Environmental federalism: a survey of the empirical literature.
Generally, the debate over environmental federalism strongly focuses on anecdotal evidence and intuition. Empirical facts have not been the focus of arguments concerning the optimal allocation of environmental authority. For example, the Tiebout model, which highlights the positive side of decentralization as jurisdictions efficiently compete for mobile residents, relies on seven assumptions. Additionally, the group of models relying on the interjurisdictional competition framework, which have highlighted both the positive and negative outcomes of decentralized environmental authority, also rely on a number of assumptions. This Article assesses the empirical validity of many of these assumptions, concluding that the data may necessitate a rethinking of these assumptions.
CONTENTS INTRODUCTION I. ENVIRONMENTAL FEDERALISM IN THEORY II. ENVIRONMENTAL FEDERALISM IN PRACTICE A. Resource Mobility 1. Population Mobility 2. Capital Mobility B. Preference Heterogeneity C. Interjursidictional Externalities 1. Resource Externalities 2. Pecuniary Externalities 3. Fiscal Externalities D. Political Economy 1. Leviathan 2. Lobbying and Corruption 3. Political Participation 4. Knowledge E. Policy Instruments 1. Laboratory Federalism 2. Bottom-Up Federalism CONCLUSION
Professor Bednar defines federalism as a "system of government characterized by semiautonomous states in a regime with a common central government" where "governmental authority is allocated between levels of government." (1) Professor Gordon notes that within a federal system "each unit of government decides independently how much of each type of public good to provide, and what types of taxes, and which tax rates, to use in funding the public goods." (2) While many countries contain a federal political system, the "optimal" allocation of authority across levels of government is the subject of constant research and debate. (3) While much of this research and debate focuses on fiscal policy (taxation and expenditures), the allocation of authority over environmental decision making within a federal political system has also been discussed, deliberated, and agonized over for decades. (4) The puzzle concerning the optimal allocation of environmental authority across levels of government is commonly referred to as environmental federalism.
Although the issue of environmental federalism receives widespread attention across the globe, there is no resolution in sight. (5) Even some of the most prominent researchers in this area have a diverse set of beliefs. Professor Gordon concludes:
Competition among communities should lead to greater efficiency and innovation. However, this paper has shown the many ways in which decentralized decision-making can lead to inefficiencies, since a local government will ignore the effects of its decisions on the utility levels of nonresidents.... In light of these costs arising from lack of coordination, it may be preferable to have the central government take responsibility for particular activities, in spite of the lost diversity. (6)
Professor Oates states: "My own sense is that where environmental quality is basically a local public good, the case for the setting of environmental standards at an appropriately decentralized level of government is quite compelling. At the same time, one can envision an essential informational and guidance role for the central authority." (7) Professor Wilson concludes:
As such, competition among governments has both good and bad aspects, the importance of which vary across the attributes of the goods and services that the governments provide. This assessment suggests a role for intervention by a central authority, but both political considerations and information problems should be carefully addressed. (8)
Professor Adler states: "In sum, there is a strong case for a general presumption in favor of decentralization--a presumption that can be overcome in any specific policy context by demonstrating the need for federal intervention." (9) Professor Levinson writes: "The conclusion must be that under most practical circumstances, local environmental authority will lead to inefficient regulations." (10)
The diversity of views concerning the appropriate allocation of environmental authority also plays out in practice as different federations have "resolved" the issue differently. For example, the well-known Principle of Subsidiarity emanating from the Maastricht Treaty of 1992 constitutionalizes the delegation of environmental authority by dictating that centralized action is only allowed in situations where policy objectives cannot be sufficiently achieved through decentralized action. (11) This is consistent with the so-called Decentralization Theorem put forth by Professor Oates:
[I]n the absence of cost-savings from the centralized provision of a [local public] good and of interjurisdictional external effects, the level of welfare will always be at least as high (and typically higher) if Pareto-efficient levels of consumption of the good are provided in each jurisdiction than if any single, uniform level of consumption is maintained across all jurisdictions. (12)
In contrast, the delegation of authority is not constitutionalized in the United States except insofar as decentralized policymaking is not allowed to interfere with interstate commerce. (13) As a result, environmental authority in the United States has oscillated between periods of relatively greater centralized and decentralized control. (14)
Given this backdrop, the objectives of this article are twofold. The first objective is to provide a very brief summary of the main theoretical models put forth in the literature. The reason for doing so is to illuminate the issues that play a fundamental role in conclusions regarding the optimal allocation of environmental authority. The second objective is to then provide a comprehensive survey of the relevant empirical literatures for the first time in the legal literature. By doing so, the goal is to limit the scope of the debate over environmental federalism moving forward, as well as make clear where the gaps in empirical knowledge exist.
I. ENVIRONMENTAL FEDERALISM IN THEORY
The two primary theoretical frameworks used to explore the effects of the decentralization of policy decisions such as taxes, expenditures, environmental standards, etc. derive from Professor Tiebout and the literature on interjurisdictional competition. (15)
The Tiebout model highlights the positive side of decentralization as jurisdictions compete for mobile residents in such a way that yields outcomes that are efficient. (16) As laid out by Professor Revesz, the model relies on "seven important assumptions." (17) First, individuals are perfectly mobile across jurisdictions and have heterogeneous preferences over jurisdictional attributes. Second, individuals have perfect knowledge concerning the attributes of all jurisdictions, which include the "head tax" levied on residents and the level of public goods and services provided. Third, there exist a large number of jurisdictions. Fourth, employment does not affect individual residential choice as income is derived via dividends. Fifth, there are no inter jurisdictional externalities. Sixth, every jurisdiction has a (known) optimal size where the average cost of services provided is minimized. Seventh, jurisdictions below their optimal size seek to attract new residents. (18)
Professor Brueckner provides a concise summary of the model:
Tiebout argued that, in attempting to attract residents, fiscally autonomous subnational governments will tailor public spending to suit individual preferences, leading consumers to sort across jurisdictions according to their demand for public goods. With each individual able to exactly fulfill his or her demand in some jurisdiction, the economy achieves a market-like outcome in the provision of public goods. (19)
Thus, under the assumptions of the model, this market-like outcome is efficient.
In contrast, the interjurisdictional competition framework nests both the positive and negative sides of decentralization as jurisdictions compete for mobile resources, typically taken as capital. Dating back at least to Oates's 1972 model, the framework has led to a variety of theoretical models. (20) The 1988 model of Professors Oates and Schwab provides a useful starting point in the literature. (21)
Oates and Schwab find that it is possible for decentralized environmental authority to be efficient even with interjurisdictional competition for capital. (22) However, this result requires numerous assumptions. (23) First, individuals are homogeneous and immobile across jurisdictions. Second, capital is perfectly mobile across jurisdictions, seeking to maximize after-tax returns, and all production profits are earned locally. Third, capital has perfect knowledge concerning the attributes of all jurisdictions, which includes the tax rate on capital and level of public goods and services provided. Fourth, there exist a large number of jurisdictions that take the after-tax return on capital as given. Fifth, there are no interjurisdictional externalities. Sixth, governments maximize the (known) social welfare of their jurisdiction. Oates summarizes this model, stating that "the invisible hand works in much the same way as in the private sector to channel policy decisions in individual jurisdictions into an efficient outcome from a national perspective." (24) In sum, the Tiebout and interjurisdictional competition frameworks provide a definitive answer to the environmental federalism debate under certain assumptions. Failure of these assumptions, however, may reverse this conclusion. For example, Oates stresses the limitations of his earlier work, stating: "The problem is that these models make some strong assumptions.... [Violations of any of these conditions can lead to distorted outcomes." (25) Indeed, many theoretical models have extended the Oates and Schwab model by relaxing various assumptions, finding that decentralized environmental policy making with interjurisdictional competition may lead to environmental standards that are inefficiently stringent or lax. (26) Thus, the central takeaway message from this lengthy theoretical literature is eloquently provided by Oates:
The theoretical literature thus generates some diverse findings on this issue. There seem to be some basic efficiency-enhancing aspects of interjurisdictional competition, but there are clearly a range of "imperfections" that can be the source of allocative distortions. The real issue here is the magnitude of these distortions. Are we dealing with minor deviations from efficient outcomes--or does such competition produce major welfare losses? The pure theory can't help us much in answering this question. (27)
The debate, then, over environmental federalism cannot be settled using theory alone. The range of possible theoretical outcomes can only be limited by an empirical understanding of the magnitudes of any violations of the various assumptions invoked in the Tiebout and interjursidictional competition frameworks. Surprisingly, a comprehensive review of what is known and unknown does not exist to the Author's knowledge. The result is that the debate over environmental federalism focuses too much on anecdotal evidence and intuition and not enough on empirical facts. The remainder of this Article seeks to remedy this by assessing our current knowledge of the imperfections alluded to above by Oates.
Prior to continuing, several comments are warranted. First, while violation of any of the assumptions noted above may cause decentralized environmental policy making to be inefficient, this does not imply that centralized policy making is efficient (or even less inefficient). Thus, interest lies in not just the validity of the prior assumptions, but also the contrast in social welfare under local versus central environmental authority because neither system is likely to yield the efficient outcome in practice. (28) Such differences, to the extent possible, will be discussed in Part II.
Second, neither environmental policy nor regulation are homogeneous concepts. There are important sources of heterogeneity across environmental issues (for example, air pollution, water pollution, hazardous waste disposal, and energy), across stages of environmental policy (for example, scientific research, standard setting, measurement, and enforcement), and across environmental instruments (for example, emissions taxes, cap-and-trade, subsidies, and command-and-control policies). In all likelihood there is no single answer to the question of optimal allocation of environmental authority across levels of government. Differences may arise across specific environmental issues or stages of the policy process. Moreover, certain policy instruments may not be available to all levels of governments. Such differences will also be emphasized when possible in the remainder of this Article.
Finally, the quest to understand the optimal level of government at which to assign environmental authority depends on, among other things, one's definition of optimal. Economists often equate optimality with efficiency, which requires the equating of marginal social costs and benefits in all locations. Others may wish to incorporate equity or political considerations into the notion of optimality. (29) Still others may consider the optimal allocation of governmental authority as that which maximizes environmental quality. (30) As evidenced by the theoretical discussion earlier in this Part, the structure of this Article is guided by the factors affecting the efficiency of policymaking in theoretical models. That said, the efficient outcome is generally unknown, making the "first-best policy.... typically impracticable." (31) Thus, empirical evidence regarding the impact of local versus central control is often expressed in terms of the level or variance of environmental quality or the resulting nature of the political process. Professor Adler reminds us, however, to bear in mind that stricter regulation is not always efficient. (32)
II. ENVIRONMENTAL FEDERALISM IN PRACTICE
This Part discusses the empirical literature relating to several of the assumptions invoked in the Tiebout and interjursidictional competition frameworks. First, in Part II.A, the empirical literature on the mobility of capital and individuals is examined. Second, in Part II.B, preference heterogeneity is assessed. Third, in Part II.C, the importance of interjurisdictional externalities is evaluated. Fourth, in Part II.D, political economy issues are examined. Finally, in Part II.E, issues related to policy instruments are discussed.
A. Resource Mobility
The Tiebout framework has as its linchpin the perfect mobility of individuals. (33) The interjurisdictional competition framework has its linchpin the perfect mobility of capital. (34) Here, I review the empirical evidence on the mobility of these two resources, starting first with individual mobility.
1. Population Mobility
Efficiency in the Tiebout model is predicated upon the population to be perfectly mobile across jurisdictions in order for individuals to sort within communities that choose policies aligned with their preferences. Molloy, Smith, and Wozniak write: "The notion that one can pick up and move to a location that promises better opportunities has long been an important part of the American mystique." (35) Clearly, however, individuals are not perfectly mobile. Revesz states that "much of the legal literature has dismissed as unrealistic the assumption of perfect mobility by individuals. There may, indeed, be substantial transaction costs in exiting one jurisdiction and moving to another, particularly in a world in which individuals have jobs and do not live solely on dividend income." (36) That said, how mobile are individuals? And, does mobility differ across socioeconomic groups?
Evidence on residential mobility in the United States and other countries in the 1970s and 1980s is presented by Long. (37) Defining mobility as simply changing residential address, he finds that the United States has the second highest one-year residential mobility rate at 17.5% (behind New Zealand) and the third highest five-year residential mobility rate at 46.4% (behind Canada and Australia) in the early 1980s among the handful of countries analyzed. (38) However, changing residences within a jurisdiction is not the type of mobility envisioned in the Tiebout model. Upon further examination, Long reports that roughly sixty percent of one-year and five-year residential mobility is due to moves within the same county; (39) roughly eighty percent is due to moves within the same state. (40) Thus, very little internal migration crosses jurisdictional boundaries that might entail significant changes in the provision of environmental quality.
Molloy, Smith, and Wozniak provide a more current, in-depth analysis of internal migration in the United States. The crux of their analysis points to an interesting puzzle: despite having one of the highest rates of internal migration in the world, internal migration in the United States has declined since roughly 1980. (41) Specifically, five-year, cross-state migration rates in the United States fell from 9.9% of the population in 1980 to 8.9% in 2000. (42) Five-year, cross-county migration rates declined from 19.8% in 1980 to 18.6% in 2000. (43) Molloy, Smith, and Wozniak state:
This decrease marks a noticeable departure from the longer-run trend, as migration shows a secular rise from 1900 to 1990.... Not only are migration rates lower in levels than at any point in the post-war period, they have also entered a period of continuous decline that is longer than any recorded in the twentieth century. (44)
Beyond simply documenting rates of internal migration, several studies have assessed the determinants of migration and destination choice. Among different socioeconomic groups, the largest differences in migration rates appear across education levels and homeownership status. For example, Molloy, Smith, and Wozniak estimate the annual interstate migration rate for individuals with at least a college degree to be 2.1% over the period from 2001 to 2010 in comparison to only 1.2% for high school graduates and 1.0% for high school dropouts. (45) Similarly, the migration rate for renters is 3.5%, but only 0.9% for homeowners. (46) Surprisingly, there is little difference in migration rates by race or income status over this period. (47)
Several studies also investigate the role of economic factors on migration decisions. Professors Barro and Sala-I-Martin document a statistically significant, but small, effect of per capita income differentials on interstate migration. (48) Davies, Greenwood, and Li use a different statistical technique and obtain larger effects. Specifically, the authors find that one standard deviation increase in the ratio of the unemployment rate at a potential destination state to one's current state reduces the likelihood of migration to that destination by about nineteen percent. (49) Similarly, one standard deviation increase in the ratio of per capita income at a potential destination state to one's current state increases the likelihood of migration to that destination by about sixteen percent. (50) Finally, Professors Borjas, Bronars, and Trejo find that workers migrate to maximize their relative standard of living. In particular, high-skilled workers move to states where the return-to-skill rate is high (and, hence, income inequality is greater), while low-skilled workers move to states where the return-to-skill rate is low (and, hence, income inequality is lower). (51)
To the Author's knowledge, there is little empirical evidence on the role of environmental quality in explaining internal migration decisions. Barro and Sala-I-Martin explore interstate migration rates in the United States from 1900 to 1987. To proxy for environmental amenities, the authors include an independent variable measuring the average number of days requiring heating. This is admittedly an extremely crude proxy for environmental amenities. Nonetheless, the authors find it to be a meaningful determinant of net migration rates into particular states. (52) Similarly, Professor Kahn explores the role of smog--measured as days per year exceeding the Clean Air Act's one-hour standard for ozone--on county-level population growth in California between 1980 and 1994. He finds that a county experiencing a ten-day decline in the number of "high ozone" days over this time period registered a nearly 7.8% increase in population on average. (53)
Professors Banzhaf and Walsh undertake a detailed analysis using more spatially disaggregate data than these prior studies. (54) The authors use data from California over the period from 1990 to 2000. The authors divide urban areas of the state into mutually exclusive half-mile diameter circles. They then assess the relationship that changes in pollution have with changes in population and mean household income over this period. (55) Toxicity-weighted air emissions from the Toxic Release Inventory (TRI) are used as the measure of pollution. The results support
the role of environmental amenities in household location decisions. Specifically, locations with the average level of emissions in 1990, as opposed to no emissions in 1990, experienced about a twelve percent reduction in the population growth rate between 1990 and 2000. (56) The average location experiencing an increase in emissions over the time period suffered a 6.9% reduction in the population growth rate. (57) The average location experiencing a decline in emissions witnessed a 6.3% increase in the population growth rate. (58)
Lastly, Professor Konisky, in tangentially related research, finds that better educated individuals are more likely to favor federal control over most environmental issues, including local issues such as protecting community drinking water. (59) If the higher mobility rates of the better educated documented by Molloy, Smith, and Wozniak are related to environmental issues, one might expect the reverse to be true.
In sum, labor mobility is relatively high in the United States compared to many other developed countries, and there exists some evidence that location decisions are impacted by environmental amenities at a very spatially disaggregate level. But, mobility is on a "historically unprecedented" downward trajectory in the United States. (60) Whether this low (and declining) mobility invalidates the Tiebout assumption, or simply reflects a population close to equilibrium where most individuals are currently residing in communities that maximize their utility is not clear. Moreover, whether the "threat" of migration exists and is sufficient to satisfy the assumptions of the Tiebout model is equally unclear. (61)
However, three facts work against the Tiebout model. First, migration patterns appear to be strongly related to employment, not the provision of public goods. Second, the survey evidence provided in Konisky indicates that more mobile segments of the population are more likely to favor federal authority over even local environmental issues. Finally, the fact that mobility rates are significantly lower for the less educated runs the risk that jurisdictions do not compete for residents of this type, but instead focus on the competition for more mobile, educated residents. If so, the Tiebout model would predict that the preferences of lower educated residents will be ignored. Provision of even purely local goods, such as drinking water, may not reflect the preferences of all residents. Bednar states: "If only a categorical portion of the population is immobile--the poor or ethnic minorities--then outcomes are even worse; governments compete for the wealthiest and are free to ignore these minority categories." (62)
2. Capital Mobility
The empirical literature on capital mobility is much more extensive. Not only are there a number of studies assessing the impact of taxation on the location of industrial activity, but there are many studies focused solely on the impact of environmental regulation on the location of industrial activity. In the interest of relative brevity, this Article focuses solely on the latter and starts with the 1995 survey article by Jaffe, Peterson, Portney, and Stavins. (63) The authors survey the literature at the time on the effects of the early period of environmental regulation in the United States (post-1970) on country-level competitiveness. (64) To measure competitiveness, the authors focus on empirical studies examining trade patterns, both generally and specifically in pollution-intensive industries, domestic firm location decisions, and foreign direct investment (FDI). (65) Jaffe, Peterson, Portney, and Stavins summarize the literature, concluding:
Overall, there is relatively little evidence to support the hypothesis that environmental regulations have had a large adverse effect on competitiveness, however that elusive term is defined.... [S]tudies attempting to measure the effect of environmental regulation on net exports, overall trade flows, and plant-location decisions have produced estimates that are either small, statistically insignificant, or not robust to tests of model specification. (66)
Surveying the literature at roughly the same time, Levinson similarly notes that "the empirical literature suggests that ... economic activity does not respond significantly to the different taxes and regulations in competing jurisdictions." (67)
However, there are a number of statistical problems that plague these early empirical studies. In their concluding remarks, Jaffe, Peterson, Portney, and Stavins discuss the issue of measuring environmental stringency in practice. (68) Environmental regulation is a complex concept which does not lend itself to precise measurement. Regulations differ across pollutants and depend not just on legislation (de jure regulation) but also enforcement (de facto regulation). Reliance on poor proxies for environmental regulation- referred to as measurement error in the econometric literature--might explain the lack of meaningful effects found in these early studies. A second problem with this early literature is that the results lacked a convincing causal interpretation. The authors note: "[T]he choice of a new plant location is obviously a complex one.... Hence, isolating the effect of environmental regulations on the decision will inevitably be difficult." (69) In particular, the choice of environmental stringency in a particular jurisdiction may be endogenous to the level (or expected level) of economic activity, or may be correlated with other location-specific attributes that determine the location of economic activity but are unobserved by the researcher.
The empirical literature beginning in the late 1990s addresses these criticisms. As noted by Professors Jeppesen, List, and Folmer, this "second wave" of studies has consistently found meaningful, detrimental effects of environmental regulation on industrial activity. (70) Professors Brunnermeier and Levinson conclude:
The early literature based on cross-sectional analyses typically tended to find that environmental regulations had an insignificant effect on firm location decisions. However, several recent studies that use panel data to control for unobserved heterogeneity, or instruments to control for endogeneity, do find statistically significant pollution haven effects of reasonable magnitude. Furthermore, it does not appear to matter whether these studies look across countries, industries, states, or counties, or whether they examine plant location, investment, or international trade patterns. (71)
One branch of this second wave of studies uses the spatial and temporal variation in U.S. county-level environmental regulation induced by the Clean Air Act (72) (CAA) to address the question of capital mobility. Specifically, beginning in 1972, every U.S. county is designated as either in attainment or out of attainment (nonattainment) of the federally designated standard for each of the criteria air pollutants established under the statute. (73) Counties that are in nonattainment are subject to more stringent regulation. (74) Thus, nonattainment status is synonymous with greater regulatory stringency.
Professor Henderson uses data on the ozone attainment status of urban counties over the period from 1977 to 1987, along with data on the number of establishments in each county during each year from the U.S. Census Bureau's County Business Patterns, to examine the effects of the more stringent regulations imposed in nonattainment counties. (75) He focuses on five pollution-intensive manufacturing industries: industrial organic chemicals, petroleum refining, miscellaneous plastics, plastic materials and synthetics, and blast furnace and primary steel. The results indicate that counties in attainment over the prior three years contain seven to ten percent more establishments than other counties. (76)
In a subsequent study, Professor Henderson revisits the same data using a different statistical methodology and focusing on a binary indicator of the presence of at least one of the pollution-intensive establishments in a county during a given year. (77) He finds that the probability of a given industry being located in a county increases by at least fourteen percent in four of the five industries considered after a county switches from nonattainment to attainment. (78)
Becker and Henderson continue this line of inquiry, using plant-level data over the period from 1963 to 1992 for the industrial organic chemicals, miscellaneous plastic products, metal cans and barrels, and the wood furniture sectors. (79) The data used by Becker and Henderson came from the Longitudinal Research Database, administered by the Center for Economic Studies of the U.S. Census Bureau. The results are in line with the prior studies; nonattainment status reduces the expected number of new plant births by twenty-six to forty-five percent. (80) However, the authors find that new plants locating in nonattainment counties are initially larger in size (relative to new plants locating in attainment areas). This size discrepancy disappears after roughly ten years. (81)
Greenstone builds on this research by examining approximately .75 million plant-level observations over the period from 1967 to 1987, obtained from the Census of Manufactures, to assess the separate impacts of carbon monoxide, ozone, sulfur dioxide, and total suspended particulate nonattainment status across all counties in the United States. (82) With such rich data, the author is able to identify the effect of nonattainment status on plants in sectors deemed to be high emitters of each of the four criteria pollutants using only the temporal variation within plants, controlling for industry-level and county-level unobserved attributes in an unrestricted way. The results indicate that over the period from 1972 to 1987, nonattainment counties (relative to in attainment counties) lost roughly 590,000 jobs, $37 billion in capital stock, and $75 billion in output. (83)
In a series of papers by Professor List and co-authors, the impacts of ozone nonattainment status on county-level industrial activity in New York State are examined over the period from 1980 to 1990. Annual data on new plant births by domestic and foreign firms, as well as plant relocations, are provided by the Industrial Migration File maintained by the New York State Department of Economic Development. (84) List, McHone, and Millimet focus on plant relocations and find that being in nonattainment costs a county roughly 0.50 fewer relocating plants in pollution-intensive sectors per year relative to being in attainment. (85) Given that the average county in the sample only receives about 0.2 relocating plants per year, this represents a sizeable decrease. List, Millimet, McHone, and Fredriksson analyze new plant births using similar statistical techniques. Here, the authors find even larger effects: nonattainment counties lose out on approximately one new pollution-intensive plant per year relative to counties in attainment. (86) The average county in the sample obtains 0.4 new plant births per year. Finally, List, McHone, and Millimet revisit the data on new plant births, distinguishing between new plant births by foreign and domestic firms. The authors find that the prior deterrent effects of nonattainment status on new plant births is driven entirely by domestically owned plants; foreign-owned plants are unresponsive to spatial variation in environmental regulatory costs due to differences in ozone attainment status. (87)
The other branch of this second wave of the empirical literature on the effects of environmental regulation on the location of industrial activity analyzes patterns of international trade and FDI (the so-called pollution haven hypothesis). The criticisms of the early literature are particularly worrisome when examining the effects of cross-country differences in environmental regulation. First, one needs a consistent measure of environmental regulation across multiple countries. Second, one needs extensive data on country (or industry) characteristics; otherwise any association between environmental stringency and the patterns of global trade or investment may simply reflect unobserved attributes of countries (or industries) correlated with both regulation and the location of industrial activity. The traditional econometric approach used to overcome measurement error in the proxy for environmental regulation and unobserved, country-level heterogeneity is the method of instrumental variables. While not all studies employing instrumental variables are necessarily sound (due to the use of weak or potentially invalid instruments), this Article focuses only on studies that employ this technique as the alternative--such as cross-sectional or fixed-effects estimation-- because this approach is not likely to produce causal estimates of the impacts of environmental regulation in this Article's author's view (due to the requirement that all independent variables in such models be strictly exogenous).
The first set of studies focus on patterns of FDI. Professors Xing and Kolstad assess the pattern on U.S. outbound FDI in six manufacturing sectors across twenty-two host countries using data from 1985 and 1990. (88) As a proxy for environmental regulation, the authors utilize sulfur dioxide (S[O.sub.2]) emissions. The authors find more lax regulation (as measured by higher S[O.sub.2] emissions) leads to greater investment in the United States within two pollution-intensive industries: chemicals and primary metals. (89) There is no meaningful effect for the other, less pollution-intensive, sectors considered.
Professors Fredriksson, List, and Millimet examine the impact of environmental stringency on U.S. inbound FDI across the forty-eight contiguous states for the period from 1977 to 1986. (90) To measure environmental regulation, the authors utilize an index based on the ratio of actual pollution abatement expenditures incurred by plants located in the state to the predicted level of expenditures based on the state's industrial composition. The results indicate a meaningful deleterious effect of stringency on inbound FDI. For example, the effect of one standard deviation in stringency in California in 1984 is predicted to lower employment in the foreign-owned chemical plants by six percent--or 2500 jobs. (91)
Professors Cole and Elliott analyze the patterns of U.S. outbound FDI in Brazil and Mexico across thirty-one and thirty-six industries respectively from 1989-1994. (92) The authors' goal is to disentangle the effects of environmental regulatory costs--measured by industry-level pollution abatement costs per unit of value added--on firm behavior from the effects of endowments of physical capital. Specifically, the authors contend that since many pollution-intensive sectors are also capital-intensive, the benefit to firms from more lax environmental standards in less developed countries may be offset by lower levels of physical capital in these same countries. (93) Brazil and Mexico, the authors argue, are relatively well endowed with physical capital but have lax environmental standards, thus making it more likely that investment in these countries will be more sensitive to environmental costs. (94) Indeed, the authors find evidence supporting the claim that sectors experiencing relatively high environmental regulatory costs in the United States do invest more in each country.
Cole and Fredriksson utilize data on inbound FDI to thirteen countries that are members of Organisation for Economic Cooperation and Development (OCED) and twenty less developed countries over the period from 1982 to 1992. (95) The proxy variable for environmental regulation is the allowable lead content in gasoline. The results point to a sizeable, deleterious effect of environmental regulation on the amount of inbound FDI enjoyed by a country. For example, roughly one standard deviation increase in the authors' measure of regulation leads to approximately a 0.6 standard deviation reduction in a country's stock of FDI. (96)
Professor Kellenberg focuses on effects of country-level environmental stringency on the value added of majority-owned U.S. affiliates in fifty countries across nine industries over the period from 1999 to 2003. (97) To measure environmental policy, Kellenberg uses data from the Global Competitiveness Report (GCR) on the stringency and enforcement of environmental regulations. The GCR creates separate indices related to stringency and enforcement based on survey responses elicited from executives. (98) The results point to a meaningful effect of environmental regulation on the location of production abroad by majority-owned U.S. affiliates. Specifically, for the top quintile of countries in terms of the increase in economic activity by majority-owned U.S. affiliates over the sample period, 8.6% of this increase is attributable to more lax environmental policies in these countries. (99)
Finally, Professors Millimet and Roy analyze the pattern of inbound FDI across the forty-eight contiguous U.S. states from 1977 to 1986 and 1988 to 1994 using recently developed econometric techniques designed for situations where valid instrumental variables are difficult to envision. (100) The measure of state-level environmental regulation is identical to that used by Fredriksson, List, and Millimet. (101) The authors find an adverse impact of environmental regulation on the amount of FDI locating in a state for the pollution-intensive chemical sector, but no effect on the level of FDI for manufacturing as a whole. (102)
The second set of studies focus on patterns of international trade. Cole and Elliot examine the impact of environmental regulation on the net exports of sixty countries in four pollution-intensive sectors in 1995. (103) The sectors studied include iron and steel, chemicals, pulp and paper, and non-ferrous metals. Two measures of environmental stringency are employed. The first is an index based on country reports concerning environmental policies and enforcement compiled under United Nations Conference on Environment and Development guidelines. The second is a proxy variable computed using information on the energy intensity of production. The authors find no effects of environmental stringency on net exports, but they do find effects on the composition of trade: stricter regulation leads to a greater fraction of trade that is inter-industry. (104)
Several studies utilize U.S. industry-level measures of environmental costs based on pollution abatement expenditures to assess the impact of regulation on industry-level measures of U.S. imports or exports. Professors Ederington and Minier analyze data on the net imports of 374 U.S. industries over the period of 1978, 1980 to 1986, and 1988 to 1992. (105) The authors find extremely large effects of environmental costs on net imports. Specifically, they find that a one percent increase in pollution abatement costs results in a thirty percent increase in net imports scaled by the total value of U.S. shipments in the industry. (106) Similarly, Cole, Elliott, and Shimamoto analyze the effect of environmental costs on U.S. net exports across ninety-six industries over the same time period. (107) The authors also control for the human and physical capital intensity of sectors, finding a detrimental effect of environmental costs on net exports. However, the magnitude of the effect is much smaller than that attributable to human or physical capital considerations. (108)
Finally, Professors Levinson and Taylor examine U.S. net imports from Mexico and Canada across 132 industries over the period from 1977 to 1986. (109) The authors also find a large, adverse effect of environmental costs on domestic production. For example, net imports from Canada increased by $601 million for the average industry in the sample over this period; $79 million of this increase is attributable to the rise in pollution abatement costs. (110) Among the twenty industries in the U.S. that experienced the largest rise in pollution abatement costs, the numbers are $595 million and $453 million, respectively. (111)
The final two studies utilize data across several countries. Professors Jug and Mirza analyze the import patterns of twelve countries in the EU15 and export patterns of nineteen countries from the EU15, central Europe, and eastern Europe across nine industries over the period from 1996 to 1999. (112) Environmental regulation is measured using data from the Eurostat Environmental Expenditures and Environmental Taxes database on total current expenditures related to environmental protection activities for all manufacturing. The results indicate a meaningful, negative effect of environmental stringency on domestic production, with results of greater magnitude for eastern Europe. (113)
Professors Mulatu, Gerlagh, Rigby, and Wossink utilize data on sixteen industries across thirteen countries to assess the role of environmental regulation on the share of production in each industry that occurs in each country. The objective is to determine whether countries with relatively less stringent regulation are responsible for a greater share of production in pollution-intensive sectors over the period from 1990 to 1994. The country-level measure of regulation is given by the Environmental Sustainability Index (collected in 2001) developed by the World Economic forum, Yale Center for Environmental Law, and Center for International Earth Science Information Network at Columbia University. As with the prior studies, the findings point to significantly greater domestic production in pollution-intensive industries in countries with lax regulation. (114)
In sum, second generation studies--utilizing better data to identify the causal effect of environmental policy--have consistently documented a meaningful effect of environmental stringency on the location of economic activity. That said, these findings must be interpreted carefully. First, environmental costs are a small fraction of the total production costs for most industrial sectors. (115) Thus, the effects documented in the literature typically apply to only the most pollution-intensive industries. For the vast majority of industries, environmental costs are a small fraction of overall costs and location decisions are dominated by other factors. For example, Henderson and Millimet allow for heterogeneous effects of more stringent environmental policy on the amount of foreign investment and find the effects to be negative for some locations and positive for others. (116) Consistent with this finding, Henderson and Millimet had previously found no overall effect of environmental stringency on the Gross State Product of U.S. states. (117)
Second, the effects estimated by econometric models are ceteris paribus effects. In other words, they indicate the impact of more stringent environment regulation with "other things being equal" or with the assumption that all other variables are being held constant. (118) In practice, other important determinants of the location of economic activity are not held fixed such as a location's endowment of physical and human capital. This fact is highlighted in several analyses. Studies by Ederington, Levinson, and Minier and Levinson show that the rate of increase in total U.S. imports has risen faster over the past few decades than the rate of imports of pollution-intensive goods. (119) As such, the pollution content of U.S. imports has fallen over a period where U.S. regulatory stringency has increased markedly. (120) Professors Grether, Mathys, and de Melo analyze the pollution content of imports across forty-eight countries and find that environmental standards play a small role in the overall pattern of trade. (121) In this vein, Cole and Elliott conclude: "We do not suggest that pollution havens are widespread." (122)
Finally, while the studies reviewed in this Part make use of instrumental-variables or fixed-effects strategies in an attempt to isolate the causal effect of environmental regulation, not all identification strategies are convincing. (123) Accordingly, readers should be cognizant that simply because researchers employ panel-data-fixed-effects or instrumental-variables methods does not mean that such strategies produce credible inferences.
B. Preference Heterogeneity
The Tiebout framework relies on heterogeneous individuals sorting themselves across communities offering different combinations of taxes and public goods. In equilibrium, the policies of each community will reflect the preferences of its homogeneous residents. The literature on environmental federalism, and fiscal federalism more generally, has long touted the ability of communities to synchronize policy choices with individual preferences as the primary advantage of decentralized policy making. Gordon states: "One of the key advantages of decentralization is the resulting diversity of policies." (124) Oates writes: "The hope is that state and local governments, being closer to the people, will be more responsive to the particular preferences of their constituencies...." (125) Adler asserts that "localized control of environmental policy will produce environmental measures that are more likely to reflect the preferences and needs of those who will be most affected by them." (126)
In practice, the advantages to decentralized policy making depend on three factors: (1) the extent of preference heterogeneity in the population; (2) the degree to which individuals act on such preferences to sort themselves into homogeneous communities; and (3) the ability of local governments to better respond to community preferences than the central government. I discuss the empirical evidence on each in turn.
For the purposes of evaluating the benefits of decentralization, empirical evidence on the extent of heterogeneous preferences over environmental issues is scant. Several papers utilize survey or voting data to examine associations between socioeconomic characteristics and preferences. Elliott, Seldon, and Regens use data from the General Social Survey, which is administered by the U.S. National Opinion Research Center, to analyze public attitudes toward environmental spending over the periods from 1974 to 1978, 1980, and 1982 to 1991. (127) The authors find that liberalism, lower age, being female, being nonwhite, urban status, education, and income are positively associated with preferences for environmental spending. (128) However, one is not able to discern how much overall variation in preferences exists, nor how much of this variation is explained by these attributes.
Additional analyses have been undertaken using individual-level, cross-country data. Professors Israel and Levinson utilize data spanning thirty-three countries from the World Values Survey during the mid-1990s. (129) Consistent with the prior study, the authors find that lower age, being female, education, and income are positively associated with willingness to pay for environmental improvements. (130) However, the vast majority of the variation in willingness to pay (seventy-two percent) is explained by the country one resides in, not by the socioeconomic attributes included in the analysis. (131) As such, over two-thirds of the international variation in willingness to pay occurs across countries, with the rest representing within-country variation. (132)
Lorenzoni and Pidgeon discuss data collected by the European Opinion Research Group in 2002 across EU15 member states on individual concerns regarding climate change. (133) The authors report the percentage of survey respondents in each country reporting that they are "very worried" about climate change. The percentages varied from about twenty-one percent in The Netherlands to about sixty-three percent in Greece. (134) When asked about concern over future trends in climate change, the percentage responding "very much" or "quite a lot" varied from roughly forty-nine percent in the Netherlands to roughly eighty-five percent in Greece and Italy. (135) Thus, there is variation in concern over climate change both within and across countries.
Professors Kahn and Matsusaka examine county-level voting on sixteen environmental ballot initiatives in California spanning 1970 to 1994. (136) The authors report the county with the lowest and highest fraction of votes in favor of each ballot initiative. (137) For example, Proposition 1986-65 that sought to restrict the release of chemicals into drinking water sources received 32.7% votes in favor in the least favorable county and 78.3% votes in favor in the most favorable county. (138) Other ballot measures yielded greater variation. Votes in favor of Proposition 1990-130 that sought to ban clear-cutting of forests and authorize a $742 million bond issue to buy forest land ranged from 15.3% in the least favorable county to 70.7% in the most favorable county. (139) This is indicative of significant variation in preferences across counties. Kahn and Matsusaka further find that much of this cross-county variation is explained by differences in per capita income, urban status, education, and per capita income derived from specific industries (construction, farming, forestry, and manufacturing). (140)
Rather than relying on survey response data, several studies utilize information on home values and environmental amenities (or disamenities) to assess the demand for environmental quality. Professors Zabel and Kiel combine data from the American Housing Survey, the U.S. Census Bureau, and the EPA across four cities (Chicago, Denver, Philadelphia, and Washington, D.C.) from 1974 to 1991 to estimate household-level marginal willingness to pay for air quality by relating local air quality to home values. (141) The authors then examine the association between household characteristics and this value. In contrast to the results above, Zabel and Kiel find lower marginal willingness to pay for nonwhites and no meaningful association with gender. (142)
Professors Brasington and Hite utilize data on home sales across six metropolitan areas in Ohio (Akron, Cincinnati, Cleveland, Columbus, Dayton, and Toledo) in 1991 to first estimate the implicit price of distance from the nearest environmental hazard and then examine the effect of price, income, and other attributes on the demand. (143) The authors find a small, negative price elasticity of demand for distance from the nearest environmental hazard and a positive income elasticity of demand for distance from the nearest environmental hazard. (144) Carruthers and Clark follow the strategy utilized by Brasington and Hite and examine data on home sales in King County, Washington in 2004. (145) The authors obtain similar results in terms of the price and income elasticity of demand for distance from the nearest environmental hazard. (146)
The second factor affecting the advantages to decentralization relates to the degree to which individuals act on variation in environmental preferences to sort themselves into homogeneous communities. The preceding studies on preferences over environmental quality provide little evidence in this regard. With the exception of Kahn and Matsusaka, the studies discussed using survey responses or voting behavior are geographically aggregated. Thus, while preferences vary across individuals, there is no information concerning the level of variation in the total population versus the level of variation in a single community. Kahn and Matsusaka, however, do provide evidence of significant variation in voting behaviors across California counties driven by differences in income and industrial composition.
Studies utilizing housing prices to infer something about the willingness to pay for environmental amenities or how the prices of environmental amenities affect its demand also fail to provide insights into the amount of overall variation in preferences or whether preferences are more homogeneous at the city level than at the state or country level. Lastly, the empirical evidence on population mobility discussed in Part II.A.1 indicates that residential location choices, at least in the United States, are driven primarily by employment prospects.
The final factor impacting the advantages to decentralized policy making pertains to the ability of local governments to better respond to community preferences than the central government. If preferences concerning the environment do vary across individuals and individuals do sort into homogeneous communities, both of which are far from certain, are local policy makers better able to align environmental policies with these preferences? While this is typically asserted, as evidenced by some of the quotes above, there is no evidence to support this contention. Wilson states:
[W]e do not have a good understanding of how information asymmetries occur between different levels of government, and what exact form these asymmetries take. Rather, we have vague ideas, such as the understanding that local officials know more because they are "closer to the people." ... [I]t seems difficult to justify why the central authority cannot easily obtain the information that is assumed absent. (147)
In sum, the empirical evidence regarding preference heterogeneity and its implications on environmental federalism is limited and incomplete. While it seems likely that preferences do vary, perhaps much of this variation is across countries. The main drivers of within-country variation are income, education, and industry. Because these attributes are easily observed, Wilson's point about central governments being as capable as localities of understanding local preferences appears valid. That said, it is not clear that communities are particularly homogeneous with respect to these attributes. For example, in terms of income, the level of income inequality in 1999 across all households in the United States was 0.463 as measured by the Gini coefficient, a common measure of inequality. (148) The Gini coefficient ranges from zero to one, where zero indicates perfect equality (all households have identical incomes) and one indicates perfect inequality (one household possesses all income). The corresponding state-level Gini coefficients ranged from 0.402 (Alaska) to 0.499 (New York). (149) County-level Gini coefficients are available pooling data from 2006 to 2010. Between counties, the Gini coefficients vary from 0.207 (Loving, TX) to 0.645 (East Carroll Parish, Louisiana), while the Gini coefficient for the United States as a whole was 0.467 over this time period. (150) Among the twenty-five most populous counties in the United States, the Gini coefficients range from 0.417 (Suffolk County, New York) to 0.601 (New York County, New York). (151)
This suggests that there is not much sorting at the state or county level according to income, which is a strong predictor of environmental preferences. In any event, much more research is needed on the extent of preference heterogeneity and, more importantly, whether individuals sort themselves such that localities within a country are less heterogeneous than the country as a whole.
C. Interjursidictional Externalities
Efficiency of decentralized policy making in both the Tiebout and inter jurisdictional competition frameworks requires local governments to internalize all externalities. (152) A lengthy empirical literature has emerged assessing the practical relevance of spillovers across jurisdictions. This is crucial because the failure of local governments to internalize all externalities is often cited as the main argument against decentralized policy making. For example, Engel writes:
The interstate spillover rationale is the classic economic efficiency argument that federal intervention is necessary to prevent the environmental, social, and economic losses that accrue when air and water pollution originating in one state are carried by natural forces into other states. States from which the pollution originates have little incentive to curb interstate pollution because they benefit from having the harmful effects of pollution externalized while they enjoy the economic benefits of the polluting activity. (153)
Professor Sigman notes that the size of "spillovers across jurisdictions ... is a central question in the literature on the problem of assigning functions to different levels of government." (154) Adler states: "The strongest case for federal involvement comes in the context of interstate spillovers." (155) Hall writes: "One of the foundational justifications for the federalization of environmental law is the problem of interstate environmental harms." (156) Dijkstra and Fredriksson state: "Transboundary pollution is a standard and well-known reason for preferring centralized environmental policy making...." (157) Professor Esty notes: "While a few environmental harms (some waste problems, for example) are geographically localized, many forms of pollution (surface water contamination and most air pollutants, for example) spread across the land. Because state boundaries often do not fully encompass airsheds and watersheds, interjurisdictional externalities arise." (158)
The empirical literature can be categorized by the type of externality considered. This Article will refer to the first category as resource externalities. This is the case where the actions of one jurisdiction affect the resource quantity or quality available to other jurisdictions. (159) These situations are discussed in Part II.C. (1). The second category is referred to as "pecuniary externalities]" by Wilson. (160) This refers to situations where the actions of one jurisdiction affect prices in other jurisdictions. These situations are discussed in Part II.C.2. A final category of externalities is denoted by Wilson as "fiscal externalit[ies]." (161) This is the case where the policy choices of one jurisdiction have effects on the policy choices of other jurisdictions through strategic policy making. (162) This will be discussed in Part II.C.3.
1. Resource Externalities
The empirical literature on resource externalities focuses on whether jurisdictions fail to consider the effects of their actions on the quality or quantity of a resource available in other jurisdictions. Transboundary pollution is the canonical example. With transboundary pollution, one jurisdiction fails to consider the full environmental consequences of its actions as some of the costs--typically in terms of greater air or water pollution--fall outside one's jurisdiction. Another type of spillover may occur when resources are shared across multiple jurisdictions. Common examples of such shared resources are watersheds, fisheries in the ocean, and endangered species that are present in multiple jurisdictions.
Professors Murdoch, Sandler, and Sargent examined emissions of sulphur and nitrogen oxides in twenty-five European countries during the 1980s and early 1990s. (163) The authors explore the reductions in these types of emissions achieved after the formulation of the 1979 Convention on Long-Range Transboundary Air Pollution (LRTAP). In particular, the authors are motivated by the fact that mandated reductions for sulfur were more likely to be met than for nitrogen oxides (N[O.sub.x]). In an attempt to understand this, the authors investigate the importance of transboundary spillovers in explaining cross-country variation in emissions reductions achieved under LRTAP. Specifically, the authors estimate separate econometric models for the two types of emissions and incorporate two measures of spillovers as independent variables in the model. The first measure captures the fraction of a country's emissions that stay within its own borders. The second measure captures the amount of emissions originating outside of one's own jurisdiction that ends up in one's jurisdiction. (164)
Since sulfur emissions are less likely to spill across jurisdictional boundaries, if these variables are meaningful (in a statistical and economic sense) determinants of emissions, then externalities not only play an important role in determining emissions, but also help explain the disparate trends in sulfur and nitrogen oxide emissions in Europe over this time period. The results point to a very meaningful effect of the second measure of spillovers: as the amount of emissions entering one's jurisdiction from other countries declines, a country's own emissions rise. The authors interpret this finding as indicative of free riding behavior by countries. (165) However, the first measure is found to have at best a weak statistical relationship with emissions. (166)
Kahn examines the importance of transboundary pollution using data at the county-level from the United States. (167) Specifically, he assesses the impact of manufacturing activity in one's own county, as well as adjacent counties, on ambient concentrations of total suspended particulates (TSP). The data are from 1977, 1982, and 1987, thus spanning a period during which manufacturing activity experienced large declines. However, the spatial and temporal variation in this decline allows Kahn to assess its impact on own and neighboring county air quality.
Kahn's results suggest that manufacturing activity in adjacent counties has meaningful effects on a county's own ambient concentrations. (168) For example, a full standard deviation increase in a county's own activity in the primary metal industry raises ambient concentrations of TSP by 3.5%. A standard deviation increase in activity in the primary metals industry in a county's adjacent neighbors raises a county's own ambient concentrations of TSP by 1.1%. A standard deviation increase in activity in the stone, clay, and glass industry in a county's adjacent neighbors raises a county's own ambient concentrations of TSP by 4.1%. (169) For the other industries Kahn considers, the discrepancy between the effects of a county's own activity and that of its neighbors is smaller.
A series of studies by Sigman investigates pollution in rivers in order to assess the empirical relevance of transboundary spillovers. In the first study, Sigman uses international data on water quality obtained from monitoring stations on rivers administered by the United Nations' Global Environmental Monitoring System Water Quality Monitoring Project (GEMS/Water). (170) The data span 291 river monitoring stations across forty-nine countries over the period from 1979 to 1996. Sigman measures water quality using biochemical oxygen demand (BOD). The results are consistent with free riding behavior by countries. (171) Pollution is meaningfully higher at upstream locations (locations before a river flows into another country), as well as in rivers that form the political border between two countries. Interestingly, these effects apply predominantly to non-European Union (EU) countries. (172) Thus, institutional arrangements within the EU are hypothesized to limit the extent of free riding in terms of river pollution.
In her next study, Sigman undertakes a similar analysis with U.S. data obtained from the National Stream Quality Accounting Network, administered by the U.S. Geologic Survey. (173) The data come from 501 monitoring stations and span the period from 1973 to 1995. Sigman uses changes in the "authorization" or "primacy" status of neighboring states under the Clean Water Act (174) to assess the extent of upstream states to free ride on their downstream neighbors. (175) Since the statistical procedure only utilizes temporal variation within states arising from changes in authorization status, the results are more likely to capture the causal effect of decentralized control. (176) Measuring water pollution using the EPA's water quality index based on five major pollutants (dissolved oxygen, fecal coliform, total suspended solids, phosphorous, and nitrogen), the results indicate a four percent reduction in water quality at sites downstream from an authorized state. (177) In addition, rivers forming the border between states suffer a six percent reduction in water quality if at least one state is authorized. (178)
In a more recent study, Sigman continues to examine pollution levels in rivers at the international level using data from GEMS/Water spanning forty-seven countries over the period from 1979 to 1999. (179) Two measures of pollution are examined: BOD and fecal coliform. Compared to BOD, fecal coliform is considered to be more of a local pollutant. (180) In contrast to his earlier studies, here Sigman assesses the impact of decentralization, defined at the country-level, on subnational variation in pollution. (181) Decentralization is either measured using a binary indicator for a federalist system, or as a continuous measure of the ratio of subnational government expenditures to total government expenditures (net of any intergovernmental transfers). (182) The most convincing results that utilize only temporal variation in the continuous measure of decentralization suggest a harmful effect of decentralization on BOD but not fecal coliform. (183) Moreover, decentralization is also found to increase the subnational variation in both BOD and fecal coliform, consistent with a tailoring of decentralized policies to local preferences. (184)
Other papers pursue a similar strategy to Sigman and assess the impact of proximity to political boundaries on pollution. Professors Helland and Whitford use reported emissions by U.S. establishments from the Toxic Release Inventory (TRI) spanning between 1987 and 1996 to determine if emissions are higher from establishments located in counties that border neighboring states (either any border or only along the eastern edge). (185) The results point to significantly higher air and water emissions in establishments located on any border, and effects even larger among establishments located on eastern borders. (186) Interestingly, the results point to greater spillovers when the authors allow for the possibility that establishments located in border counties may systematically differ in unobserved dimensions from other establishments. (187)
Professors Gray and Shadbegian utilize data on 409 U.S. pulp and paper mills, from thirty-eight states, over the period from 1985 to 1997. (188) The authors examine plant-level air emissions of particulates (PM10) and sulfur dioxide (S[O.sub.2]) and water emissions of BOD and total suspended solids. The importance of spillovers is measured by assessing the sensitivity of a plant's emissions to the distance to the nearest state or Canadian border, as well as by the marginal benefits to reductions in air and water emissions enjoyed by neighboring jurisdictions. (189) For plants located within fifty miles of the Canadian border, only BOD emissions are higher than other plants further from the Canadian border; S[O.sub.2] emissions are lower (attributable to the focus on acid rain near the U.S.-Canadian border). (190) The results also suggest that plants reduce their emissions by less when the marginal benefits from such reductions are enjoyed by neighboring jurisdictions. For example, the authors find that out-of-state benefits to a reduction in S[O.sub.2] emissions have only one-third the impact of in-state benefits. (191)
Professors Lipscomb and Mobarak assess pollution in rivers at upstream and downstream locations in Brazil. (192) The authors analyze quarterly data on BOD levels from several hundred monitoring stations during the timeframe from 1990 to 2007. The importance of spillovers on pollution levels are investigated by measuring the distance of each station from its nearest upstream and downstream border. While generally such distances may be endogenous, Brazil frequently redraws its jurisdictional boundaries allowing the authors to exploit only the temporal variation in the distance of a given station to its nearest borders. (193) Lipscomb and Mobarak find meaningful evidence that spillovers matter. As the distance to the nearest downstream border falls from ten to nine kilometers, say, pollution increases by 1.3%; pollution rises by 1.9% as distance falls from one kilometer to zero (zero represents the point at the actual border). (194) Thus, BOD levels rise at an increasing rate as the river approaches its downstream border. In further analysis, the authors examine pairs of station monitors and find that BOD levels at the downstream monitor, relative to BOD levels at the upstream monitor, are 3.1% higher per jurisdictional boundary that the river crosses between the two stations. (195)
In more recent work, Hatfield and Kosec exploit the variation in the number of counties spanned by U.S. metropolitan areas to determine if areas divided across more counties experience greater pollution. (196) Because division into multiple jurisdictions may not be random, the authors exploit jurisdictional boundaries caused by the presence of streams in order to isolate the causal effect of the number of jurisdictions on pollution. (197) Pollution is measured using the average of the EPA's Air Quality Index over the years between 1999 and 2002, as well as the ambient concentrations of several individual pollutants. (198) Finally, the authors also examine the effect of increasing the number of jurisdictions within a metropolitan area on drinking water quality using data from the EPA's Safe Drinking Water Information System database. Since drinking water is a local good, the authors did not expect to find any effect of jurisdictional boundaries on this outcome. (199) The results indicate a sizeable effect of increasing the number of counties within an area on air pollution. (200) For example, changing a metropolitan area from being entirely contained in one county to split among two counties worsened air quality by half a standard deviation and added an additional thirteen days per year where the air is considered unhealthy. (201) Moreover, Hatfield and Kosec find that a similar increase from one to two counties within a metropolitan area increases the concentrations of carcinogenic pollutants by anywhere from nineteen percent to 250%, depending on the pollutant. (202) However, the authors found no effect on drinking water quality, consistent with their results being driven by interjurisdictional externalities. (203)
Professors Kahn, Li, and Zhao assess river pollution in China, taking advantage of a unique natural experiment. (204) Using data from 2004 to 2010, the authors assess relative pollution levels at internal versus border locations. (205) Beginning in 2005, the central government began monitoring local compliance with environmental targets related to the chemical oxygen demand (COD) of rivers. Compliance became a criterion upon which the political promotion of local officials is based. Prior to 2005, local officials had little incentive to reduce river pollution near jurisdictional boundaries. After 2005, this is no longer the case for officials seeking promotion. (206) However, other pollutants besides COD are not a part of the promotion criteria. Using pollution data from 499 river monitoring stations located in China's seven major rivers, the authors find that the 2005 change reduced COD levels. Moreover, the decline was greater at border locations, consistent with significant transboundary pollution prior to 2005. (207) Finally, the authors find no impact of the promotion criteria on other measures of pollution. (208)
Other studies pursue very different strategies for assessing the importance of interjurisdictional externalities. Banzhaf and Chupp present a detailed simulation model of the U.S. electricity sector, incorporating N[O.sub.x] and S[O.sub.2] emissions. (209) They then compare the level of welfare achieved under a first-best policy where each state fully internalizes all transboundary pollution damages, a second-best uniform policy across all states, and the decentralized case where each states acts only in its best interest. The results indicate that social welfare is only 0.2% lower under the second-best uniform policy compared to the first-best policy. (210) However, the decentralized case with self-interested states results in a 31.5% reduction in social welfare. (211) This discrepancy does not arise due to a lack of preference heterogeneity across states. Rather, the welfare cost of ignoring such heterogeneity under the second-best uniform policy is swamped by the welfare cost of failing to internalize the pollution externalities. (212) Evidence of the magnitude of the externality is further provided by Chupp in separate work. He illustrates for two states, Arizona and North Carolina, that the in-state marginal benefit per ton of S[O.sub.2] reduction is only about one-fourth the marginal benefit to the nation as a whole. (213) Banzhaf and Chupp conclude that "inter-jurisdictional spillovers appear to be a bigger problem in this application than heterogeneous benefits." (214)
Lastly, Perino and Talavera assess the determinants of state sulfur emissions rate standards prior to the Acid Rain Program in 1995. (215) In particular, the authors are interested in the relative importance of the internal costs and benefits of reduced emissions versus the external benefits on the state emissions standard. The state sulfur standard is measured by pounds of S[O.sub.2] per million British thermal units (MMBtu). The marginal cost of a more stringent standard is proxied by the transportation costs incurred to import low-sulfur coal from Wyoming. Finally, internal and external benefits to emissions reductions are measured by the state's own acidity of rainwater and the average acidity of rainwater in northeastern states (for states located in the Midwest). The results indicate that marginal abatement costs and internal and external benefits matter in a statistical sense. However, the effect of a state's own acidity is more than twenty times larger than the effect of the acidity of rainwater in the northeast. (216)
Rather than focusing on pollution, several empirical studies test for the presence of spillover effects on enforcement of environmental regulations. Gray and Shadbegian, in the study of 409 paper and pulp establishments in the United States discussed earlier in this Part, (217) also examine determinants of the number of air and water pollution inspections and enforcement actions. (218) The results are generally weaker than their reported results pertaining to emissions. (219) Nonetheless, the authors find some evidence that air pollution enforcement actions are higher against establishments near the Canadian border (consistent with the lower S[O.sub.2] emissions found in the analysis), but fewer water pollution inspections take place against such plants. (220)
In a subsequent study, Gray and Shadbegian use emissions data from 1997 on 521 U.S. manufacturing plants located within fifty miles of the center of three cities located near state borders (St. Louis, Cincinnati, and Charlotte). (221) The primary purpose of the study is to examine the effect of prior environmental enforcement--against either oneself (so-called specific deterrence) or against other plants within ten miles (so-called general deterrence)--on subsequent plant-level emissions. However, a very interesting finding emerges: inspections of other plants within ten miles reduce a plant's own emissions of air toxics (obtained from the TRI) as long as those other plants are located within the same state. Inspections of other nearby plants, which are located in another jurisdiction, fail to produce any general deterrent effect. (222) In this case, the lack of a spillover across jurisdictional boundaries reduces welfare. If enforcement were instead the responsibility of a higher level of government, each inspection would result in a greater reduction in emissions through general deterrence.
More recently, Professors Konisky and Woods assessed the impact of proximity to jurisdictional borders on plant-level enforcement actions. (223) Data on state-led enforcement actions--inspections, total punitive actions, and formal punitive actions--under the CAA between 1990 and 2000 are aggregated to the county-level and used to determine if counties on state or international borders are subject to less enforcement. The results indicate approximate twenty-five percent and fifty percent reductions in the count of inspections in counties bordering Canada and Mexico, respectively, but no meaningful effect in counties bordering other states. (224)
In a follow-up study, Konisky and Woods utilize data from the EPA's Integrated Database for Enforcement Analysis to assess the determinants of enforcement actions--compliance monitoring and punitive actions--against roughly 6400 facilities regulated under the Clean Water Act from 1995 to 2005. (225) The importance of spillovers is assessed by examining whether a facility that discharges its effluent into an interstate river or a multi-state watershed is the subject of less enforcement action. The authors also assess, among other things, the impact of distance to the nearest downstream state from the point where a facility's effluent likely enters a river. Konisky and Woods fail to find any evidence consistent with fewer enforcement actions being taken against firms more likely to be responsible for transboundary pollution. (226) Thus, the empirical studies to date assessing the spatial and temporal variation in enforcement behavior indicate less free riding than those assessing pollution directly. Understanding the source(s) of this difference is necessary.
The final set of papers examining the empirical relevance of resource externalities assess the impact of decentralized decision making on the exploitation of shared resources. Professor McWhinnie analyzes data on the global exploitation status of various fish stock according to the Food and Agriculture Organization (FAO) in 1994 and 2002. (227) Specifically, the FAO designates each fish stock in fifteen regions as underexploited, moderately exploited, fully exploited, overexploited, depleted, or recovering. The author then examines whether the number of countries that report catching a given fish in a given region and year help predict the exploitation status of the fish stock. McWhinnie finds that exploitation is increasing in the number of countries accessing the fish stock. (228) For example, if the fish stock is shared by two countries rather than one, "it is 9% more likely to be overfished and 19% more likely to be depleted." (229) If the fish stock is shared by five countries rather than one, "it is 36% more likely to be overfished and 82% more likely to be depleted." (230)
The analysis by Burgess, Hansen, Olken, Potapov, and Sieber is similar to that by Hatfield and Kosec in that it explores the impact of dividing a given geographic area into a larger number of jurisdictions on the environment. (231) However, Burgess, Hansen, Olken, Potapov, and Sieber examine the impact of the number of administrative jurisdictions in a given Indonesian province on the rate of deforestation between 2001 and 2008. (232) The results indicate nearly a four percent increase in the annual rate of deforestation if an additional district is formed within a province. (233) The authors conclude that their analysis provides a "counterexample to those who argue that decentralization of control over natural resources in weakly governed tropical environments should enhance their conservation." (234)
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Introduction through II. Environmental Federalism in Practice C. Interjurisdictional Externalities 1. Resource Externalities, p. 1669-1713|
|Author:||Millimet, Daniel L.|
|Publication:||Case Western Reserve Law Review|
|Date:||Jun 22, 2014|
|Previous Article:||Federalism and the rise of renewable energy: preserving state and local voices in the green energy revolution.|
|Next Article:||Environmental federalism: a survey of the empirical literature.|