1264) found that most of these
categories, with the exception of police and health spending, are ill-defined
discretionary programs. They speculate that this could include patronage, although
no direct evidence of this is offered. Interestingly, the greater share of spending on
police is robust to controls for crime. They speculate that the share of health
spending may increase with ethnic diversity because it also includes transfer
payments in the form of subsidized health services.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
12
However, other researchers have found (James, 1987, 1993) that ethnic diversity is
associated with higher levels of private spending on education, possibly suggesting
that people would rather incur the extra private costs of education than have racially
mixed public education. Furthermore, an explanation for the fact that education is
usually publicly rather than privately funded is that education is part of an
intergenerational loan (Goldin and Katz, NBER, 1998, p. 10). Communities that are
more homogeneous, and in which people remain for most of their lives, would be
more likely to make this ‘loan’, by supporting publicly funded education. Evidence
in support of this hypothesis is provided by Goldin and Katz (1998, NBER) who
show that more homogeneous areas in Iowa, in terms of ethnicity, religion and
income, had greater expansion in their high school systems from 1910 to 1930.
Similarly, Poterba (NBER, 1998) shows that an increase in the percentage of elderly
people in U.S. jurisdictions from 1960 to 1990 was associated with a reduction in
education spending per child. This is especially true when the old and the young are
from different racial groups.
The use of air quality as the subject of analysis in the present paper is expected to
yield new insights in this field, as it is a pure public good which is difficult to
exclude others from. Although it is true that previous research found that air quality
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
13
is capitalized into housing prices (Chay and Greenstone, NBER, 1998), that study
covered whole cities, not local neighborhoods.
Unlike education, air quality is not voted upon directly. Most of the existing
literature offers a clear voting mechanism to allocate the good in question. However,
some of the existing papers offer evidence that their results are not dependent upon
political intermediaries, but rather stem from the uncoordinated actions of
individuals. The present paper does not explain the exact means by which air quality
might be affected by ethnic diversity. Given this, the prior literature showing that
political actors are not needed is important, as it mitigates the lack of a theoretical
framework in the present paper.
The best example in the literature that the phenomenon under study does not
necessarily require a political mechanism is the absence of trust in ethnically diverse
communities (Alesina and LaFerrara, 2000, 2002). Easterly and Levine (QJE, 1997)
use the extremely high ethnic diversity of African countries to explain their
disastrous economic performance since 1960. They find, among other things, that
African ethnic diversity is associated with a host of problems of underdeveloped
countries, such as the paucity of telephones- hardly pure public goods. Alesina, Baqir
and Easterly (QJE, 1999) find that ethnic diversity is associated with a lower share of
spending on supposed public goods such as education, roads and waste disposal, in
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
14
U.S. cities, counties and metropolitan areas. However, they themselves acknowledge
(p.1255) that different types of public goods have different jurisdictions. That is, one
metropolitan area may have dozens of school districts, and another may have just
one2.
There is evidence on the control that local elected authorities can have over air
quality. Recent research by Henderson (AER, 1996) lends credence to the idea that
local governments have substantial control over air pollution, and exercise it in their
perceived self-interest. That is, their self-interest is to avoid the harsher regulation
that the loss of ‘attainment status’ (when a county exceeds federally mandated air
quality standards) brings3. For example, the so-called ‘second annual daily
maximum’ ozone standard used by the EPA did not necessarily bear any strong
relationship to annual mean levels of ozone. The EPA standard used only the second
highest daily level of ozone in a whole year to determine if a county was in or out of
attainment. In other words, it was based only upon a single extreme observation, and
not upon average annual measures such as the annual mean or median. Henderson
showed that localities can improve their compliance with EPA regulations relating to
pollution maxima by changing the pattern of polluting activity across the day,
2 This fact was exploited by Hoxby (AER, 2000). She found that the fact that some areas had more
school districts than others dues to natural barriers such as the existence o f streams, which were much
greater obstacles to geographical mobility in the past than today, had created a ‘natural experiment’ in
educational competition. This was then used to quantify the projected effects o f reforms to the
American high school educational system.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
15
without improving measures of typical conditions, such as daily medians. There is
even evidence that after the passage of the CAA (Clean Air Act) states encouraged
their counties to attain compliance status by building taller smokestacks just upwind
from a neighboring state (Revesz, 1996, p.2353).
None of this is to suggest that the topics of the previous literature do not rely at all
upon political actors. Clearly, ethnic diversity has caused political instability in
Africa, and education, roads and waste disposal are voted upon at some level of
American local government. Similarly, city and local authorities do have influence
over air quality. But the important point is that no direct political mechanism is
shown in the present paper, and that the literature suggests that that may not be
necessary.
BACKGROUND AND FRAMEWORK
The purpose of this section is to outline some testable hypotheses concerning the
distribution of air quality as a public good. Some background information is also in
order. Five main hypotheses to explain the distribution of air quality are considered;
1. Industrial Base and Electricity Generation- Since air pollution is generated
disproportionately more by manufacturing industry, than by services or agriculture, it
3 Further research by Henderson (JPE, 2000) quantifies the effect o f the loss o f attainment status on
lower births o f new firms and on the exiting o f existing firms.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
16
is obviously important to consider the different bases of the local economy. Not only
is the size of the manufacturing sector important, in terms of its total employment
and output, but also its composition is important. For example, Antweiler et al (AER,
2001, in footnote or reference) show that some industrial sectors, as defined by three
digit Standard Industrial Classification (SIC) codes, are more than a hundred times as
pollution-intensive as others4. Their differences in pollution-intensity are attributed
to their different capital-labor ratios. Antweiler et al (p.893-4) find that a 1% increase
in the capital-labor ratio leads to a 1 % increase in pollution. It is reasonable to
assume that both the size of the manufacturing sector, and its composition, vary
according to the racial composition of the surrounding area. That is to say, it may be
reasonable to imagine that largely white, ethnically homogenous areas, have
disproportionate concentrations of highly capital-intensive and highly polluting
industry. To test for this, the capital-labor ratio in industry is used as an independent
variable in the analysis. Another control is used for the type of coal used in
electricity generation. This will be explained in greater detail later.
2. Selective Monitoring of Air Pollution- For the purposes of measuring ambient air
quality without any biases, monitoring stations should be either randomly distributed
to provide representative measures of background air quality, or systematically
4 Also used as a control for the industrial base o f the area was the percentage o f the workforce
employed in manufacturing. It is unreported in the present paper as it lacks the better theoretical and
empirical underpinning o f the capital-labor ratio.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
17
located according to set criteria, such as always being exactly one mile downwind of
a power station. In fact, they tend to be located in areas with severe pollution
problems, often around a site that has been the subject of complaints (Grossman,
Krueger and Laity, 1994, p.7). This raises the possibility that actual air quality may
not be objectively measured by monitoring stations. To counter this, analysis is
performed both on air quality as measured by the monitoring stations, and on
emissions of pollutants into the atmosphere. NET (National Emissions Trends) data
is used for the emissions of air pollutants. This is a database of projected emissions
of air pollutants from both point and non-point sources by counties in the United
States. It is available online at
http://www.epa.gov/air/data/nettier.html7us~USA~United%20States The data is
obtained not from measurements of actual emissions, but rather from the
measurement of inputs to the process of pollution generation, such as industrial
processes and usage of fossil fuels.
3. Political Institutions- Some recent theoretical and empirical work (Persson &
Tabellini, 1999; Lizzeri & Persico, 2001) suggests that public goods may be
underprovided in majoritarian as opposed to proportional electoral systems. This is
hypothesized to be because winner-take-all majoritarian electoral systems focus
political competition on a few key districts, unlike proportional electoral systems
where the spoils of office are distributed more evenly in relation to the size of the
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
18
vote. So in winner-take-all systems, such as the U.S. electoral college used to elect
the President, there is little incentive to invest in public goods that benefit everybody,
when instead ‘pork-barrel’ spending could be directed to ‘swing’ states whose votes
decide elections. Unfortunately, proportional representation does not exist on a large
scale in U.S. cities5. Instead, the present paper makes an effort to capture possible
collective action problems between different governments. The present paper uses
Metropolitan Statistical Areas (MSAs) as the unit of geographic aggregation. These
are sprawling areas, often covering many cities, counties and states. To that end, the
number of states that the MSA straddles is used as an independent variable in the
regression analysis6.
4. Distribution of Population- Although air quality is a public good, it is possible to
imagine certain circumstances whereby some areas within a city could be affected by
it more than others. Since the present paper is focussed on the analysis of air quality
as a public good, the distribution of the population that uses this public good is
important. That is to say, if the population is highly segregated by race, as most
American cites are, then that implies that the ethnic diversity of the city as a whole
could be misleading. It may be, under those circumstances, more accurate to consider
5 There is, however, an institutional difference between cities which elect their city council members
‘at large’ from a single city-wide district as opposed to from single-member districts. This is by no
means a form o f proportional representation, but it does have one characteristic in common with it.
That is, it is possible that transboundary pollution across political district boundaries could be more o f
a problem in single-member district cities rather than ones which elect all o f their representatives at
large. This is elaborated upon further in the appendix.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
19
each of the segregated neighborhoods as ethnically homogeneous cities unto
themselves. To that end, it is important to consider the distribution of the population
across space. This is done by using an index of segregation explained later.
5. Income Distribution- One may argue (Alesina et al, QJE, 1999, p. 1257) that
polarization of preferences is a function of heterogeneity in income, not ethnicity.
Therefore, income inequality is also modeled. This is measured by mean household
income divided by median household income. Alesina et al (QJE, 1999) used the
same measure of inequality, as it is theoretically the most appropriate measure in any
model based upon the median voter theorem.
DATA AND EMPIRICAL MODEL
Data
This section will, firstly, defend the choice of geographic aggregation, and then
secondly, justify the choice of dependent variable, and finally, explain and motivate
the importance of the independent variables in the empirical analysis.
Firstly, the empirical results are generated using Metropolitan Statistical Areas
(MSAs) as the unit of geographic aggregation. These are agglomerations of cities,
such as Los Angeles-Long Beach MSA. The reason that MSAs are used, instead of
6 Controls were also used for the number o f cities and counties in the MSA. These were found not to
be as useful as the number o f states, and so these results were included only in the appendix.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
20
counties, which are the basis for attainment status, is because air pollution does not
respect the arbitrary human-defined political borders of a county, but is rather
generated by anthropogenic economic activity which is centered around an urbanized
area such as an MSA. MSAs are also generally made up of groups of counties,
except in New England, where they are also made up of subdivisions of counties7.
Secondly, since the aim of this paper is to analyze air quality as a public good in
American cities, and there are many types of pollution, it is important to focus on
particular types of pollution that have characteristics suitable for this analysis. In
particular, the source should be locally emitted, have strong local negative effects
(which rules out ozone and carbon monoxide), be subject to regulation because of
these deleterious effects, have well-known abatement technologies, and be the
product of fixed, as opposed to mobile sources. The last requirement is important, as
in the United States, motor vehicle pollution and other mobile sources are generally
subject to federal control, rather than local control, (although California does have
n
more stringent emission requirements than the rest of the country ). Particularly
important are the requirements that the pollutant must originate largely from point
sources and that it must not be very mobile. That is, it must not drift far from its
source. These two points are elaborated on below.
7 However, even in N ew England, the present analysis still assumes that MSAs are made up o f groups
o f counties. This means that there is a slight mismeasurement in pollution emissions compiled from
the NET database. NET data is available only by counties, not MSAs. It is unlikely to make much o f a
difference, as the deviation from the county level o f aggregation is slight.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
21
1. Point Source- More than any of the other pollutants in my data, sulfur dioxide is
generated from point rather than non-point sources, mainly from power plants and
industrial facilities;
TABLE 1
DIVISION OF AIR POLLUTANTS BY POINT AND NON-POINT SOURCES
% of air pollutant source;
Point Non-Point
Sulfur Dioxide (SO2 ) 80.90 19.10
Nitrogen Oxides (NOx ) 35.87 64.13
Particulate Matter (PM 10 <2.5 micrometers) 13.63 86.37
Ammonia (NH3 ) 13.47 86.53
Volatile Organic Compounds (VOC) 13.38 86.62
Particulate Matter (PM 10 <10 micrometers) 6.28 93.72
Carbon Monoxide (CO) 6.06 93.94
In other words, for the purposes of my study, sulfur dioxide is the best pollutant to
study. This is because it is generated mainly from stationary point sources (as
8 The AQMDs (Air Quality Management Districts) in California, such as the SCAQMD also impose
regulations on point source pollution. Most states also have state level Environmental Protection
Agencies.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
22
opposed to non-point sources) and so is likely to vary much more substantially from
one city to the next. Much of the pollutants other than sulfur dioxide are generated
by cars, whose pollution is less subject to local control.
2. Low-Mobility- Distinct from the observation that some pollutants are likely to be
more evenly distributed across the neighborhoods of a city (due to the fact that they
are generated by mobile as opposed to stationary sources) is the question of the
mobility of the pollutant. Flinterman, Kwiatkowska, and Lammers (1986, p.3) report
the following residence times, that is the length of time that the pollutant is typically
suspended in the atmosphere, for these air pollutants;
TABLE 2
AIR POLLUTANT RESIDENCE TIMES
Residence time (days')
Minimum Maximum
Ozone 0.40 90.00
Nitrogen Dioxide 2.00 8.00
Sulfur Dioxide 0.01 7.00
Nitric Oxide 4.00 5.00
Sulfate 3.00 5.00
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
23
The fact that ozone can remain suspended in the air for 90 days means that it can
travel great distances from its source, and so the level of ozone is not likely to bear
much relation to local emissions or abatement efforts. Sulfur dioxide is one of the
more desirable pollutants as its residence time is generally short.
The best fit according to these criteria is therefore sulfur dioxide. For much the same
reasons, this was the pollutant chosen as the subject of analysis by Antweiler,
Copeland and Taylor (1998, p. 19). As the main focus of the present paper, it is worth
examining the identity of sources of S02 pollution, as well as the chemistry of its
formation.
S02, or sulfur dioxide, belongs to the family of sulfur Oxide gases (SOx). These
gases are easily soluble in water. Sulfur is found in all raw materials, especially coal,
crude oil and metals such as iron. Sulfur Oxide gases are formed when fossil fuels
are burned or when metals are extracted from ore. More specifically, S02 dissolves
in water vapor to form acid. Other interactions with other gases and air particles
forms sulfates and other damaging pollutants.
S02 has a number of detrimental effects on the environment and on human health.
These include contributing to acid rain, visibility impairment, and aggravation of
respiratory conditions such as asthma.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
24
A detailed breakdown of the sources of S02 is;
TABLE 3
SOURCES OF SULFUR DIOXIDE AIR POLLUTION
% of air pollutant source9:
Point Non-Point Total
Fuel Combustion-
Electricity Generation - Coal 0.00 64.23 64.28
Fuel Combustion-
Industrial - Coal 3.16 4.93 8.08
Fuel Combustion-
Industrial - Other 2.40 4.51 6.91
Highway and Off-Highway Vehicles 6.23 0.00 6.23
Fuel Combustion-Other 2.65 0.87 3.51
Metals Processing 0.00 3.06 3.06
Fuel Combustion-
Electricity Generation - Other 0.00 2.91 2.91
Petroleum and Related Industries 0.01 1.81 1.81
Other_________________________________ 0T4__________ T05_________ 3T9
Total 14.58 85.42 100.00
9 Note that the discrepancy with the other table (which gave point sources o f S 0 2 as 80.90%, not
85.42% as in this table) is because the previous table referred only to the pollution emissions in
MSAs; the same data as used in the empirical results in Table 5. Both tables use only 1990 data.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
25
Clearly, the burning of coal is by far the largest contributor to the creation of S02.
Coal varies a great deal in terms of how much sulfur it has. This was the subject o f a
book researching the political economy of high sulfur coal producers (‘Clean Coal-
Dirty Air’, Ackerman and Hassler, 1981). To control for this, a variable is included
in the empirical results which proxies for the sulfur intensity of coal used.
Despite the focus on S02, results are also produced for three of the other ‘criteria’
air pollutants regulated by the EPA (Environmental Protection Agency) that also
have NET (National Emissions Trends) data available; carbon monoxide, particulate
matter (PM 10) and nitrogen. Because of its lower mobility, as anticipated, the
strongest and most consistent results are obtained for sulfur dioxide. For comparative
purposes, the results for carbon monoxide are also presented. However, the nitrogen
and particulate matter results are not discussed in as much detail This is because, in
the case of nitrogen, the air pollution emissions data does not refer to the same thing
as the observed air pollution data. For PM 10, the problem is that the emissions of
PM10 are not the only source of ambient PM10. Murdoch, Sandler and Sargent
(1997, p.283) report that up to 20% of PM is attributable to sulphates. Similarly,
some Oxides of Nitrogen are converted to nitrates, which contribute up to 5% of PM
mass. Other pollutants, such as ozone, are also formed by complex photochemical
reactions between pollutants and atmospheric conditions.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
26
Finally, this section explains and motivates the choice of independent variables.
Descriptive statistics are presented for Metropolitan Statistical Areas in table 5.
AIRS (Aerometric Information Retrieval System) as used in Grossman, Krueger and
Laity (1994, p.4, 7-9) is used for the measured air pollution data for this paper. This
consists of readings of air pollution levels from monitoring stations across the U.S. It
is available as public information from the EPA. NET data is used for the emissions
of sulfur dioxide, as explained earlier. The income and racial explanatory variables
are generated from U.S. census data. They are explained in greater detail below.
The central independent variable is ETHNIC, which has already been defined. In my
results, it is defined using three groups; White, Black and Other. The introduction of
other ethnic groups, e.g., Hispanic or Oriental, would be unlikely to make much of a
difference to ETHNIC, and hence in the results, as in my data 95% of the population
is either White or Black1 0 . Indeed, ETHNIC based of the existence of five groups has
a correlation coefficient of more than 0.99 with ETHNIC based on the existence of
just two groups.
1 0 Results are also produced using ETHNIC defined over two groups (White and Non-White). This
did not change the results. ETHNIC is compiled from U.S. decennial census data, and so it is
available by county and by MSA.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
27
TABLE 4
DESCRIPTIVE STATISTICS
OBS. MEAN STDEV MIN MAX
SULFUR DIOXIDE AIR POLLUTION, 170 0.0075 0.0045 0.0010 0.0245
OBSERVED, IN PARTS PER MILLION
TOTAL SULFUR DIOXIDE EMISSIONS, 72 34.4 42.4 0.8 254.5
IN POUNDS, DIVIDED BY AREA OF MSA
ETHNIC 170 0.2729 0.1416 0.0260 0.5400
% OF POPULATION BLACK 170 0.1208 0.1020 0.0033 0.4168
HOUSING SEGREGATION 170 0.5553 0.1190 0.1672 0.7854
CAPITAL/LABOR RATIO 170 0.0011 0.0007 0.0002 0.0038
IN MANUFACTURING
SULFUR CONTENT OF COAL USED FOR 170 1.1495 0.7018 0.0000 2.1300
ELECTRICITY GENERATION BY STATE
NUMBER OF STATES IN MSA 170 1.1765 0.5143 1 4
MEAN DIVIDED BY MEDIAN 170 1.1773 0.0771 1.0463 1.4347
HOUSEHOLD INCOME
All data is is from 1980 and 1990 equally (85 observations each), except for the S02 emissions,
which are from 1990 only. Note that Metropolitan Statistical Areas (MSAs) in the New England
area are not necessarily coterminous with county boundaries, and although the present analysis
assumed that they were. This would affect only the S02 emissions.
Residential segregation in the city is also included in the model, as sulfur dioxide
usually does not travel far from its source. The index of segregation, or dissimilarity,
used is the same as Cutler and Glaeser (1997, p. 83 6-7), which has the intuitive
interpretation of being the share of the black population which would need to move
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
28
their living accomodations in order to induce an even distribution of black people
across the metropolitan statistical area1 1 ;
where Black j is the number of blacks in census tract i, Black is the number of Blacks
in the city as a whole. Similarly, non-Black j is the number of non-Blacks in census
tract i, and non-Black is the number of non-Blacks in the city as a whole. If Blacks
and non-Blacks never reside in the same census tracts, the measure of housing
segregation will be 1. If they are distributed evenly, the measure will be zero. Cutler
and Glaeser report that the level of segregation varies dramatically across
metropolitan statistical areas, from the least segregated MSA of Jacksonville, North
Carolina (0.21) to the most segregated, Detroit, Michigan, (0.87). This measure
could be criticized in that it does not use the same three groups as does the definition
of ETHNIC. As noted above, however, only 5% of the population in my sample is
neither White nor Black. Indeed, Massey and Denton (1988) report that most
measures of residential segregation are highly correlated with each other, so a more
1 1 This data is available from the NBER at http://www.nber.org/data/ or directly from
http://trinity.aas.duke.edu/~jvigdor/segregation/index.html It is available only by MSAs, not by
counties.
1 2 Note that in my usage, the formula for housing segregation is defined as Whites and Non-Whites,
not Blacks and Non-Blacks.
Housing
Segregation
Black i Nonblack i
Black Nonblack 1 2
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
29
sophisticated measure of segregation that allowed for the existence of more than the
two groups would be unlikely to produce different results.
The capital-labor ratio in manufacturing is used as a control for the pollution
intensity of the industrial base of the MSA. This variable was constructed from data
on the value of output in manufacturing at the three digit level of aggregation in each
MSA. Since the Census of Manufactures and the Annual Survey of Manufactures do
not gather information about the stock of capital in industrial sectors, but rather only
about the investment in new capital, new capital investments were aggregated to
form a measure of the stock of capital. Data from the Census and Surveys from 1987
to 1992 inclusive was gathered. Since this six year period covered both a recession
and a boom, it is assumed that the fluctuations in capital investment would cancel
each other out, and a representative measure of the capital stock would be obtained.
The measure of capital was then divided by the labor force in the MSA to obtain the
capital-labor ratio. The capital-labor ratio is considered to be superior to the simple
volume of output or employment in manufacturing for reasons outlined earlier; that
is, that pollution intensity of manufacturing varies enormously. Across the various
subsectors of manufacturing and as mentioned above, capital intensity of
manufacturing has been found to be a significant determinant of air pollution.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
30
Given the fact that the large majority of S02 air pollution is attributable to coal fired
electric power generation, and also that there is a great deal of variation in the sulfur
content of coal, a control for this is necessary. The U.S. Energy Information
Administration compiles data on the sulfur content of coal used for electricity
1
generation . The exact variable used is the sulfur % by weight used in coal for
power stations by state. This is not available for MSAs, but rather only at the state
level of aggregation. Nevertheless, it is potentially an important control. This
variable is not used for any other pollutant other than S02, since it is not relevant for
them.
The number of states that the MSA straddles is compiled from 1990 NET data. As
mentioned, it is intended to capture collective action problems in the control of air
pollution by governments.
Income inequality is measured by the median divided by the mean household income
in 1990. This measure is taken from the 1980 and 1990 censuses (comparable data
1 3 This data is gathered from the page 45 of'Energy Information Administration/Electric Power
Annual 1999 Volume II', available online at; http://tonto.eia.doe.gov/ftproot/electricity/0348992.pdf
Most o f the data used was for 1998, but later figures are used if that was missing. For the six states
that are missing (CA,DC, ID, ME, RI, VT), I have been told (by U.S. Electric Utility Environmental
Statistics: Natalie Ko (Natalie.Ko@eia.doe.gov) that this means that they do not qualify for Energy
Information Administration, Form EIA-767, Steam-Electric Plant Operation and Design Report. I
have assigned these values o f zero.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
31
was not available from the 1970 census). It is meant to control for heterogeneity in
income, which is often closely related to heterogeneity in ethnicity in American data.
To summarize, the expected signs of the controls are as follows. Housing segregation
is expected to have a positive effect on air pollution. This is because the more
segregated an MSA is by race, the easier it would be to ‘direct’ air pollution to
certain neighborhoods which lack economic and political power. This would be
made more difficult in an integrated MSA. The capital-labor ratio in manufacturing
is expected to have a positive effect on air pollution. This is unambiguously
suggested in the literature, and should be especially true in the present paper, given
that nearly all S02 comes from point sources, which are largely industrial. The sulfur
content of coal used for electricity generation by state is expected to be positive. This
is because the large majority of S02 air pollution can be traced to coal fired power
stations, and it is known that the coal that they use is not homogenous. The number
of states in the MSA is expected to be positive, which would reflect the difficulties
of different states in agreeing on an appropriate level of air pollution. Finally, income
inequality, as measured by mean divided by median household income, is expected
to have a positive effect on air pollution. This is because heterogeneity in income has
the same theoretical effect as heterogeneity in ethnicity. That is, it is expected to
discourage investment in public goods because it leads to heterogeneity in
preferences.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
32
Empirical Model
Two OLS (Ordinary Least Squares) regression models, and one fixed effects model
are used. The purpose of the first regression model is to explain the actual, observed,
levels of ambient air pollution. The null hypotheses of the theoretical model are
tested by the addition of further explanatory variables to control for ethnic divisions,
income distribution and other controls. The first model uses as a dependent variable
the mean of the annual means of all the readings from the air pollution monitoring
stations in each metropolitan statistical area in 1980 and 1990. The form of the
regression is as follows;
Measured Air = a + Pi ETHNIC j t + ...+ pN XN + s i t (1)
Pollution; t
where i refers to the individual MSAs, and t refers to the year, which is either 1980 or 1990.
The variables X2.. .Xn refer to the controls for housing segregation, the industrial base etc.
The second model tests if the relationships between observed levels of air pollution and the
hypotheses of the theoretical model hold true when pollution emissions, rather than observed
air quality, is used as the dependent variable;
Air Pollution = a + Pi ETHNIC j t +...+ Pn Xn i t + s i t (2)
Em issions, t
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
33
Only 1990 data is available for this estimation. As far as possible, the specifications
are the same for both models. This is done to ensure maximum comparability
between them. Also to ensure comparability, the emissions of S02 are divided by the
land area of the MSA1 4 . Assuming no measurement biases in the positioning of air
pollution monitoring stations and an even distribution of population everywhere, this
should result in the same results in models (1) and (2). In fact, the results are quite
different. This can be ascribed to differences in the nature of the different air
pollutants.
A disadvantage of using an OLS model is that it does not allow for the use of fixed
effects for MSAs, such as the adverse topography and climate of the Los Angeles
basin that does not allow air pollution to disperse easily. There are many other
effects not controlled for in (1) and (2) above. A fixed-effects panel data model is
used for the observed levels of air pollution of S02 and CO as follows;
Measured Air = a , + Pi ETHNIC j t +...+ Pn X n i t + e i t (3)
Pollution j t
where the term a i represents a fixed effect for each individual MSA.
1 4 Emissions o f S 0 2 were also divided by the population o f the MSA in earlier experimental
regressions. This gave similar results.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
34
Model (1) uses observations drawn equally from 1980 and 1990 for all air pollutants.
Unfortunately, NET data is available only from 1985-1999, and so model (2) can
only be used with 1990 census data. Model (3) is estimated only for S02 and CO.
For CO, the same data is used as for model (1). For S02, slightly different data is
used than for model (1). This is because the results were found to be sensitive to the
inclusion (results in table 6) or exclusion (results in table 7) of data from 1970. As
table 3 shows, levels of S02 air pollution in 1970 were substantially higher than in
1980. This may be due to measurement error (as suggested by Grossman, Krueger
and Laity in a 1994 unpublished paper) or simply to undoubtedly higher levels of
pollution at that time. The different results dependent upon the inclusion or exclusion
of the 9 observations from 1970 are elaborated upon below.
Descriptive statistics for all the variables are given in table 2. Correlation coefficients
are given in table 2. The correlation between the observed S02 pollution and S02
emissions divided by land area is surprisingly low at only .40. This probably reflects
the dispersal of pollutants by atmospheric conditions, but it may also reflect the
selective placement of monitoring stations. This is the reason why regressions using
both actual emissions data and ambient air quality are used. The capital-labor ratio is
also surprisingly uncorrelated with S02 emissions, at only .09. Controlling for other
variables, as we will see, the relationship between them is stronger. ETHNIC and the
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
35
percentage of the population that is black, at .72 are highly correlated. This point will
be returned to below to address concerns that ETHNIC is essentially just a proxy for
TABLE 5
MATRIX OF CORRELATION COEFFICIENTS
1 2
1 SULFUR DIOXIDE 1
AIR POLLUTION, OBSERVED,
IN PARTS PER MILLION
2 TOTAL SULFUR DIOXIDE 0.40 1
EMISSIONS, IN POUNDS, 0.00
DIVIDED BY AREA OF MSA
3 ETHNIC -0.35 -0.11
<.0001 0.37
4 % OF POPULATION BLACK -0.16 0.05 0.72
0.03 0.68 <.0001
5 HOUSING SEGREGATION 0.27 0.04 -0.06 0.35
0.00 0.76 0.46 <.0001
6 CAPITAL/LABOR RATIO 0.26 0.09 -0.11 0.16 0.30 1
IN MANUFACTURING 0.00 0.43 0.14 0.04 <.0001
7 SULFUR CONTENT OF COAL 0.41 0.21 -0.43 -0.04 0.45 0.23 1
USED FOR ELECTRICITY <.0001 0.07 <.0001 0.61 <.0001 0.00
GENERATION BY STATE
8 NUMBER OF STATES 0.07 0.10 0.07 0.21 0.05 0.02 0.03
IN MSA 0.38 0.41 0.34 0.01 0.50 0.76 0.67
9 MEAN DIVIDED BY MEDIAN -0.36 0.03 0.21 0.12 -0.27 -0.13 -0.09 0.10
HOUSEHOLD INCOME <.0001 0.83 0.01 0.11 0.00 0.08 0.24 0.21
Each figure is a Pearson correlation coefficient. Underneath the correlation coefficients are Prob > |r| under H0:Rho=0.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
36
the percentage of the population that is black. Table 3 shows the change in some of
the key variables over time. The most obvious point of the table is that levels of
pollution have fallen sharply over time. This is especially true of S02 from 1970 to
1980. Ethnic diversity has stayed approximately constant over time. These points
will be returned to in the discussion of the fixed effects models below.
TABLE 6
CHANGES IN VARIABLES OVER TIME
OBS. MEAN STDEV MIN MAX
1970 9 0.0257 0.0222 0.0050 0.0743
1980 89 0.0093 0.0057 0.0011 0.0335
1990 89 0.0064 0.0038 0.0010 0.0199
ETHNIC IN - 1970 9 0.2332 0.1387 0.0449 0.4708
1980 89 0.2695 0.1497 0.0265 0.5400
1990 89 0.2581 0.1406 0.0260 0.5053
CARBON MONOXIDE AIR
POLLUTION, 1980 85 1.5304 0.6661 0.3500 3.0000
OBSERVED,
IN PARTS PER
MILLION, IN - 1990 85 1.0755 0.4042 0.3500 2.4700
The number of observations for sulfur dioxide in 1980 and 1990 is slightly different from
tables 1 and 2 (89 observations each as opposed to 85) because the measure of income
inequality, Mean Divided by Median Household Income, was not available in the 1970
census.
SULFUR DIOXIDE AIR
POLLUTION,
OBSERVED,
IN PARTS PER
MILLION, IN -
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
37
The first specification includes ETHNIC as a measure of ethnic diversity. For S02 in
(1), the coefficients expressing the impact of ETHNIC on observed pollution is
negative and significant. In (2), the coefficient of ETHNIC is negative but
insignificant. The results of (1) imply that greater ethnic diversity is associated with
cleaner air. It is also encouraging for the relevance of the results of the present paper
that the coefficient of ETHNIC in model (1) is not only statistically significant, but
that it is also large in relation to the mean of the dependent variable. In model (1), it
is generally at least as large as the mean of the dependent variable. Since ETHNIC is
defined to be between 0 and %, this suggests that ethnic divisions have a substantial
impact on the level of pollution. The results of model (2) for the coefficient of
ETHNIC on emissions of S02 are not as strong. Although generally negative in sign,
consistent with the results of model (1), the coefficient of ETHNIC is generally
insignificantly different from zero. Interesting results are found for model (3) for
S02. As alluded to above, although there are only 9 additional observations from
1970, they have a dramatic effect on the sign of the coefficient of ETHNIC. The four
specifications of table 6 have a consistently negative and significant coefficient of
ETHNIC, whereas in table 7, without the data from 1970, it is generally
insignificant, and even positive and significant in specification 2. This gives rise to
concerns that the counterintuitive sign of the coefficient of ETHNIC for the S02
regressions is not robust to the addition of fixed effects for MSAs, at least not
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
38
without the addition of the (possibly anomalous) 1970 data1 5 . For CO, the coefficient
of ETHNIC is significant in models (2) and (3), where it is positive. Model (1) has
generally insignificant results for ETHNIC, possibly reflecting the fact that CO air
pollution is often not locally generated.
An obvious concern with the results from specification one is that they may not be
being driven by ETHNIC, but rather by some other variable which is closely related
to it. Most obviously, this might be the percentage of the population that is Black. To
this end, specification four includes the percentage of the population that is Black as
well as ETHNIC. The reasoning behind this was outlined by Alesina, Baqir and
Easterly (1999, p.1270). Since the largest minority group in the U.S. in this time
period is blacks, ETHNIC is highly correlated with BLACK, with a correlation
coefficient in my data of 0.72. However, from the point of view of my theoretical
section, the two variables are quite different. ETHNIC treats as equivalent two
observations one of which is 70% white and 30% black, and the other 30% white and
70% black. In fact, the results on ETHNIC turn out not to be driven by their
relationship to BLACK. For S02, the significance of the coefficient of ETHNIC is
unchanged, and indeed, in model (2), the size of the ETHNIC coefficient even
increases substantially. The coefficient of BLACK itself is generally positive
1 5 However, when the number observations o f the explanatory variables from 1980 and 1990 census
data is artificially increased by linear interpolation, the counterintuitive finding on the coefficient of
ETHNIC remains.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
39
whenever it is significantly different from zero. The results for CO are more
ambiguous. In model (2), the addition of BLACK renders the coefficient of ETHNIC
insignificant.
Specification three adds the variable denoting housing segregation. For S02, as
hypothesized, its coefficient is positive and significant, as expected, in model (1) and
(3) including 1970 data. It is insignificant in model (2). It was expected to be
positive, as given greater residential segregation, it would be easier to ‘direct’
pollution to a segregated neighborhood that lacks a political voice. It does not,
however, change the sign or significance of the coefficient of ETHNIC. Less clear
results are found for CO. The coefficient of housing segregation changes sign from
being negative and significant in model (1) to positive and significant when fixed
effects are added in model (3). It also causes the coefficient of ETHNIC to become
insignificantly different from zero in model (3).
Specification four introduces the capital-labor ratio. This coefficient has the
anticipated positive and significant sign for S02 in model (1). It is insignificant in
models (2) and (3). Notice that in both models the size of the coefficient is by far the
largest out of any of the controls used, reflecting the importance of this control for
the generation of S02. For CO, the coefficient of the capital-labor ratio is negative
and significant in model (1), and insignificantly different from zero in models (2)
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
40
and (3). These unexpected results may be because the capital-labor ratio is less well
suited to CO than to S02, as CO is less a product of industry.
Specification five uses the control for the sulfur content of coal used in electricity
generation in that state. It is used only for the S02 regressions as it pertains only to
sulfur generation. Further, since the data is available only for a point in time, and not
as a time series, it cannot be used in the fixed effects regressions as it would be
perfectly collinear with the MSA fixed effect. Thus, this explanatory variable
appears only in tables 4 and 5, for models (1) and (2) respectively. The coefficient
has the anticipated positive and significant sign in table 4. It is still positive, although
insignificantly different from zero, in table 5, for model (2). Note that in table 4, it
has the greatest effect of all of the specifications in reducing the size of the
coefficient of ETHNIC.
Specification six attempts to control for collective action problems in reducing air
pollution. This is done through the introduction of an independent variable denoting
the number of states that the MSA straddles. Although the coefficient generally has
the anticipated positive sign, it is usually insignificantly different from zero, if only
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
41
TABLE 7
1 2 3 4 5 6 7
NUMBER OF OBSERVATIONS 170 170 170 170 170 170 170
R2 0.1260 0.1440 0.1864 0.1749 0.2104 0.1348 0.2125
ADJUSTED R2 0.1208 0.1338 0.1767 0.1651 0.2009 0.1244 0.2031
MEAN OF DEPENDENT VARIABLE 0.0075 0.0075 0.0075 0.0075 0.0075 0.0075 0.0075
INTERCEPT 0.0106 *** 0.0108 *** 0.0053 *** 0.0088 — 0.0070 *** 0.0097 *** 0.0307
0.0007 0.0007 0.0017 0.0009 0.0011 0.0010 0.0047
ETHNIC -0.0113 *** -0.0157 *** -0.0108 *** -0.0105 * “ -0.0069 * " -0.0115 *** -0.0092
0.0023 0.0033 0.0022 0.0022 0.0024 0.0023 0.0022
, OF POPULATION BLACK 0.0086 ’
0.0046
HOUSING SEGREGATION 0.0093 '
0.0026
CAPITAL/LABOR RATIO IN MANUFACTURING 1.4911 '
0.4738
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
0.0021 '
0.0005
NUMBER OF STATES IN MSA 0.0008
0.0006
MEAN DIVIDED BY MEDIAN
HOUSEHOLD INCOME
-0.0176 ‘
0.0041
Each column In the table reports coefficients from an OLS regression. T he unit of observation Is a metropolitan
statistical area (city). The dependent variable is a m ean of annual observation m eans from monitoring stations.
Standard errors are underneath the coefficient estim ates. O bservations from the years 1980 and 1990 only
are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
42
TABLE 8
1 2 3 4 5 6 7
NUMBER OF OBSERVATIONS 72 72 72 72 72 72 72
R2 0.0115 0.0430 0.0117 0.0185 0.0459 0.0241 0.0162
ADJUSTED R2 -0.0027 0.0153 -0.0169 -0.0100 0.0182 -0.0042 -0.0123
MEAN OF DEPENDENT VARIABLE 34.36 34.36 34.36 34.36 34.36 34.36 34.36
INTERCEPT 43.92 '
11.74
46.81 1
11.79
40.18
29.12
36.40 1
15.92
18.64
19.79
34.69 '
15.27
-33.08
133.45
ETHNIC -34.27
38.05
-91.38 '
53.44
-33.25
39.00
-31.46
38.40
0.86
43.74
-38.90
38.40
-42.85
41.00
, OF POPULATION BLACK 105.54
69.99
HOUSING SEGREGATION 6.45
45.90
CAPITAL/LABOR RATIO IN MANUFACTURING 6044.38
8615.05
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
13.06
8.28
NUMBER OF STATES IN MSA 8.91
9.43
MEAN DIVIDED BY MEDIAN
HOUSEHOLD INCOME
63.77
110.08
Each column in th e table reports coefficients from an OLS regression. The unit of observation is a metropolitan
statistical a re a (city). The dependent variable is the em issions of sulfur dioxide in that MSA in pounds, from
both point and non-point sources, divided by the land area of the MSA. Standard errors are underneath the
coefficient estim ates. O bservations from the year 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
TABLE 9
FIXED EFFECT PANEL DATA REGRESSIONS INCLUDING 1970 DATA: SULFUR DIOXIDE AIR POLLUTION- OBSERVED
1 2 2 4 5 6
NUMBER OF OBSERVATIONS 187 187 187 187
R2 0.6723 0.6839 0.6870 0.6738
MEAN OF DEPENDENT VARIABLE 0.0087 0.0087 0.0087 0.0087
INTERCEPT 0.0425 •
0.0056
0.0329 *
0.0075
0.0256 1
0.0097
0.0410 '
0.0060
ETHNIC -0.1219 '
0.0159
-0.1349 '
0.0171
-0.1160 *
0.0158
-0.1249 ‘
0.0166
> OF POPULATION BLACK 0.1149 •
0.0612
HOUSING SEGREGATION 0.0213 '
0.0100
CAPITAL/LABOR RATIO IN MANUFACTURING 1.4145
2.1311
Each column in the table reports coefficients from a fixed effects panel data regression. The unit of observation
is a metropolitan statistical a re a (city). The dependent variable is a m ean of annual observation m eans from
monitoring stations. Standard errors are underneath the coefficient estim ates. O bservations from th e years
197 0,1980 and 1990 are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1 % level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
44
TABLE 10
FIXED EFFECT PANEL DATA REGRESSIONS EXCLUDING 1970 DATA: SULFUR DIOXIDE AIR POLLUTION- OBSERVED
1 2 2 4 5 6
NUMBER OF OBSERVATIONS 178 178 178 178
R2 0.7774 0.7827 0.7774 0.7801
MEAN OF DEPENDENT VARIABLE 0.0079 0.0079 0.0079 0.0079
INTERCEPT 0.0122 *** 0.0156*** 0.0122 ** 0 .0 1 3 2 '
0.0039 0.0045 0.0057 0.0040
ETHNIC 0.0158 0.0306 * 0.0158 0.0201
0.0142 0.0175 0.0152 0.0148
% OF POPULATION BLACK -0.0608
0.0420
HOUSING SEGREGATION 0.0001
0.0072
CAPITAL/LABOR RATIO IN MANUFACTURING -1.2133
1.1848
Each column In th e table reports coefficients from a fixed effects panel data regression. T he unit of observation
Is a metropolitan statistical area (city). The dependent variable is a m ean of annual observation m ean s from
monitoring stations. Standard errors are underneath the coefficient estim ates. O bservations from the years
1980 and 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
45
TABLE 11
1 2 3 4 5 6 7
NUMBER OF OBSERVATIONS 170 170 170 170 170 170
R2 0.0106 0.0145 0.0464 0.0595 0.0114 0.1276
ADJUSTED R2 0.0047 0.0027 0.0349 0.0483 -0.0005 0.1171
MEAN OF DEPENDENT VARIABLE 1.3030 1.3030 1.3030 1.3030 1.3030 1.3030
INTERCEPT 1 .4 2 5 0 " * 1.4201 * " 1.8206 *** 1.6612 *** 1.4641 *** 4 .5 4 1 2 '
0.1019 0.1021 0.1871 0.1278 0.1467 0.6653
ETHNIC -0.4471 -0.2543 -0.4338 -0.6472 * -0.4465 -0.1617
0.3340 0.4085 0.3289 0.3335 0.3348 0.3203
% OF POPULATION BLACK -0.4851
0.5908
HOUSING SEGREGATION -0.7750 **
0.3095
CAPITAL/LABOR RATIO IN MANUFACTURING -194.98 ***
66.1212
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
NUMBER OF STATES IN MSA -0.0341
0.0918
MEAN DIVIDED BY MEDIAN -2.7057 *
HOUSEHOLD INCOME 0.5716
Each column In th e table reports coefficients from an OLS regression. The unit of observation is a
metropolitan statistical area (city). T he dependent variable is a m ean of annual observation m eans from
monitoring stations. Standard errors are underneath the coefficient estim ates. O bservations from the years
1980 and 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
46
TABLE 12
1 2 3 4 5 6 7
NUMBER OF OBSERVATIONS 68 68 68 68 68 68
R2 0.0743 0.1080 0.0942 0.0813 0.1115 0.1133
ADJUSTED R2 0.0603 0.0806 0.0663 0.0530 0.0841 0.0860
MEAN OF DEPENDENT VARIABLE 120.595 120.595 120.595 120.595 120.595 120.595
INTERCEPT 80.6586 *“ 80.5370 * " 32.7093 6 6 .6 2 0 6 '
19.1318 18.9246 44.4726 27.6890
53.9799 '
24.8739
449.634 '
219.019
ETHNIC 147.496 ’
64.0793
80.6427
76.4081
166.784 '
65.8871
161.799 '
67.4597
139.873 '
63.4305
187.517 '
67.4820
> OF POPULATION BLACK 182.990
116.792
HOUSING SEGREGATION 82.9145
69.4732
CAPITAL/LABOR RATIO IN MANUFACTURING 10481
14891
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
NUMBER OF STATES IN MSA 24.7404
15.0094
MEAN DIVIDED BY MEDIAN
HOUSEHOLD INCOME
-305 '
180.462
E ach column in th e table reports coefficients from an OLS regression. T he unit of observation is a metropolitan
statistical area. The dependent variable is the em issions of carbon monoxide in that MSA in pounds, from
both point and non-point sources, divided by the land area of the MSA. Standard errors are underneath the
coefficient estim ates. O bservations from the year 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
47
TABLE 13
FIXED EFFECT PANEL DATA REGRESSIONS: CARBON MONOXIDE AIR POLLUTION- OBSERVED
1 2 3 4
NUMBER OF OBSERVATIONS 170 170 170 170
R2 0.6363 0.6371 0.6939 0.6413
MEAN OF DEPENDENT VARIABLE 1.3030 1.3030 1.3030 1.3030
INTERCEPT 0.7464 * 0.7815 * -2.2183 *** 0 .8 9 8 5 '
0.3906 0.4011 0.8324 0.4150
ETHNIC 4.6214 ** 5.0974 ** 2.1451 5 .2 2 2 8 '
1.8798 2.1979 1.8447 1.9592
% OF POPULATION BLACK -2.6679
6.2971
HOUSING SEGREGATION 5.1882 ***
1.3131
CAPITAL/LABOR RATIO IN MANUFACTURING -183.46
170.30
MEAN DIVIDED BY MEDIAN
HOUSEHOLD INCOME
Each column in the table reports coefficients from a fixed effects panel data regression. The unit of observation
is a metropolitan statistical a re a (city). The dependent variable is a m ean of annual observation m eans from
monitoring stations. S tandard errors are underneath the coefficient estim ates. O bservations from the years
1980 and 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
7
170
0.7621
1.3030
4.9706 ***
0.7126
0.3715
1.6588
-3.4598
0.5224
48
TABLE 14
1 2 3 4
NUMBER OF OBSERVATIONS 100 100 100 100
R2 0.0509 0.0925 0.0572 0.0747
ADJUSTED R2 0.0412 0.0738 0.0378 0.0557
MEAN OF DEPENDENT VARIABLE 30.0766 30.0766 30.0766 30.0766
INTERCEPT 26.3208 *** 25.7705 *** 29.2834 “ * 28.4905
1.8104 1.7984 4.0921 2.2609
ETHNIC 13.8234 ** 25.1179 *** 13.1431 ** 12.8678
6.0294 7.9888 6.0986 6.0143
% OF POPULATION BLACK -22.46 **
10.6545
HOUSING SEGREGATION -5.3453
6 7
100 100
0.0510 0.0548
0.0314 0.0353
30.0766 30.0766
26.5077 *‘* 12.5020
2.5283 21.8844
13.8749 ** 12.2584
6.0793 6.5328
6.6183
CAPITAL/LABOR RATIO IN MANUFACTURING -1745.4
1104.03
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
NUMBER OF STATES IN MSA -0.1763
1.6551
MEAN DIVIDED BY MEDIAN 11.4430
HOUSEHOLD INCOME 18.0594
Each column in the table reports coefficients from an OLS regression. The unit of observation is a metropolitan
statistical a re a (city). T he d ependent variable is a m ean of annual observation m ean s from monitoring stations.
Standard errors are underneath the coefficient estim ates. O bservations from the years 1980 and 1990 only
are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
49
TABLE 15
OLS REGRESSIONS: PM10 AIR POLLUTION- EMISSIONS FROM ALL SOURCES DIVIDED BY AREA
§ 6
100
0.0270
0.0070
25.7084
1 2 2 4
NUMBER OF OBSERVATIONS 100 100 100 100
R2 0.0266 0.0276 0.0292 0.0578
ADJUSTED R2 0.0166 0.0076 0.0092 0.0384
MEAN OF DEPENDENT VARIABLE 25.7084 25.7084 25.7084 25.7084
INTERCEPT 19.7117 *** 19.5190 *** 15.4754 * 14.2238 *** 18.8729 '
4.0518 4.1140 9.1767 5.0420 5.6577
ETHNIC 22.0709 26.0265 23.0437 * 24.4880 * 21.8399
13.4944 18.2750 13.6764 13.4124 13.6037
% OF POPULATION BLACK -7.8668
24.3729
HOUSING SEGREGATION 7.6433
14.8418
CAPITAL/LABOR RATIO IN MANUFACTURING 4414.65 *
2462.10
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
NUMBER OF STATES IN MSA 0.7909
3.7035
MEAN DIVIDED BY MEDIAN
HOUSEHOLD INCOME
Each column in the table reports coefficients from an OLS regression. T he unit of observation is a metropolitan
statistical area (city). The dependent variable is the em issions of PM10 in that MSA in pounds, from both
point and non-point sources, divided by the land area of the MSA. Standard errors are underneath the
coefficient estim ates. O bservations from the years 1980 and 1990
only are used (50 observations from each year).
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1 % level
7
100
0.0490
0.0294
25.7084
92.8237 *
48.5123
30.3512 **
14.4817
-60.54
40.0332
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
50
TABLE 16
OLS REGRESSIONS: NITROGEN AIR POLLUTION- OBSERVED
1 2 3 4 5 6 7
NUMBER OF OBSERVATIONS 56 56 56 56 56 56
R2 0.0079 0.0436 0.0157 0.1035 0.0427 0.0131
ADJUSTED R2 -0.0105 0.0076 -0.0214 0.0697 0.0066 -0.0241
MEAN OF DEPENDENT VARIABLE 0.0165 0.0165 0.0165 0.0165 0.0165 0.0165
INTERCEPT 0.0153*** 0.0149*** 0.0181 *** 0.0194*** 0.0130 *** 0.0036
0.0020 0.0020 0.0048 0.0026 0.0026 0.0221
ETHNIC 0.0042 0.0114 0.0031 0.0016 0.0037 0.0032
0.0063 0.0081 0.0066 0.0062 0.0063 0.0066
% OF POPULATION BLACK -0.0151
0.0107
HOUSING SEGREGATION -0.0049
0.0075
CAPITAL/LABOR RATIO IN MANUFACTURING -3.2517 * *
1.3678
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
NUMBER OF STATES IN MSA 0.0020
0.0014
MEAN DIVIDED BY MEDIAN 0.0096
HOUSEHOLD INCOME 0.0180
Each column In the table reports coefficients from an OLS regression. The unit of observation is a metropolitan
statistical a re a (city). T he dependent variable is a m ean of annual observation m eans from monitoring stations.
Standard errors are underneath the coefficient estim ates. O bservations from the years 1980 and 1990
only are used (28 observations from each year).
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
51
TABLE 17
OLS REGRESSIONS: NITROGEN AIR POLLUTION- EMISSIONS FROM ALL SOURCES DIVIDED BY AREA
1 2 3 4 5 6
NUMBER OF OBSERVATIONS 56 56 56 56 56
R2 0.0031 0.2687 0.1942 0.0763 0.1058
ADJUSTED R2 -0.0153 0.2411 0.1638 0.0414 0.0720
MEAN OF DEPENDENT VARIABLE 34.1516 34.1516 34.1516 34.1516 34.1516
INTERCEPT 31.1527 *** 34.6331 *** -23.80 17.0399 16.0813
8.0098 6.9700 17.1213 10.3929 9.7967
ETHNIC 10.2149 -67.17 ** 31.1144 18.8941 6.9708
24.8759 27.8143 23.3320 24.5389 23.8180
% OF POPULATION BLACK 161.42 ***
36.7930
HOUSING SEGREGATION 94.7260 ***
26.7192
CAPITAL/LABOR RATIO IN MANUFACTURING 11148.0 * *
5441.04
SULFUR CONTENT OF COAL USED FOR
ELECTRICITY GENERATION BY STATE
NUMBER OF STATES IN MSA 13.3930*
5.4300
MEAN DIVIDED BY MEDIAN
HOUSEHOLD INCOME
Each column in the table reports coefficients from an OLS regression. The unit of observation is a metropolitan
statistical area (city). The dependent variable is the em issions of nitrogen In that MSA In pounds, from both
point and non-point sources, divided by th e land area of the MSA. Standard errors are underneath the coefficient
estim ates. O bservations from the year 1990 only are u sed (28 observations from each year).
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
7
56
0.0066
-0.0308
34.1516
-6.4118
86.9093
7.2768
25.9627
30.7299
70.7897
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
52
marginally so1 6 . The anticipated result was found for nitrogen pollution in table 14.
Given that some states have different attitudes to air quality than others do
(Henderson, 1994 unpublished paper), this variable was anticipated to be positive,
reflecting the difficulty of agreeing upon common standards.
Finally, specification seven controls for income inequality. Most surprisingly, this
coefficient is always negative whenever it is significantly different from zero. That is
to say, more unequally distributed income is associated with lower levels of air
pollution, not higher levels as hypothesized. This is a very unexpected result.
Possibly, the reason this coefficient associated with this variable has the opposite
sign than those found in earlier studies (Alesina et al, QJE, 1999) is because they
dealt with goods directly taxed for by the government, such as education and roads.
Air quality does not involve such direct fiscal mechanisms. Note that in tables 6 and
7, for model (3) for S02, this variable is not used. That is because the data is not
available from the 1970 census, and an effort was made to ensure comparability
between the data used for tables 6 and 7.
1 6 Arguably more impressive results are found for the other controls for governmental collective
action problems considered in the appendix. These are the number o f cities and counties in the MSA.
These were relegated to the appendix because missing data meant that the sample size was reduced.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
53
SUMMARY AND SUGGESTIONS FOR FURTHER RESEARCH
The results of the present paper are consistent with the existing literature in the
case of Carbon Monoxide, in that emissions of CO appear to be greater in ethnically
diverse American MSAs. Counterintuitive results are found for S02. That is, S02
appears to be inversely associated with ethnic diversity, implying that the provision
of the public good of air quality is actually better in these areas.
These pollutants can be distinguished from each other along two dimensions, as
follows;
TABLE 18
SOURCE AND MOBILITY OF AIR POLLUTANTS
Source of Pollutant
Point Non-Point
Non-Mobile S02 PM 10
Mobile CO
Since S02 is the only pollutant whose emissions overwhelmingly come from point
sources, it appears that this, and not the fact that S02 is also a relatively immobile
pollutant, is the distinguishing characteristic of S02.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
54
No convincing explanation for the unexpected result on S02 is offered here.
However, some conjecture is in order. The most likely explanation for the result
appears to be that point and non-point air pollution emissions are regulated at
different levels of government in the U.S. Because of the complexity of regulation,
no easy generalizations are possible, but it seems that point sources are generally
subject to more local, rather than federal, oversight. For example, SIP (State
Implementation Plans) devised in order to bring counties in the U.S. back into
compliance with the Clean Air Act, are devised at the State level, but subject to local
implementation. Earlier research cited, such as Henderson (JPE, AER) notes the
control that localities can have over air pollution.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
55
CHAPTER 2- POLITICAL EFFECTS IN ENVIRONMENTAL LITIGATION
INTRODUCTION
The ideal of democracy is the rule of law, not the arbitrary rule of a dictator. To that
end, much of the mundane day-to-day business of governing is carried out by a vast
bureaucracy, which is supposedly not subject to capricious political whims.
Bureaucrats supposedly have a clear apolitical task that they pursue with no political
interference. Politicians can change this mission through new legislation, but only if
they explicitly set out the new mission. In this view of the nature of the relationship
between politicians and bureaucrats, political changes should lead to changes in the
behavior of the bureaucracy only through new legislation.
Plainly, this view is at odds with reality. There is an extensive literature documenting
the effects of political interference in bureaucratic agencies. Much of this deals with
the enforcement of antitrust disputes as mediated by the FTC (Federal Trade
Commission). A probable explanation for this is that there is such a clear economic
justification for antitrust, unlike for much other economic regulation of the economy.
There is a similarly strong economic case to be made for environmental regulation to
internalize the externalities of pollution, but a corresponding literature on the
political control of environmental bureaucracy does not exist. The present paper aims
to fill this gap. It uses the change in the price of publicly quoted companies that have
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
56
civil judicial litigation outstanding with the EPA (Environmental Protection Agency)
when political events take place, such as Presidential and Congressional elections, to
assess the market’s reaction to these events. The assumption is that the stock market
reflects all available information, in the form of the EMH (Efficient Markets
Hypothesis). Furthermore, it is assumed that the market can distinguish differences
between candidates, which is not at all obvious to the lay observer, given the fact that
politicians often confuse the electorate and make unrealistic promises to win votes.
The effect of political events on environmental regulation is especially interesting
given the current controversy concerning variations in the aggressiveness of the EPA
from one administration to another. There is a perception that it could vary in
strictness from one administration to the next1 7 . There is plenty of anecdotal
evidence that this view may be correct. For example, there are newspaper accounts
of staff attrition and deliberate non-enforcement of environmental law under the
current Republican administration1 8 . The present paper aims to subject these
anecdotal impressions to empirical tests.
1 7 ‘The real special-interest payoffs come via less showy policies, like the way the [present]
administration is undermining enforcement o f the Clean Air Act’ (Paul Krugman, March 15th 2002,
page 23, N ew York Times).
1 8 Examples o f this are the resignation o f Eric Schaeffer, the EPA’s director o f regulatory
enforcement, who said that he had been “fighting a White House that seems determined to weaken the
rules we are trying to enforce”, and complained about the cutting o f more than 200 staff positions
from the civil enforcement program (Los Angeles Times, March 1s t 2002, p. 17); also legislation
making it harder to sue older coal-fired power plants and refineries (Washington Post, June 14th 2002,
p.A l).
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
57
LITERATURE REVIEW
Yet, the idea that political and legislative changes cause changes in market
valuations is by no means obvious. Indeed, several studies have found that there does
not seem to be any strong relationship between the timing of elections and the type
of political leadership, on the one hand, and aggregate movements in the stock
market, on the other hand. Stovall (1992, 5) said that there had been virtually no
difference between the growth of the Dow Jones Industrial Average during
Democratic and Republican administrations over the course of the twentieth century,
although the market did prefer Republicans since 1945. Niederhoffer et al (1970)
reached a similar conclusion as to the indifference of market aggregates to the party
of presidential administrations. On the other hand, Riley and Luksetich (1980) did
find some evidence that, in the short term at least, markets prefer Republicans.
These studies make the assumption of stationarity on the behavior of political parties
over time. That is to say, they pool data across elections. This assumption is often
unwarranted, especially in the American political system, where the Democratic and
Republican parties are closer to each other in terms of policy and are much less
ideologically cohesive than European political parties. There is little justification for
assuming that market aggregates should have performed similarly under Reagan and
Nixon, except for their common membership of the same loosely defined political
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
58
party (Beck, 1982). Indeed, some of the feistiest political battles of the 1990s have
been fought between members of the same political party, such as the Republican
party Presidential primary elections in 1992 and 1996, which pitted the protectionist
economic nationalist Pat Buchanan against the eventual nominees, George Bush
senior and Bob Dole. Also, there is the problem of causality. Do people elect
Republicans in good economic times, or Republicans cause good economic times?
The biggest problem in doing empirical studies of the effect of political and
legislative changes on stock market valuations is the fact that these changes happen
slowly over a period of months or years. Given the assumption of efficient markets,
stock markets should immediately reflect all available information, including the
possibility of political changes. It is thus impossible to obtain statistically significant
results for the market’s valuation of the political change, as the date that the market
learned the information is either unknown or the information reached the market
only slowly (Salinger, 1992). The problem of the gradual release of information is
the reason why studies of the effect of legislation and political change on market
valuations have generally found insignificant results (Binder, 1985). There have also
been studies on the effect of legislation on security prices, such as the 1986 Tax
Reform Act (Cutler, 1988; Downs and Tehranian, 1988; Graddy et al, 1992), mostly
with disappointing results.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
59
Yet, there have been some exceptions. Gilligan and Krehbiel’s study (1988) that
showed that the 1974 Energy Tax Act of 1974 increased the returns to oil and gas
stocks. A macabre exception to the inconclusive results in this literature is Roberts’
(1990a) quantification of the importance of Congressional seniority. When Henry
‘Scoop’ Jackson, the ranking Democrat on the Senate Armed Services Committee
died suddenly of a ruptured coronary artery on September 1st 1983, the share prices
of defense contractors based in his district fell, and those of Sam Nunn, the next in
seniority to assume his position, simultaneously rose. The fact that his tragic death
was unexpected was the reason they found significant results. An excellent recent
paper on the gradual release of information from political changes to the stock
market is Ellison and Mullin (2001). They use a new technique, isotonic regression,
to quantify the loss of value of pharmaceutical stocks because of the healthcare
reforms of President Clinton in 1992-93.
Most of the existing literature in this field has used opinion poll data (Manning,
1989; Brander, 1991; Langohr and Viallet, 1986; Gemmill, 1992). The disadvantages
of this are well-known; they are usually conducted over several days, and so do not
represent the probability of the election of a candidate on a single day; they are not
usually released daily; they have wide margins of error, especially long before an
election when many people have not yet made up their minds who to support; they
are subject to manipulation and questionable methodology, especially as people often
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
60
engage in ‘preference falsification’ by wanting to vote for the winner (Kuran, 1995),
and so politicians have an incentive to show themselves in the lead, and finally, they
have often been wrong (Herron et al, 1999, 58-9).
These are also some of the reasons that economists are generally skeptical of
surveys, preferring to rely instead on revealed preferences. Another problem with
opinion polls is that they cannot be used to measure the probability of which party
will control Congress, as that is based on a multitude of separate elections.
The only other possibility is the use of bookmaker’s quoted odds. Such a measure
was used by Roberts (1990b) to measure who was likely to win the 1980 U.S.
Presidential election. This is certainly better than opinion polls, but it still has several
disadvantages. Odds are not quoted daily, and so must be interpolated. Roberts also
had no quoted odds for the probability of a Republican Senate majority in 1980, and
so he simply used a dummy variable which assumes the value of 1 on the first
trading day after the election, and zero otherwise.
The portion of the literature that analyzes the effect of elections on specific
companies or sectors of the economy is quite small. The distinction between this
literature and that focusing on legislative as opposed to political changes is not
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
6 1
arbitrary, although they are closely related. Legislation is specific, whereas
politicians make careers out of being vague.
The present paper has a number of advantages over this existing literature.
Firstly, it uses an objective measure of the probability of a political event. The
measure of the probability of the outcome of an election is obtained from the Iowa
Electronic Market (IEM). Only Herron et al. (1999) and Knight (2004) have used
this data source for similar purpose before1 9 . The IEM is a set of continuous double
auction markets operated by the University of Iowa Henry B. Tippie School of
Business since 1988. They have traded on the outcome of political events, such as
elections in the U.S. and elsewhere, including celebrity elections such as Hillary
Clinton’s run for a Senate seat in New York, as well as on such diverse topics as the
probability of Federal Reserve interest rate changes and the box office receipts of
movies. Nine IEM markets are used in the present paper. The IEM has usable data
only from 1992 onwards. It had a market for the outcome of the 1988 U.S.
Presidential election, but that was a VS (vote share) rather than a WTA (winner takes
all) market, and thus could not be used as an estimate of the probability of one
candidate or the other winning. The IEM did not have markets for the 1998
Congressional U.S. elections.
1 9 However, the IEM has already been used in dozens o f published papers in academic journals (listed
here: http://www.biz.uiowa.edu/iem/archive/references.html) for the purposes o f different studies, and
so it is a respectable source o f data
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
62
Secondly, it attempts to quantify the effects of the election of the President, U.S.
Senate and U.S. House of Representatives, separately and together. Previous papers
have estimated the effect of the President alone (Herron et al, 1999; Homaifar et al,
1988), or the President and the U.S. Senate (Roberts, 1990b), but none have analyzed
the election of the President and both houses of Congress together. This is important,
since any of these three can exercise effective veto power over the others, implying
that any study analyzing them individually would be incomplete. Recent research
indicates that such ‘checks and balances’ are beneficial for economic growth
(Henisz, 2000). Indeed, Henisz constructed an index of political constraints from
data on the number of institutional players in a given polity, partisan alignments
across institutions and data on the party composition of legislatures. Data is not
pooled across elections, as the literature has suggested that that may be
inappropriate. For example, it is generally acknowledged that the House of
Representatives Republican leadership elected in 1994 was more radical than that
elected in 1996. There is some previous research that examines the effect of political
changes in Congress and the Presidency on the output of bureaucracies (Wood and
Waterman, 1991; Moe, 1985). However, such authors use other political events, such
as appointments and hearings, not elections. Also, they only look at agency outputs,
such as actions, litigation and violations found, not the effect of those actions in the
eyes of the market as measured through stock market valuations. Their methodology
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
63
may be questionable, as it relies on the subjective selection of politically important
events and then their quantitative analysis, so there may be a sample-selection
problem.
Finally, this paper makes a contribution by attempting to quantify the importance of
the non-transparent exercise of political power, that is to say, political power which
is not confined to simply increasing or decreasing the EPA’s budget. Much of the
previous literature deals with defense (Herron et al., 1999; Roberts, 1990b; Homaifar
et al., 1988). Unlike defense contractors, who have a clear interest in the stated
defense policy aims of politicians, as they can vote for bigger defense budgets, it is
not at all clear how politicians affect the prosecution of environmental litigation.
While the present paper makes no serious effort to determine exactly how politicians
might affect the EPA, but it does suggest that they can affect it in ways that go
beyond merely the appropriation of funds. In that sense, this paper is similar to
Roberts’ (1990a) quantification on the importance of Congressional seniority
through the sudden death of a senator. Commentators had long trumpeted the
importance of seniority. Roberts’ contribution was to empirically demonstrate and
quantify it, although he did too not measure exactly how its influence was exerted.
What are the institutional mechanisms through which politicians might control the
EPA? It is to this point that this paper now turns.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
64
INSTITUTIONS AND POLICY
This section attempts to suggest possible ways in which politicians can exert control
over bureaucracies. Three mechanisms of political control are outlined.
The first and probably most important political control mechanism is that both
Congress and the President control who is appointed and reappointed to the federal
bureaucracy. In the case of the EPA, the most important of these is the administrator.
A superficial glance at Congressional confirmation hearings may seem to suggest
that they simply ‘rubber-stamp’ nominees (Weingast and Moran, 1983, 769).
However, the real decisions of who the nominees would be are less public, but are
subject to close scrutiny by politicians, both the President and Congress. For
example, the official confirmation hearings of ICC (Interstate Commerce
Commission) commissioners were an average of only 17 minutes during the period
of 1949-74, but the earlier vetting process before the nominees were named was far
more thorough.
Furthermore, there is evidence that politicians design the organizational structure of
bureaucracies with a view to maximizing their political influence over them.
Shughart, Tollison and Goff (1986, 969) show that bureaucratic agencies which are
given wide-ranging responsibilities are more likely to be set up as commissions with
multiple administrators rather than to be placed under the control of a single
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
65
administrator, as that structure facilitates Congressional oversight. The EPA operates
under a single administrator, appointed subject to Congressional veto. The output of
agencies operating under commissions has a smaller variance than those under a
single administrator.
The second political control mechanism is that Congressional committees exert
power over bureaucracies. Most of the empirical work in this area focuses on
Congressional committees. Probably this is because such committees are amenable
to analysis, and studies based on the analysis of committees have yielded stronger
results. For example, Moran and Weingast (1982) find that mean Congressional
ADA (Americans for Democratic Action, a liberal pressure group) scores are
correlated with the type of cases that the FTC (Federal Trade Commission) pursues.
They also find (JPE, 1983, 790) that this effect becomes stronger the better placed
the politician is within Congress; a member of a subcommittee with oversight over
the FTC has around 2‘ A times the influence of an ordinary member of Congress, and
a subcommittee chairman has around 12 times the influence.
There is also evidence that politicians can direct the benefits of being on a committee
to their districts in the same way as ‘pork barrel’ spending, or locally beneficial
spending which would not pass a cost-benefit test as a globally useful public good.
For example, Faith, Leavens and Tollison (1982) find that antitrust complaints
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
66
against a company based in the district of a politician who is on a committee with
oversight over the FTC are more likely to be dismissed. This is especially true when
the complaint comes from outside the politician’s district. In other words, when the
beneficiary of the complaint is not also in the politician’s home district, dismissal is
more likely. Similarly, Coate, Higgins and McChesney (1990) find that challenges to
proposed horizontal mergers by the FTC are made more likely by Congressional
hearings and newspaper stories in the Wall Street Journal. This is because merger
challenges prevent the exit of resources, and with them, votes, from a politician’s
district.
Not all studies have found that Congress has influence over bureaucratic agencies.
Anderson (1993) found that the ITC (International Trade Commission) is not
influenced by political considerations in their analysis of the granting of import
protection. They qualify this by saying that it does not imply that politics plays no
role in the granting of anti-dumping protection, only rather that their analysis did not
find it. Clearly, politics is important in the selection and confirmation of nominees.
The Congressional committees with the most responsibility for oversight over the
EPA are the House Energy and Commerce Committee and the Senate Environment
and Public Works Committee. These are the most active in terms of hearings,
although the responsibility is widely dispersed (Helland, 2001, 282). The little
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
67
previous research into their effect on policy output has not yielded firm conclusions
(Lazarus, 1991).
The third political control mechanism is that Congress and the President jointly
decide on the level of funding for the EPA, along with all the other federal
bureaucracies20. Several thousand federal agencies face only 13 subcommittees of
the Appropriations committee. Each subcommittee develops a bill that is amended to
become the federal budgetary allocation (Weingast and Moran, 1983, 769). It is
possible to obtain “Congressional Justifications” that the EPA prepares when
requesting money. Unfortunately, sophisticated analyses of which politicians were in
favor of greater or lesser funds for the EPA is not possible, as it is voted upon in an
omnibus bill called the “VA/HUD Bill”. This allocates funding to the Departments
of Veterans Affairs and Housing and Urban Development as well as to many
independent agencies, of which the EPA is only one.
2 0 At the beginning o f the Reagan administration there was a drastic change in the tactics o f the EPA.
140 cases were brought in 1980 (prior to Reagan), but only 45 in 1981, after Reagan became
President. This switch in strategy may have been driven at least partly by cuts in funding. It may also
have been because o f the reorganization o f the EPA office o f enforcement in 1981 under six program
areas, which some environmentalists charged was an effort to diffuse enforcement by putting it under
the control o f political appointees (Helland, 2001, 292).
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
68
FIGURE 1
EPA BUDGET AND WORKFORCE
9
8
7
6
5
4
3
2
1
0 0
1970 72 74 76 78 '80 '82 '84 '86 '88 '90 '92 '94 '96 '98 '00 '02 2004
I Budget ($Bn
left scale)
■Workforce
(thousands,
right scale)
In any case, the simple level of funding may not necessarily be a good indicator of
political support for the EPA itself, let alone for the efficiency with which the funds
are spent. Alesina, Baqir and Easterly (1998) show that many government jobs are
created more as political favors than to perform any useful tasks. The same may be
true of employment and expenditures by the EPA. They may not be equally well
spent under different administrators. There is also a substantial lag between the
timing of a political change caused by an election, and the voting and appropriation
of funds, and finally the spending of those funds. There does not appear to be any
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
69
obvious pattern in the EPA budget from 1992-2000 in response to shifts in political
0 1
power in the graph below.
DATA AND EMPIRICAL MODEL
Data
The dependent variable used in this paper is the abnormal daily returns to the stock
market price of a company that has environmental litigation outstanding on the day
of an election. The idea is that if the market considers the company to be politically
sensitive, then the returns to the company’s shares should respond to changes in the
probability of a political event as reported by the IEM. The stock market data comes
from the Center for Research in Security Prices (CRSP)22. The paper uses individual
company returns, as opposed to the construction of portfolios of stocks that are then
combined with data on returns to the portfolio, as used by Roberts (1990b) and
Herron et al (1999). The present paper uses interaction terms between the likelihood
of political events and individual characteristics of the companies, such as the
2 1 Source; U.S. Environmental Protection Agency, at Source; U.S. Environmental Protection Agency,
at http://www.epa.gov/history/org/resources/budget.htm
2 2 One problem with the use o f the IEM and CRSP data together is the fact that the two data sources
are non-synchronous. In both cases the last price o f the day’s trading is used. However, for the IEM
this occurs at 12 midnight CST (10pm PST), whereas for the CRSP this can occur at 4pm EST (1pm
PST) if the company’s stock is traded only on the East Coast, such as on the NYSE, or it could occur
at 5pm PST if the company is traded on the PACX (Pacific Stock Exchange). In other words, the
CRSP data lags the IEM data by between five and nine hours. Because neither the CRSP nor the IEM
have data on the time o f day their trades were executed, it is not possible to make the data perfectly
synchronous. The IEM is also open on weekends and holidays, unlike real stock exchanges. Thus on
Mondays, the change in the probability o f a political event is the change between the last quoted price
on that day and that on Friday, not Sunday.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
70
industry that they are in and the number of times they have been sued before. This is
not possible when aggregate portfolio returns are used, as that would not allow the
isolation o f disaggregated individual company effects.
Data is used from August 6 until the day before the election, which was on Tuesday
November 7th 2000. If the company only had litigation filed against it after August
6 th, then only its daily returns after the filing date were used. The reason the day of
the election itself was not used is because the IEM payout for the presidential
election is based on the winner of the popular vote, not on whoever actually became
the President. Usually these two things are the same, but the 2000 election was the
first since 1888 where this was not true. The data on environmental litigation was
obtained from the database of U.S EPA judicial civil actions that were outstanding at
the time of the election.
Empirical Model
As explained, the biggest problem in doing empirical studies of this type is the
problem of the gradual release of information to the market. This means that
traditional event study methodology is unlikely to be able to detect significant effects
even if they exist.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
71
Instead, this paper uses a two-stage methodology similar to Knight (2004). First,
abnormal returns to corporate defendants are calculated by estimating the following
market model;
r it = ct i + p i r m t + e it (1 )
where is the rate of return to company i on day t, and R m t is the equally weighted
return to the market with dividends at time t, as calculated by CRSP. This market
model is estimated over the year before the election, from August 6 th 1999 to August
5th 2000. Based upon the company specific regression coefficients above, the
abnormal returns, abr jt, are calculated for the period from August 6 th to November
6th 2000. These abnormal returns are then used as the dependent variable in the
model (2 ) below.
abr it = a j + p ! A Pr(Bush) t + P 2 A Pr(H_Rep) t
+ P 3 APr(S_Rep)t + u it (2 )
where;
A Pr(Bush) = the change in the probability of the election of George W. Bush23,
defined as Pr(Bush) t - Pr(Bush) t. 1
2 3 All o f the A probabilities are random walks. This is tested for by Lo and MacKinlay’s variance ratio
test (1988).
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
72
A Pr(H_Rep) = the change in the probability of the election of a Republican
majority in the House of Representatives, defined as Pr(H_Rep) t
-Pr(H _Rep)t.i
A Pr(S_Rep) = the change in the probability of the election of a Republican
majority in the Senate, defined as Pr(S_Rep) t - Pr(S Rep) t - 1
The fixed effects, a 1... 125 capture firm-specific trends for the 125 corporate
defendants during the election period. It is hypothesized that all of the companies
stand to gain from Republican election victories, and by implication, stand to lose
from the election of Democrats. This hypothesis therefore predicts that all three
capitalization parameters, /? i,/?2 and/? 3 >0. They have the interpretation of being the
percentage difference in market capitalization under a counterfactual President Gore,
Democratic House of Representatives and Democratic Senate, respectively, as
opposed to the actual aftermath of the 2 0 0 0 election, which (eventually) produced a
clean Republican sweep of the Presidency and both houses of Congress.
The first specification is exactly as given in (2), as shown in column 1 of table 1,
using all the variables that are available. As hypothesized, all three capitalization
parameters are positive, although only the two Congressional coefficients are
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
73
TABLE 19
ABNORMAL DAILY RETURNS TO COMPANIES WITH EPA LITIGATION OUTSTANDING DURING THE 2000 ELECTION
1 2 3 4
NUMBER OF OBSERVATIONS 6,778 6,778 6,778 6,778
NUMBER OF COMPANIES 125 125 125 125
TIME SERIES LENGTH 65 65 65 65
R2 0.017 0.017 0.017 0.017
MEAN OF DEPENDENT VARIABLE 0.0006 0.0006 0.0006 0.0006
CHANGE IN PROBABILITY OF GEORGE W. BUSH 0.006 0.005 -0.002 0.012
WINNING PRESIDENTIAL ELECTION 0.013 0.013 0.014 0.014
CHANGE IN PROBABILITY OF A REPUBLICAN 0.043 **
MAJORITY IN THE HOUSE OF REPRESENTATIVES 0.018
CHANGE IN PROBABILITY OF A REPUBLICAN 0.034 *
MAJORITY IN THE SENATE 0.020
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CONGRESS 0.056 " * 0.065 **’ 0.058
0.016 0.017 0.017
CHANGE IN PROBABILITY OF GEORGE W. BUSH'S ELECTION 0.033
* DUMMY FOR POLLUTION-INTENSIVE INDUSTRIES 0.033
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CONGRESS -0.050
* DUMMY FOR POLLUTION-INTENSIVE INDUSTRIES 0.040
CHANGE IN PROBABILITY OF G EORGE W. BUSH'S ELECTION -0.054
* DUMMY FOR COMPANIES WITH TWENTY CASES OR MORE 0.038
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CO NG RESS -0.019
* DUMMY FOR COMPANIES WITH TWENTY CASES OR MORE 0.046
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CONGRESS
AND GEORGE W. BUSH WINNING PRESIDENTIAL ELECTION
Each column In the table reports coefficients from a fixed effects regression. The com pany fixed effects are not
reported here to conserve sp ace. T he unit of observation Is the abnorm al daily returns to a com pany with EPA litigation
outstanding on date of the election, on Tuesday N ovem ber 7th 2000. The daily returns are betw een A ugust 6th
and until the day before the election. Standard errors are underneath the coefficient estim ates.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
5
6,778
125
65
0.016
0.0006
0.057
0.020
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
74
significantly different from zero. They imply that Republican control over the House
of Representatives and the Senate respectively is worth 4% and 3% of the value of
environmental offenders.
It is possible that the market would only react to the overall control of the bicameral
legislature, as bills must be approved by both the House and Senate in order to
become law. To that end, the probability of a Republican majority in both houses of
Congress is used in specification two. This variable is constructed by simply
multiplying the probability of a Republican majority in the House of Representatives
by the probability of a Republican majority in the Senate24. This shows that a unified
Republican Congress was worth 5% of market capitalization to environmental
offenders.
Not only are these results statistically significant, they are also very significant in
terms of the magnitude of economic value transferred. The total market
capitalization of all 125 defendant companies is nearly $3 trillion dollars, and so a
different set of election results would have resulted in a reduction of approximately
$150 billion dollars of shareholder value from the sample. The economic
significance of these results is at least as great as most of the results found in the
literature. It is also worth noting that the shareholder value transferred is many orders
2 4 Note that the correlation coefficient between Pr(H Rep) and Pr(Senate_Rep) is 0.13093.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
75
of magnitude greater than the direct, observable costs of the litigation. Only in
exceptional cases does EPA litigation result in fines or penalties of greater than a
million dollars. Other costs of the litigation, such as consent decrees, are difficult to
assess directly.
These first two specifications make a strong case that there are significant effects on
the stock market valuation of companies with environmental litigation outstanding
because of political events. The most serious concern is in their interpretation. What
is driving these results? Is it the fact that these companies had litigation outstanding,
or litigation in the past, or probable litigation in the future, or bad records of
environmental compliance independent of any EPA litigation? Or are the results
purely incidental to environmental regulation, and are instead being caused by
something else, such as the prospect of government contracts on entirely unrelated
issues?
Given the diversity of industries to which these companies belong, the last point
seems extremely unlikely. Around 80% of the companies are in manufacturing, but
they are spread thinly across many different sectors. Nor does it seem likely that the
results are being driven by the case law the litigation is filed under. More than 95%
of the cases are CERCLA (Comprehensive Environmental Response, Compensation
and Liability Act, or Superfund) cases, intended to clean up toxic waste dumps.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
76
Much litigation is filed under several different laws. Interaction terms of the political
variables and dummies for the type of case law were constructed, but they were not
reported as there did not appear to be any consistent, significant relationship between
them and the abnormal company returns.
Most of the companies are in manufacturing, and these firms, since they produce
most of the pollution, by their very nature bear the brunt of environmental regulation.
It could be that these ‘dirty’ firms are driving the results, and the fact that they
happen to have environmental litigation outstanding at the time of the election is
purely incidental. To that end, the political variables are interacted with a dummy
variable; this assumes the value of one when the company is in a four digit SIC
(Standard Industrial Classification) code that is pollution-intensive, and zero
otherwise25.
This interaction term is used in specification three. The results do not support the
hypothesis that the pollution-intensive industries are driving the results. Neither of
the two interaction terms are statistically significant. However, this result should be
2 5 A pollution-intensive industry is defined as any one o f 37 four digit SIC codes that are in the top
five polluters by four digit SIC code from the World Bank data at
http://www.worldbank.org/nipr/data/ippsdown.htm, as measured by the amount o f pollution produced
relative to value-added, value o f output and employment, o f air, water, metals and toxics pollution.
There are 170 companies quoted on the Stock Exchange during this period in these industries.
Approximately 20% o f my sample o f companies are in these pollution-intensive industries. These are
listed in full in the appendix at the end o f this paper.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
77
interpreted with caution. It could also be that the market has already discounted the
value of these pollution-intensive companies, and so further litigation will not reduce
their values. It could also be that they have excellent records of environmental
compliance in spite of the industry they are operating in.
Specification four tests whether the number of previous cases filed against the
company has any effect. This is intended as another test of whether or not it is indeed
the fact that these companies have environmental litigation outstanding that is
driving the results, and not the fact that they may have poor records of environmental
compliance as measured in some other way, which could be purely incidental to the
litigation, although they may be collinear.
The political variables are interacted with a dummy variable that assumes the value
of 1 if the company has had twenty or more cases filed against it in the past, and zero
otherwise. This represents about the most frequently sued 20% of the companies.
These interaction terms are used in specification four. They are never significant, and
seem to have virtually no effect on the size and significance of the original political
variables in each case. Also, in results not reported in Table 1, similar dummies were
also used for the number of cases outstanding against the company at the time of the
election, not including those in the past. Perhaps surprisingly, these were also always
insignificant. The results of specification four must be interpreted with caution. The
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
78
interpretation that the number of cases filed against the company is an indicator of its
future vulnerability to such litigation assumes that this measure of poor
environmental compliance is not collinear with other unspecified measures of
misbehavior. Also, it could be that the litigation can be made to become toothless by
political pressure, perhaps by reducing fines or by making compliance voluntary. In
that case, the number of times a company is sued might be irrelevant.
Taken together, the results suggest that the market did perceive differences in
environmental policies in the 2000 election under different prospective political
masters. Two interpretations of these results are possible; that they are caused by
differences in the enforcement of the law caused by political pressure on the EPA
bureaucracy (the bureaucratic explanation), or that they are caused by the prospect of
new legislation under new political masters (the legislative explanation). To
distinguish between these two competing explanations, matched samples were
constructed. These were constructed in accordance with the literature, as in Farrell
and Whidbee (2000). Virtually all of these matches (hereafter referred to as the
Matched sample) are within 50% and 150% of the market capitalization of the
company with which they are matched (hereafter referred to as the EPA sample),
with an average market capitalization of 94% of the EPA sample. They are also
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
79
TABLE 20
ABNORMAL DAILY RETURNS TO MATCHED COMPANIES DURING THE 2000 ELECTION
1 2 3
NUMBER OF OBSERVATIONS 7,779 7,779 7,779
NUMBER OF COMPANIES 125 125 125
TIME SERIES LENGTH 65 65 65
R2 0.012 0.012 0.013
MEAN OF DEPENDENT VARIABLE 0.0003 0.0003 0.0003
CHANGE IN PROBABILITY OF GEORGE W. BUSH 0.002 0.002 0.004
WINNING PRESIDENTIAL ELECTION 0.013 0.012 0.013
CHANGE IN PROBABILITY OF A REPUBLICAN 0.011
MAJORITY IN THE HOUSE OF REPRESENTATIVES 0.017
CHANGE IN PROBABILITY OF A REPUBLICAN 0.012
MAJORITY IN THE SENATE 0.019
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CONGRESS 0.016 0.018
0.015 0.016
CHANGE IN PROBABILITY OF GEORGE W. BUSH'S ELECTION -0.020
* DUMMY FOR POLLUTION-INTENSIVE INDUSTRIES 0.040
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CONGRESS -0.013
* DUMMY FOR POLLUTION-INTENSIVE INDUSTRIES 0.048
CHANGE IN PROBABILITY OF G EORGE W. BUSH'S ELECTION
* DUMMY FOR COMPANIES WITH TWENTY CASES OR MORE
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CONGRESS
* DUMMY FOR COMPANIES WITH TWENTY CASES OR MORE
CHANGE IN PROBABILITY OF REPUBLICAN MAJORITY IN CONGRESS
AND GEORGE W. BUSH WINNING PRESIDENTIAL ELECTION
Each column In the table reports coefficients from a fixed effects regression. The com pany fixed effects are not
reported here to conserve sp ace. The unit of observation is the abnorm al daily returns to a com pany. T he daily returns are
betw een August 6th and until the day before the election. Standard errors are underneath the coefficient estim ates,
coefficient estim ates.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
5
7,779
125
65
0.012
0.0003
0.010
0.019
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
80
matched very closely by Standard Industrial Classification (SIC) code, usually at the
* ) ( \
2 or 3 digit level . Finally, all companies that were ever sued by the EPA are
excluded from the matched sample. The results of these matched sample regressions
are given in table 2. The matched sample results mirror the empirical results for the
companies that actually were sued as closely as possible. The only difference is that
the interaction terms for the companies that have been sued twenty or more times
cannot be used in the matched sample, and so column 4 is left blank. The results
strongly support the hypothesis that the effects found are caused by the EPA
litigation. That is, the results support the bureaucratic explanation. No significant
results are found in the matched sample.
SUMMARY AND SUGGESTIONS FOR FURTHER RESEARCH
Several avenues of further research are suggested by the literature. They are beyond
the ambitions of the present paper, but they could confirm the exact mechanism by
which politicians influence the EPA. For example, a study could test to see if
companies based in districts of politicians on Congressional committees with EPA
oversight responsibility are less likely to be sued. Similarly to Faith, Leavens and
Tollison (1982), a researcher could test to see if this effect is stronger if the pollution
2 6 They are matched as closely as possible by SIC (46% at the 4 digit SIC level; 73% at the 3 digit SIC
level; 90% at the 2 digit SIC level and 100% at the 1 digit SIC level). The matched sample has a
larger number o f daily returns than the sample o f companies that were actually sued at the time o f the
2000 election, even though it has the same number o f companies. This is because the returns o f the
companies that were being sued by the EPA were only included after the litigation was filed. For the
matched sample o f companies, by contrast, all o f the returns after August 6th are included.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
81
is transboundary, so that the people adversely affected live outside the favored
politician’s district. The effect of reorganization of the EPA such as that pointed out
by Helland could also be analyzed. Given the results of Shughart, Tollison and Goff,
it seems that such changes could be intended to keep the EPA on a shorter political
leash, with less room for its own discretion. The breakdown of cases could also be
used, in a similar way to Weingast and Moran (1983). Perhaps the EPA is more
likely to sue big businesses when the ideological composition of its Congressional
oversight committees becomes more liberal.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
82
CONCLUSION
As ever, much further research remains to be done. The present dissertation is no
exception. There is a lot of further work that could be done, but it is more suited to a
lifetime of study and an army of research assistants than to one person.
The interpretation of the first paper’s results is at present inconclusive. However,
there is an intuitively appealing explanation that could be empirical tested. That is
that there is a political explanation for the results. There are several ways of solving
the problems that the existence of ethnic divisions presents. One of these is the
separation and segregation of ethnic groups. Unfortunately, this can lead to an
aggravation of inequality and to the worsening of relations between ethnic and racial
groups (Cutler and Glaeser, QJE, 1997). Another solution is power sharing. A good
recent example of this is the ‘Good Friday’ agreement in Northern Ireland in 1997.
For fifty years until the early 1970s Northern Ireland was ruled by a simple
majoritarian system similar to the U.S. electoral system. Since Protestants formed a
majority of the population, and political parties were all based on this ethnic division,
they formed a permanent majority and monopolized power. The Catholic minority’s
30-40% of the votes did not translate into any real say in how the province was
governed. This lead to frustration, and eventually to the boycotting of the system by
Catholics. The 1997 Good Friday agreement institutionalized power sharing between
the two communities, ensuring that the concerns of minorities would be addressed.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
83
There is evidence from the present paper that institutional political factors do play a
role in reducing emissions. That is, that pollution is reduced when representatives on
city councils are elected from districts ‘at large’ as opposed to single member
districts. However, for reasons explained in the paper, this is a very inadequate
means of controlling for institutional means of power sharing in American
metropolitan areas. The most likely explanation for the counterintuitive results found
for sulfur dioxide is that, because S02 is mainly attributable to point, as opposed to
non-point sources, its abatement is controlled by local government to a greater extent
that non-point source pollutants. Unlike for other so-called ‘public goods’, such as
education, roads and so on, which are also controlled by local governments, and
where research has shown that ethnic divisions lead to relative lack of provision, air
quality is a pure public good. It is therefore much harder to ensure that the members
of the favored or dominant ethnic group get the benefits of this good, as air pollution
is difficult to stop from blowing from one part of a metropolitan area to another.
Because of the fact that ethnic diversity has lead to tension in American metropolitan
areas in the past, institutional arrangements were changed in these areas to facilitate
compromise over conflict, exactly as has happened in Northern Ireland. These
institutional changes in the political systems of these areas resulted in more careful
control of S02 emissions than in other, ethnically homogenous areas, that
presumably did not have the same institutions.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
84
Proving that the above narrative is indeed correct would require the gathering of
detailed information on such institutional arrangements. These would include, for
example, whether the mayor is directly elected, if direct elections can be held on
referenda or initiatives, etc. It would also probably be instructive to examine exactly
how the SIP, (State Implementation Plans) to bring high pollution areas back into
compliance, are formed and implemented by local officials. It is also possible, of
course, that there is a simpler explanation for the results found. What this could be is
not clear, but, for example, emissions from power plants could be a candidate. They
are not controlled for in the capital-labor ration in the empirical results, and they are
known to be heavy S02 emitters. However, the emissions (as opposed to ambient)
data does include emissions of S02 from power plants.
There are many ways in which the second paper could be extended and improved
upon. Unfortunately, these are all very data intensive, and far beyond the scope of
paper as it stands, in terms of both time and resources. The focus of these extensions
would be on the biggest question left unanswered by the paper. That is, on a deeper
interpretation of the results; are they caused by changes in the bureaucratic
enforcement of existing laws and regulations, or rather are they caused by the
expectation o f legislative changes, in either the introduction of repeal of new laws? I
will refer to these as explanations as the bureaucratic and legislative theories,
respectively. The existing literature suggests many ways that these two theories
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
85
could be separated from each other. Unfortunately, the effective ones are very data-
intensive.
Relating the actions of the EPA to the Congressional committees that oversee it
would mean that the IEM would no longer be used. The IEM would not reveal
sufficient information about the preferences of the committee members, as it deals
with the overall control of Congress, not the exact composition of certain committees
[the party in control of Congress has the Chairmanships of all of the committees, but
that still leaves substantial scope for ideological variation- parties are not
homogenous]. The least data-intensive approach would be to relate the ideological
leanings of committee members as measured by a pressure group such as Americans
for Democratic Action (ADA) scores. It could be that when the committee members
become more liberal, the EPA is more likely to sue large corporations; or to change
to composition of its cases to favor a different political constituency, such as
favoring toxic waste cleanups over the preservation of public parks; or to become
more active in litigation in general. More sophisticated analysis would be permitted
if the facilities of the companies being sued were mapped to Congressional districts.
This would allow one to analyze whether having a local Congressman on the
relevant committee makes a company less likely to be sued.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
86
BIBLIOGRAPHY
Ackerman, B. A. and Hassler, W. T., “Clean Coal, Dirty Air,” Yale University Press,
1981
Alesina, A. and La Ferrara, E., “Participation in Heterogenous Communities,”
Quarterly Journal of Economics, 115:3, 2000, 847-904.
Alesina, A., Baqir, R. & Easterly, W., “Redistributive Public Employment,” NBER
working paper 6746, 1998.
Alesina, A., Baqir, R. and Easterly, W., “Public Goods and Ethnic Divisions,”
Quarterly Journal of Economics, 114:4, 1999,1243-84.
Alesina, A., Baqir, R. and Easterly, W., “Redistributive Public Employment,” NBER
working paper 6746, 1998.
Anderton, D. L., Anderson, A. B., Oakes, J. M. and Fraser, M. R., “Environmental
Equity: The Demographics of Dumping,” Demography, 31:2, 1994, 229- 248.
Antweiler, W., Copeland, B. and Taylor, M. S., “Is Free Trade Good for the
Environment?” NBER working paper 6707,1998.
Arora, S. and Cason, T. N, “Do Community Characteristics Influence Environmental
Outcomes? Evidence from the Toxics Release Inventory,” Journal of Applied
Economics, 1:2, 1998, 413- 453.
Beck, N., “Parties, Administrations and American Macroeconomic Outcomes
[Political Parties and Macroeconomic Policy],” American Political Science Review,
76:1, 1982, 83- 93.
Been, V. and Gupta, F., “Coming to the Nuisance or Going to the Barrios? A
Longitudinal Analysis of Environmental Justice Claims,” Ecology Law Quarterly,
24:1, 1997, 1-56.
Been, V., “Locally Undesirable Land Uses in Minority Neighborhoods:
Disproportionate Siting or Market Dynamics?” The Yale Law Journal, 103:6, 1994,
1383-1422.
Berman, E. and Bui, L. T. M., “Environmental Regulation and Labor Demand:
Evidence from the South Coast Air Basin,” Journal of Public Economics, 79:2, 2001,
265-95.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
87
Binder, J. J., “Measuring the Effects of Regulation with Stock Price Data,” Rand
Journal of Economics, 16:2, 1985, 167- 183.
Brander, J. A., “Election Polls, Free Trade, and the Stock Market: Evidence from the
1988 Canadian General Election,” Canadian Journal of Economics, 24:4, 1991, 827-
843.
Brooks, N. and Sethi, R., “The Distribution of Pollution: Community Characteristics
and Exposure to Air Toxics,” Journal of Environmental Economics and
Management, 32, 1997, 233-250.
Brown, S. J. & Warner, J. B., “Measuring Security Price Performance,” Journal of
Financial Economics, 8:3, 1980, 205- 258.
Chay, K. Y., and Greenstone, M., “Does Air Quality Matter? Evidence from the
Housing Market,” NBER Working paper 6826, 1998.
Cutler, D. M. and Glaeser, E. L., “Are Ghettos Good or Bad?” Quarterly Journal of
Economics, 112:3,1997, 827-72.
Cutler, D. M., “Tax Reform and the Stock Market: An Asset Price Approach,”
American Economic Review, 78:5, 1988, 1107- 1117.
Cutler, D. M., Elmendorf, D. W., Zeckhauser, R. J., “Demographic Characteristics
and the Public Bundle,” NBER Working paper W4283, 1993.
Downs, T. W. & Tehranian, H., “Predicting Stock Price Responses to Tax Policy
Changes,” American Economic Review, 78:5, 1988, 1118- 1130.
Easterly, W. and Levine, R., “Africa’s Growth Tragedy: Policies and Ethnic
Divisions,” Quarterly Journal of Economics, 112:4,1997,1203-50.
Ellison, S. F. & Mullin, W. P., “Gradual Incorporation of Information:
Pharmaceutical Stocks and the Evolution of President Clinton’s Health Care
Reform,” Journal of Law and Economics, 44:1, 2001, 89- 129.
Faith, R. L., Leavens, D. R. & Tollison, R. D., “Antitrust Pork Barrel,” Journal of
Law and Economics, 25:2, 1982, 329- 342.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
8 8
Fischel, W., “The Homevoter Hypothesis: How Home Values Influence Local
Government Taxation, School Finance and Land Use Policies,” Draft of a book to be
published by Harvard University Press in 2001. April 2001 version available online
at http ://www.dartmouth.edu/~wfische 1/WP .html
Flinterman, C., Kwiatkowska, B., and Lammers, J. G. (eds.), “Transboundary Air
Pollution,” 1986.
Fraga, L., “Policy Consequences and the Change from At-Large Elections to Single
Member Districts,” Stanford Center for Chicano Research, 1991.
Gemmill, G., “Political Risk and Market Efficiency: Tests Based on British Stock
and Options Markets in the 1987 Election,” Journal of Banking and Finance, 16:1,
1992,211-231.
Gilligan, T. W. & Krehbiel, K., “Complex Rules and Congressional Outcomes: An
Event Study of Energy Tax Legislation,” Journal of Politics, 50:3, 1988, 625- 654.
Goldin, C. and Katz, L. F., “Human Capital and Social Capital: The Rise of
Secondary Schooling in America, 1910 to 1940,” NBER working paper 6439, 1998.
Graddy, D. B., Homaifar, G., & Hollman, K. W., “Tax Reform’s Impact on
Insurance Industry Stock Returns,” Journal of Risk and Insurance, 59:2, 1992, 284-
290.
Grossman, G. M. and Krueger, A. B., “Economic Growth and the Environment,”
Quarterly Journal of Economics, 110:2,1995, 353-77.
Grossman, G. M., Krueger, A. B. and Laity, J., “Determinants of Air Pollution in US
counties,” Unpublished paper, Princeton University, Woodrow Wilson discussion
paper in economics, 1994.
Helland, E., “Prosecutorial Discretion at the EPA: Some Evidence on Litigation
Strategy,” Journal of Regulatory Economics, 19:3,2001,271- 94.
Henderson, J. V., “Effects of Air Quality Regulation,” American Economic Review,
86:4, 1996, 789-813.
Henderson, J.V., “State Attitudes Towards Air Quality Regulation,” Unpublished
Working Paper, Brown University, 1994.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
89
Henisz, W. J., “The Institutional Environment for Economic Growth,” Economics
and Politics, 12:1, 2000, 1-31.
Herron, M. C., Lavin, J., Cram, D. & Silver, J., “Measurement of Political Effects in
the United States Economy: A Study of the 1992 Presidential Election,” Economics
and Politics, 11:1,1999,51-81.
Holtz-Eakin, D. and Selden, T. M., “Stoking the Fires? C02 Emissions and
Economic Growth,” Journal of Public Economics, 57:1,1995, 85-101.
Homaifar, G. et al, “American Presidential Elections and Returns of Defence
Industry Stocks,” Applied Economics, 20:7, 1988, 985- 993.
Hoxby, C. M., “Does Competition among Public Schools Benefit Students and
Taxpayers?” American EconomicReview, 90:5, 2000, 1209-38.
James, E., “The Public-Private Division of Responsibility for Education: An
International Comparison,” Economics of Education Review, 6, 1987, 1-14.
James, E., “Why Do Different Countries Choose a Different Public-Private Mix of
Educational Services?” Journal of Human Resources, 1993.
Khanna, M., Quimio, W.H., and Bojilova, D., “Toxics Release Information: A Policy
Tool for Environmental Protection,” Journal of Environmental Economics and
Management, 36,1998, 243- 266.
Knight, B. 2004. “Are Policy Platforms Capitalized into Equity Prices? Evidence
from the Bush/Gore 2000 Presidential Election.” NBER Working Paper 10333.
Kuran ,T., Private truths, public lies: The social consequences of preference
falsification, Cambridge and London: Harvard University Press, 1995
Langohr, H. M. & Viallet, C. J., “Compensation and Wealth Transfers in the French
Nationalizations: 1981-1982,” Journal of Financial Economics, 17:2, 1986, 273- 312.
Levitt, S., “Using Electoral Cycles in Police Hiring to Estimate The Effect of Police
on Crime,” American Economic Review, 87:3, 1997, 270-290.
Lizzeri, A. and Persico, N., “The Provision of Public Goods under Alternative
Electoral Incentives,” American Economic Review, 91:1, 225-39.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
90
Lo, Andrew, and MacKinlay, Craig, “Stock Prices Do Not Follow Random Walks:
Evidence from a Simple Specification Test.”, Review of Financial Studies, Vol. 1,
41-66.
Luttmer, E.F. P., “Group Loyalty and the Taste for Redistribution,” Unpublished
paper, Harris School of Public Policy, University of Chicago, 1998.
Manning, D. N., “The Effect of Political Uncertainty on the Stock Market: The Case
of British Telecom,” Applied Economics, 21:7, 1989, 881- 889.
Massey, D. S. and Denton, N. A., “The Dimensions of Residential Segregation,”
Social Forces, 67:2, 1988, 281-315.
McCarty, T. M., “Demographic Diversity and the Size of the Public Sector,” Kyklos,
46:2, 1993,225-40.
McConnell V.D and Schwab, R. M., “The Impact of Environmental Regulation on
Industry Location Decisions: The Motor Vehicle Industry,” Land Economics, 66:1,
1990, 67-81.
Murdoch, J. C., Sandler, T. and Sargent, K., “A Tale of Two Collectives: Sulphur
versus Nitrogen Oxides Emission Reduction in Europe,” Economica, 64, 1997, 281 -
301
Niederhoffer, V., Gibbs, S. & Bullock, J., “Presidential Elections and the Stock
Market,” Financial Analysts Journal, 26, 111-113.
Oakes, J. M., Anderton, D. L. and Anderson, A. B. “A Longitudinal Analysis of
Environmental Equity in Communities with Hazardous Waste Facilites,” Social
Science Research, 25, 1996, 125-148.
Pashigian, B. Peter. 1985. "Environmental Regulation: Whose Self-interests Are
Being Protected?" Economic Inquiry 23:551-584.
Pastor, M., Sadd, J. and Hipp, J., “Which Came First? Toxic Facilities, Minority
Move-In and Environmental Justice,” Journal of Urban Affairs, Forthcoming.
Persson, T. and Tabellini, G., “The Size and Scope of Government: Comparative
Politics with Rational Politicians,” European Economic Review, 43:4-6, 699-735.
Poterba, J., “Demographic Structure and the Political Economy of Public
Education,” NBER working paper 5677, 1996.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
91
Revesz, R. L., “Federalism and Interstate Environmental Externalities,” University
of Pennsylvania Law Review, 144:6, 1996, 2341- 2416.
Riley, W. B., & Luksetich, W. A., “The Market Prefers Republicans: Myth or
Reality,” Journal of Financial and Quantitative Analysis, 15:3,1980, 541- 560.
Roberts, B. E., “Political Institutions, Policy Expectations, and the 1980 Election: A
Financial Market Perspective,” American Journal of Political Science, 34:2,1990,
289-310.
Salinger, M., “Standard Errors in Event Studies,” Journal of Financial and
Quantitative Analysis, 27:1, 1992, 39- 53.
Selden, T. M. and Song, D., “Environmental Quality and Development: Is there a
Kuznets Curve for Air Pollution Emissions. Journal of Environmental Economics
and Management, 27, 1994, 147-162.
Shukla, V. and Parikh, K., “The Environmental Consequences of Urban Growth:
Cross-National Perspectives on Economic Development, Air Pollution and City
Size,” Collective Volume Article, Urbanization and Economic Growth, 1996, 361-
95.
Stovall, R. H., “Forecasting Stock Market Performance via the Presidential Cycle,”
Financial Analysts Journal, 1992, 48, 5- 8.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
92
EXPLANATION OF AIR QUALITY REGRESSION RESULTS IN APPENDICES
Five additional sets of explanatory variables are reported on here. They were not
reported on in the main empirical results for a number of reasons. These were mainly
that the results themselves were insignificantly different from zero, had questionable
relevance due to missing data, or were supplanted by a better measure in the main
empirical results. These results are reported in tables 15 through 18. They are only
for S02 and CO, and only for models (1) and (2). Since the data was available only
for a point in time, and not as a time series, model (3) was not possible to estimate.
The same five specifications were used in each of tables 15 through 18. Note that the
number of observations is often different from one specification to the next in the
same table, unlike for earlier tables.
The first specification uses a set of dummy variables denoting groups of U.S. states.
The inspiration behind this was some research by Pashigian (1985). This analyzed
the political economy of the PSD (Prevention of Significant Deterioration) air quality
policy in the United States. This policy adversely affected the competitiveness of
states with clean air by not permitting them to sully their environment to attract
industry. Certain groups of states were found to engage in rent-seeking.
Unfortunately, no consistent pattern is found in tables 15 through 18, either
consistent for the same pollutant (a different pattern is found in the two S02
regressions in tables 15 and 16; similarly, there is a different pattern in table 17 from
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
93
table 18, both of which are for CO). Also, no pattern is found that is consistent with
Pashigian’s research. Regressions were also estimated using individual states instead
of groups of states as dummy variables. This was motivated in part by anecdotes and
research suggesting that some states have different attitudes to clean air than others
(Henderson, 1994). These results were not reported here. Since many states have
only one MSA, or even none, the results are not informative, and do not have a clear
interpretation.
The second specification uses data on the frequency of pollution inspections. Given
the fact that ethnic diversity is associated with higher levels of government
employment (Alesina, Baqir and Easterly, NBER, 1998), it could be hypothesized
that environmental regulations are more strictly enforced due to the extra
government employees. The literature had implied that these additional employees
were not productive, but no direct tests of this are given. Unfortunately, data on the
number of inspectors for compliance with air quality standards is unavailable.
Instead, we must rely on data on inspections, not inspectors. Under reasonable
assumptions, the number of inspectors and inspections would be closely related.
Further, data contemporaneous with the other data is unavailable. The data also
suffers from the same problem as air pollution emissions; that is, it is available only
for counties, which are not exactly equivalent to MSAs in New England. The
percentage of facilities subjected to full air pollution inspections is expected to be
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
94
negative, as the less chance a polluter stands of getting away with a violation, the
more careful they are likely to be to ensure their own full compliance. The data used
77
is from 2001 to 2003 . As expected, the coefficient is negative, although it is
significantly different from zero only in table 18.
The third specification represents another effort to quantify the institutional political
incentives for air pollution control faced by local governments. It relies on the
institutional difference between cities that elect their city council members ‘at large’
from a single city-wide district as opposed to from single-member districts. This is
inspired by the research recent (Persson & Tabellini, 1999; Lizzeri & Persico, 2001)
suggesting that public goods may be underprovided in majoritarian as opposed to
proportional electoral systems. This is because there is little incentive to invest in
public goods that benefit everybody, when instead ‘pork-barrel’ spending could be
directed to ‘swing’ states whose votes decide elections. The existence of ‘at large’
voting districts is by no means a form of proportional representation, but it does have
one characteristic in common with it. That is, it is possible that transboundary
pollution across political district boundaries could be more of a problem in single
member district cities rather than ones that elect all of their representatives at large.
2 7 The source o f the data is http://www.epa.gov/echo/compliance_report_air.html. The exact measure
used is the percentage o f facilities which have been fully inspected over the April 2001 to March 2003
period. This is defined as ‘For Clean Air Act (CAA) permits only, indicates that at least one full
compliance evaluation (FCE) has occurred within the eight most recent complete quarters o f data (2
years), or within the current quarter.’ The designation ‘other minor’ is ignored for consistency across
States.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
95
To that end, the percentage of city council seats elected from ‘at large’ districts is
used as an independent variable in the regression analysis. This is a very imperfect
empirical test, because the unit of geographic aggregation used in the present paper
is MSAs, not cities, it was necessary to use the percentage of at large council seats
for the largest city in the MSA. For example, for the Los Angeles-Long Beach MSA
the city government of Los Angeles is coded in this variable. There is also a large
number of missing observations for this variable. To the extent that ‘at large’
districts offer incentives for local government to reduce transboundary air pollution,
the sign of the coefficient is expected to be negative. It is, in fact, always negative in
tables 15 through 18, although it is significantly different from zero only in the last
table, for CO.
The fourth and fifth specifications can be seen as additional attempts to control for
the collective action problems that specification six in the main empirical results tries
to do. That is, they use variables denoting the number of counties and cities in the
MSA. They are expected to have the same positive sign that the number of states in
the MSA has in specification six of the main empirical results. Indeed, the
coefficients are always positive in tables 15 through 18, although they are often not
significantly different from zero.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
96
TABLE 21
OLS REGRESSIONS: SULFUR DIOXIDE AIR POLLUTION- OBSERVED
NUMBER OF OBSERVATIONS
R2
ADJUSTED R2
MEAN OF DEPENDENT VARIABLE
INTERCEPT
ETHNIC
DUMMY VARIABLE FOR EAST NORTH CENTRAL STATES
DUMMY VARIABLE FOR MID ATLANTIC STATES
DUMMY VARIABLE FOR MID W EST STATES
DUMMY VARIABLE FOR MOUNTAIN PLAINS STATES
DUMMY VARIABLE FOR NEW ENGLAND STATES
DUMMY VARIABLE FOR PACIFIC SOUTHWEST STATES
1 2 3 4 5
170 142 110 154 154
0.2644 0.1696 0.1333 0.1816 0.1660
0.2279 0.1576 0.1171 0.1708 0.1550
0.0075 0.0073 0.0067 0.0074 0.0074
0.0075 *** 0.0117'** 0.0102 *** 0.0104 **"1 0.0105
0.0012 0.0010 0.0010 0.0007 0.0008
-0.0040 -0.0127 * * * ’ -0.0101 *** -0.0133 * **’ -0.0127
0.0029 0.0026 0.0026 0.0023 0.0023
0.0020 *
0.0010
0.0036 ***
0.0010
0.0000
0.0014
0.0094 ***
0.0029
0.0033 **
0.0016
-0.0018
0.0011
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
97
TABLE 21 CONTINUED
OLS REGRESSIONS: SULFUR DIOXIDE AIR POLLUTION- OBSERVED
1 2 3 4 5
NUMBER OF OBSERVATIONS 170 142 110 154 154
R2 0.2644 0.1696 0.1333 0.1816 0.1660
ADJUSTED R2 0.2279 0.1576 0.1171 0.1708 0.1550
MEAN OF DEPENDENT VARIABLE 0.0075 0.0073 0.0067 0.0074 0.0074
DUMMY VARIABLE FOR SOUTH CENTRAL STATES -0.0004
0.0011
PERCENTAGE OF COUNCIL SEATS
ELECTED FROM 'AT LARGE' DISTRICTS
-0.0014
0.0009
PERCENTAGE OF FACILITIES FULLY
INSPECTED FROM 2001-2003
-0.0013
0.0014
NUMBER OF CITIES IN MSA 0.0001
0.0000
NUMBER OF COUNTIES IN MSA 0.0002 '
0.0001
Each column in the table reports coefficients from an OLS regression. The unit of observation is a m etropolitan statistical a re a (city).
T he dependent variable is a m ean of annual observation m ean s from monitoring stations. Standard errors are underneath the
coefficient estim ates. O bservations from th e years 1980 and 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
98
TABLE 22
OLS REGRESSIONS: SULFUR DIOXIDE AIR POLLUTION- EMISSIONS FROM ALL SOURCES DIVIDED BY AREA
12 2 4
NUMBER OF OBSERVATIONS 72 71 47 72
R2 0.1007 0.0144 0.0454 0.0597
ADJUSTED R2 0.0023 -0.0146 0.0020 0.0324
MEAN OF DEPENDENT VARIABLE 34.3595 34.0605 30.6834 34.3595
45.0103 *“ 38.2085
13.3951 11.9221
-21.215 -48.380
38.344 38.126
DUMMY VARIABLE FOR EAST NORTH CENTRAL STATES -5.1541
16.2249
DUMMY VARIABLE FOR MID ATLANTIC STATES 2.9820
18.8779
DUMMY VARIABLE FOR MID W EST STATES -19.887
28.718
DUMMY VARIABLE FOR MOUNTAIN PLAINS STATES
INTERCEPT 45.5665 ** 47.7778
21.3472 15.2673
ETHNIC -11.882 -31.816
54.205 39.219
DUMMY VARIABLE FOR NEW ENGLAND STATES -12.946
24.556
DUMMY VARIABLE FOR PACIFIC SOUTHWEST STATES -39.139 **
17.100
5
72
0.0115
-0.0172
34.3595
43.7991 ***
12.6077
-34.418
38.683
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
99
TABLE 22 CONTINUED
OLS REGRESSIONS: SULFUR DIOXIDE AIR POLLUTION- EMISSIONS FROM ALL SOU RCES DIVIDED BY AREA
12 2 4 5
NUMBER OF OBSERVATIONS 72 71 47 72 72
R2 0.1007 0.0144 0.0454 0.0597 0.0115
ADJUSTED R2 0.0023 -0.0146 0.0020 0.0324 -0.0172
MEAN OF DEPENDENT VARIABLE 34.3595 34.0605 30.6834 34.3595 34.3595
DUMMY VARIABLE FOR SOUTH CENTRAL STATES -6.216
17.808
PERCENTAGE OF COUNCIL SEATS
ELECTED FROM 'AT LARGE' DISTRICTS
-16.910
12.455
PERCENTAGE OF FACILITIES FULLY
INSPECTED FROM 2001-2003
-8.330
20.030
NUMBER OF CITIES IN MSA 0.0196
0.7056
NUMBER OF COUNTIES IN MSA 2.4036 '
1.2775
Each column in the table reports coefficients from an OLS regression. T he unit of observation is a metropolitan statistical area (city).
T he dependent variable Is the em issions of sulfur dioxide in that MSA in pounds, from both point and non-point sources, divided by
the land a re a of the MSA. Standard en o rs are underneath the coefficient estim ates. O bservations from the year 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
100
TABLE 23
OLS REGRESSIONS: CARBON MONOXIDE AIR POLLUTION- OBSERVED
1 2 3 4 5
NUMBER OF OBSERVATIONS 170 114 114 112 118
R2 0.062 0.017 0.012 0.008 0.009
ADJUSTED R2 0.016 -0.001 -0.006 -0.010 -0.009
MEAN OF DEPENDENT VARIABLE 1.303 1.225 1.330 1.238 1.243
INTERCEPT 1.647 *** 1.016 “ * 1.496 *** 1.137 *** 1.131
0.142 0.165 0.159 0.124 0.124
ETHNIC -0.744 * 0.217 -0.354 0.293 0.362
0.393 0.398 0.424 0.412 0.407
DUMMY VARIABLE FOR EAST NORTH CENTRAL STATES -0.271 *
0.154
DUMMY VARIABLE FOR MID ATLANTIC STATES -0.252 *
0.148
DUMMY VARIABLE FOR MID W EST STATES -0.286
0.178
DUMMY VARIABLE FOR MOUNTAIN PLAINS STATES 0.085
0.258
DUMMY VARIABLE FOR NEW ENGLAND STATES -0.479 **
0.227
DUMMY VARIABLE FOR PACIFIC SOUTHWEST STATES -0.078
0.142
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
101
TABLE 23 CONTINUED
OLS REGRESSIONS: CARBON MONOXIDE AIR POLLUTION- OBSERVED
1 2 3 4 5
NUMBER OF OBSERVATIONS 170 114 114 112 118
R2 0.062 0.017 0.012 0.008 0.009
ADJUSTED R2 0.016 -0.001 -0.006 -0.010 -0.009
MEAN OF DEPENDENT VARIABLE 1.303 1.225 1.330 1.238 1.243
DUMMY VARIABLE FOR SOUTH CENTRAL STATES -0.139
0.182
PERCENTAGE OF COUNCIL SEATS
ELECTED FROM 'AT LARGE' DISTRICTS
-0.125
0.139
PERCENTAGE OF FACILITIES FULLY
INSPECTED FROM 2001-2003
0.252
0.207
NUMBER OF CITIES IN MSA 0.002
0.007
NUMBER OF COUNTIES IN MSA 0.006
0.013
Each column In the table reports coefficients from an OLS regression. T he unit of observation is a metropolitan statistical area (city).
The dependent variable is a m ean of annual observation m eans from monitoring stations. Standard errors are underneath the
coefficient estim ates. O bservations from the years 1980 and 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
102
TABLE 24
1 2 3 4 5
NUMBER OF OBSERVATIONS 68 57 44 54 57
R2 0.412 0.085 0.248 0.327 0.205
ADJUSTED R2 0.333 0.051 0.211 0.300 0.176
MEAN OF DEPENDENT VARIABLE 120.595 121.990 119.241 119.050 117.483
124.683 *** 78.356 *** 78.926 ***
27.361 16.462 17.705
125.871 24.761 17.746
77.783 56.172 60.028
DUMMY VARIABLE FOR EAST NORTH CENTRAL STATES 13.798
23.190
DUMMY VARIABLE FOR MID ATLANTIC STATES 24.350
24.374
DUMMY VARIABLE FOR MID W EST STATES -7.860
31.119
DUMMY VARIABLE FOR MOUNTAIN PLAINS STATES -69.183
57.787
DUMMY VARIABLE FOR NEW ENGLAND STATES -18.313
32.055
DUMMY VARIABLE FOR PACIFIC SOUTHWEST STATES -106.25 ***
21.18
INTERCEPT 57.157 ** 111.491
26.510 28.042
ETHNIC 304.714 *** 137.932
71.927 69.736
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
103
TABLE 24 CONTINUED
OLS REGRESSIONS: CARBON MONOXIDE AIR POLLUTION- EMISSIONS FROM ALL SOURCES DIVIDED BY AREA
1 2 3 4 5
NUMBER OF OBSERVATIONS 68 57 44 54 57
R2 0.412 0.085 0.248 0.327 0.205
ADJUSTED R2 0.333 0.051 0.211 0.300 0.176
MEAN OF DEPENDENT VARIABLE 120.595 121.990 119.241 119.050 117.483
DUMMY VARIABLE FOR SOUTH CENTRAL STATES -33.601
25.144
PERCENTAGE OF COUNCIL SEATS -73.154 ***
ELECTED FROM 'AT LARGE' DISTRICTS 24.044
PERCENTAGE OF FACILITIES FULLY -44.350
INSPECTED FROM 2001-2003 35.411
NUMBER OF CITIES IN MSA 3.528 '
1.011
NUMBER OF COUNTIES IN MSA 7.832 ***
1.651
Each column in the table reports coefficients from an OLS regression. The unit of observation is a metropolitan
statistical area. The dependent variable Is the em issions of carbon monoxide in that MSA in pounds, from both
point and non-point sources, divided by the land area of the MSA. Standard errors are underneath the coefficient
estim ates. O bservations from th e year 1990 only are used.
* denotes significance at 10% level
** denotes significance at 5% level
*** denotes significance at 1% level
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
104
LIST OF POLLUTION INTENSIVE INDUSTRY SIC CODES
SIC SIC Description Pollutant
2046 Com oil AIR
2085 DISTILLED AND BLENDED LIQUORS AIR
2295 COATED FABRICS, NOT RUBBERIZED AIR
2611 PULP MILLS AIR
2754 COMMERCIAL PRINTING, GRAVURE AIR
2816 INORGANIC PIGMENTS AIR
2861 GUM AND WOOD CHEMICALS AIR
2873 Urea AIR
2911 PETROLEUM REFINING AIR
2951 ASPHALT PAVING MIXTURES AND BLOCKS AIR
2999 PETROLEUM AND COAL PRODUCTS, N.E.C. AIR
3241 CEMENT, HYDRAULIC AIR
3251 BRICK AND STRUCTURAL CLAY TILE AIR
3274 LIME AIR
3275 GYPSUM PRODUCTS AIR
3281 CUT STONE AND STONE PRODUCTS AIR
3295 MINERALS, GROUND AND TREATED AIR
3334 PRIMARY ALUMINUM AIR
3493 STEEL SPRINGS, EXCEPT WIRE AIR
2611 PULP MILLS METAL
2812 ALKALIS AND CHLORINE METAL
2816 INORGANIC PIGMENTS METAL
2819 INDUSTRIAL INORGANIC CHEMICALS, N.E.C. METAL
2823 CELLULOSIC MAN-MADE FIBERS METAL
3312 BLAST FURNACES AND STEEL MILLS METAL
3313 ELECTROMETALLURGICAL PRODUCTS METAL
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
105
3331 PRIMARY COPPER METAL
3339 PRIMARY NONFERROUS METALS, N.E.C. METAL
3341 SECONDARY NONFERROUS METALS METAL
3366 COPPER FOUNDRIES METAL
3479 METAL COATING AND ALLIED SERVICES METAL
2261 FINISHING PLANTS, COTTON TOXIC
2611 PULP MILLS TOXIC
2812 ALKALIS AND CHLORINE TOXIC
2816 INORGANIC PIGMENTS TOXIC
2819 INDUSTRIAL INORGANIC CHEMICALS, N.E.C. TOXIC
2823 CELLULOSIC MAN-MADE FIBERS TOXIC
2873 Urea TOXIC
2874 Phosphoric acid TOXIC
2895 CARBON BLACK TOXIC
3313 ELECTROMETALLURGICAL PRODUCTS TOXIC
3331 PRIMARY COPPER TOXIC
3339 PRIMARY NONFERROUS METALS, N.E.C. TOXIC
2022 CHEESE, NATURAL AND PROCESSED WATER
2023 DRY, CONDENSED, EVAPORATED PRODUCTS WATER
2611 PULP MILLS WATER
2631 PAPERBOARD MILLS WATER
2874 Phosphoric acid WATER
2899 Fatty acids WATER
3312 BLAST FURNACES AND STEEL MILLS WATER
3341 SECONDARY NONFERROUS METALS WATER
3914 SILVERWARE AND PLATED WARE WATER
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.
106
Source: IPPS (Industrial Pollution Projection System) Pollution Intensity and
Abatement Cost Datasets, available from the World Bank at
(http://www.worldbank.org/nipr/data/ippsdown.htm - ‘This is a zipped wkl version
using US industrial codes (SIC) (pounds of pollution)'). These are the top five most
pollution-intensive industries as measured by the amount of pollution produced
relative to value-added, value of output and employment, of air, water, metals and
toxics pollution.
R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Dependence on foreign labor, quality of education and unemployment in the GCC countries: In search of solutions

PDF

Analysis of the role of government in Taiwan's industrialization and economic development

PDF

Globalization and the decline of the welfare state in less developed countries

PDF

A comparative examination of the impact of business -government relations on labor market reform in Egypt and Mexico, 1975--1995

PDF

Before and after reform: The effects of structural adjustment on women's employment in Latin America

PDF

An empirical analysis of foreign aid and the provision of international public goods

PDF

Economic transition, micro- and small -scale enterprise in Zimbabwe, and the significance of participatory development to institutional economics

PDF

By sea: The port nexus in the global commodity network (the case of the West Coast ports)

PDF

Dialectical feminism and contradictory economic systems: Campesina daily life from Somoza to Aleman

PDF

Carrots or sticks: The Taiwanese government's contrasting policy approaches to banking sector liberalization

PDF

Political effects in environmental litigation

PDF

Three essays on democracy and economic policy

PDF

The implementation of endangered species policy: An analysis of the 'takings' issue in specific contexts

PDF

Context is everything: An empirical study of institutions, instability and economic growth

PDF

The political economy of procyclical fiscal policy in Mexico, 1970--1988

PDF

Buscando la prosperidad: Migration as long -term investment in El Salvador

PDF

Assessment of prognostic comorbidity in hospital outcomes research: Is there a role for outpatient pharmacy data?

PDF

Compliance study of second-generation antipsychotics on patients with schizophrenia

PDF

Trends in mortality in Turkey, 1960--1995

PDF

The political economy of Turkish Islamism and secularism: The role of preference falsification in the establishment of Kemalist secularism and the polarization of Turkish public opinion on the pu...

Asset Metadata

Creator Hughes, Paul Jeremy (author)

Core Title Essays in environmental policy

School Graduate School

Degree Doctor of Philosophy

Degree Program Political Economy and Public Policy

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag economics, general,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Nugent, Jeffrey B. (committee chair)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-348156

Unique identifier UC11335687

Identifier 3180356.pdf (filename),usctheses-c16-348156 (legacy record id)

Legacy Identifier 3180356.pdf

Dmrecord 348156

Document Type Dissertation

Rights Hughes, Paul Jeremy

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA