UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Formation and political consequence of environmental enforcement policy : evidence from the US industries Sun, Meng 2010

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2010_fall_sun_meng.pdf [ 444.02kB ]
Metadata
JSON: 24-1.0071414.json
JSON-LD: 24-1.0071414-ld.json
RDF/XML (Pretty): 24-1.0071414-rdf.xml
RDF/JSON: 24-1.0071414-rdf.json
Turtle: 24-1.0071414-turtle.txt
N-Triples: 24-1.0071414-rdf-ntriples.txt
Original Record: 24-1.0071414-source.json
Full Text
24-1.0071414-fulltext.txt
Citation
24-1.0071414.ris

Full Text

Formation and Political Consequence of Environmental Enforcement Policy Evidence from the US Industries by Meng Sun  B.A., The Jilin University, 2000 M.Phil., The Hong Kong University of Science and Technology, 2001  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in The Faculty of Graduate Studies (Economics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) October 2010 c Meng Sun 2010  Abstract This thesis investigates the determinants and political consequences of environmental enforcement policy. Three related issues are addressed using the U.S. manufacturing data at the industry level. In Chapter Two, I estimate the elasticity of the violation rate with respect to the inspection rate for the manufacturing industries. This elasticity reflects the power of monitoring to alter an industry’s violation status. I conduct an estimation of the relationship between the violation rate and the inspection rate at the industry level, but the specification is based on an individual firm’s dichotomous choice in a logit model. The inspection rate is instrumented with the average of the inspection rates across the other industries that belong to the same sector to deal with the endogeneity problem. I find a substantial variation in the elasticity of the violation rate across industries. Chapter Three addresses how the inspection rate is determined by the elasticity of the violation rate when the enforcement agency is constrained with a hard monitoring budget. Given the limited monitoring resources, an enforcement agency targets industries where inspections are more likely to be effective in reducing violations to reduce inefficiency, all else equal. Empirical results confirm the positive effect from the absolute elasticity on the inspection rate. But the magnitude of this effect is conditional on the pollutants’ damage level to the environment. In Chapter Four, I focus on the political consequence of the enforcement policy. I employ the campaign contribution presented to the Congress exclusively for environmental issues (EPC) to measure the environmental political activity. The variation of EPC across industries is explained by the enforcement policy stringency. Although the EPC presented to the Congress is ii  Abstract targeted directly on environmental regulation rather than the enforcement policy, a stricter enforcement policy increases the marginal benefit from contributing the congress for relaxing the regulation. Empirically, I use the elasticity of the violation rate to isolate the effect of the enforcement policy on EPC. An industry with a larger elasticity of the violation rate is more likely to face a higher inspection rate, it therefore is more likely to engage in the political activity.  iii  Table of Contents Abstract  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ii  Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iv  List of Tables  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  v  List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vi  Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Dedication  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii  1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  1  2 Estimating the Elasticity of the Violation Rate at the Industry Level  . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  5  2.1  Introduction  . . . . . . . . . . . . . . . . . . . . . . . . . . .  5  2.2  Econometric model of the industry-level violation rate . . . .  8  2.2.1  Individual firm’s welfare  8  2.2.2  Industry-level violation rate function  2.3  2.4  . . . . . . . . . . . . . . . . . . . . . . . . .  9  . . . . . . . . . . . . . . . . . . . . .  11  2.3.1  Endogeneity of enforcement policy . . . . . . . . . . .  11  2.3.2  Endogeneity of emission intensity  . . . . . . . . . . .  12  2.3.3  Elasticity of violation . . . . . . . . . . . . . . . . . .  13  Empirical specification  Data sources and variables 2.4.1  . . . . . . . . . . . . . . . . . . .  14  Measuring compliance behavior and enforcement policy  . . . . . . . . . . . . . . . . . . . . . . . . . . . .  14  iv  Table of Contents 2.4.2  Other variables  . . . . . . . . . . . . . . . . . . . . .  15  2.5  Empirical results . . . . . . . . . . . . . . . . . . . . . . . . .  20  2.6  Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .  24  3 The Formation of Environmental Enforcement Policy: Evidence from the U.S. Industry-level Data . . . . . . . . . . .  25  3.1  Introduction  25  3.2  A simple theoretical model - enforcement with heterogeneous  . . . . . . . . . . . . . . . . . . . . . . . . . . .  industries . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  30  Empirical specification and data . . . . . . . . . . . . . . . .  33  3.3.1  The inspection rate equation . . . . . . . . . . . . . .  33  3.3.2  Endogeneity of the emission intensity . . . . . . . . .  35  3.3.3  Data description . . . . . . . . . . . . . . . . . . . . .  36  3.4  Empirical results . . . . . . . . . . . . . . . . . . . . . . . . .  42  3.5  Robustness and sensitivity analysis  . . . . . . . . . . . . . .  44  3.6  Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .  46  4 Determinants of Environmental Political Contribution . .  48  3.3  4.1  Introduction  . . . . . . . . . . . . . . . . . . . . . . . . . . .  48  4.2  Theoretical perspective  . . . . . . . . . . . . . . . . . . . . .  55  4.3  Descriptive analysis of environmental political contribution .  58  4.3.1  Measuring environmental political activity  . . . . . .  58  4.3.2  Trend analysis . . . . . . . . . . . . . . . . . . . . . .  60  4.4  Empirical specification and data sources . . . . . . . . . . . .  62  4.5  Empirical results . . . . . . . . . . . . . . . . . . . . . . . . .  71  4.6  Robustness checks . . . . . . . . . . . . . . . . . . . . . . . .  73  4.7  Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . .  75  Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  78  Appendices A Appendix to Chapter 2 . . . . . . . . . . . . . . . . . . . . . . A.1 The BLP model  . . . . . . . . . . . . . . . . . . . . . . . . .  87 87 v  Table of Contents A.2 Econometric model of the industry-level violation rate . . . .  88  A.3 Data of the inspection rate and the violation rate  . . . . . .  89  . . . . . . . . . . . .  91  B Appendix to Chapter 3 . . . . . . . . . . . . . . . . . . . . . .  92  A.4 First stage results of 2SLS in Table 2.3  B.1 Proof of equation 3.4  . . . . . . . . . . . . . . . . . . . . . .  92  B.2 Risk-screening environmental indicators . . . . . . . . . . . .  92  C Appendix to Chapter 4 . . . . . . . . . . . . . . . . . . . . . .  94  C.1 Background  . . . . . . . . . . . . . . . . . . . . . . . . . . .  C.2 Data of environmental political contribution  . . . . . . . . .  94 97  C.2.1 Members of key House/Senate environmental committee/subcommittee . . . . . . . . . . . . . . . . . . . .  97  C.2.2 Campaign contribution . . . . . . . . . . . . . . . . .  98  C.2.3 SIC code . . . . . . . . . . . . . . . . . . . . . . . . .  98  C.2.4 Data summary . . . . . . . . . . . . . . . . . . . . . .  99  vi  List of Tables 2.1  Variable Definition and Data Sources I . . . . . . . . . . . . .  18  2.2  Summary Statistics I . . . . . . . . . . . . . . . . . . . . . . .  19  2.3  Second-Stage Results of Violation Rates Regressions (2003’2005’) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  2.4  21  Industries Ranked by the Estimated Elasticities of the Violation Rate (111 Industries) . . . . . . . . . . . . . . . . . . . .  22  3.1  Variable Definition and Data Sources II . . . . . . . . . . . .  40  3.2  Summary Statistics II . . . . . . . . . . . . . . . . . . . . . .  41  3.3  Inspection Rates Regressions . . . . . . . . . . . . . . . . . .  45  3.4  Robustness of Inspection Rates Regressions . . . . . . . . . .  47  4.1  Environmental Political Contribution to Congressional Candidates (124 Industries in Manufacturing Group) . . . . . . .  4.2  61  Industries in Manufacturing Group Ranked by Environmental Political Contribution Amounts, 2005-2006 . . . . . . . . . .  63  4.3  Variable Definition and Data Sources III . . . . . . . . . . . .  69  4.4  Summary Statistics III . . . . . . . . . . . . . . . . . . . . . .  70  4.5  Results of Environmental Political Contribution Intensities Regressions: Election Cycles of 2000, 2002, 2004 and 2006 . .  4.6  74  Robustness of Environmental Political Contribution Intensities Regressions . . . . . . . . . . . . . . . . . . . . . . . . . .  76  A.1 Comparision with the BLP Model . . . . . . . . . . . . . . .  88  vii  List of Tables B.1 Industries at the Four-Digit SIC Level Ranked by Average Values of the Damage Indicator over 2003’-2005’ (410 Industries) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  93  C.1 Selected Vote Description . . . . . . . . . . . . . . . . . . . .  95  viii  List of Figures 4.1  The Variation of the EPC Intensity . . . . . . . . . . . . . . .  49  ix  Acknowledgements I woulk like to show my deep gratitude to my thesis supervisor, Prof. Brian Copeland. His continuing guidance and support with great patience enabled me to have stumbled forward a little on the research path. This thesis would not have been possible without his help. Prof. Werner Antweiler, Prof. Matilde Bombardini and Prof. Summet Gulati provided valuable comments that led to substantive improvements. Their input has been appreciated. I am also grateful to Prof. John Hartwick for his comments at the 2006 Canadian Economics Association Annual Meeting, and all those who helped me to complete this thesis.  x  Dedication To my parents, Qichun and Siyi  xi  Chapter 1  Introduction The enforcement of environmental legislation is crucial to achieving desirable environmental outcomes. This thesis explores the determinants and consequences of the environmental enforcement policy. Specifically, I address the following three questions using the U.S. manufacturing data: (1) How do industries react to a change in environmental enforcement policy? (Chapter Two) (2) How is the environmental enforcement policy determined at the industry level? (Chapter Three) (3) How does the environmental enforcement policy affect an industry’s political activity that is designated for environmental issues? (Chapter Four) In Chapter Two, I estimate the elasticity of the violation rate with respect to the inspection rate for industries in the manufacturing group. This elasticity reflects the power of monitoring to alter an industry’s violation status and is regarded as an industry characteristic. I conduct an estimation of the relationship between the violation rate and the inspection rate at the industry level, but the specification is based on individual firm’s dichotomous (violation/compliance) choice in a logit model. The potential endogeneity problem of the inspection rate is dealt with by using an instrument, which is the average of the inspection rates across the other industries that belong to the same sector. I find a substantial variation in the elasticity of the violation rate across industries. This estimated elasticity is employed to investigate its role in determining the environmental enforcement policy in Chapter Three. The elasticity could be fundamental to the making of environmental enforcement policy when an enforcement agency has limited monitoring resources. In the United States, the Environmental Protection Agency (EPA) explicitly faces a hard budget constraint in monitoring firms to make sure they abide by environmental 1  Chapter 1. Introduction regulatory policies such as the standard and the emission limit. Each year, the EPA develops a proposed budget defining the funding for the goal of compliance and related priorities. Then, the Congress passes bills to enact amended budgets into law.1 For example, the FY (fiscal year) 2010 Budget includes approximately $600 million for EPA’s Enforcement and Compliance Assurance program, and a $32 million increase over the FY 2009 Enacted level. In particular, for the Agency’s Compliance Monitoring program2 , in FY 2010, its proposed budget is $101.1 million. FY 2010 Compliance Monitoring activities will be both environmental media- and sector-based.3 Given the limited monitoring resources, an enforcement agency targets industries where inspections are more likely to be effective in reducing violations to reduce inefficiency, all else equal. This implies the important role of the elasticity of the violation rate in determining the inspection stringency. I develop a simple model where the regulator faces a hard budget constraint for its inspection and enforcement activities. If the agency maximizes social welfare subject to this constraint, the model predicts that the inspection rate will be higher in those industries with a higher elasticity of the violation rate with respect to the inspection rate. Empirical results confirm a positive effect from the absolute elasticity on the inspection rate. But the magnitude of the effect of the elasticity of the violation rate is conditional on the pollutants’ damage level to the environment. In Chapter Four, I focus on the political consequence of the environmental enforcement policy. I first construct an index of environmental political contribution (EPC) – the campaign contribution presented to the Congress exclusively for environmental issues, which serves as the direct evidence of political pressure on the environmental regulations – to measure the environmental political activity. Then, I explore the role of the enforcement policy stringency in explaining the variation of the EPC across industries. In a 1  See details in the U.S. EPA website: http://www.epa.gov/ocfo/budget/. This Compliance Monitoring program reviews and evaluates the activities of the regulated community to determine compliance with applicable laws, regulations, permit conditions and settlement agreements, and to determine whether conditions presenting imminent and substantial endangerment exist. (FY2010 EPA Budget in Brief, U.S. EPA.) 3 Source: FY2010 EPA Budget in Brief, U.S. EPA. 2  2  Chapter 1. Introduction multi-agent system, the environmental legislation/regulation is determined by the Congress, but the enforcement policy is set by the EPA4 . Although the EPC presented to the Congress is targeted directly on environmental legislation/regulation rather than the enforcement policy, the enforcement policy affects the marginal benefit from contributing to the congress for relaxing the legislation/regulation. For example, for an industry with a higher inspection rate (which represents a stringent enforcement policie), its identified pollution amount is larger. This implies that this industry pays more pollution taxes than the industry with a lower inspection rate at the same pollution tax rate. One unit relaxation of the pollution tax rate (which represents an environmental regulation) after the congress receives certain political contribution amount brings more benefits (a larger reduction in pollution tax payment) to this industry. Therefore, industries facing stricter enforcement policies are more likely to engage in the political activity all else equal. There are two features in the empirical examination. First, I employ EPC to characterize the environmental political activity. This is in contrast to existing works that use total political campaign contributions which may be the result of multiple purposes. Second, I use the elasticity of the violation rate to infer the effect of the enforcement policy on the environmental political contribution. This is because an industry with a larger elasticity of the violation rate is more likely to face a higher inspection rate, as shown in Chapter Three. The reason that I do not use the enforcement policy as a regressor directly is to distinguish the effect of the environmental enforcement policy from that of the environmental regulation. Although the enforcement policy and the environmental policy are determined by separate agencies, the two kinds of policies have many common determinants (e.g., the emission intensity). If the relationship between the enforcement policy and the environmental political contribution arises as a result of these 4  In the U.S., environmental enforcement takes place in the context of a system (the environmental federalism) consisting of local and state jurisdiction and the Federal enforcement agency. Because this work focuses on the variation across industries instead of that across regions, I partly control for the role of the state and local communities with the geographical concentration variable and focus on the role of EPA.  3  Chapter 1. Introduction common determinants, it is hard to isolate the enforcement policy’s effect. I find a positive relationship between the absolute elasticity of the violation rate and the EPC. The results are consistent with the predictions. In sum, this thesis joins the literature on the monitoring and enforcement of environmental policy with that on the political formation of economic policy. The rest of the thesis is organized as follows: Chapter Two empirically investigates environmental policy’s deterrence effect on firms’ compliance behavior and estimates the elasticities of the violation rate with the inspection rate at the industry level. Chapter Three develops a simple model to explain the variation of environmental enforcement policy when the enforcement agency is constrained with a hard budget. The predictions are empirically examined. Chapter Four provides a descriptive analysis of EPC, focuses on the amounts that industries contribute and uncovers the pattern of environmental political action across industries.  4  Chapter 2  Estimating the Elasticity of the Violation Rate at the Industry Level 2.1  Introduction  How do industries react to the changes in environmental enforcement policy? Evidence from industry studies suggests that there is variation across industries in the response of the violation rates to inspections. For example, Gray and and Shadbegian (2005) find that pulp mills are less sensitive to inspections than paper mills. This is not surprising because heterogeneities in abatement costs and other industry characteristics are likely to affect violation rates. However, to date there has not been a systematic cross-industry analysis of how violation rates respond to inspections. This chapter estimates the elasticity of an industry’s violation rate with respect to its inspection rate. This elasticity reflects the power of monitoring to alter an industry’s violation status. It is fundamental to the design of environmental enforcement policy.5 When an enforcement agency has limited monitoring resources (economically or politically), it can target industries where inspections are more likely to be effective in reducing violations to reduce inefficiency (all else equal). The next chapter (Chapter three) investigates how the elasticity of the violation rate plays a role in allocating the monitoring resources across industries for an enforcement agency. The 5 This estimation of the elasticities is also useful for the policy’s effectiveness evaluation and environmental consequence forecasting.  5  2.1. Introduction estimates from this chapter will play a key role in that analysis. The estimation of the elasticities for all industries is still absent in the literature. Existing works focuses on certain industries (e.g. pulp/paper6 /oil spills) and estimate the effectivess of the enforcement policy on firms/plants/facilities within the industries (Magat and Viscusi, 1990; Earnhart 2004a, 2004b; Shimshack and Ward, 2005, 2008; Laplante and Rilstone, 1996; Gray and Deily, 1996; and Telle, 2009). I conduct an estimation at the aggregate level (using the industry-level data) but the specification is based on individual firm’s dichotomous (violation/compliance) choice using a logit model. I thus avoid having to collect the data of firm characteristics for all industries, which is difficult. This chapter applies the BLP demand function estimation technique (Berry, Levinsohn, & Pakes, 1995). Each firm is a rational economic decisionmaker who violates its effluent standard when the net benefit of doing so exceeds that of compliance. Each firm’s welfare is determined by the regulatory policies imposed on violating/complying firms and individual firm characteristics. Within each industry, firms are subject to the same compliance/violation policies, such as the unit penalty for violation, and the minimum unit abatement investment for compliance. In this work, the enforcement policy I consider is the inspection rate. Individual firm characteristics are idiosyncratic ones across firms, such as the abatement efficiency, the production technology, and the managers’ preferences. I assume the individual firm characteristics follow an i.i.d. logit draw. Given the inspection rate imposed on violating firms, the probability of choosing violation is modelled as a realization from the logit distribution with the firm characteristics. I assume further that there are infinitely many firms in each industry. Then, this logit probability can be regarded as the industry violation rate, depending on the policy – the inspection rate. In order to facility the empirical estimation, I incorporate an outside alternative choice (the shutdown choice). It can be deived that the difference between the logarithm of the 6  Nadeau (1997) focuses on air pollution, and estimates the elasticity of the violation time with respect to the EPA monitoring activity for the pulp and paper industry which is between -0.4 and -0.47.  6  2.1. Introduction violation rate and that of the shutdown rate (the log odds ratio) is a linear function of the inspection rate at the industry level. This approach allows me to apply the linear instrumental variable (IV) approach to combat the endogeneity problem. This endogeneity problem arises when there may exist a reverse causality from the violation rate to the inspection rate. The instrument of the inspection rate I use is the average of the inspection rates across the other industries7 that belong to the same sector8 . The instrument is highly correlated with the inspection rate because they pick up common government-regulation shocks (e.g. a budget cut). Meanwhile, the instrument is uncorrelated with the unobserved industry-specific factors. In the literature at the disaggregate level, some ignore the endogeneity problem (e.g. Gawande and Wheeler, 19999 ). For others, there are two types of instruments10 used. But the employment of these two different instruments may lead to opposite results about the enforcement policy’s deterrent effect: “models using lagged regulatory activity continue to find a negative impact of enforcement on compliance, while models using predicted activity yield positive coefficients, which regulatory activity increasing compliance” (Gray and Shadbegian, 2005).11 Empirically, I construct the industry-level inspection rate (proxying the enforcement policy) and the industry-level violation rate (proxying the pollution noncompliance level) using data from “Enforcement & Compliance History Online (ECHO)” provided by the U.S. Environmental Protection 7  At the 4-digit SIC level. At the 2-digit SIC level. 9 Gawande and Wheeler (1999) addresses the need to improve efficiency in governmental organizations. It analyzes the Maritime Safety Program of the U.S. Coast Guard, which is responsible for monitoring the quality of vessels that sail in U.S. waters, and present measures of effectiveness for the Program. The estimation is based on a Poisson regression model. 10 When policy is regarded to be targeted on individual firm’s behavior, the effect of enforcement policy on firm’s compliance may encounter reverse causality problem. This can be directly alleviated with instrument variables such as the lagged policy (Magat and Viscusi, 1990; Earnhart 2004a, 2004b; Shimshack and Ward, 2005, 2008) or predicted policy probabilities (e.g., Laplante and Rilstone, 1996; Gray and Deily, 1996; and Telle, 2009) or both (Gray and Shadbegian, 2005). 11 The other reason that I do not use the lagged value of the inspection rate as the instrument is that the time series is very short (three years). 8  7  2.2. Econometric model of the industry-level violation rate Agency (EPA). I find that the inspection rates significantly reduce the industry-level violation rates after the endogeneity problem of inspection rate is overcome with the IV approach. The elasticities of the violation rate with respect to the inspection rate at the industry level are calculated, which vary greatly between inspection “elastic” industries (e.g., 1865 and 3313) and inspection “inelastic” industries (e.g., 3845 and 3661). The mean value of the elasticity is about -0.34, implying that a 10% increase in the inspection rate yields a 3% reduction in the violation rate. The rest of the chapter is organized as follows. In Section 2.2, I derive the industry-level violation rate function for estimation based on the BLP econometric model. In Section 2.3, I discuss how to deal with some econometric issues and present the empirical specification. Section 2.4 describes the data sources and the construction of the variables. Empirical results are presented in Sections 2.5. Section 2.6 concludes. Appendices A contains proofs, data descriptions and the first stage results of the two-stage least square (2SLS) regressions.  2.2  Econometric model of the industry-level violation rate  This section employs the BLP econometric model (Berry, Levinsohn, and Pakes, 1995)12 to derive the industry-level violation rate function.  2.2.1  Individual firm’s welfare  There are infinitely many heterogeneous firms in a polluting industry and each firm makes choices between violation and compliance. Each firm’s welfare is determined by choice (compliance vs. violation) characteristics and individual firm characteristics. The equation for each firm’s welfare denoted by Fljt can be specified as Fljt = Rjt + ξjt + γljt 12  See Appendix A.1 for details.  8  2.2. Econometric model of the industry-level violation rate where the subscript l represents firms, j ∈ (c=compliance, υ=violation) is the choice index, and t is the time index in years. Rjt and ξjt represent the characteristics of choice j that may affect welfare Fljt . That is, they do not vary across firms that make the same choice within an industry. Specifically, Rv is the enforcement policy imposed by the regulator on firms in violation, such as the unit violation penalty or inspection rate13 ; Rc is the regulation that a complying firm has to meet, such as the minimum unit abatement investment. ξv represents unobserved sources of welfare resulting from violation (e.g. pollution disutility to violating firms’ own employees); ξc represents unobserved characteristics for firms in compliance (e.g. a good reputation with their consumers and the community). Besides unobserved choice characteristics, ξjt may also capture violation/compliance shocks in an industry14 or measurement errors. The last term, γljt , contains the characteristics of individual firm l that affect welfare Fljt , such as the input amount, the abatement technology and the manager’s preference.  2.2.2  Industry-level violation rate function  When γljt is assumed to follow an i.i.d. logit draw, Wjt = Rjt + ξjt can be regarded as the mean welfare from choice j in year t in industry i. Violation is chosen whenever it leads to a higher welfare than compliance, that is, Flvt > Flct . Given any welfare in violation Flvt , the proportion of firms that choose to violate in industry i is the following according to γc ’s density function, −(Wvt −Wct +γlvt )  P ri,lv = Pr(Flvt > Flct ) = Pr(γlct < Wvt − Wct + γlvt ) = e−e  Since the welfare in violation Flvt is random, characterized by the i.i.d. 13  This increases violating firms’ cost because more violations would be identified as firms are inspected more often. 14 For example, the price of an industry-specific abatement equipment that prevents firms from violation becomes higher. Because this price may enter the welfare function of choosing compliance only, the welfare of choosing violation is not affected as much as the welfare of choosing compliance. And this is a shock common to all firms in this industry.  9  2.2. Econometric model of the industry-level violation rate logit distribution of γv , the violation rate of industry i is the following, ∞  −γlvt  Pr(Flvt > Flct )e−γlvt e−e  dγlvt =  −∞  1 e−(Wvt −Wct ) + 1  =  eWvt eWct + eWvt  Proof. See Appendix A.2. The desirable property of this logit probability (the violation rate, V R) is as follows. The numerator is the exponential function to the power of the welfare of this choice; the denominator is the sum of the exponential functions over all alternative choices. The implication is that the aggregate violation rate at industry level depends on the difference of the welfares, Wvt − Wct , instead of individual firm characteristics. Then, I add an outside alternative that is the shutdown of the facility and normalize the welfare of this alternative choice to zero. According to the property shown above, this implies that the violation rate is V Rit =  eWvt 1 + eWct + eWvt  and the closing firm share is Closeit =  e0 . 1+eWct +eWvt  Intuitively, the violation  rate is the probability that the choice of violation has a higher welfare than the choices of compliance and the shutdown of the facility. Taking logarithms of the two equations and canceling out the denominator 1 + eWct + eWvt that can not be controlled for, I have ln V Rit − ln Closeit = Rivt + ξivt  (2.1)  This allows us to consider a linear relationship between the logarithm of the violation rate, lnV Rit , and the enforcement policy on violating firms, Rivt , at the industry level.  10  2.3. Empirical specification  2.3  Empirical specification  The following empirical specification based on equation (2.1) is established,  ln V Rit − ln Closeit = β0 + β1 IRit + β2 ln EIit + β3 HHIit + β4 DIit +β5 ln EM Pit + β6 GCit + G + Y + ξit  (2.2a)  where i is the industry index, t is the time index, V Rit is the industry-level violation rate representing the degree of environmental non-compliance, and IRit is the industry-level inspection rate proxying the enforcement policy for violation (i.e., Rivt in equation (2.1)). The general industry conditions to be controlled for include the following: ln EI: the natural log of the level of emission intensity HHI: the Herfindahl-Hirschman Index measuring the concentration of firms in an industry (industry concentration) DI: “diverse industry” is a measure describing the number of different segments in which a firm operates EM P : the number of paid employment as a measure of industry size GC: “geographic concentration” describes the dispersion of firms according to their locations. In addition, I also add an election year dummy Y and a group dummy G (Group 2 or 3) to control for the election year effect15 and the group fixed effect respectively. βi s are coefficient scalars, and ξ is zero-mean error term.  2.3.1  Endogeneity of enforcement policy  When estimating the effect of the inspection rate on the violation rate using the specification above, I adopt the IV approach to overcome the potential endogeneity problem of the inspection rate. This endogeneity problem arises because the violation rate may have an influence on the choice of the 15 In List and Sturm (2006), it is argued that the incumbent finds it worthwhile to manipulate the environmental policy to attract additional votes to his platform. Evidence of this effect of electoral incentives on environmental policy is shown in that work.  11  2.3. Empirical specification inspection rate by the enforcement agency as shown below, IRit = ϕ0 + ϕ1 V Rit + ϕ2 controls2 + eit  (2.2b)  In addition, the endogeneity problem of inspection rate may arise if there are unobserved variables (e.g., local political factors) that influence the industry-level violation rate and are correlated with the inspection rate. In order to produce unbiased estimated coefficients, I choose valid instruments for IR. The instrument of the inspection rate is the average of the inspection rates across the other 4-digit industries that belong to the same 2-digit sector. For example, there are three industries “2111”, “2131” and “2141” at the 4-digit level in the 2-digit sector of “21” in the dataset. The instrument for the inspection rate of “2111” is the average of the inspection rates of “2131” and “2141”. The instrument is highly correlated with the inspection rate. This is because they pick up common governmentregulation shocks. For example, when an enforcement agency’s budget is reduced, the inspection rates across all industries in a sector are expected to be lower. Meanwhile, the instruments are uncorrelated with the unobserved industry-specific factors (e.g., the greater community pressure as a result of an pollution accident or an abatement technology improvement in this 4-digit industry) in the violation rate equation (2.2a) and those in the emission intensity equation (2.2c) below.  2.3.2  Endogeneity of emission intensity  For the emission intensity, I employ a reduced-form specification in equation (2.2c) to show its potential endogeneity problem. ln EIit = γ0 + γ1 IRit + γ2 ln wit + γ3 ln kit + γ4 ln M aterialit +γ5 ln Energyit + γ6 (controls) + ϕit  (2.2c)  where EI is the emission intensity, IR is the inspection rate, w is the labor share, k is the capital share, M aterial is the material share, Energy is the energy share, and (controls) include other industry characteristics. 12  2.3. Empirical specification Equation (2.2c) links EI to the enforcement policy (IR) and input shares of production. This specification is motivated by Hettige et al. (2000) [43]: “The marginal cost of abating pollution from industrial sources is a function of the scale of activity, pollutant concentration in process influent, the degree of abatement, and local input prices”. The endogeneity problem of EI could arise when the error term in equation (2.2a), ξ, is correlated with the error term in equation (2.2c), ϕ. On the one hand, V R could have an effect on EI through IR based on Equations (2.2b) and (2.2c). On the other hand, some unobserved violation characteristics or shocks may affect the emission intensity. This is because both the violation rate and the emission intensity are aggregated outcomes of firms’ problems at the industry level. As shown above, the full model is supposed to be a simultaneous equation system that consists of equations (2.2a), (2.2b), and (2.2c). However, in this chapter, I only estimate the first equation. Specifically, I adopt the 2SLS estimation: (i) inspection rates and emission intensities are regressed on the instruments of inspection rates, input shares and other exogenous variables to obtain their predicted values respectively; (ii) I estimate Equation (2.2a) with both the predicted emission intensity and the predicted inspection rates.  2.3.3  Elasticity of violation  Last, I discuss the calculations of the marginal effect of the inspection rate on the violation rate and the elasticities of the violation rate with respect to the inspection rate. The marginal effect depends on the estimated coefficient β1 and the violation rate16 : β1 IRi,vt +ξivt  e ∂ λIRi,ct +ξict ∂V Ri +eβ1 IRi,vt +ξivt = 1+e = V Ri (1 − V Ri ) β1 ∂IRi,vt ∂IRi,vt  (2.3)  Because the violation rate is not larger than 1, the calculated marginal effect is negative as long as the estimate of β1 is negative. A negative 16  Please see the appendix A.2 for proof.  13  2.4. Data sources and variables marginal effect implies a deterrent effect of the inspection rate. Further derivation provides the elasticity estimates as follows, ei =  2.4  ∂V Ri IRi,vt = IRit (1 − V Ri ) β1 ∂IRi,vt V Ri  (2.4)  Data sources and variables  Combining information from several sources, the dataset is described in the following and summarized in Table 2.1 and Table 2.2.  2.4.1  Measuring compliance behavior and enforcement policy  The major data source is the ECHO publicly provided by the U.S. EPA. This source provides the definition of formal enforcement action (FEA) with its data by facility within the last five years17 : in practice, for firms with very high levels of pollution violation, civil or administrative enforcement action that is taken against them in the U.S. is called the formal enforcement action.18 ECHO also gives us information on the number of inspections by facility within the same time period (See Appendix A.3 for details). The number of formal enforcement actions and the number of inspections at the facility level were aggregated into the SIC 4-digit industry level respectively.19 Meanwhile, the numbers of total facilities inspected for each 4-digit industry were collected. 17 ECHO has kept being updated. This paper uses the data last updated on Nov, 2008 at ECHO, so the inspection period starts from 2003. 18 “In general, a violation at a facility means the facility has been noted as out of compliance with an environmental requirement set forth by the Clean Air Act, Clean Water Act, or Resource Conservation and Recovery Act statutes and their respective regulations...A violation may indicate that the facility released excessive pollutants, that a hazardous waste handling requirement was not met, or that a facility failed to submit a required report...Not all violations receive formal enforcement action. Violations that are minor, short in duration, or quickly corrected by the facility may not warrant formal enforcement action.”–cited from the EPA website. 19 One thing to notice is that I do not use minor facilities which are not required to report because reported facilities may tend to be more environmentally compliant and could lead to bias. The other thing is that industries whose total number of inspected facilities is less than 5 are dropped because the sample is too small.  14  2.4. Data sources and variables I constructed the industry-level violation rate (VR) as the ratio of formal enforcement action numbers at the 4-digit industry level to the inspection numbers at the 4-digit level. It represents the probability of serious violation per inspection and is regarded as a measure of the noncompliance level for an industry. In this paper, I distinguish between serious and minor violations and use serious violations to capture the degree of violation. I use this because it is more likely that serious violations are chosen by facilities on purpose. That is, it is profit-maximizing for facilities to choose to be in non-compliance. On the contrary, minor violations are sometimes caused by accidents or other unexpected shocks. According to EPA, these minor violations are usually short in duration or quickly corrected by the facility, which avoids further formal enforcement actions.20 In order to measure the stringency of the enforcement policy, I divided the inspection number at the 4-digit level by the number of total facilities for each industry. The ratio is defined as the industry-level inspection rate (IR) describing the inspection probability that an industry is subject to.  2.4.2  Other variables  Here I describe the other variables used. The data source of the chemical released amounts at the 4-digit level is the U.S. EPA’s Risk-Screening Environmental Indicators (RSEI) (Version 2.1.5),21 which use annually updated Toxic Release Inventory (TRI)22 data. I constructed emission intensity (EI) by dividing the chemical released amounts at the 4-digit industry level by the value of shipments. The data of the values of shipments are from the 1992 Census of Manufactures report MC92-S-2, “Concentration Ratios in 20  Nevertheless, in so doing, one would under-estimate the effect of inspection on environmental outcome. This is because minor violations cost little for facilities to correct. Therefore, one can argue that minor violations are more sensitive or responsive to inspections. We leave this for future research. 21 See Appendix B.2 for details. 22 “The Toxic Release Inventory (TRI) is a publicly-available EPA database that contains information on toxic chemical releases and other waste management activities reported annually by certain covered industry groups and federal facilities.” – User’s Manual for RSEI.  15  2.4. Data sources and variables Manufacturing”. This MC92 is also the data source for HHI (Herfindahl Hirschmann Index) that characterizes industry concentration.23 The proportion of firms that choose to shutdown is proxied by the variable ‘close’. This is the ratio of the number of Establishments Deaths24 in 1998 to that of Initial Year Establishments in 1997.25 The data are recorded in the U.S. government census and are only available at the 3-digit SIC level.26 The data of inputs were obtained from the 1996 Annual Survey of Manufactures.27 The values of “production workers wages”, “cost of materials”, “new capital expenditures”, and “cost of purchased fuels and electric energy” are scaled by “value of industry shipments” to proxy the input shares w, k, Material and Energy respectively. The number of paid employees (EMP) is used to measure industry size and is from the 1997 Economic Census.28 Measures of Diverse Industry (DI) and Geographic Concentration (GC) were constructed following Grier, Munger and Roberts (1994). The number of segments is reported by each firm, up to a maximum of 10 in COMPUSTAT industrial segment file. The data were averaged by industry (mean=1.7, s.d.=0.65) and a DI is defined as “one with mean number of products more than two s.d. above the sample average, that is, all industries whose firms average more than three products line are coded 1.0 and the others 0.00”. Because the data vary little at 4-digit level, I compiled them into 3-digit level data. And the data cover the period of 1998-2006. GC is calculated as  i  Salesi Sales  2  for each industry where i denotes state. The data of sales  23 The 1997 data are available but are classified by NAICS (North American Industry Classification System) instead of SIC. 24 Only enterprises whose sizes (the employment of the controlling enterprise in the initial year) are over 20 are included. 25 Starting with the 2003-2004 tabulations, the industry classification is based on 2002 NAICS. Previous years data are based on 1997 NAICS (1999-2003) and SIC (1990-1998). 26 Please see http://www.census.gov/csd/susb/susbdyn.htm. 27 Please check http://www.census.gov/mcd/asm-as1.html. 28 The Economic Census profiles American business every 5 years and the 2002 data are available. However, the 2002 Economic Census is based on NAICS. Given the violation rate is based on SICs, the conversion of employment based on NAICS into that based on SICs may lead to some bias. The 1997 data are available in both NAICS and SICs. And since the employment number is used to control for industry size that is relatively stable over short periods, I choose the 1997 data.  16  2.4. Data sources and variables (DATA12) in each state were collected from COMPUSTAT as well. Combining non-missing observations from these sources yields an unbalanced panel of 421 observations over the period 2003-2006, which is the sample used in this study.  17  Variable  Table 2.1: Variable Definition and Data Sources I Description Sources of formal enforcement actions = The number The total number of inspections  ECHO, 2003’-2008’  Close Rate (close)  The number of Establishments deaths = The number of initial year establishments total number of inspections = The The total number of facilities chemical released amounts = TheThe value of shipments  U.S. gov’t census,the initial yr:97’,3-digit  Industry Size (EM P )  Herfindahl Hirschmann Index =1 for a large number of product lines =0 for a small number of product lines The number of paid employment  1992 Economic Census, 1992’ Industrial segment file in COMPUSTAT Grier et al.(1994), 1998’-2006’, 3-digit 1997 Economic Census, 1997’  Geographic Concentration (GC )  =  Inspection Rate (IR) Emission Intensity (EI ) Industry Concentration (HHI ) Industry Diversity (DI )  Labor Share (w) Capital Share (k ) Material Share (M aterial) Energy Share (Energy )  2 Salesi (i: state) i Sales workers wages = TheProduction value of industry shipments capital expenditures = TheNew value of industry shipments Cost of materials = The value of industry shipments of purchased fuels and electric energy = Cost The value of industry shipments  ECHO, 2003’-2008’ RSEI, 1992 Economic Census, 1996’-2005’  COMPUSTAT, Grier et al.(1994), 95’-06’ 1996 ASM, 1996’ 1996 ASM, 1996’ 1996 ASM, 1996’ 1996 ASM, 1996’  2.4. Data sources and variables  Violation Rate (V R)  18  I  Variable Minimum Indicator VR .013 Close .01 IR .03 ln EI -5.29 HHI 0.000056 DI 0 ln EM P 6.20 GC .09 w .01 k .001 M aterial .17 Energy .002 a :IR could be larger than 1 because each facility could be inspected multiple times.  Maximum 1 0.16 1.69a 19.16 0.297 1 13.18 1 .32 .34 .90 .21  2.4. Data sources and variables  Table 2.2: Summary Statistics Number of Mean Standard Observations Deviation 803 .12 .16 155 .05 .027 914 .39 .29 4115 12.68 2.57 447 0.066 0.06 3950 .04 .20 419 9.95 1.11 5302 .40 .24 456 .11 .05 450 .03 .03 455 .50 .13 455 .02 .03  19  2.5. Empirical results  2.5  Empirical results  The second-stage estimation results of equation (2.2a) are reported in Table 2.3.29 Column (1) of Table 2.3 contains the second-stage empirical results for the 2SLS estimation of the baseline specification with random effects. I use the predicted values of inspection rates (IR) and emission intensities (EI) to estimate specification (2.2a), controlling for the group heterogeneity and the changes in the election year by including Group2 and year2004. The included random effects account for the unobserved time-invariant industry heterogeneity. Inspection rate Our main interest is to find out how the inspection rate (IR) affects the violation rate. The estimated coefficient on the inspection rate is negative and significant at the 1 percent level.30 It corresponds to a mean marginal effect of -0.087 31 across 111 industries. The elasticities of the violation rate, evaluated at the mean values of the inspection rate and the mean values of the violation rate over time, vary from -0.089 to -1.03 as shown in Table 2.4. The mean value of the elasticity is about -0.34, implying that a 10 percent increase in the inspection rate yields a 3 percent reduction in the violation rate. The results suggest that a more stringent inspection policy improves the compliance, but the magnitudes vary greatly between inspection “elastic” industries (e.g., 2865 and 3313) and inspection “inelastic” industries (e.g., 3845 and 3661). Other variables 29  The first stage results of estimating equation (2.2a) are reported in the appendix A.4. The subsample during period 2003-2005 is used because data of the emission intensity for 2006 are not available. 31 The calculation is as follows based on the expression (2.3) in page 13: V R(1−V R)β = 0.107 × (1 − 0.107) × (−0.91) = −0.087. Here, the mean value of VR differs a bit from that in Table 2.2. This is because this mean value is across 111 industries whose elasticities can be calculated while the mean value in Table 2.2 is across the whole sample. 30  20  2.5. Empirical results  Table 2.3: Second-Stage Results of Violation Rates Regressions (2003’2005’) Independent Variable IR lnEI lnEMP  HHI DI GC Group2 Year2004  IV Random (1) −.91∗∗∗ (−3.52) .19∗∗∗ (3.27) −.1∗∗∗ (−1.68) 1.9 (1.58) .23 (.80) .59∗∗ (2.29) 0.34∗∗ (2.20) .02 (.24)  Second-Stage Dependent Variable: lnVR-lnClose IV Pooled IV fixed Random Pooled (2) (3) (4) (5) −.93∗∗ (−2.26) .15∗∗∗ (2.82) −.1∗∗ (−2.20) 1.48 (1.48) .23 (.74) .62∗∗∗ (2.97) .28∗ (1.92) −.02 (−0.21)  −1.04∗∗∗ (−3.95)  .26 (.94) .54 (.63)  .02 (0.25)  −.44∗∗ (−2.53) .02 (.72) −.14∗∗∗ (−2.66) 1.1 (1.03) .32 (1.35) .62∗∗ (2.59) .05 (.45) −.08 (−1.00)  Fixed (6) −.92∗∗∗ (−4.16) .04 (.45)  −.09 (−.53) .002 (.11) −.12∗∗∗ (−3.20) .42 (.51) .37 (1.40) .64∗∗∗ (3.47) −.01 (−.15) −.14 (−1.54)  .25 (.95) .48 (.57)  .006 (.08)  R2  0.026 – 0.001 0.09 0.10 Obs. 319 319 319 323 323 Note: both IR and EI are endogenous and instrumented. *** Significant at the 1% level, ** at the 5% level, and * at the 10% level (t statistics are in parentheses)  0.01 323  21  2.5. Empirical results  Table 2.4: Industries Ranked by the Estimated Elasticities of the Violation Rate (111 Industries) 5 Most Inspection “Elastic” Industries SIC Description Elasticity 2865 Cyclic Crudes and Intermediates 3313 Electrometallurgical Products 2621 Paper Mills 2812 Alkalies and Chlorine 2631 Paperboard Mills 5 Most Inspection “Inelastic” Industries SIC Description 3271 2711 3825 3845 3661  Concrete Block and Brick Newspapers Instruments for Measuring and Testing of Electricity and Electrical Signals Electromedical and Electrotherapeutic Apparatus Telephone and Telegraph Apparatus  -1.03 -.88 -.82 -.81 -.79 Elasticity -.10 -.10 -.098 -.094 -.089  22  2.5. Empirical results  In Table 2.3, the estimated coefficient on the emission intensity is positive and significant at the 1% level, after both the endogeneity problems of the inspection rate and the emission intensity are addressed. Coefficient estimates on the other control variables are also reported in Column (1) of Table 2.3. Their signs and magnitudes fail to provide much information about their marginal effects since the dependent variable is lnV R − lnclose. However, their statistical significance does show that whether they are important in explaining the variation of the violation rates across industries. In particular, the estimated coefficient on GC 32 is significantly different from zero at the p<0.05 confidence level. The group dummy, Group2, and the election year dummy, Year2004 are also controlled for. The group effect shows that the unobserved fixed differences between industries whose first digit SIC code is 2 and those whose first digit is 3 are captured. The election year effect is found to be insignificant. Column (2) of Table 2.3 presents the second stage results for the 2SLS estimation using the pooled data.33 Column (3) contains the second-stage results for the 2SLS estimation of the baseline specification with fixed effects. In columns (4), (5) and (6), they are the results from the random-effect, pooled OLS, and fixed-effect regressions respectively. But IR and EI are not instrumented. As I implement the fixed-effect estimation, some primary 32  According to the Pollution Haven Hypothesis in existing literature, firms prefer location where environmental regulation/enforcement is weak. Firms, especially polluting firms whose abatement costs are large, agglomerate in places where enforcement is weak. GC can be regarded as a proxy of local strictness of environmental enforcement policy and has an influence on the violation rate. 33 This estimate reports a negative R-squared using STATA. According to “Negative and missing R-squared for 2SLS/IV” by Scribney et. al. (1999), “for two-stage least squares, some of the regressors enter the model as instruments when the parameters are estimated. However, since our goal is to estimate the structural model, the actual values, not the instruments for the endogenous right-hand-side variables, are used to determine the MSS. The model’s residuals are computed over a set of regressors different from those used to fit the model. This means a constant-only model of the dependent variable is not nested within the two-stage least squares model, even though the two-stage model estimates an intercept, and the RSS is no longer constrained to be smaller than the TSS”. Thus, only standard errors matter for the estimates of these parameters produced by the two-stage model regardless of the value of R-squared.  23  2.6. Conclusions control variables are dropped. This is because the time series for these primary control variables are not available. The results are similar both in terms of magnitudes and significance levels across the first three alternate techniques of estimation in Table 2.3.  2.6  Conclusions  This chapter assesses the impact of the inspection rate on the environmental pollution noncompliance level when the endogeneity problem of the enforcement policy is considered. The elasticity of an industry’s violation rate with respect to its inspection rate is estimated. I started with an econometric BLP model to build this work at the industry level while keeping the microfoundations on firms’ optimal decisions. A linear relationship between the logarithsm of the industry-level violation rate and the inspection rate is derived, which is convenient for the application of the IV approach. In the empirical section, I estimated the structural empirical specification with the instruments for the inspection rate. The measures of the key variables at the industry level are firstly introduced and a new dataset is used. It complements the firm-level analysis in the literature. The 2SLS results indicate a significant deterrent effect of the inspection rate on the industry-level violation rate after the endogeneity of the inspection rate is considered. More importantly, I calculate the elasticities of the violation rate with respect to the inspection rate. And a substantial variation of the elasticity across industries is found. These estimates of elasticity would serve as a basis in the next two chapters for identifying the determinants of the environmental enforcement policy and the environmental political activity.  24  Chapter 3  The Formation of Environmental Enforcement Policy: Evidence from the U.S. Industry-level Data 3.1  Introduction  The enforcement of environmental legislation is crucial for achieving a desirable environmental outcome. How does the environment enforcement agency such as the environmental protection agency (EPA) enforce environmental regulatory policies? This paper intends to address this question in an environment where the environmental protection agency is explicitly constrained with a hard budget in monitoring polluting firms. Constrained by the hard budget, the enforcement agency has to target industries where inspections are more likely to be effective in achieving its goal, all else equal. This implies the important role of the elasticity of the violation rate with respect to the inspection rate in determining the inspection stringency. For example, when the agency’s goal is to maximize the social welfare, if there exists no budget constraint, for the first-best case, inspection rates should be just set to ensure that the expected penalty (inspection rate multiplied by the penalty) equals the marginal damages caused by the pollution violation.34 But the hard budget leads the inspection rates to deviate from the marginal damages. Then, similar to the Ramsey rule in 34  At certain maximum penalty.  25  3.1. Introduction the optimal tax literature, the distortion is determined by the elasticities of the violation rate with respect to inspection rates to minimize the efficiency loss. Specifically, all else equal, a higher equilibrium inspection rate is set in an industry with a large absolute elasticity of the violation rate. This is because, if one industry is more elastic than the other industry, the deadweight loss at the same inspection rate from inspecting the elastic industry is lower. This paper tries to empirically investigate whether the enforcement agency allocates the limited monitoring resources across industries based on the industry elasticity of the violation rate with respect to the inspection rate. EPA35 explicitly faces a hard budget constraint in monitoring firms36 to make sure they abide by environmental regulatory policies such as standard and limit. Each year, the EPA develops a proposed budget defining the funding for the goal of compliance and related priorities. Then, the Congress37 passes bills to enact amended budgets into law.38 For example, the FY (fiscal year) 2010 Budget includes approximately $600 million for EPA’s Enforcement and Compliance Assurance program, and a $32 million increase over the FY 2009 Enacted level. In particular, for the Agency’s Compliance Monitoring program, in FY 2010, its proposed budget is $101.1 million39 . As stated in the FY 2010, Compliance Monitoring activities will be both environmental media- and sector-based40 . And the allocation of the budget involves multiple steps. In designing the allocation rule, the en35  In the U.S., environmental enforcement takes place in the context of a system (the environmental federalism) consisting of local and state jurisdiction and the Federal enforcement agency. Because this work focuses on the variation across industries instead of that across regions, I partly control for the role of the state and local communities with the geographical concentration variable and focus on the role of EPA. 36 The monitoring process is to discover firms that are in violation of the law and to produce evidence. 37 In particular, Jones and Scotchmer (1990) states that the enforcement budget is set by the Congressional appropriations committees. 38 See details in the U.S. EPA website: http://www.epa.gov/ocfo/budget/. 39 This Compliance Monitoring program reviews and evaluates the activities of the regulated community to determine compliance with applicable laws, regulations, permit conditions and settlement agreements, and to determine whether conditions presenting imminent and substantial endangerment exist. (FY2010 EPA Budget in Brief, U.S. EPA.) 40 Source: FY2010 EPA Budget in Brief, U.S. EPA.  26  3.1. Introduction forcement agency sets some priorities at the industry level first41 . Then, given the resources allocated to certain industry, more specific amount of resources directed to each firm within this industry would be decided42 . In this work, I study the allocation rule at the industry level only. The elasticity of an industry’s violation rate with respect to the inspection rate reflects the power of monitoring to alter an industry’s violation status. I regard this elasticity as an important characteristic of an industry and estimated the elasticities in Chapter two. As shown by Gray and Shadbegian (2005) that pulp mills are less sensitive to inspections than paper mills, the estimation results of Chapter Two also show the great variation across industries in the elasticity. The role of the elasticity of the violation rate with respect to the inspection rate in determining the enforcement policy stringency has not been explored in the literature. Existing works empirically paid attention to two groups of factors43 (e.g., Laplante and Rilstone, 1996; Lear, 1998; Helland, 1998; Deily and Gray, 1991, 1996; Dion, Lanoie and Laplante, 1998) in explaining the pattern of inspections. One is the past violation history, which could lead to a higher probability of being inspected in subsequent periods. This is consistent with the dynamic targeting approach described by Harrington (1988) and others44 which ensures compliance at reduced monitoring costs. For the other group, the role of the factors in determining the 41  Or, when the priorities are set based on the environmental media, some industries that release these regulated pollutants intensively are subject to more stringent policy than others. That is, priorities on some industries are implicitly defined in this case. 42 “The priority system adopted by the Michigan Occupational Safety and Health Act (MIOSHA) for conducting scheduled, programmed inspections in private sector workplaces involves two major steps. In the first step, MIOSHA designated target industries. In the second step, MIOSHA generates a priority list of establishments to be inspected based on the targeted industries.” Source: “General Industry Inspection Priority System Programmed Scheduled Inspections” by General Industry Safety & Health Division, MOSHA, March 2010 43 For details, see Cohen (1999). 44 Cohen (1999) states that this prior history of violation environmental laws can also be explained by the economic theory of regulation (if the public demands it) or compliance maximization (since the agency is most likely to find violations at these firms) besides the social cost minimization (based on the dynamic models) empirically. I emphasize the explanation under social cost minimization in this work.  27  3.1. Introduction enforcement policy are explained from the political perspective. The factors include the per-capita level of pollution, the affluence of the community, the affected labor force’s percentage of the local population, the compliance cost45 , the firm size46 and others. Thus, this study provides an additional explanation -the elasticity of the violation rate- for the observed differences in the stringency of enforcement policy.4748 I exploit the variation in the inspection rate and the elasticities of the violation rate (constructed in Chapter Two) from the United States manufacturing industries to test the theoretical predictions. The inspection rate at the industry level is the percentage of inspection number in the number of inspected facility for each industry. That is, it is the inspection number that an inspected facility is subject to. This is used to proxy the enforcement policy stringency. Moreover, I control for an important industry characteristic – the pollution damage indicator. The damage indicator is defined as a measure of industry-specific pollutants’ potential damage to the environment and human health per unit of emission in this work. That is, it describes the toxicity of the chemical and the number of people affected. Pollutants with potential damage have been known to vary across industries for a while (Hettige, Mani and Wheeler, 2000; Olewiler and Dawson, 1998). For instance, Olewiler and Dawson (1998) argue: “they can highlight industries that produce compounds that are more toxic-intensive than others, even though some of these industries may not have very large emissions intensity 45 In Gray and Deily (1996), plants with a high cost of compliance are more likely to spend money fighting against a stricter enforcement action. 46 Lear (1998) finds a “deeper pockets” effect with higher fines for larger firms. 47 Although different theories (along with predictions about factors that affect the enforcement policy) arises based on various assumptions about the enforcement agency’s goal (such as the net political support maximization, the bureaucratic benefit maximization or the compliance maximization), the factors are not exclusive to each other in the empirical test. 48 Theoretical works in this line about the priorities for enforcement are based on how firms differ. For example, firms’ prior compliance history or expected compliance behavior (Harrington 1988, Russell 1990 and harford 1991); firms’ opportunities to evade and private benefits from polluting (Macho-Stadler and Perez-Castrillo 2006); firms’ compliance costs (Jones and Scotchmer 1990).  28  3.1. Introduction ratios.” It was not rigorously incorporated in a theoretical model and empirically tested until recently (see Michida and Nishikimi, 200749 ; Antweiler, 2003). Because one unit of violation by an industry with a large damage indicator leads to a high social cost, I predict that a larger inspection rate is set in this industry, all else equal. Data collection of the damage indicator is adopted from the EPA’s Risk-Screening Environmental Indicators (RSEI) Version 2.1.5.50 I find that both of (the reciprocal of) the elasticity of the violation rate and the damage indicator have significant effects on the inspection rate level in both of the IV and the OLS estimations. In particular, a strong complementary effect between the damage indicator and the elasticity of the violation rate on the inspection rate is found. The net effect of the absolute elasticity is shown to be positive and significant. This supports the prediction that the elasticity of the violation rate determines the inspection rate in a modified Ramsey pricing rule in the presence of a hard budget constraint. The net effect of the damage indicator is conditional on the elasticity of the violation rate. The intuition is as follows: although the violation of an industry with a larger damage indicator may lead to a larger social cost, each inspection of this damaging industry’s behavior may be more costly. When the elasticity of the violation rate is small, that is the inspection is not effective, an enforcement agency may choose to inspect this industry less. In contrast, when the elasticity of the violation rate is large enough, an industry with a larger damage indicator is subject to a larger inspection rate. The rest of the chapter is organized as follows. In Section 3.2, I develop a simple theoretical model to show how the damage indicator and the elasticity of the violation rate determine the environment enforcement policy in equilibrium. Section 3.3 presents the empirical specification and describes the data. In Section 3.4, I present the empirical results. Section 3.5 con49  Industry-specific pollutants are modeled by Michida and Nishikimi (2007)[54]: “in terms of industrial pollution, there exist a large variety of toxic substances, some of which are not common across different industries.” 50 see Appendix B.2 for details.  29  3.2. A simple theoretical model - enforcement with heterogeneous industries ducts the robustness checks. Conclusion is provided in Section 3.6. Proofs of propositions and some supplementary data descriptions are contained in Appendix B.  3.2  A simple theoretical model - enforcement with heterogeneous industries  Suppose firms in an industry are identical and polluting. A representative firm could be in compliance or in violation51 . The firm can control its violation probability by changing the environmental expenditure, e.g., the level of investment in pollution abatement equipment, the number of employees in the pollution management, and the employment of cleaner input in production. It decides its violation probability v ∈ (0, 1] to maximize its profit. The violation probability decreases as the firm devotes more environmental expenditure, but the environmental expenditure has to be infinitely large to keep the violation probability at zero. The net benefit from violation –B (v, k)– is increasing in violation probability, v. The functional form of B (v, k) is assumed to be v θ k 1−θ  52 .  The  benefit may come from saved abatement cost and related gains from more pollution (e.g. a lower output price as the cost becomes lower). This net benefit also depends on a sector-specific input k. The sector-specific input is limited and inelastically supplied, and the technology shows constant return in both factors but diminishing return in individual factors. This production 51 The extent of the violation is neglected to focus on the violation rate (as a result of dichotomous compliance/non-compliance determination) in this work. As Oljaca et al. (1998) indicate that the violation size has also been neglected in Viscusi and Zeckhauser (1979), Jones and Scotchmer (1990), Harrington (1988), Harford and Harrington (1991), Jost (1997), Malik (1993). In this line of literature, it is assumed that the agency chooses the maximum fine level within certain externally-imposed limit (e.g. the wealth of the offender, the social norms of fairness) to achieve the greatest deterrence. In practice, the U.S. EPA calculates its civil penalty with a ‘gravity’ component which is based on the harm from the offense. This component of the EPA penalty is based on qualitative descriptions of harm and is not directly related to any quantitative measure of harm (Cohen, 1999). 52 This power function form is necessary to derive a constant elasticity of the violation rate with respect to the inspection rate in the following.  30  3.2. A simple theoretical model - enforcement with heterogeneous industries structure leads to a quasi-rent π for the fixed factor. At the industry level, v is regarded as the industry-level violation rate because firms are identical in an industry. An agency that enforces a pollution standard monitors and imposes a sanction once a violation is detected. A sanction could be a fine or some other costly penalty and its level is assumed to be exogenous  53 .  The en-  forcement agency decides the inspection rate p ∈ [0, 1] to maximize the social welfare. This is a Stackleberg game: the enforcement agency announces the inspection rate and the firms decide their violation probabilities afterwards. Therefore, a firm’s choice of violation probability is influenced by the inspection rate announced by the enforcement agency. The optimization problem of a representative firm can be written as M ax π = B (v, k) − ptv v  (3.1)  where v is the violation probability, and B (v, k) is the benefit from violating the pollution standard. The detected violation rate is pv. The total fine is ptv which depends on both the detected violation rate and the penalty t54 . The first order condition is: Bv (v, k) − pt = 0  (3.2)  This shows that the marginal benefit of increasing the violation probability is equal to its marginal penalty. A firm’s violation probability is decreasing in the inspection rate. The enforcement agency has to incur costs in monitoring firms so as to discover and verify a violation. And the agency’s inspection expenditure is increasing with a larger violation probability for each firm. I assume that the inspection expenditure on a representative firm in industry i is epi vi , 53  In Garvie and Keeler (1994), it is pointed out that the size of financial penalties is dependent on institutional factors (e.g. judicial attitudes, higher level support from the executive branch, and public opinion) largely outside the control of both firms and regulators. 54 A relatively more general functional form, a power function of v, is also easy to be used to derive the same propositions and predictions.  31  3.2. A simple theoretical model - enforcement with heterogeneous industries where e is the monitoring expenditure coefficient and it is the same across industries.  55  The total inspection budget is assigned by the Congress and  is limited. The enforcement agency chooses an industry-specific inspection strategy to maximize the social welfare subject to the budget constraint and firms’ reaction functions in equation (3.2), pi tvi −  M ax Ω = pi  i  ρi vi + i  πi  (3.3)  i  epi vi ≤ E s.t.  i  Bv (v, k) − pt = 0 where ρj is the pollution damage indicator in industry i so that ρi vi represents the pollution disutility from this industrial pollution violation. Here, the pollution damage indicator is an important characteristic of industryspecific pollutant. It represents different levels of risks to the environment and human health, including information such as the toxicity or the proximity to consumers. This corresponds to the sector-specific input. pi tvi is the penalty collected from industry i and is redistributed lump-sum to all individuals. Thus, the social welfare is defined as the sum of the pollution disutility, the expected collected penalty and the total profits across all industries. In the budget constraint, epi vi is the inspection cost function. E is the total budget assigned by the congress. ∂vi pi 1 56 I define εi = as the elasticity of the violation rate with = ∂pi vi θ−1 respect to the inspection rate. In equilibrium, the industry i’s inspection 55 The inspection expenditure depends on both the inspection probability and the violation probability in this thesis. It can be expected that the inspection cost is higher for a serious and large amount violation because the workload is high. But the extent of the violation has been neglected to focus on the violation rate as a result of dichotomous compliance/noncompliance decision (The explanation is in footnote 47 in page 29.). I use the violation probability to proxy the expected violation size assuming that the mean violation amount is the same across industries. That is, if an industry has a larger violation probability, its expected violation size is larger. Each inspection may incur a higher cost and the total inspection expenditure is larger. In addition, there may exist additional enforcement actions required for firms that are found violating, such as control orders, phone calls, meetings etc., so the total inspection costs are higher when more firms are found in violation. 56 This is derived from Equation (3.2) when B (v, k)=v θ k1−θ  32  3.3. Empirical specification and data frequency the agency chooses satisfies: pi =  1 t − λe 1 +  1 εi  ρi  (3.4)  Proof. See Appendix B.1 where the Lagrangian multiplier λ is positive because an increase in the budge E implies a gain in social welfare. It reflects the opportunity cost of a dollar in terms of gaining the social welfare. This gives the following relationships for the inspection rate: Proposition 1 In equilibrium, all else equal, the enforcement agency inspects the following industries more frequently: (i) the industries with larger absolute values of the elasticity of the violation rate with respect to the inspection rate |ε|; (ii) the industries whose damage indictor ρ is larger. If one industry is more elastic than the other industry, the deadweight loss at the same inspection rate from inspecting the elastic industry is lower. Then the enforcement agency may devote more resources to inspecting the industry whose absolute elasticity of violation rate is large. And a high equilibrium inspection rate is set in this industry. Equation (3.4) provides a suitable platform for empirical analysis and identification with data from the USA.  3.3 3.3.1  Empirical specification and data The inspection rate equation  Equation (3.4) suggests the central roles of two variables (the damage indictor ρ and the reciprocal of the elasticity of the violation rate  1 ε)  in de-  termining the inspection rate that represents the enforcement policy level. I estimate three specifications below from a simplified linear approximation  33  3.3. Empirical specification and data to a general nonlinear functional form.57 Specification one I incorporate way as follows,  1 and ln ρ as regressors in Specification One in a linear ε  1 + β2 ln ρi + β3 ln EIi + β4 ln EM Pi + β5 HHIi + ε i β6 DIi + β7 GCi + β8 Groupi + i (3.5a)  ln IRi = β0 + β1  where IR is the industry-level inspection rate, ρ is the pollution damage indicator,  1 ε  is the reciprocal of the elasticity of the violation rate, Group is  a group dummy (the first digit of SIC code is 2 or 3),  is a zero-mean error  term, i is the industry index. The general industry conditions to be controlled for include the following: EI: the level of emission intensity. HHI: the Herfindahl-Hirschman Index measuring the concentration of firms in an industry (the industry concentration). DI: “diverse industry” is a measure describing the number of different segments in which a firm operates. EM P : the number of paid employment as a measure of industry size. 57 The elasticity of the violation rate is regarded to be exogenous in the estimations. On the one hand, it is regarded as an industry characteristics. As shown by the theoret∂vi pi 1 ical work, εi = = in a simplified framework. At the micro-level(the facility ∂pi vi θ−1 level), the elasticity of violation may be endogenous, while we aggregate the data to the industry level, which lessens the degree of endogeneity of the violation elasticity. Moreover, in the empirical papers that test the predictions of Grossman and Helpman (1994), elasticity (although different from that in my paper) is treated as exogenous (as an industry characteristic) to explain the differences in tariff across industries (see Gawande and Bandyopadhyay, 2000). Nevertheless, the estimated coefficients on the elasticity of violation may be biased to some extent. However, since the estimated coefficients are generally significant at the 1% level, the potential bias might not affect the main predictions much.  34  3.3. Empirical specification and data GC: “geographic concentration” describes the dispersion of firms according to their locations. Specification two I amend Specification One by adding an interaction term between ln ρ 1 and as Specification Two. This is to catch the possibility that the effect e of the elasticity of violation may be conditional on an industry’s damaging level (represented by the value of the damage indicator ρ). I therefore use the following as Specification Two, 1 1 + β2 ln ρi + β1,2 ln ρi × + β3 ln EIi + β4 ln EM Pi ε i εi +β5 HHIi + β6 DIi + β7 GCi + β8 Groupi + i (3.5b)  ln IRi = β0 + β1  Specification three I consider a general nonlinear functional form by adding higher order 1 terms for ln ρ, and their interaction term as follows, e ln IRi = β0 + β1 ln ρi + β11 (ln ρi )2 + β2  1 ε  + β22 i  1 εi  2  +  1 2 1 β1,2 ln ρi × + β11,22 ln ρi × + β3 ln EIi + εi εi β4 ln EM Pi + β5 HHIi + β6 DIi + β7 GCi +β8 Groupi +  3.3.2  i  (3.5c)  Endogeneity of the emission intensity  When estimating the three specifications above, I adopt the IV approach to overcome the potential endogeneity problem of the emission intensity. This endogeneity problem arises because the inspection rate has an influence on  35  3.3. Empirical specification and data the emission intensity as shown below in a reduced-form specification58 .  ln EIi = γ0 + γ1 IRi + γ2 ln wi + γ3 ln ki + γ4 ln M ateriali  (3.6)  +γ5 ln Energyi + γ6 (controls) + ϕit where EI is the emission intensity, w is the labor share, k is the capital share, M aterial is the material share, Energy is the energy share, (controls) include industry characteristics, and Groupi is the group dummy. In estimating equations (3.5a), (3.5b) and (3.5c), ln EI is instrumented with the input shares (ln w, ln k, ln material, ln energy).  3.3.3  Data description  This section describes the data sources and how they are utilized to construct measures of the interested variables. Table 3.1 provides definitions and sources of the variables, and Table 3.2 summarizes the descriptive statistics of the variables used in the empirical analysis.59 Regulatory variable The dependent variable that measures the environmental enforcement policy is the inspection rate (IR). The data source of IR is the “Enforcement & Compliance History Online (ECHO)” publicly provided by the U.S. EPA. This source provides information on the number and dates of inspections by facility within last five years60 (See Appendix A.3 for details). The number of inspections at the facility level was aggregated into the SIC 4-digit industry 58 This is similar to the Equation (2.2c) in Chapter Two. But in estimation of Equation (2.2c), I use a panel data and time varies from 2003’ to 2005’. In estimating Equation (3.6), I conduct a cross-industry analysis only. 59 Some variables are used for estimation both in this chapter and Chapter Two. Thus, the description of the measurement and data sources for these variables are the same as those in Chapter Two. 60 ECHO has been keeping updated. This paper uses the data last updated on Nov, 2008 at ECHO. So the inspection period starts from 2003.  36  3.3. Empirical specification and data level respectively.61 Meanwhile, the total number of facilities for each 4-digit industry was collected. I divided the inspection number at the 4-digit level by the number of facilities for each industry. The ratio is defined as the industry-level inspection rate IR describing the inspection number that an industry is subject to. I also use the abatement cost per unit emission (AC emission) as a proxy of the enforcement policy in the robustness checks in Section 3.5. The data source of AC is the file of “Pollution abatement costs and expenditures: 1994” from U.S. Bureau of the Census. AC emission is defined as the ratio of the sum of operating costs for all media (PAOC) and capital expenditure for pollution abatement (PACE) to the “TRI pounds”.62 “TRI pounds” refer to the pounds released or transferred by industries, and are reported in the U.S. EPA’s Risk-Screening Environmental Indicators (RSEI Version 2.1.5). Damage indicator and elasticity of the violation rate The damage indicator ρ and the elasticity of the violation rate ε are two important independent variables. Data source of the damage indicator is also the U.S. EPA’s Risk-Screening Environmental Indicators (RSEI Version 2.1.5)63 . RSEI contains “risk-related results”, which imposes a risk score on industries at different SIC levels. The toxicity, surrogate dose, and population components are multiplied to calculate this risk score. It is used to assess the potential impact of industrial releases. The data are available for 1996’-2005’. The risk score (“risk-related results”) is defined as the damage indicator in this chapter. This dataset has been seldom explored to my knowledge. The damage indicator is similar to the measure of environmental 61  I do not use minor facilities which are not required to report because reported facilities may tend to be more environmentally compliant and could lead to bias. And industries whose total number of inspected facilities is less than 5 are dropped because the sample is too small. 62 The data in some industries are withheld to avoid disclosure, and no industries are shown where PAOC or PACE is less than $1.0 million. 63 See Appendix B.2 for details. Data of the damage multiplier is available with the SIC code instead of NAICs.  37  3.3. Empirical specification and data exposure in Antweiler (2003) to some extent64 . Elasticity of the violation rate with respect to the inspection rate is the measure of responsiveness in the industry-level pollution violation rate as a result of change in the enforcement policy. It was calculated in Chapter Two. Other variables I describe the other variables used in the analysis as follows. The number of paid employees (EMP) is used to measure industry size and is from the 1997 Economic Census.65 The 1992 Census of Manufactures report MC92S-2, “Concentration Ratios in Manufacturing”, is the data source for HHI (Herfindahl Hirschmann Index) that characterizes the industry concentration66 . I constructed the emission intensity (EI) by dividing the chemical released amounts at the 4-digit industry level in 1996’67 by the value of shipments. The data source of the chemical released amounts at the 4-digit level is the U.S. EPA’s Risk-Screening Environmental Indicators (RSEI) (Version 2.1.5)68 that uses annually updated Toxic Release Inventory69 data. The 64  It combines a firm’s composition and toxicity of emissions and its location that determines the size of the population at risk with the firm’s size and intensity of pollution. This measure of environmental exposure is utilized as one of firm characteristics to investigate its effect on firm’s abatement effort. This effect is used to back out the effect of green regulatory threat which is not directly measurable. Because environmental exposure triggers regulation, it has been shown that the effectiveness of threat depends on environmental exposure of a firm. It also depends on abatement ladder rung of a firm and government’s incentive scheme as argued in this paper. 65 The Economic Census profiles American business every 5 years and the 2002 data are available. However, the 2002 Economic Census is based on North American Industry Classification System (NAICS). Given the violation rate is based on SICs, the conversion of employment based on NAICS into that based on SICs may lead to some bias. The 1997 data are available in both NAICS and SICs. And since the employment number is used to control for industry size which is relatively stable over short periods, I choose the 1997 data. 66 The 1997 data are available but are classified by NAICS instead of SIC. 67 I use the data in 1996’ because the data of inputs are in 1996’. 68 See Appendix B.2 for details. 69 “The toxic Release Inventory (TRI) is a publicly-available EPA database That contains information on toxic chemical releases and other waste management activities reported annually by certain covered industry groups and federal facilities.” – User’s Manual for RSEI  38  3.3. Empirical specification and data data of the values of shipments are from the 1992 Census of Manufactures report MC92-S-2, “Concentration Ratios in Manufacturing”. The data of inputs were obtained from the 1996 Annual Survey of Manufactures.70 The values of “production workers wages”, “cost of materials”, “new capital expenditures”, and “cost of purchased fuels and electric energy” are scaled by “value of industry shipments” to proxy the input shares w, k, Material and Energy respectively. The value of industry shipments is from the 1996’ ASM. Measures of Diverse industry (DI) and Geographic Concentration (GC) were constructed following Grier, Munger and Roberts (1994). The number of segments is reported by each firm, up to a maximum of 10 in COMPUSTAT industrial segment file. The data were averaged by industry (mean=1.7, s.d.=0.65) and a diverse industry is defined as “one with mean number of products more than two s.d. above the sample average, that is, all industries whose firms average more than three products line are coded 1.0 and the others 0.00”. Because the data vary little at 4-digit level, I compiled them into 3-digit level data. And the data cover the period of 1998-2006. Geographic concentration is calculated as  i  Salesi Sales  2  for each industry where  i denotes state. The data of sales (DATA12) for each state were collected from COMPUSTAT as well. Combining non-missing observations from these sources yields an unbalanced panel of 331 observations over the period 2003’-2005’. I take the mean values over the three years as the sample used in this chapter in estimation.  70  http://www.census.gov/mcd/asm-as1.html.  39  Description  Sources  Inspection Rate (IR) Industry Concentration (HHI )  total number of inspections = The The total number of facilities Herfindahl Hirschmann Index  ECHO, 2003’-2008’ 1992 Economic Census, 1992’ Industrial segment file in COMPUSTAT, 3-digit Grier et al.(1994), 98’-06’ 1997 Economic Census, 97’ COMPUSTAT,95’-06’ Grier et al.(1994) RSEI, 1996’-2005’  Industry Diversity (DI )  =1 for a large number of product lines =0 for a small number of product lines  Industry Size (EM P )  The number of paid employment  Geographic Concentration (GC )  =  Damage Indicator (ρ)  Risk-related results A measure of responsiveness in the violation rate as a result of change in inspection rate  Elasticity of Violation Rate (ε)  i  Salesi Sales  2  (i: state)  Variables that appear in the robustness checks only AC costs+Capital expenditure Abatement Cost (ln = Abatement operating ) TRI pounds Emission  Chapter Two, no time series  “PACE1994” and 1996 ASM, 1994’  3.3. Empirical specification and data  Table 3.1: Variable Definition and Data Sources II Variable  40  Variable Indicator IR 914 .39 .29 .03 ln EI 4115 12.68 2.57 -5.29 HHI 447 0.066 0.06 0.000056 DI 3950 .04 .20 0 ln EM P 419 9.95 1.11 6.20 GC 5302 .40 .24 .09 ln ρ 4134 7.91 3.50 -10 1/ε 111 -.4.0 2.19 -11.3 Variable that appears in the robustness checks only AC ln Emission 109 -11.76 2.11 -16.18 a: IR could be larger than 1 because each facility could be inspected multiple times.  Maximum 1.69a 19.16 0.297 1 13.18 1 16 -0.9 -2.64  3.3. Empirical specification and data  Table 3.2: Summary Statistics II Number of Mean Standard Minimum Observations Deviation  41  3.4. Empirical results  3.4  Empirical results  The results are reported in Table 3.3. Columns (1)-(3) correspond to the three specifications 3.5a-3.5c with the OLS estimation. In Column (4), ln EI is instrumented with the input shares (ln w, ln k, ln material, ln energy). This is a cross-industry analysis. In all regressions, the observed industry heterogeneity (ln EM P , HHI, GC, DI), and the group effect are controlled for. In column (1), the result shows that the estimated coefficient on the inverse elasticity 1/ε is positive and significant at the 1% level. This positive sign is consistent with the prediction implied by equation (3.4). That is, the enforcement policy is more stringent (a higher level of inspection rate) for industries with large absolute elasticities. The natural log of the pollution damage indicator ln ρ has a positive effect on the inspection rate, but the coefficient is not statistically significant. ln EI has a positive and significant estimated coefficient. Column (2) reports the results when the interaction between ln ρ and 1/ε is included. It shows that the estimated coefficients on ln ρ is still positive and becomes significant. The interaction term ln ρ × 1/ε has a positive and statistically significant estimated coefficient at the 1% level. The strong significance level of the estimated coefficient on the interaction term suggests the existence of a complementary effect between ln ρ and 1/ε. I conducted two other tests for additional evidences about the existence of this interaction term. First, because Specification one is nested in Specification two, I test the restrictions imposed in Specification one via a likelihood ratio (LR) test. The LR test statistic is presented at the bottom of Column (1) and is very significant. The restrictions are rejected and the result is in favor of Specification two. Second, in Column (2), I report the result of the joint significance test of ln ρ and the interaction term and the result of the joint significance test of 1/ε and the interaction term. The null hypothesis that the coefficients of both ln ρ and the interaction term are zero is strongly rejected. The result of the joint significance test for 1/ε and the interaction term is similar. 42  3.4. Empirical results I explore their effects in determining the inspection level further by calculating the net effect. The net effect of the reciprocal of the elasticity of violation 1/ε on the inspection rate, ln IR, is positive for around 98.8 percent of industries. In particular, at the sample mean of ln ρ71 , one standard deviation increase in 1/ε would approximately increase the inspection rate level ln IR by 0.49.72 The net effect of the damage indicator is conditional on the elasticity. At a larger absolute elasticity of violation rate, the net effect of the damage indicator is positive. But when the absolute elasticity is small enough, the net effect becomes negative. This is because each inspection of an industry with a larger damage indicator has to incur more inspection cost. When the inspection is not effective at deterring violation for this industry, the enforcement agency chooses to inspect less.73 In Column (3), I estimate Specification three which is a more general nonlinear function by adding (ln ρ)2 ,  1 2 ε ,  and ln ρ ×  1 2 ε  to Specification Two.  The estimated coefficients on ln ρ, 1/ε and the interaction term ln ρ × 1/ε are positive. The estimated coefficient on 1/ε is still significant. This is similar to the results of Specification two in Column (2). The estimation finds significantly effects of the quadratic term of 1/ε and that of the interaction term ln ρ ×  1 ε  . This serves as further evidence that the interaction terms  are necessary to be considered. As shown at the bottom of Column (2), the result of the LR test that tests the restrictions on Specification Two is very significant. Moreover, R2 increases from 0.88 to 0.95. Thus, considering the role of higher orders for the interaction term may improve the performance of the regression. However, there exists concern that the specification with 71 The summary statistics of ln ρ in this sample: mean=7.91, s.d.=3.50, min=-10, max=16. 72 The calculation is as follows: [0.04 + (0.023) × 7.91] × 2.2 = 0.49, where the standard deviation of 1/ε is 2.2. 73 This result can be obtained by slightly modifying the theoretical model. In the enforcement agency’s maximization problem (6), I assume the monitoring expenditure coefficient e is an increasing function of the damage indicator ρ. Then Equation (7) can be rewritten 1 1 as pi = ρi = ρi . It shows clearly that the effect 1 t − λe (ρi ) 1 + εi t + λe (ρi ) |ε1i | − 1 of the damage indicator is conditional on the absolute elasticity. When the absolute elasticity is small enough, an increase in the damage indicator may lower the inspection rate due to a higher inspection expenditure.  43  3.5. Robustness and sensitivity analysis higher orders for the interaction term may be subject to the multicollinearity problem to some extent. And the estimated coefficients in Specification (3) are not directly interpretable. Thus, of the three specifications in the first three columns, I put more emphasis on Specification (2). Based on Specification Two (Equation 3.5b), I control for the endogeneity of ln EI using the IV method and report the results in Column (4). The estimated coefficients in Column (4) are similar to those in Column (2) both in terms of the magnitude and the significance level. The only big difference is that the estimated coefficient on ln EI becomes bigger after its endogeneity is controlled for. In sum, it confirms the hypothesis that the inspection rate structure across industries depends on the inverse elasticity of the violation rate. Moreover, there exists a strong complementary effect between ln ρ and 1/ε in determining the inspection rate.  3.5  Robustness and sensitivity analysis  To investigate the robustness of the findings, I run another group of regressions, with the results reported in Table 3.4. Table 3.4 contains estimates based on the second measure of the enforcement policy, the abatement cost per unit emission. The abatement cost per unit emission replaces the inspecAC tion rate and the logarithm of the abatement cost (ln Emission ) is employed  as the dependent variable. Columns (1)-(4) correspond to the four columns in Table 9 respectively. The abatement cost data are one year (1994’) crossindustry data. The mean values over time are used for time varying independent variables (ln ρ, DI, GC) and the number of observations reduces to 108. Results in Columns (2)-(3) show the existence of the interaction term’s effect on the enforcement policy level. The estimated coefficient on the inAC verse elasticity 1/ε is negative. This is perhaps because ln Emission captures  not only the stringency of the enforcement policy, but also that of the environmental regulation (e.g. the emission limit). ln EI has a negative and significant estimated coefficient in contrast to the positive estimated coefficient in Table 3.3. 44  3.5. Robustness and sensitivity analysis  Table 3.3: Inspection Rates Regressions Dependent variable: ln IR Independent Variable  lnρ 1/ε  OLS (1)  OLS (2)  OLS (3)  IV (4)  .013 (1.07) .19∗∗∗ (11.50)  .10∗∗∗ (6.8) .04∗ (1.86) .023∗∗∗ (7.73)  .10∗∗∗ (5.38) .04∗ (1.84) .023∗∗∗ (7.56)  .02 (1.08) −.04 (−1.56) .54 (1.10) .22 (0.90) −.03 (−0.29) .11∗ (1.93)  .01 (0.62) −.02 (−1.22) .15 (0.38) −0.06 (−0.31) 0.05 (0.54) .10∗∗ (2.06)  .12∗∗ (2.09) .35∗∗∗ (3.45) .04∗∗∗ (2.78) −.004 (−1.65) .02∗∗∗ (3.94) .0003∗∗∗ (3.31) −0.002 (−0.28) .008 (0.63) .08 (0.32) −.09 (−0.75) .06 (1.04) .07∗∗ (2.35)  0.007 (0.25) −.02 (−1.16) .15 (0.37) −.06 (−0.30) .05 (0.51) .09∗ (1.76)  0.95 110  0.88 110  lnρ× (1/ε) (ln ρ)2 (1/ε)2 (ln ρ × 1/ε)2 ln EI ln EM P HHI DI GC Group2 R2  0.81 0.88 Obs. 110 110 F-statistic for the joint significance test (Prob>F) lnρ, lnρ× (1/ε) 30.74∗∗∗ 1/ε, lnρ× (1/ε) 134.37∗∗∗ 2 2 lnρ, lnρ× (1/ε) , (ln ρ) , (ln ρ × 1/ε) 1/ε, lnρ× (1/ε) , (1/ε)2 , (ln ρ × 1/ε)2 LR test /χ2 51.48∗∗∗ 102.39∗∗∗  4.41∗∗∗ 199.82∗∗∗  *** Significant at the 1% level, ** at the 5% level, and * at the 10% level (t statistics in parentheses)  45  3.6. Conclusions In addition, for the four columns in Table 3.3, I run regressions for each year from 2003’-2005’ instead of using the mean value over the three years. The results are similar and not reported here.  3.6  Conclusions  This chapter investigates the determinants of the environmental enforcement policy when the enforcement agency is constrained by a hard monitoring budget. I started with a simple theoretical model to emphasize two industry characteristics: the damage indicator and the elasticity of the violation rate with respect to the inspection rate. The damage indicator measures the industry-specific pollutants’ potential damage to the environment and human health. Industries with large damage indicators may be subject to more inspections because of the high social costs resulting from their violation, all else being equal. The elasticity of violation reflects the responsiveness in the industry-level pollution violation rate as a result of change in the enforcement policy. Deadweight loss from inspection is smaller in industries with large absolute elasticities of the violation rate so that the enforcement agency tends to direct more resources to these industries and inspect them more frequently, all else being equal. In the empirical section, I estimated the linearized specification and its more general forms. The results identified the effects of the damage indicator and (the reciprocal of) the elasticity of the violation rate on the inspection rate level. In particular, a strong complementary effect between the damage indicator and the elasticity of the violation rate on the inspection rate is found. Existing works have not empirically linked these two industry characteristics to the enforcement level. The dataset about the damage indicator has been seldom explored, and the data of the elasticities that are calculated in Chapter Two are unique. This chapter calls for more attention on the indicators that measure the overall damage to the environment and human health in addition to the traditional pollution measurement – the emission intensity. 46  3.6. Conclusions  Table 3.4: Robustness of Inspection Rates Regressions a AC Dependent variable: ln Emission Independent Variable  lnρ 1/ε  OLS (1)  OLS (2)  OLS (3)  IV (4)  .001 (.02) .23∗∗∗ (2.87)  .17∗ (1.78) −.03 (−.24) .04∗∗ (2.32)  .15 (1.16) −.04 (−.26) .04∗∗ (2.20)  −.85∗∗∗ (−11.11) −.06 (−.47) 1.33 (.56) .078 (.07) .14 (.26) .33 (1.17)  −.87∗∗∗ (−11.53) −.03 (−.27) .65 (.28) −.40 (−.35) .28 (.54) .29 (1.06)  .73 (1.35) −.75 (−.80) .20∗ (1.71) −.02 (−1.06) −.03 (−.64) .001∗ (1.67) −.88∗∗∗ (−11.73) .01 (.08) 1.38 (.57) −.51 (−.45) .41 (.76) .25 (.92)  −.83∗∗∗ (−4.36) −.02 (−.13) .76 (.31) −.48 (−.40) .31 (.57) .33 (.95)  0.73 108  0.72 108  lnρ× (1/ε) (ln ρ)2 (1/ε)2 (ln ρ × 1/ε)2 ln EI ln EM P HHI DI GC Group2 R2  0.71 0.72 Obs. 108 108 F-statistic for the joint significance test (Prob>F) lnρ, lnρ× (1/ε) 2.68∗ 1/ε, lnρ× (1/ε) 6.98∗∗∗ 2 2 lnρ, lnρ× (1/ε) , (ln ρ) , (ln ρ × 1/ε) 1/ε, lnρ× (1/ε) , (1/ε)2 , (ln ρ × 1/ε)2 LR test /χ2 5.76∗∗ 4.95  .79 4.42∗∗∗  *** Significant at the 1% level, ** at the 5% level, and * at the 10% level (t statistics in parentheses) a: The dependent variable is used to proxy the enforcement policy.  47  Chapter 4  Determinants of Environmental Political Contribution 4.1  Introduction  Is environmental protection for sale? There have been overwhelming concern and evidence that pro-environment policies are obstructed by politics (e.g., Pittman, 1977; Kraft and Vig, 200974 ; Mixon, 1995; Cropper et al., 1992; Pashigian, 198575 ; Damania et al., 200376 ). In this paper, I first construct an index of environmental political contribution (EPC) – the campaign contribution presented to the congress exclusively for environmental issues, which serves as the direct evidence of environmental protection for sale (EPFS) – to measure the environmental political activity. The data show that there 74 Kraft and Vig (2009) argue that “intense opposition to environmental and natural resource policies arose in the 104th Congress (1995-1997), when the Republican Party took control of both the House and Senate for the first time in forty years...pitched battles over environmental and energy policy continued in every Congress through the 110th (2007-2009), and they were equally evident in the executive branch as the Bush White House sought to rewrite environmental rules and regulations to favor industry...” 75 Mixon (1995) provides evidence that the ratio of the number of registered lobbyists in the state population has an effect on the degree of an environmental protection agency (EPA) citation for carbon emissions standards; Cropper et al. (1992) indicate that comments in the regulatory political process by grower organizations significantly reduced the probability of cancellation of certain pesticide; Pashigian (1985) points out that votes from the South and the West with higher growth rates and superior air quality oppose the environmental policy of prevention of significant deterioration in the locational competition. 76 Damania et al. find that corruption reduces environmental policy stringency in a cross-country analysis.  48  4.1. Introduction is a substantive variation across industry and time in the intensity of EPC (EPC/employment) as shown in Figure 4.1.77 I intend to identify what explains the variation across industries in the intensity of EPC.  Figure 4.1: The Variation of the EPC Intensity I explore the role of the enforcement policy stringency in explaining the variation. The intuition is as follows. Although the EPC presented to the Congress is targeted directly on environmental legislation/regulation rather than the enforcement policy set by the environmental protection agency (EPA), the enforcement policy affects the marginal benefit from contributing the congress for relaxing the legislation/regulation.78 For example, an industry that has higher expected sanctions/penalties from violating an emission standard has more to gain from spending a dollar to get the standard lowered than an industry with lower expected penalties. In this way, I predict that industries facing stricter enforcement policies are more likely to engage in the political activity all else equal. And the enforcement policy has a po77  Because industries vary a lot in size, the environmental political contribution (EPC) is divided by the number of paid employment. EPC/employment is defined as the EPC intensity in this chapter implying the amount of EPC in dollars made by each employee. The unit is the the U.S. Dollar. The details will be discussed in Section 4.3. 78 One reason to test EPFS indirectly this way is that I do not have data on regulation (i.e., the elasticity of violation with respect to regulation). Instead, I have data on the elasticities of violation with respect to enforcement estimated in Chapter 2.  49  4.1. Introduction litical consequence. We then test the prediction on the industry-level panel data of the U.S.A. This chapter builds on prevous empirical work that studies the determinants of the industrial political activity. They are based on the theory of regulation (Stigler, 1971; Olson 1965): well-organized groups can exert political influence on a government which maximizes its political support to relax the environmental policy stringency that imposes a deadweight loss on the general voter.79 The industrial political activity is related to the environmental regulations as well as other government regulations empirically. For example, Gawande and Bandyopadhyay (2000) shows that the tariff barrier is a factor that determines the trade-related political contribution. In Pittman (1977), environmental regulations are found to be positively related with total political contributions. In sum, the campaign contributions are used to measure the industrial political activity and the attention is paid to the government regulations themselves as determinants. This work makes two key contributions to the literature. One is the characterization of the environmental political activity using EPC. In the literature, the total campaign contribution has been employed to proxy the industrial political activity (e.g., Pittman, 1977; Grier and Munger, 1991, 1993; Gopoian, 1984)80 . However, there are concerns about using the total contributions to evaluate the impact on a certain policy (such as trade policy81 or environmental policy) since total amounts of contributions may be the result of multiple purposes. As argued by Gawande and Krishna (2001), “Taking trade policy as an example, lobbying data come as a bundle so it is not easy to disentangle the purely trade-related component of lobbying data. Although this problem may be alle79  For the theoretical works that characterize the formation of environmental policy as a result of the interaction between the government and organized lobbying groups, see Fredriksson (1997) and Damania (2001). 80 This is because, although payments for political favors can take many forms, campaign contributions from the political interest groups are the most visible and contentious, as argued by Doren et al. (1999). 81 In Gawande and Bandyopadhyay (2000) and Gawande (1998), the trade-related political contribution is calculated and employed. But the approach is different from the one that this chapter employs.  50  4.1. Introduction viated by considering a set of industries whose primary lobbying concern is trade protection (e.g., Lopez & Pagoulatos, 1996), it requires more care to do a full cross-sectional study for all of manufacturing, both in the measurement of lobbying as well as its econometric treatment.” I calculate EPC as follows: key members of House/Senate environmental committee/subcommittee were identified. Then, campaign contributions from each corporate PAC (political action committee) to those members were collected and summarized. Finally, individual corporate PAC contributions with their SICs were aggregated up to the SIC four-digit industry level. This calculation is based on the insight that PACs target incumbents who have seats on committees with jurisdiction over the PACs’ areas of interest or committee membership, as pointed out by many existing works (Grier and Munger, 1991, 1993; Riddel, 2003; Gardner, 1995). A firm’s EPC is supposed to be a better measure of its environmental political activity than its total political contribution. The other contribution is to isolate the effect of the environmental enforcement policy from that of the environmental regulation on the environmental political contribution using the elasticity of the violation rate. Although the enforcement policy and the environmental policy are determined by separate agencies, the two kinds of policies have many common determinants (e.g. the emission intensity). If the relationship between the enforcement policy and the environmental political contribution arises as a result of these common determinants, it is hard to distinguish the enforcement policy’s effect from the environmental policy’s effect. In this work, I use the elasticity of the violation rate with respect to the inspection rate to infer the enforcement policy’s effect on the environmental political contribution. The elasticity of the violation rate is a measure of responsiveness in the industrial pollution violation rate as a result of change in the inspection rate. It is regarded as an exogenous industry characteristic and has been estimated in Chapter Two. When an enforcement agency is constrained by a hard enforcement budget, it tends to target industries with 51  4.1. Introduction larger absolute elasticities of the violation rate. This is because inspections in these industries are more likely to be effective in achieving the enforcement’s goal all else equal. I regard the elasticity of the violation rate as a major determinant of the inspection rate and found empirical evidence in Chapter Three. Based on these, I predict that industries with larger absolute elasticities of the violation rate are subject to stricter enforcement policies and therefore present more political contributions. Empirically, I also highlight the risk-weighted emission intensity and its interaction term with the inverse elasticity of the violation rate, as well as the industry concentration as determinants of the environmental contribution expenditures. First, the risk-weighted emission intensity is defined as a measure of pollutants’ potential damage to the environment and human health per unit of production. It can be regarded as the emission intensity weighted by the damage indicator82 . The risk-weighted emission intensity has an effect on the environmental political contribution through two channels: the environmental regulation (the direct effect) and the enforcement policy (the indirect effect). For the direct effect, a damaging industry (an industry with a larger risk-weighted emission intensity) is subject to more stringent environmental regulations (e.g. the pollution tax). Then, this industry has incentives to contribute to the congress to relax the environmental policies. For the indirect effect, whether a damaging industry is subject to a strict enforcement policy may be conditional on the elasticity of the violation rate. This is because the damage indicator is a key component of the risk-weighted emission intensity. And I find a complementary effect between the damage indicator and the inverse elasticity in determining the inspection rate in Chapter Three. Therefore, I include an interaction term of the risk-weighted emission intensity and the inverse elasticity in the estimation. The underlying mechanism is as follows: when the absolute elasticity 82  The damage indicator can be regarded as the damage level per unit of emission, while the risk-weighted emission intensity represents the damage level per unit of output. See Chapter Three for details about the damage indicator.  52  4.1. Introduction of violation rate is large, the inspection of a damaging industry is effective. Although its inspection has to incur a higher inspection cost compared with the inspection of a less damaging industry, the inspection rate is higher for a damaging industry. In other word, the risk-weighted emission intensity is positively associated with the inspection rate. That is, a damaging industry is subject to a higher inspection rate. Furthermore, for this industry, a oneunit decrease in the pollution limit as a result of the political contribution can lead to a larger reduced sanction amount and this industry tends to present more environmental political contribution. Otherwise, when the absolute elasticity of violation rate is small, the inspection of an industry with a large damage indicator is not effective and has to incur a higher cost. The enforcement agency may be reluctant to impose a strict enforcement policy. As a result, a damaging industry is not inspected more frequently than other industries and does not contribute more. That is, the indirect effect of the risk-weighted emission intensity on the EPC through the enforcement policy is negative. Thus, the overall effect of the risk-weighted emission intensity is ambiguous depending on the magnitudes of the indirect effect and the direct effect. But it is predicted to be positive when the elasticity of the violation rate is large. Second, the industry concentration measuring obstacles to collective action is an important factor of the environmental political contribution. Polluting firms that have common interest in changing certain environmental policy usually organize and form a lobby group to present the contributions. Based on the lobbying function in Gawande (1998), I predict that industries that are less concentrated are hard to overcome the difficulty in collective action and thus present less environmental political contributions. The empirical findings using the environmental political contribution as the dependent variable are summarized as follows. (1) The net effects of the absolute elasticity of violation rate on the environmental political contribution, evaluated at most industries’ risk-weighted emission intensity levels, are positive. This implies that a damaging industry is subject to a higher inspection rate and tend to make more contributions. In particular, at the 53  4.1. Introduction sample mean level of the risk-weighted emission intensity, one standard deviation increase in the inverse elasticity would approximately increase the natural log of the environmental political contribution intensity by 0.35. (2) The net effect of the risk-weighted emission intensity is positive as long as the absolute elasticity of the violation rate is large enough. One example is that one standard deviation increase in the natural log of the risk-weighted emission intensity would lead to a 0.024 increase in the natural log of the environmental political contribution intensity, evaluated at the median value of the inverse elasticity. (3) The industry concentration affects EPC (intensity) significantly and positively which is consistent with existing literature. In addition, I repeat the estimations, but replace the environmental political contribution by the House environmental political contribution and the total political contribution as the dependent variables respectively. The results using the House environmental political contribution are similar with those using the environmental political contribution, both in terms of the estimated magnitude and the significance level. But when I use the total political contribution as a measure of the environmental political activity, the significance levels of the estimated coefficients on some key independent variables are reduced. This suggests that the environmental political contribution constructed in this work serves as a better measure than the total political contribution in finding evidence for the roles of the elasticity of the violation rate and the risk-weighted emission intensity. The rest of the chapter is organized as follows. In Section 4.2, I derive the lobbying function to identify the determinants of EPC. In Section 4.3, I describe how to construct the environmental political contribution and summarize its statistics. Section 4.4 introduces the empirical specification and describes data sources. Section 4.5 presents the empirical results. In Section 4.6, I conduct the robustness checks. Conclusion is provided in Section 4.7. Some supplementary data descriptions are contained in Appendix C.  54  4.2. Theoretical perspective  4.2  Theoretical perspective  The following simple theoretical model is to shed light on the factors that affect each lobby group member’s private contribution amount in a multiagent system. I emphasize the roles of the enforcement policy stringency and the lobbying organization. A political economy is composed of a congress, an enforcement agency and polluting firms. The congress formulates the environmental regulation which is the emission limit z. I assume, for simplicity, that the emission limit is the same for all firms in each industry. The emission limit z depends on the total campaign contribution amounts presented to the congress. This is because the congress cares about both the social welfare and contributions83 , and so it is willing to adjust the emission limit away from the value that maximizes social welfare in exchange for more political contributions. The enforcement agency sets the enforcement policy – the sanction level s for each unit of violation (the emission amount z that exceeds the emission limit z). Because the campaign contributions are presented to the congress, their amounts do not affect the enforcement agency’s decision. Both of the emission limit and the sanction rate are the same for all firms in each industry. Polluting firms that have common interests in changing the emission limit z could organize and form a lobby group. The total contribution amount presented by each lobby group to the congress is the sum of its members’ private provisions of campaign contribution. I assume z = log ( where Λi is the campaign contributions presented by firm i and  Λ−i + Λi ),  Λ−i is the  total campaign contributions by contributing firms except firm i. This function shows that the emission limit level depends on the total contribution amount (  Λ−i + Λi ). Given the contribution amount  Λ−i by other firms  in this lobby group, firm i independently decides its individual contribution amount Λi . An increase in Λi can lead to a larger z. The net cost to contributing firm i, Gi , is composed of the total sanction 83  This welfare function is common in the political economy literature (e.g. Grossman and Helpman, 1994).  55  4.2. Theoretical perspective and its own contribution amount. It takes the following form based on Gawande (1998), M in Gi = s z − log Λi  Λ−i + Λi  + Λi  where z is the total pollution level. I assume that the contribution decision is independent of the production decision so that z is regarded as a constant in this political contribution minimization problem. And it is the same for all firms in an industry.84 The emission limit z is also the same for all firms in each industry. Therefore, an individual firms’ net cost of making a contribution depends on its excess pollution (z − z) which is symmetric for all firms in an industry. Given the sanction rate s, each member of a lobby group decides its contribution amount. Taking first order condition and assuming each protectionist firm is of the same size (the number of equal-size-equivalent firms in an industry n can be approximately measured as the inverse of the Herfindahl index of that industry 1/H), individual firm’s contribution is shown as the following85 , 1 = s/nΛi In the equation above, the LHS is the marginal cost of contribution, while the RHS is the marginal benefit of contribution. The marginal benefit is a linear increasing function of s. That is, for an industry with a larger s, the marginal benefit of contribution is larger at the same amount of EPC, Lambdai . Given the constant marginal cost of contribution, the firm tends to contribution more. The variant of this lobbying function by log84 I am assuming firms in each industry producing the same amount of output and emissions so that z is not indexed by i. In this setup, I assume that firms make emission decisions before they make the contribution decisions. When they make the emission decision, they are symmetric and I do not consider this step here. When they make the contribution decision, their behaviors are strategic, as shown by the minimization problem above. s 85 The first order condition with respect to Λi is the following, − + 1 = 0. Λ−i + Λi Because firms are assumed to be symmetric in making the contribution decision and there are n firms, Λ−i + Λi is equal to nΛi . Therefore, this first order condition is rearranged to be 1 = s/nΛi .  56  4.2. Theoretical perspective transformation and linear approximation is as follows86 , ln Λi = ln S + H  (4.6)  Proposition 2 summarizes the prediction. Proposition 2 All else equal, firms in an industry that is subject to a more stringent enforcement policy tend to contribute more to relax the emission limit. Then, I have the following Lemma based on Prosposition 1 in Chapter Three and Proposition 2. Lemma 1 Industries with larger absolute elasticities of the violation rate with respect to the inspection rate tend to contribute more. Here, the elasticity of the violation rate is a measure of responsiveness in the industrial pollution violation rate as a result of change in the inspection rate. As stated in the Proposition 1 in Chapter Three, if the enforcement agency maximizes social welfare subject to a hard budget constraint for its enforcement activities, the inspection rate will be higher in those industries with a higher elasticity of the violation rate. And as the industries are subject to more frequent inspections, they tend to contribute more. Consequently, the absolute elasticity of the violation rate is predicted to be positively associated with EPCI all else equal. Equation 4.6 also shows clearly that the industry concentration measuring obstacles to collective action as indicated by H is an important factor of contribution, which is summarized in the following proposition. Proposition 3 The industry concentration affects the average contribution positively. 86  I estimate the log-log version of the first order condition as shown below, ln Λi = 1 1 ln s + ln . Because firms are assumed to be symmetric, H = . Then, the equation n n above becomes ln Λi = ln s + ln H. In this work, I allow the industry concentration to enter the specification linearly as most empirical works do, and the specification becomes ln Λi = ln s + H.  57  4.3. Descriptive analysis of environmental political contribution  4.3  Descriptive analysis of environmental political contribution  This section provides the methodology for constructing the environmental political contribution and summarizes its variation across industries. As argued, this serves as the evidence that the political pressure to distort the environmental policy exists, and will be used for estimating Equation 4.6 as the dependent variable.  4.3.1  Measuring environmental political activity  Many existing works have given insights into the candidate characteristics that political action committees prefer when they make funding decisions. As stated by Riddel (2003), many reduced-form models of PAC funding choice include legislator characteristics such as party affiliation, voting records, committee membership, seniority, and state-specific data pertaining to each legislator to explain contributions to candidates from various special interest groups such as agriculture, labor, and oil. In particular, Gardner (1995) argues that agriculture PACs target funds to where they have the highest payoffs. His results demonstrate the importance of the senate key committees. Agricultural PACs will contribute more to senators on key committees.87 Similarly, according to Grier and Munger (1991, 1993), corporate, union, and trade association PACs target incumbents who win by moderate margins, have voted sympathetically, and have seats on committees with jurisdiction over the PACs area of interest. Based on these insights, I first obtain the key House/Senate environmental committee/subcommittees whose jurisdictions are related with Environmental issues from the League of Conservation Voters. They include Subcommittee on Labor, Health and Human Services, Education, and Related Agencies in Committee on Appropriations; Subcommittee on Environment and Hazardous Materials in Committee on Energy and Commerce; 87  Cotton interest groups contribute more to agricultural appropriations subcommittee members than they contribute to nonmembers.  58  4.3. Descriptive analysis of environmental political contribution Subcommittee on Energy and Resources in Committee on Government Reform; Committee on Resources; Subcommittee on Environment, Technology, and Standards in Committee on Science and Subcommittee on Water Resources and Environment in Committee on Transportation and Infrastructure. Then, from the U.S. Government Printing Office, I find the name lists of the members of each key House/Senate environmental Committee/subcommittee. Second, each PAC is then matched against the COMPUSTAT database and various data sources to obtain its SIC code. This will be used for aggregation to the industry level. Third, I collect the campaign contributions presented by each corporate PAC88 to each member in the key committees/subcommittees. The amount of aggregated contribution by each corporate PAC to all the members in the key committees/subcommittees indicates the level of each corporate’s political activity targeting on environmental issues. These data of campaign contributions are available in the Federal Election Commission. Finally, I merged individual corporate PAC contributions with their SICs and then aggregated up to the four-digit industry level. Thus, the total environmental political contribution for each industry (the sum of the contributions received by all the candidates in the key environmental committees/subcommittees presented by the corporate PACs in this industry) is obtained. The total environmental political contribution is then divided by the number of PAC to calculate the environmental political contribution per PAC. Because industries vary a lot in size, the environmental contribution is scaled by the number of paid employment. The data source of the number of 88  Grier et al. (1994)[36] state: “It is useful to distinguish PACs sponsored by corporations and the trade association PACs that represent the same industry. corporate PAC contributions at the industry level are important in their own right for at least three reasons: first, trade associations are subject to the statutory limit of $10,000 per election cycle per candidate. Second, we address collective action directly, by explaining differences in industries’ ability to overcome free-rider problems and induce corporations to act.; Since the pattern of contributions observed just one firm at a time could be consistent with either pure cooperation or purely atomistic behavior, we use the industry as the unit of analysis.”  59  4.3. Descriptive analysis of environmental political contribution paid employment proxying the industry size is the 1997 Economic Census. This scaled environmental contribution is defined as Environmental Political Contribution Intensity (EP CI) in this paper and will be used as the main dependent variable. The details of the data sources and methodology are presented in Appendix C.2. Meanwhile, I calculate the total political contribution intensity (P CI) similarly. Although EP CI is a better proxy of environmental political activity, P CI contains the information of EP CI and will be used for comparison. EP CI H is the intensity of the environmental political contribution to the candidates in the key House environmental committees/subcommittees. The statistic analysis in this section is based on data covering four election cycles 1999’-2006’. It is noteworthy that I focus on the manufacturing sector (2000 to 3999 SIC code) only. This is not only because of the data richness, but also because a large share of the contributions are presented by the PACs in this sector. Specifically, the corporate PACs in manufacturing sector captures around 35 percent of the corporate PACs listed in the Federal Election Commission data and recorded with a SIC code in one election cycle.  4.3.2  Trend analysis  Table 11 shows the growth of campaign contribution per PAC.89 The sample comprises 124 industries in the manufacturing group. Average contributions per PAC across industries have increased from around $50 thousand in 1999 to more than $80 thousand in 2006. Both the average environmental contributions per PAC, and those to the House candidates only, grow significantly. The largest environmental contribution per PAC by a single industry rises from $133500 to $22974090 . 89  The environmental campaign contribution per PAC is not the dependent variable that will be used later. The dependent variable is the environmental campaign contribution intensity, which is the ratio of the EPC over the paid employment. 90 The expenditures are not adjusted for inflation. This chapter focuses on the variations across industries, not those over time.  60  Election Cycle (year)  Average Nonzero Contributions per PAC($)  Average Nonzero Environmental Contributions per PAC($)  1999-2000 2001-2002 2003-2004 2005-2006  53768.99 52605.98 64551.49 82203.08  20225.72 23266.94 27344.44 33345.63  Largest Single-Industry Environmental Contributions per PAC($) 133500 170730.6 202603.2 229740  EPC per PAC to Members in House Committees($) 14447.62 15736.58 18741.86 21640.89  4.3. Descriptive analysis of environmental political contribution  Table 4.1: Environmental Political Contribution to Congressional Candidates (124 Industries in Manufacturing Group)  61  4.4. Empirical specification and data sources There are 160 industries with EP C data available in the election cycle 2005-2006.91 Table 4.2 lists the ranks of industries according to the total environmental contribution (EP C). These show the large variation across industries. Meanwhile, The mean value of EP C per PAC is 33345.63, and that of P C per PAC is 82203.08. This implies that about 40% of the total political contribution presented by a PAC is accounted as the contributions for environmental purposes. This is considered as the upper limit of the proportion of environmental political contribution in total political contribution. The reason is that the key environmental committee members may have positions in other committees so that the contributions could be presented for other purposes as well. In sum, this section constructs the data on EPC to provide evidence that government cares about both the social welfare and EPC and trades political protection for EPC. Given this, the following studies what drives the variation of the environmental political contribution across industries.  4.4  Empirical specification and data sources  The reduced-form specification is established based on Equation 4.6, which suggests the central roles of the inverse elasticity of violation rate with re1 spect to inspection rate (via its effect on the inspection rate) and the ε industry concentration H in determining the environmental political contribution. The estimation is based on the following specification: 1 1 + α12 ln scorej,t × + α3 HHIj εj εj +α4 P T ratej + α5 ISOj + α6 ln EM Pj + α7 GCj,t  ln EP CIj,t = α1 ln scorej,t + α2  +α8 DIj,t + α9 Groupj + α10 Y earj + εj  (4.7)  where ln score represents the natural log of the risk-weighted emission in1 tensity, is the reciprocal of the elasticity of the violation rate, HHI is the ε 91  The ranks in other election cycles do not differ from those much.  62  4.4. Empirical specification and data sources  Table 4.2: Industries in Manufacturing Group Ranked by Environmental Political Contribution Amounts, 2005-2006 5 Industries with the Most EPC (160 industries) Industry Name 2834 Pharmaceutical Preparations 2911 Petroleum Refining 3721 Aircraft 3812 SDNGAN Systems and Instruments 3711 Motor Vehicles and Passenger Car Bodies 5 Industries with the Least EPC Industry Name 3949 Toys and Sporting Goods 2741 Miscellaneous Publishing 3699 Misc. Electrical Equipment&Supplies 3421 Cutlery, Handtools, and Hardware 3569 General Industrial Machinery  EPC  2529469 1621337 1148700 1118461 1024220 EPC  49 700 1000 1000 1000  *1987 US SIC division classification 5 Industries with the Most EPCI (142 industries) Industry Name 2021 Creamery Butter 2076 Vegetable Oil Mills, Except Corn, Cottonseed, and Soybean 2111 Cigarettes 2911 Petroleum Refining 2834 Pharmaceutical Preparations 5 Industries with the Least EPCI Industry Name 3949 Toys and Sporting Goods 2741 Miscellaneous Publishing 3089 Plastics Products, NEC 3569 General Industrial Machinery 2657 Folding Paperboard Boxes, Including Sanitary  EPCI  81.28 43.55 28.3 24.76 22.50 EPCI  0.0007 0.0089 0.01 0.02 0.04  *1987 US SIC division classification  63  4.4. Empirical specification and data sources industry concentration, P T rate is the ratio of the facilities that participate the voluntary program (performance track), ISO is the number of firms that voluntarily adopt the international standard ISO14001, EM P is the paid employment, GC is the geographical concentration, DI is the diverse industry, and Group is the group fixed effect and Y ear is the year dummies to capture the fixed year effects. According to Lemma 1, I predict the coefficient of the inverse elasticity of violation rate  1 ε  is positive. It is noteworthy that I am using the elasticity  of the violation rate to infer the relationship between the enforcement policy and the political contribution level. That is, the enforcement policy does not appear in the regression directly. If I include the enforcement policy in the regression, it is hard to isolate the enforcement policy’s effect from the environmental regulation’s effect. Although the enforcement policy and the environmental regulation are determined by separate agencies, the two kinds of policies have many common determinants (e.g. the emission intensity). The correlation between the enforcement policy and the environmental political contribution may arise as a result of these common determinants. As shown by the specification 4.7 above, I also control for the natural log of the risk-weighted emission intensity ln score and include its interaction term with the elasticity of the violation rate ln score × 1ε . The reasons are as follows. The risk-weighted emission intensity score is defined as a measure of pollutants’ potential damage to the environment and human health per unit of production. In the US EPA’s Risk-Screening Environmental Indicators (RSEI Version 2.1.5),92 RSEI contains “risk-related results”, which imposes a risk score on industries at different SIC levels. The toxicity, surrogate dose, and population components are multiplied to calculate this risk score. It is used to assess the potential impact of industrial releases. The “riskrelated results” are scaled by paid employment93 and is denoted by “score” 92  See Appendix B.2 for details. Data of the risk-weighted emission intensity is available with the SIC code. 93 It can also be scaled by value added. The regression results are similar if EPC is also scaled by value added to calculate EPCI.  64  4.4. Empirical specification and data sources in this chapter94 . The risk-weighted emission intensity can be regarded as the emission intensity weighted by the damage indicator (which is the potential damage per unit of emission). The risk-weighted emission intensity has an effect on the environmental political contribution through two channels: the environmental regulation (the direct effect) and the enforcement policy (the indirect effect). For the direct effect, a damaging industry (an industry with a larger risk-related emission intensity) is more likely to be subject to a stringent environmental regulation (e.g. the pollution tax) directly. Then, this industry has incentives to contribute to the congress to relax the environmental regulation. For the indirect effect, if a damaging industry is subject to strict enforcement policies, a one-unit decrease in the pollution limit as a result of the political contribution can lead to a larger reduced sanction amount for this industry, so that this industry tend to contribute more. In this case, the overall effect of the risk-weighted emission intensity on the environmental political contribution is positive. However, if a damaging industry is subject to lax enforcement policies, this industry tend to make less contribution. And the overall effect of the risk-weighted emission intensity is ambiguous. In order to investigate further when the effect of the risk-weighted emission intensity on the inspection rate is positive so that the effect of the risk-weighted emission intensity on the environmental political contribution can be identified, I add an interaction term of the risk-weighted emission intensity and the inverse elasticity in the estimation. In Chapter Three, it has been found that there exists a complementary effect between the damage indicator and the inverse elasticity in determining the inspection rate. Because the damage indicator is a key component of the risk-weighted emis94  This is similar to the measure of environmental exposure in Antweiler (2003). It combines a firm’s composition and toxicity of emissions and its location that determines the size of the population at risk with the firm’s size and intensity of pollution. This measure of environmental exposure is utilized as one of firm characteristics to investigate its effect on firm’s abatement effort. This effect is used to back out the effect of green regulatory threat which is not directly measurable. Because environmental exposure triggers regulation, it has been shown that the effectiveness of threat depends on environmental exposure of a firm. It also depends on abatement ladder rung of a firm and government’s incentive scheme as argued in this paper.  65  4.4. Empirical specification and data sources sion intensity, it is hypothesized that the effect of the risk-weighted emission intensity on the inspection rate is conditional on the elasticity of the violation rate. The underlying mechanism is as follows: when the absolute elasticity of violation rate is large, the inspection of a damaging industry is effective. Although its inspection has to incur a higher inspection cost compared with the inspection of a less damaging industry, the inspection rate is higher for a damaging industry. In other word, the risk-weighted emission intensity is positively associated with the inspection rate, hence the environmental political contribution, when the absolute elasticity of violation rate is large. Otherwise, when the absolute elasticity of violation rate is small, the inspection is not effective and has to incur a higher cost. The enforcement agency may be reluctant to impose a strict enforcement policy. As a result, a damaging industry is not inspected more frequently than the less damaging industry. That is, the indirect effect is negative. As the indirect effect dominate the direct effect, a damaging industry may not make a larger contribution. I describe the other variables used in the analysis as follows. The number of paid employees (EMP) is used to measure industry size and is from the 1997 Economic Census.95 The 1992 Census of Manufactures report MC92S-2, “Concentration Ratios in Manufacturing”, is the data source for HHI (Herfindahl Hirschmann Index) that characterizes the industry concentration96 . DI and GC may also impact the collective action to seek political influence because they are measures of the extent to which the industry has common interest. Measures of Diverse industry (DI) and Geographic Concentration (GC) were constructed as described by Grier, Munger and Roberts (1994). The number of segments is reported by each firm, up to a maximum 95  The Economic Census profiles American business every 5 years and the 2002 data are available. However, the 2002 Economic Census is based on North American Industry Classification System (NAICS). Given the violation rate is based on SICs, the conversion of employment based on NAICS into that based on SICs may lead to some bias. The 1997 data are available in both NAICS and SICs. And since the employment number is used to control for industry size which is relatively stable over short periods, I choose the 1997 data. 96 The 1997 data are available but are classified by NAICS instead of SIC.  66  4.4. Empirical specification and data sources of 10 in COMPUSTAT industrial segment file. The data were averaged by industry (mean=1.7, s.d.=0.65) and a diverse industry is defined as “one with mean number of products more than two s.d. above the sample average, that is, all industries whose firms average more than three products line are coded 1.0 and the others 0.00”. Because the data vary little at 4-digit level, I compiled them into 3-digit level data. And the data cover the period of 1998-2006. Industries with greater product diversity have weaker common interests among firms within the industry, so we expect contributions to be lower. Geographic concentration is calculated as  i  Salesi Sales  2  for each  industry where i denotes state. The data of sales (DATA12) in each state were collected from COMPUSTAT as well. GC is a measure of whether an industry has good alternatives to contributions as means of gaining political influence (e.g., lobbying, direct appeals to voters). If an industry is located entirely in a single state, its employees are an important voting block for legislators in that state. The expected sign for the coefficient on GC is negative, since votes can substitute for contributions. I include P T rate and ISO to take into account of the competition between opposing lobbies and pressures from other groups. Performance-track is a program that firms can voluntarily participate. The “Enforcement & Compliance History Online (ECHO)” publicly provided by the U.S. EPA records the facilities that are performance track members. The P T rate is calculated as the ratio of the number of facilities the are performance track members in total facilities inspected within the last 5 years. Because the variation at the 4 digit is too small, I aggregate the data according to the SIC code to the 3-digit level. ISO is the number of firms that certify their environmental management system (EMS) to the international standard of ISO14001, scaled by the number of paid employees, and the data source is Darnall (2003).97 The participation of the voluntary program and the adoption of voluntary standard are both indicators of firms’ 97  It is described in this paper that “lists of all ISO 14001-certified facilities (1996-1999) were obtained from Global International Quality Group and McGraw-Hill. In addition, the websites of state environmental agencies were searched for additional ISO 14001 certified facilities.” However, I do not have the time series, and the data used in my work are the average values across industries only found in the descriptive statistics table of Darnall’s.  67  4.4. Empirical specification and data sources voluntary abatement behavior. It has been argued that conflicts between industry and environmental group’s pressure (such as green consumers and ecologist groups) are incentives, at least partly, for voluntary abatement behavior in the literature (e.g., Arora and Gangopadhayay, 1995; Vidovic and Khanna, 2007; and Blackman and Bannister, 1998). Thus, higher P T rate and ISO are employed to measure higher level of political strategic competitions among lobby groups and greater pressure from environmental groups and the community. It implies that industries with more intense competitions and greater pressure have to contribute more. The dataset combining information from several sources is summarized below in Table 4.3 and Table 4.4.  68  Table 4.3: Variable Definition and Data Sources III Sources  Description  EP CI  campaign contribution = Environmental The number of paid employment  P CI  Total campaign contribution = The number of paid employment  EP CI H  in the key House environmental committee = PC to candidates The number of paid employment  score  Risk-related results = The number of paid employment  ε w k M aterial Energy EM P HHI DI GC P T rate ISO  A measure of responsiveness in the pollution violation rate as a result of change in enforcement policy workers wages = TheProduction value of industry shipments  FEC and various sources 4 election cycles over 99’-06’ FEC and various sources 4 election cycles over 99’-06’ FEC and various sources 4 election cycles over 99’-06’ RSEI, 1996’-2005’ Chapter Two, no time series 1996 ASM, 1996’  capital expenditures = TheNew value of industry shipments Cost of materials = The value of industry shipments Cost of purchased fuels and electric energy = The value of industry shipments  1996 ASM, 1996’  The number of paid employment Industrial concentration normalized between 0 and 1 =1 for a large number of product lines =0 for a small number of product lines  1997 Economic Census, 1997’ 1992 Economic Census, 1992’ Industrial segment file in COMPUSTAT Grier et al.(1994), 1998’-2006’, 3-digit  2 Salesi (i: state) i Sales The number of facilities that are Performance Track members = The number of facilities of firms that certify their EMSs to ISO14001 = The number The number of paid employment  =  1996 ASM, 1996’ 1996 ASM, 1996’  COMPUSTAT, Grier et al.(1994), 1995’-2006’ ECHO, 3-digit, no time series Darnall (2003), 2-digit, no time series  4.4. Empirical specification and data sources  Variable  69  Variable  Industry Concentration (HHI ) Industry Diversity (DI ) Industry Size (ln EM P ) Geographic Concentration (GC ) Damage Score (ln score) Inverse Elasticity of the Violation Rate(1/ε) Environmental Political Contribution Intensity (EP CI ) Contribution Intensity (P CI ) House EPCI (EP CI H ) Performance-Track Member (P T rate) ISO14001 (ISO ) Labor Share (w) Capital Share (k ) Material Share (M aterial) Energy Share (Energy )  Minimum  Maximum  447 3950 419 5302 3709 111  0.066 .04 9.95 .40 18.78 -4.0  0.06 .20 1.11 .24 3.28 2.19  0.000056 0 6.20 .09 1.02 -11.3  0.297 1 13.18 1 27.57 -0.9  563  2.86  7.00  .0007  81.28  607 540  7.26 1.77  18.65 4.12  .0007 .0009  213.74 50.08  94  1.0  1.26  .019  7.17  20  .08 .11 .03 .50 .02  .05 .05 .03 .13 .03  .01 .01 .001 .17 .002  .18 .32 .34 .90 .21  456 450 455 455  4.4. Empirical specification and data sources  Table 4.4: Summary Statistics III Number of Mean Standard Observations Deviation  70  4.5. Empirical results  4.5  Empirical results  The results are presented in Table 4.5. The three columns of Table 4.5 use different political activity measures as the dependent variables. They are EP CI (Environmental Political Contribution Intensity: the sum of the campaign contribution amounts made to candidates in the key environmental committees weighted by the number of paid employment), EP CI H (Environmental Political Contribution Intensity: the sum of the campaign contribution amounts made to candidates in the key HOUSE environmental committees weighted by the number of paid employment), and P CI (Political Contribution Intensity: the sum of the campaign contribution amounts weighted by the number of paid employment) respectively98 . Although the dataset covers four election cycles for the dependent variable, 1999-2000, 2001-2002, 2003-2004 and 2005-2006, many of the independent variables are not time-varying. I use the pooled OLS and control for the year dummies. In all regressions, the political pressure (P T rate and ISO), some industry heterogeneities (ln EM P , GC, DI), and the group effect (Group) are controlled for. In Column (1), I use the environmental political contribution intensity to measure the political activity and its natural log is the dependent variable (ln EP CI). The result shows that the risk-weighted emission intensity ln score has a positive estimated coefficient on the environmental political contribution intensity and is strongly significant at the 1% level. The esti1 mated coefficient on the reciprocal of the elasticity of violation is negative ε 1 and significant at the 1% level. Their interaction term ln score × has a ε positive and significant estimated coefficient. I calculate the net effect of the reciprocal of the elasticity of violation on EPCI first. The net effects are positive evaluated for most of the industries (around 87 percent). This is consistent with Lemma 1. In particular, at 98  EP CI H is a subset of EP CI because the campaign contribution amounts made to candidates in the key Senate environmental committees are excluded. EP CI is a subset of P CI because the campaign contribution amounts made to candidates that are not in the key environmental committees are excluded.  71  4.5. Empirical results the sample mean level of ln score99 , one standard deviation increase in the 1 would approximately increase the environmental political contribution ε intensity by 0.35.100 The effect of ln score on ln EP CI depends on the magnitude of the elasticity of the violation rate. That is, the net effect of ln score is conditional. It 1 is calculated by using the estimated coefficients on ln score and ln score × . ε When the absolute elasticity of violation rate is large, the effect of ln score is positive. For example, one standard deviation increase in ln score at the 1 median value of would approximately increase ln EP CI by 0.024.101 This ε shows that the net effect of the risk-weighted emission intensity at the median value of the inverse elasticity is positive. It implies that, for industries whose inverse elasticities are at the median value, the damaging industries are subject to more stringent environmental policy and inspections. They have incentive to make more contribution. When the absolute elasticity is small, the effect of ln score may become negative. For example, The esti1 mated coefficient of ln score at the mean value of (= −4.0, lower than its ε median value) is negative. HHI’s estimated coefficient is significantly positive at the 1% level, which is as predicted. When HHI is low, it implies this industry is less concentrated, so that the behavior of each firm is less noticeable. Hence, the free-rider problem is more serious and each firm contributes less. Furthermore, the large number of similar-sized firms complicates formation and enforcement of agreements over contribution strategies. That is, the cost of collective action is high. In addition, less concentrated industries do not have as much market power to earn maximal profits as highly concentrated industries. Less concentrated industries are not easy to overcome the difficulty in collective action and thus present little environmental political contribution. This result is consistent with the existing literature. 99  The risk-weighted emission intensity has been scaled to keep its natural log positive. The calculation is as follows: [−0.6 + 0.04 × 19] × 2.2 = 0.352, where the sample mean 1 of ln ρ is 19, the standard deviation of is 2.2 e 101 The net effects were calculated as follows: [0.15 + 0.04 × (−3.57)] × 3.28 = 0.024, 1 where median( ) = −3.57 and the standard deviation of ln score is 3.28. e 100  72  4.6. Robustness checks In Column (2), I use the environmental political contribution presented to the key House environmental committees/subcommittees per PAC (scaled by the number of paid employment) to measure the political activity and its natural log is the dependent variable (ln EP CI H). The House environmental committee members are 40% of the total committee members that accepted contributions from the listed PACs in the manufacturing group. The Senate has a higher ratio that is about 50% because many candidates are performing duties in multiple committees. That is, the contributions to the key Senate committee members are more likely to target on nonenvironmental issues than the contributions to the key House committee members. Nevertheless, the results are similar to that in Column (1) of Table 4.5 both in terms of the estimated magnitude and the significance level. Column (3) reports the results when the total political contribution intensity is used as a measure of the environmental political activity. The 1 dependent variable is ln P CI. The estimated coefficients on and ln score ε are not as significant as those in Columns (1) and (2). This suggests that EPCI is a better measure than PCI for finding evidence for Propositions 2 in this framework.  4.6  Robustness checks  I conduct five robustness checks based on the regression in Column (1) of Table 4.5 and report the results in Table 4.6. In column (1) of Table 4.6, 1 the main determinant of the inspection rate is omitted in the regression ε to focus on the effects of the industry concentration (HHI) and the riskweighted emission intensity (ln score) on the environmental contribution intensity (ln EP CI). And the risk-weighted emission intensity (ln score) is instrumented with the damage indicator (ln ρ) and the input shares (ln wage, ln material, ln k and ln energy) to alleviate its potential endogeneity problem. The estimated coefficients on ln score and HHI are both significant and positive. The magnitude of HHI’s estimated coefficient is stable compared with those in Table 4.5. The R2 is about 0.3, which is 0.1 less than 73  4.6. Robustness checks  Table 4.5: Results of Environmental Political Contribution Intensities Regressions: Election Cycles of 2000, 2002, 2004 and 2006 Dependent Variable is Independent Variable  ln score 1/ε ln score× 1ε ln EM P HHI DI GC P T rate ISO Group Year  R2 Obs.  ln EP CI  ln EP CI H  ln P CI  (1)  Pooled OLS (2)  (3)  .15∗∗∗ (3.04) −.60∗∗∗ (−3.23) .04∗∗∗ (4.03) −.61∗∗∗ (−8.43) 4.37∗∗∗ (2.92) .18 (0.56) −.15 (−0.42) .35∗∗∗ (5.52) 4.00∗∗∗ (2.65) .16 (.83) Y  .15∗∗∗ (2.99) −.54∗∗∗ (−2.93) .04∗∗∗ (3.67) −.61∗∗∗ (−8.26) 4.80∗∗∗ (3.17) .12 (0.37) −.28 (−0.80) .34∗∗∗ (5.34) 4.72∗∗∗ (3.08) .26 (1.34) Y  .08 (1.61) −.45∗∗ (−2.26) .03∗∗∗ (2.86) −.56∗∗∗ (−7.28) 6.31∗∗∗ (3.94) .54 (1.56) .09 (0.25) .35∗∗∗ (5.12) 4.30∗∗∗ (2.68) .33 (1.59) Y  0.402 372  0.398 360  0.4583 180  *** Significant at the 1% level, ** at the 5% level, and * at the 10% level (t statistics in parentheses)  74  4.7. Conclusions that in Column (1) of Table 4.5. This implies the important explanatory power of the inverse elasticity. Column (2) presents the results using the random-effect estimation. I keep the key control variables only in the regression. The signs and magnitudes of the estimated coefficients on ln score, 1ε , and ln score× 1ε are similar compared with those in Column (1) of Table 4.5. This is not used as the baseline regression because many of the control variables do not have time series. The estimation in Column (3) is the same as that in Column (1) of Table 4.5 except that P T rate and ISO are omitted. The damaging industries may be more likely to participate the voluntary programs. This is to avoid their collinearity with ln score. In Column (4), the sample is reduced to contain two election cycles (2003’-2004’ and 2005’-2006’). In Column (5), I employ the weighted least squares. The weights are the value of shipments. The overall results are stable.  4.7  Conclusions  This chapter investigates the determinants of the firms’ environmental political activity. I started with the data construction of the environmental political contribution (EPC), which is defined as the political campaign donations by firms exclusively for the purposes of environmental issues and is employed to measure firms’ environmental political activity. The constructed dataset shows that the EPC changes over time and varies significantly across industries. I highlight three factors’ roles in explaining the variation of the environmental political contribution across industries using the U.S. data: the industry concentration, the elasticity of the violation rate, and the risk-weighted emission intensity. The empirical findings are summarized as follows. (1) The net effects of the absolute elasticity of violation rate on the environmental political contribution, evaluated at most industries’ risk-weighted emission intensity levels, are positive. This implies that an elastic industry is subject to a higher inspection rate and tends to make more contributions. In particu75  4.7. Conclusions  Table 4.6: Robustness of Environmental Political Contribution Intensities Regressions Dependent Variable is ln EP CI Independent Variable  ln score  IV OLS (1)  Random (2)  OLS (3)  OLS(03’-06’) (4)  WLS (5)  .07∗∗ (2.39)  .15∗∗ (2.21) −.38∗ (−1.65) .03∗∗ (2.11) −.41∗∗∗ (−3.38) 9.77∗∗∗ (3.65)  .18∗∗∗ (3.53) −.80∗∗∗ (−4.22) .06∗∗∗ (5.10) −.45∗∗∗ (−6.47) 6.63∗∗∗ (4.37) .52 (1.57) −.08 (−0.23)  −.01 (−0.03) Y  −.24 (−1.34) Y  .11 (1.53) −.51∗∗ (−2.05) .04∗∗∗ (2.59) −.69∗∗∗ (−6.58) 1.93 (.86) .50 (.78) −.27 (−0.55) .41∗∗∗ (4.58) 3.73∗ (1.66) −.02 (−.05) Y  .11∗∗ (2.28) −.48∗∗∗ (−2.58) .04∗∗∗ (3.25) −.67∗∗∗ (−9.15) 4.69∗∗∗ (3.06) .24 (.76) −.20 (−.59) .35∗∗∗ (5.35) 2.91∗ (2.09) .03 (.13) Y  0.307 372  0.341 372  0.406 172  0.407 372  1/ε ln score× 1ε ln EM P HHI DI GC P T rate ISO Group Y ear R2 Obs.  −.51∗∗∗ (−7.39) 5.71∗∗∗ (4.44) .13 (0.43) .26 (.84) .32∗∗∗ (5.31) 2.04 (1.42) .43∗∗∗ (2.6) Y 0.291 469  *** Significant at the 1% level, ** at the 5% level, and * at the 10% level (t statistics in parentheses)  76  4.7. Conclusions lar, at the sample mean level of the risk-weighted emission intensity, one standard deviation increase in the inverse elasticity would approximately increase the natural log of the environmental political contribution intensity by 0.35. (2) The net effect of the risk-weighted emission intensity is positive as long as the absolute elasticity of violation rate is large enough. One example is that one standard deviation increase in the natural log of the risk-weighted emission intensity would lead to a 0.024 increase in the natural log of the environmental political contribution intensity, evaluated at the median value of the inverse elasticity. (3) The industry concentration affects EPC (intensity) significantly and positively which is consistent with existing literature. In addition, when I use the total political contribution as a measure of the environmental political activity, the significance levels of the estimated coefficients on some key independent variables are reduced. This suggests that the environmental political contribution constructed in this work serves as a better measure than the total political contribution in finding evidence for the roles of the elasticity of the violation rate and the risk-weighted emission intensity. This chapter could be extended in the following ways. (1) Although the contribution takes effect at the industry level, an empirical examination of individual-firm contribution about the lobbying organization would be interesting. (2) This chapter considers politically organized and contributing industries only. If the complete data were collected, a probit equation could be used to identify which industry tends to be more politically active/inactive. This is because, as argued by previous works, it is possible that variables influencing whether firms in an industry establish a PAC correlate with those determining how much money that a PAC spends.  77  Bibliography [1] Antweiler, Werner, “How effective is green regulatory threat?”, The American Economic Review, Vol.93 No.2, May 2003, pp. 436-441 [2] Arora, S. and S. Gangopadhyay, “Toward a theoretical model of voluntary overcompliance”, J. Econ. Behav. Org., 1995, Vol.28, 289-309 [3] Berry, Steven T., “Estimating discrete-choice models of product differentiation”, RAND Journal of Economics, Summer 1994, Vol.25, No.2, 242-262 [4] Berry, Steven, James Levinsohn, and Ariel Pakes, “Automobile Prices in Market Equilibrium”, Econometrica, Jul. 1995, Vol.63, 841-890 [5] Binder, Seth and Eric Neumayer, “Environmental pressure group strength and air pollution: an empirical analysis”, The Environmental Law Reporter, 4-2000, 30 ELR 10245 [6] Blackman, Allen and Geoffrey J. Bannister, “Community pressure and clean technology in the informal sector: an econometric analysis of the adoption of propane by traditional Mexican brickmakers”, Journal of Environmental Economics and Management, January 1998, Vol.35(1), 1-21 [7] Cavaliere, A., “Overcompliance and voluntary agreements”, Env. Res. Econ., Sep. 2000, Vol.17, 195-202 [8] Cohen, Mark A., “Monitoring and enforcement of environmental policy”, International Yearbook of Environmental and Resource Economics, III, Tom Tietenberg & Henk Folmer, eds. 1999 78  Bibliography [9] Copeland, Brian and Scott Taylor, Trade and the Environment: Theory and Evidence, Princeton University Press, 2003 [10] Cropper, M.L. and W.E. Oates, “Environmental Economics: A Survey”, Journal of Economic Literature, Vol.30, No.2, Jun.1992, pp. 675740 [11] Damania, Richard, “When the Weak Win: The Role of Investment in Environmental Lobbying”, Journal of Environmental Economics and Management, Vol. 42, Issue 1, July 2001, pp. 1-22 [12] Damania, Richard, Per G. Fredriksson and John A. List, “Trade Liberalization, Corruption, and Environmental Policy Formation: Theory and Evidence”, Journal of Environmental Economics and Management, Vol. 46, Issue 3, Nov. 2003, pp. 490-512 [13] Damania, Richard, Per G. Fredriksson, and Muthukumara Mani, “The Persistence of Corruption and Regulatory Compliance Failures: Theory and Evidence”, Public Choice, Feb. 2004, Vol.121(3), 363-390 [14] Damania, Richard, Per G. Fredriksson, and Thomas Osang, “Polluters and Collective Action: Theory and Evidence”, Southern Economic Journal, 2005, 72(1), 167-185 [15] Darnall, Nicole, “Why U.S. firms certify to ISO 14001: an institutional and resource-based view”, Best Paper Proceedings of the 2003 Academy of Management Conference, Seattle, Washington [16] Deily, M. and W. Gray, “Enforcement of pollution regulations in a declining industry”, Journal of Environmental Economics and Management, 1991, Vol.21, 260-274 [17] Dion, C., P. Lanoie, and B. LaPlante, “Monitoring of pollution regulation: Do local conditions matter?”, Journal of Regulatory Economics, 1998  79  Bibliography [18] Doren, Terry D. Van, Dana L. Hoag, and Thomas G. Field, “Political and Economic Factors Affecting Agricultural PAC Contribution Strategies”, American Journal of Agricultural Economics, Vol.81, No. 2, May 1999, 397-407 [19] Earnhart, D., “Regulatory factors shaping environmental performance at publicly-owned treatment plants”, Journal of Environmental Economics and Management, 2004a, Vol.48, 655-681 [20] Earnhart, D., “Panel data analysis of regulatory factors shaping environmental performance”, Rev. Econ. Stat., 2004b, Vol.86, 391-401 [21] Eicher, Theo and Thomas Osang, “Protection for sale: an empirical investigation: comment”, American Economic Review, Dec. 2002, Vol.92(5), 1702-1710 [22] Epple, D. and M. Visscher, “Environmental pollution: modeling occurrence, detection and deterrence”, Journal of Law and Economics, 1984, Vol.27, 29-60 [23] Findlay, R. and S. Wellisz, “Endogenous Tariffs, the Political Economy of Trade Restrictions, and Welfare”, in Jagdish Bhagwati (ed.), Import Competition and Response, Chicago: University of Chicago Press, 1982 [24] Fredriksson, Per G, “The Political Economy of Pollution Taxes in a Small Open Economy”, Journal of Environmental Economics and Management 33, 1997, pp. 44-58 [25] Gardner, B.D., “Plowing Ground in Washington: The Political Economy of U.S. Agriculture”, San Francisco CA: Pacific Research Institute for Public Policy, 1995 [26] Garvie, Devon and Andrew Keeler, “Incomplete enforcement wih endogenous regulatory choice ”, Journal of Public Economics 1994, Vol.55, pp.141-162  80  Bibliography [27] Gawande, Kishore, “Stigler-Olson lobbying behavior in protectionist industries: Evidence from the lobbying power function”, Journal of Economic Behavior & Organization, 1998, Vol.35, pp.477-499 [28] Gawande, Kishore and P. Krishna, “The political economy of trade policy: empirical approaches”, in J. Harrigan ed., Handbook of International Trade, 2001, New York: Basil Blackwell [29] Gawande, Kishore and Timothy Wheeler, “Measures of Effectivenss for Governmental Organizations”, Management Science, Vol. 45, No. 1, Jan. 1999, pp.42-58 [30] Gawande, Kishore and Usree Bandyopadhyay, “Is Protection for Sale? Evidence on the Grossman-Helpman Theory of Endogenous Protection”, The Review of Economics and Statistics, Feb. 2000, 82(1), pp.139-152 [31] Gopoian, D.J., “What Makes PACs Tick? An analysis of the Allocation Patterns of Economic Interest Groups”, Amer. J. Polit. Sci., May 1984, 28, pp.259-81 [32] Gray, W. and M. Deily, “Compliance and enforcement: air pollution regulation in the U.S. steel industry”, Journal of Environmental Economics and Management, 1996, Vol.31, 96-111 [33] Gray, W. and R. Shadbegian, “When and why do plants comply? Paper mills in the 1980s”, Law and Policy, 2005, Vol.27, 238-261 [34] Grier, Kevin, and Michael Munger, “Committee Assignments, Constituent Preferences, and Campaign Contributions to House Incumbents”, Economic Inquiry, 1991, 29, 24-43 [35] Grier, Kevin, and Michael Munger, “Corporate, Labor, and Trade Association Contributions to the U.S. House and Senate, 1978-1986”, Journal of Politics, 1993, 55, 614-643  81  Bibliography [36] Grier, Kevin B., Michael C. Munger, and Brian E. Roberts, ”The Determinants of Industry Political Activity, 1978-1986”, The American Political Science Review, Dec. 1994, Vol.88, No.4, 911-926 [37] Grossman, G., and E. Helpman, “Protection for Sale”, The American Economic Review, Vol. 84, No.4, Sep. 1994, pp. 833-850 [38] Grossman G. and E. Helpman, “Special Interest Politics”, 2001, Cambridge MA and London UK: The MIT Press [39] Harford, J.D., ”Measurement error and state-dependent pollution control enforcement”, Journal of environmental Economics and Management, 1991, Vol.21, 67-81 [40] Harford, J.D. and W. Harrington, ”A reconsideration of enforcement leverage when penalties are restricted”, Journal of Public Economics, 1991, Vol.45, 391-395 [41] Harrington, W., ”Enforcement leverage when penalties are restricted”, Journal of Public Economics, 1988, Vol.37, 29-53 [42] Helland, E., ”The enforcement of pollution control laws: Inspections, violations, and self-reporting”, The Review of Economics and Statistics, Feb. 1998, Vol.80, 141-153 [43] Hettige, Hemamala, Muthukumara Mani, and David Wheeler, “Industrial Pollution in Economic Development: the Environmental Kuznets Curve Revisited”, Jounral of Development Economics, August 2000, Vol. 62, 445-476 [44] Jones, C.A. and S. Scotchmer, “The social cost of uniform regulatory standards in a hierarchical government”, Journal of Environmental Economics and Management 1990, 19 No. 1, 61-72 [45] Jost, Peter-J, “Monitoring, Appeal, and Investigation: The Enforcement and Legal Process”, Journal of Regulatory Economics 1997, 12, 127-146 82  Bibliography [46] Kraft, Michael and Norman Vig, “Environmental Policy: New Directions for the Twenty-First Century”, CQ Press [47] LaPlante, B. and P. Rilstone, “Environmental inspections and emissions of the pulp & paper industry in Quebec”, Journal of Environmental Economics and Management, 1996, Vol.31, 19-36 [48] Lear, K.K., “An empirical examination of EPA administrative penalties”, Working paper, Kelley School of Business, Indiana University, March 1998 [49] List, John A., and Daniel M. Sturm, “How elections matter: theory and evidence from environmental policy”, The Quarterly Journal of Economics, Nov. 2006, Vol. 121, No.4, 1249-1281 [50] Lopez, Rigoberto and Emilio Pagoulatos, “Trade protecton and the role of campaign contributions in U.S. food and tobacco industries”, Economic Inquiry, April 1996, Vol.34, No.2, 237-248 [51] Macho-Stadler, Ines and David Perez-Castrillo, “Optimal enforcement policy and firms’ emissions and compliance with environmental taxes”, Journal of Environmental Economics and Management, Jan. 2006, Vol. 51, Iss. 1, 110-131 [52] Magat, W. and W.K. Viscusi, “Effectiveness of the EPA’s regulatory enforcement: the case of industrial effluent standards”, Journal of Law and Economics, 1990, Vol.33, 331-360 [53] Malik, A., “Avoidance, Screening, and Optimum Enforcement”, RAND Journal of Economics, Vol. 19, 1990, pp. 341-353 [54] Michida, Etsuyo and Koji Nishikimi, “North-South Trade and Industryspecific Pollutants”, Journal of Environmental Economics and Management, Vol. 54, Issue 2, Sep. 2007, pp. 229-243 [55] Mitra, Devashish, “Endogenous Lobby Formation and Endogenous Protection: A Long-Run Model of Trade Policy Determination”, The American Economic Review, Vol. 89, Issue 5, Dec. 1999, pp. 1116-1134 83  Bibliography [56] Mixon, Franklin G, “Public Choice and the EPA: Empirical Evidence on Carbon Emissions Violations”, Public Choice, Vol. 83, Issue 1-2, 1995, pp. 127-137 [57] Nadeau, L.W., “EPA Effectiveness at reducing the duration of plantlevel noncompliance”, Journal of Environmental Economics and Management, 1997, Vol.34, 54-78 [58] Olewiler, Nancy and Kelli Dawson, “Analysis of National Pollutant Release Inventory Data on Toxic Emissions by Industry”, Working paper 97-16 prepared for the Technical Committee on Business Taxation, March 1998 [59] Oljaca, Neda, Andrew Keeler and Jeffrey Dorfman, “Penalty functions for environmental violations: evidence from water quality enforcement”, Journal of Regulatory Economics 14, Nov. 1998, pp. 255-264 [60] Olson, M., “The logic of collective action”, Cambridge, MA: Harvard University Press, 1965 [61] Pashigian, B.P., “Environmental Regulation: Whose Self-interests Are Being Protected?”, Economic Inquiry 23, 1985, pp. 551-584 [62] Pittman, Russell, “Market Structure and Campaign Contributions”, Public Choice, 1977, 31, 37-51 [63] Riddel, Mary, “Candidate eco-labeling and senate campaign contributions”, Journal of Environmental Economics and Management, 2003, 45, 177-194 [64] Russell, C.S., “Game models for structuring monitoring and enforcement systems”, Natural Resource Modeling, 1990, 4(2), 143-173 [65] Shimshack, Jay P., “Monitoring, Enforcement, &Environmental Compliance:  Understanding Specific &General Deterrence”, State-of-  Science White Paper prepared for ORD and OECA, Oct. 2007  84  Bibliography [66] Shimshack, Jay P. and Michael B. Ward, “Regulator Reputation, Enforcement, & Environmental Compliance”, Journal of Environmental Economics and Management, Nov. 2005, Vol. 50, Issue3, 519-540 [67] Shimshack, J. and M.B. Ward, “Enforcement and over-compliance”, Journal of Environmental Economics and Management, 2008 [68] Stafford, S., “The effect of punishment on firm compliance with hazardous waste regulations”, Journal of Environmental Economics and Management, 2002, Vol.44, 290-308 [69] Stigler, G.J., “The theory of economic regulation”, Bell journal of economics and management science, 1971, Vol.2, 3-21 [70] Telle, Kjetil, “The threat of regulatory environmental inspection: impact on plant performance”, Journal of Regulatory Economics, 2009, Vol.35, Issue 2, 154-178 [71] Train, Kenneth, “Discrete Choice Methods with Simulation”, Cambridge University Press, 2003 [72] Trefler, D., “Trade liberalization and the theory of endogenous protection: an econometric study of U.S. import policy”,Journal of Political Economy, 1993, Vol.101, 138-160 [73] U.S. EPA, “Compliance Literature Search Results – Citations to over two hundred compliance-related books and articles from 1999 to 2007”, EPA Document No: EPA-300-B-07-001, April 2007, http://www.epa.gov/compliance/resources/reports/compliance/research/index.html [74] Vidovic, Martina and Neha Khanna, “Can voluntary pollution prevention programs fulfill their promises? Further evidence from the EPA’s 33/50 program”, Journal of Environmental Economics and Management, March 2007, Vol.53(2), 180-195 [75] Viscusi, W.K. and R.J. Zeckhauser, “Optimal standards with incomplete enforcement”, Public Policy, 1070, 27 No.4, 437-456 85  [76] Zardkoohi, A., “On the political participation of the firm in the political process”, Southern Economic Journal, 1985, Vol.51, 804-817  86  Appendix A  Appendix to Chapter 2 A.1  The BLP model  In the demand function estimation in the empirical industrial organization, the goods characteristics and price determine each goods’ mean value δ to consumers. The total value to individual consumer choosing one goods depends on this goods’ mean value and consumer’s individual taste which is assumed to follow an i.i.d. logit draw given infinite consumers. Here, the mean values for choice between compliance and violation are different. That is, when a firm is in violation, the unit penalty is Rv , otherwise, it has to meet the effluent standard by investing the minimum Rc . The total value to an individual firm choosing to violate depends on Rv with the other violation utility ξv and its individual characteristics which is assumed to follow an i.i.d. logit distribution given infinite firms in one industry. In the original BLP model, the market share for each goods is found. Then, each market share can be sorted over time or according to location as observations. Similarly, in this modified model, I find the violation rate for each industry, then I control for industry characteristics to focus on the observations sorted by industries and over time. The difference is that, one of the variation sources in the original model is the goods differentiation. Here, I just focus on one choice, the violation rate. That is, I do not study the compliance rate.  87  A.2. Econometric model of the industry-level violation rate  Table A.1: Comparision Consumer/Goods a 1 δa + v1a 2 δa + v2a .. .. . . Industry i i .. .  with the BLP Model b ··· J δb + v1b · · · δJ + v1J δb + v2b · · · δJ + v2J .. .. .. . . . δa + vnb δa + vnb · · · δJ + vnJ Firm Compliance Violation i i 1 Wci + γ1c Wvi + γ1v i i 2 Wci + γ2c Wvi + γ2v .. .. .. . . .  i  ni  n  A.2  i Wci + γnc  i Wvi + γnv  Econometric model of the industry-level violation rate  Following Train (2003)[71], I show how to derive the equation of aggregate violation rates at industry level in the difference of the welfares and to calculate the elasticities of violation below. Probability of choosing to violate for firm l in industry I is the following according to γ’s density function, Pri,lv = Pr(Flvt > Flct ) = Pr(Wvt + γlvt > Wct + γlct ) = Pr(γlct < Wvt − Wct + γlvt ) = e−e  −(Wvt −Wct +γlvt )  As a result, the violation rate in industry i is the following, ∞ −γlvt e−e−γlvt dγ lvt −∞ Pr(Flvt > Flct )e ∞ −e−(Wvt −Wct +γlvt ) −e−γlvt −γlvt = −∞ e e dγlvt ∞ −e−γlvt (e−(Wvt −Wct ) +1) −γlvt = −∞ e e dγlvt ∞ −µ(e−(Wvt −Wct ) +1) = 0 e dµ −γ lvt where µ ≡ e , as γlvt → ∞, µ → 0, and as  V Rit =  =− =  −µ e−(Wvt −Wct ) +1  e  (  eWvt −Wct +1 1 e−(Wvt −Wct ) +1  )  γlvt → −∞, µ → ∞.  |∞ 0  88  A.3. Data of the inspection rate and the violation rate = e=  eWvt eWct +eWvt ∂V R IRvt ∂IRvt V R  ∂  =  β IRi,vt +ξivt e 1 λIRi,ct +ξict β IRi,vt +ξivt e +e 1  ∂IRvt  IRvt VR  ∂Wvt (1 − V R) = IRvt ∂IR vt  = IRvt (1 − V R) β1  A.3  Data of the inspection rate and the violation rate  Inspection number  In ECHO (All Data), I search by the following cri-  teria: a four-digit SIC code in Facility Characteristics; “within 5 years since last inspection” in Inspection/Enforcement History. The result is a list of facilities that were inspected within last five years. In each facility’s detailed report, “Compliance Monitoring History” lists the dates for each inspection. A facility could be inspected from zero to multiple times in a year. I count them by years. For example, if there is a “State Conducted FCE/On-Site” on 08/22/2006, the inspection number in 2006 is increased by 1. Then, for each year, the inspection numbers of all facilities listed in the search result are summed up as the inspection number in that year for this industry. Inspection type The Detailed Facility Report includes inspections which were conducted within the last five years. Some inspection types are included in official counts, but italicized inspection types are not part of EPA official counts. In this thesis, I only consider the official counts such as FE - EPA Conducted FCE/On-Site and FZ - EPA Conducted FCE/Off-Site. For a detailed list, please refer to the section ”Compliance Monitoring History” in http://www.epa-echo.gov/echo/dfr data dictionary.html#ih. Violation number  In ECHO (All Data), search by the following criteria:  a four-digit SIC code in Facility Characteristics; “within 5 years since last inspection” in Inspection/Enforcement History; “formal enforcement actions within 5 years” in Inspection/Enforcement History. The result is a list of 89  A.3. Data of the inspection rate and the violation rate facilities that were inspected and were subject to formal enforcement actions within last five years. In each facility’s detailed report, “Formal Enforcement Actions” lists the dates for each formal enforcement action. I count them by years. For example, if there is a “State Administrative Order Issued” on 08/10/2007, the violation number in 2007 is increased by 1. Then, for each year, the violation numbers of all facilities listed in the search result are summed up as the violation number in that year for this industry. Formal enforcement actions  Enforcement actions under the Clean Air  Act, Clean Water Act, Resources Conservation and Recovery Act, Emergency Planning and Community Right-to-know Act Section 313, Federal Insecticide, Fungicide, and Rodenticide Act, and Toxic Substances Control Act are all included, such as CWA 404 Penalty AO Class I Init, Administrative Order, and Judicial. For the detailed list of the action types, please refer to http://www.epa-echo.gov/echo/dfr data dictionary.html#cea. I do not include the Notices of Violation or Informal Enforcement Actions. As indicated by the EPA, ”in many cases, a notice of violation causes a facility to correct problems and return to compliance. Many notices of violation are not escalated to formal enforcement action because the facility quickly corrects the problem(s) indicated in the notice.”.  90  A.4. First stage results of 2SLS in Table 2.3  A.4  First stage results of 2SLS in Table 2.3  Independent Variable IR IV lnEMP  lnwage lnmaterial lnk lnenergy HHI DI GC Group2 Year2004  Column (1) IR lnEI .79∗∗∗ (17.55) −.01 (−1.03) −.05∗ (−1.81) .13∗∗∗ (2.77) .07∗∗ (2.22) .05∗∗ (2.43) .63∗∗∗ (2.73) −.05 (−0.88) .11∗∗ (2.20) −.05 (−1.55) .03 (1.59)  1.18∗∗∗ (2.70) −.32∗∗ (−2.50) −.18 (−0.73) 3.48∗∗∗ (7.50) 1.69∗∗∗ (5.54) .28 (1.30) −2.81∗∗ (−2.50) .66 (1.22) .032 (0.06) −1.96∗∗∗ (−6.89) −.22 (−1.31)  Dependent variable is Column (2) IR lnEI .69∗∗∗ (13.02) −.01 (−1.27) −.05∗∗ (−2.34) .13∗∗∗ (3.02) .07∗∗ (2.51) .06∗∗∗ (3.21) .64∗∗∗ (3.10) −.06 (−0.84) .11∗∗ (2.34) −.03 (−1.35) .04∗ (1.75)  2.16∗∗∗ (3.86) −.31∗∗ (−2.50) −.15 (sl−0.61) 3.56∗∗∗ (8.06) 1.69∗∗∗ (5.79) .17 (0.82) −2.71 (−1.24) 1.20 (1.65) .08 (0.16) −2.07∗∗∗ (−7.58) −.36 (−1.46)  Column(3) IR lnEI .91∗∗∗ (24.16)  −.09 (−0.44)  −.03 (−0.63 −.12 (−0.89)  −.01 (−0.04) .50 (0.66)  .01 (0.60)  −.03 (−0.48)  *** Significant at the 1% level, ** at the 5% level, and * at the 10% level (t statistics are in parentheses)  91  Appendix B  Appendix to Chapter 3 B.1  Proof of equation 3.4  By maximizing Ω subject to the constraint, I have  ∂ΩA ∂vi ∂vi ∂vi = ti vi + pi ti − ρi − ti vi − λe vi + pi ∂pi ∂pi ∂pi ∂pi  =0  (B.1)  ∂vi pi is the elasticity of the violation rate with respect ∂pi vi to the inspection rate, it is plugged into the equation above and Equation Because ξi =  (B.1) can be simplified to the following, ξi ti vi − ρi  ξi v i = λe (vi + ξi vi ) pi  (B.2)  That is, ti pi − ρi 1 = λe 1 + pi ξi  B.2  (B.3)  Risk-screening environmental indicators  The data used to calculate ρ are from U.S. EPA’s Risk-Screening Environmental Indicators (RSEI Version 2.1.5). This model is a screening-level tool that assesses the potential impact of industrial releases from pounds-based, hazard-based, and risk-related perspectives. It uses risk concepts to screen large amounts of Toxics Release Inventory (TRI) data (1996-2005 TRI data). RSEI considers the following information: the amount of chemical released, the toxicity of the chemical, its fate and transport though the environment, the route and extent of human exposure, and the number of people affected. 92  B.2. Risk-screening environmental indicators  Table B.1: Industries at the Four-Digit SIC Level Ranked by Average Values of the Damage Indicator over 2003’-2005’ (410 Industries) Five most damaging industries SIC Industry 3312 Steel Works, Blast Furnaces (Including Coke Ovens), and Rolling Mills 2869 Industrial Organic Chemicals, NEC Air-Conditioning and Warm Air Heating Equipment 3585 and Commercial and Industrial Refrigeration Equipment 2911 Petroleum Refining 3443 Fabricated Plate Work (Boiler Shops) Five least damaging industries SIC Industry ρ 2448 Wood Pallets and Skids 0.108 3451 Screw Machine Products 0.107 2259 Knitting Mills, NEC 0.086 3952 Lead Pencils, Crayons, Artists’ Materials 0.044 2385 Waterproof Outerwear 0.017 Risk-related results: the toxicity, surrogate dose, and population components are multiplied to obtain a risk score for the Indicator Element. Pounds-based results: these results include only the pounds of releases reported to TRI. What should be noted is that the values are for comparative purposes and only meaningful when compared to other values produced by RSEI. Detailed information and related analyses can be found on the website: http://www.epa.gov/oppt/rsei.  93  ρ 3493272 2445291 1606799 1253189 955570.6  Appendix C  Appendix to Chapter 4 C.1  Background  The process of environmental policy creation is the same as those of other policies. A proposal may be introduced in Congress as a bill (or joint resolutions), which addresses the environmental issue and provides potential solutions for the problem. Each bill can be introduced by any member of Congress and has to go through several stages to come into laws. First, the bill is considered by standing committees whose jurisdictions are over this environmental issue. They investigate the details of these proposals, collect evidence and vote to decide whether the bill should be reported to the full house. Second, the full house debate and amend the bill once it is reported by the standing committees. The bill can not become law unless it is approved by both houses. Lastly, a bill that passes both houses becomes a law if the president agrees to sign it.102 Given this process of policy creation, it is regarded that special interest groups could exert political pressure on congress members to influence the policy outcome. Table 1 lists some of the environmental bills and whether they failed or passed.  102  If the President choses to take no action, the bill can also automatically becomes law after ten days. See details from http://en.wikipedia.org/wiki/United States Congress.  94  Table C.1: Selected Vote Description Senate/House Bill S S.2191 The Climate Security Act  Note  Failed H.R.5351 The Renewable Energy 2008 H Passed and Energy Conservation Tax Act 2007 S H.R.6 A Comprehensive Energy Legislation Failed H.R. 2643 The Interior2007 H Passed Environment appropriations bill 2005 S H.R. 2361 The Interior Appropriations Bill Passed 2005 S S.J. Res. 20 A resolution Failed Sources: ”National Environmental Scorecard”, League of Conservation Voters  C.1. Background  Year 2008  95  C.1. Background Description: Bill S.2191: To cut global warming pollution and drive rapid investment in the clean energy economy. The bill allowed major polluters to choose the most cost-efficient way to reduce pollution and buy pollution allowances to cover each ton of pollution that they continue to emit. Bill H.R.5351: To extend the tax credit for wind and other renewables by three years and reinstated expired credits for commercial and resident building. Bill H.R.6: To raise automobile fuel efficiency standards to 35 miles per gallon by 2020. To establish new energy efficiency standards for appliances and federal buildings etc. Bill H.R.2643: The Clean Air Act requires many of the largest emitters of toxic air pollutants to reduce their emissions by the maximum degree possible. The EPA proposed a rule change that would allow polluters to sidestep the requirement if they release less than 10 tons per year of a single pollutant or less than 25 tons per year of combined pollutants. This bill is to deny funds for implementing the proposed rule change. Bill H.R.2361: Chemical companies seeking pesticide approvals from EPA have submitted data from dozens of experiments in which human were intentionally dosed with these toxic chemicals. EPA have imposed a moratorium on considering such test. This bill is to create a one-year moratorium prohibiting EPA from using any of its funds to consider or conduct research that intentionally expose humans to pesticide. Bill S.J.Res.20: Coal-burning power plants are the largest U.S. source of mercury pollution. Rather than enforce the Clean Air Act, which requires all power plants to reduce their mercury emissions by 2008, the Bush Administration in March 2005 issued a rule that delays meaningful reductions for another two decades and encourages power plants to buy and sell mercury pollution credits. A bipartisan groups of Senators introduced a resolution to reject the EPA rule.  96  C.2. Data of environmental political contribution  C.2  Data of environmental political contribution  This appendix provides the data sources and methodology I employ in calculation of environmental political contribution. Part of descriptions are cited from the sources directly for accuracy.  C.2.1  Members of key House/Senate environmental committee/subcommittee  (1) The list of key House/Senate environmental committee/subcommittee is from League of Conservation Voters (http://www.lcv.org/president-andcongress/house and http:// www.lcv.org/president-and-congress/senate). The League of Conservation Voters is a national organization working full time to hold members of Congress accountable for their environmental votes. It provides description of the jurisdiction of the key committees/subcommittees with respect to environmental issues. (2) The U.S. Government Printing Office (http://www.gpoaccess.gov /cdirectory/browse.html) contains the congressional directory, which is the official directory of the U.S. Congress prepared by the Joint Committee on Printing. It presents short biographies of each member of the Senate and House, listed by state or district, and additional data, such as committee memberships. Two files are used to collect the names of the members of the key environmental committees/subcommittees: “Standing Committees of the House” and “Standing Committees of the Senate”. With these two data sources, I compile a dataset containing the names of the members of the key House/Senate environmental committees/subcommittees. (“Name file”) Take data in election cycle 2003-2004 for example (data over other election cycles are similar). There are 234 candidates in the key environmental committees/subcommittees out of 3553 candidates.  97  C.2. Data of environmental political contribution  C.2.2  Campaign contribution  In the website of the Federal Election Commission (http://www.fec.gov/finance/ disclosure/ftpdet.shtml), there are three files that are used in this work for each election cycle. I list them below and some descriptions are cited from the website directly for accuracy: (1) The committee master file contains one record for each committee registered with the Federal Election Commission. The file contains basic information about the committees, including the ID number the Commission assigned to the committee and the name of the committee. (2) The candidate master file contains one record for each candidate who has either registered with the Federal Election Commission or appeared on a ballot list prepared by a state elections office. The file contains basic information about the candidate, including name and the ID number assigned to the candidate by the FEC which is used in tracking campaign finance information about the campaign. (3) The itemized committee contributions file contains each contribution or independent expenditure made by a PAC, party committee, candidate committee, or other federal committee to a candidate during the two-year election cycle. It includes the ID number of the contributing committee and the ID number of the recipient. The committee master and candidate master files are used in conjunction with the itemized committee file to set up a relational database. They are merged according to the committee ID number and the candidate ID number respectively.  C.2.3  SIC code  There are various data sources used to search for the SIC code for each committee listed in the committee master file according to the name of the committee (usually a company name), e.g. COMPUSTAT, siccode.com, and websteronline.com. This data collection is difficult for the following reasons: multiple sources were used; and each committee may be assigned two or more SIC codes because of multiple production lines. However, it still 98  C.2. Data of environmental political contribution contains information that can be explored across industries. (“SIC file”) Take data in election cycle 2003-2004 for example (data over other election cycles are similar). Of the totally 1759 corporate PACS, the SIC codes are found for 1280 PACs, in which 470 PACs belong to the manufacturing group.  C.2.4  Data summary  In order to obtain the complete dataset, (a) I compile the “SIC file” with the committee master file and only keep the corporate committees. (b) I use the “name file” to label the candidates in the key committees/subcommittees in the candidate master file. These labelled candidates are kept for study. (c) The arranged files in (a) and (b) are merged with the itemized committee contributions file. Then, this dataset contains the following information for each corporate committee: the corporate committee’ ID, its SIC code, the contribution amounts of the candidates that are in the key committees/subcommittees and receive contributions from this corporate committee. This is the raw dataset that can be summarized at the four-digit industry level and scaled by the industry size to calculate the environmental political contribution intensity.  99  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0071414/manifest

Comment

Related Items