WHICH CAME FIRST, LAWS OR LOBBYISTS? AN EMPIRICAL INVESTIGATION OF ENVIRONMENTAL REGULATION AND INTEREST GROUP FORMATION. Bryan James Leonard

WHICH CAME FIRST, LAWS OR LOBBYISTS? AN EMPIRICAL INVESTIGATION OF ENVIRONMENTAL REGULATION AND INTEREST GROUP FORMATION. by Bryan Jame Leonard A thei ubmitted in partial fulfillment of the requirement for the degree of Mater of Science in Applied Economic MONTANA STATE UNIVERSITY Bozeman, Montana April 2012

COPYRIGHT by Bryan Jame Leonard 2012 All Right Reerved

ii APPROVAL of a thei ubmitted by Bryan Jame Leonard Thi thei ha been read by each member of the thei committee and ha been found to be atifactory regarding content, Englih uage, format, citation, bibliographic tyle, and conitency and i ready for ubmiion to The Graduate School. Dr. Dominic P. Parker Approved for the Department of Agricultural Economic and Economic Dr. Wendy Stock Approved for The Graduate School Dr. Carl A. Fox

iii STATEMENT OF PERMISSION TO USE In preenting thi thei in partial fulfillment of the requirement for a mater degree at Montana State Univerity, I agree that the Library hall make it available to borrower under rule of the Library. If I have indicated my intention to copyright thi thei by including a copyright notice page, copying i allowable only for cholarly purpoe, conitent with fair ue a precribed in the U.S. Copyright Law. Requet for permiion for extended quotation from or reproduction of thi thei in whole or in part may be granted only by the copyright holder. Bryan Jame Leonard April 2012

iv ACKNOWLEDGEMENTS I would like to thank the chair of my committee, Dr. Dominic Parker, for hi continual advice, upport, and encouragement over the duration of thi project, and for the hour he pent reading, editing, and reviing thi thei. I thank Dr. Randal Rucker for hi guidance on clarifying my reearch quetion, for hi ongoing guidance on the project, and epecially for hi punctiliou reading of my draft. I thank Dr. Rob Fleck for itting on my committee and making himelf available long-ditance for the defene; hi comment were invaluable. I would alo like to thank the Property Environment Reearch Center (PERC) for providing ummer upport in the form of their Summer Graduate Fellowhip. In addition to the financial upport imparted by PERC, the comment I received during eminar while there helped hape the direction of thi thei, informing both the empirical trategy and the theoretical model.

v TABLE OF CONTENTS 1. INTRODUCTION...1 2. THE LITERATURE...8 3. THEORY...17 4. NATIONAL LEGISLATION...24 5. STATE LEGISLATION...45 6. ROBUSTNESS CHECKS AND ANECDOTAL EVIDENCE...62 7. CONCLUSIONS AND FUTURE RESEARCH...69 REFERENCES CITED...71 APPENDIX A: Alternative Etimator...77

vi LIST OF TABLES Table Page 1. Variable Decription...25 2. National Summary Statitic...28 3. Wildlife Group and National ESA (Baeline)...31 4. Pollution Group and National Clean Air Act (Baeline)...33 5. National ESA Difference-in-Difference...37 6. National Clean Air Act Difference-in-Difference...39 7. Placebo Regreion of ESA on Pollution Group...41 8. Placebo Regreion of Clean Air Act on Wildlife Group...42 9. Difference-in-Difference Placebo Regreion of ESA on Pollution Group...43 10. Difference-in-Difference Placebo Regreion of Clean Air Act on Wildlife Group...44 11. State-Level Summary Statitic...48 12. ESA Difference-in-Difference...53 14. RPS Difference-in-Difference...54 15. ESA Difference-in-Difference-in-Difference...57 16. Air Difference-in-Difference-in-Difference...60 17. RPS Difference-in-Difference-in-Difference...61 18. Placebo Difference-in-Difference of ESA on Clean Group...63 19. Placebo Difference-in-Difference-in-Difference of ESA on Clean Group...64

vii LIST OF TABLES CONTINUED Table Page 20. Placebo Difference-in-Difference-in-Difference of Air on Wildlife Group...66 12A. ESA Difference-in-Difference (OLS)...78 12B. ESA Difference-in-Difference (Poion)...79 12C. ESA Difference-in-Difference (Zero-Inflated Poion)...80 13A. Air Difference-in-Difference (OLS)...81 14A. RPS Difference-in-Difference (OLS)...82 14B. RPS Difference-in-Difference (Poion)...83 14C RPS Difference-in-Difference (Zero-Inflated Poion)...84 15A. ESA Difference-in-Difference-in-Difference (OLS)...85 15B. ESA Difference-in-Difference-in-Difference (Zero-Inflated Negative Binomial)...86 15C. ESA Difference-in-Difference-in-Difference (Poion)...87 15D. ESA Difference-in-Difference-in-Difference (Zero-Inflated Poion)...88 16A. Air Difference-in-Difference-in-Difference (OLS)...89 16B. Air Difference-in-Difference-in-Difference (Zero-Inflated Negative Binomial)...90 16C. Air Difference-in-Difference-in-Difference (Poion)...91 16D. Air Difference-in-Difference-in-Difference (Zero-Inflated Poion)...92 17A. RPS Difference-in-Difference-in-Difference (OLS)...93 17B. RPS Difference-in-Difference-in-Difference (Zero-Inflated Negative Binomial)...94

viii LIST OF TABLES CONTINUED Table Page 17C. RPS Difference-in-Difference-in-Difference (Poion)...95 17D. RPS Difference-in-Difference-in-Difference (Zero-Inflated Poion)...96

ix LIST OF FIGURES Figure Page 1. Environmental Interet Group and Environmental Law 1950-2008...5 2. The Policy Space...18 3. Wildlife Group and Conervation Group...34 4. Pollution Group and Conervation Group...34 5. Timing of State Law...47 6. Hitogram of Wildlife New_Group...49 7. Hitogram of Pollution New_Group...49 8. Hitogram of Clean New_Group...50 9. Hitogram of Conervation New_Group...50 10. ESA Difference-in-Difference-in-Difference Incidence Rate Ratio...58

x ABSTRACT Nonprofit organization and interet group play a ubtantial role in the United State; there were over 260,000 nonprofit and nearly 5,000 political action committee at work in 2010. Conventional widom ugget that many of thee group have formed with the goal of influencing the paage of new legilation that i favorable to their interet. Thi thei contribute to our undertanding of interet group by providing evidence of the oppoite direction of cauation within the context of environmental legilation. I develop a theory for why new environmental regulation caue new environmental interet group to form, rather than vice vera. I tet the theory with a novel panel data et of federal and tate environmental law and of the formation date of wildlife, pollution, and conervation oriented interet group ince 1950. My empirical tet combine difference-in-difference (and difference-in-difference-in difference) etimator with event tudy method. The reult at the national level how that more group formed during a window of time before and after the paage of the Clean Air and Endangered Specie Act. The reult at the tate level, which are in many way more credible etimate, how that tate-level interet group were more likely to form during a window after new legilation wa paed than before. Overall, the reult ugget law that leave a large portion of deciion making to a bureaucracy create new lobbyit. Thi i a reult that ha been uggeted by ome environmentalit and economit but never empirically teted before thi thei.

1 INTRODUCTION Nonprofit organization and interet group play a ubtantial role in the United State. There were over 260,000 nonprofit operating in the United State in 2010, in addition to nearly 5,000 political action committee at work in 2009 (U.S. Cenu Bureau 2012). A the cope of tate and federal regulation ha expanded, group have formed to influence regulation, to capture newly-created benefit, and to help individual adapt to the ever-changing adminitrative rule and regulation of government program. Nonprofit organization collected over $1.4 trillion in revenue in 2007 alone, a 69% increae from the year 2000 (U.S. Cenu Bureau 2012). Of that $1.4 trillion, $14.8 billion went to environmental group, while $78.6 billion went to public and ocietal benefit group. Political action committee contributed a combined $378 million to federal Houe and Senate campaign in 2007 and 2008. While the number of PAC ha increaed by 46% ince 2000, contribution to congreional election have nearly doubled over that ame period (U.S. Cenu Bureau 2012). The role of interet group in our ociety i hard to ignore, given that total pending by nonprofit in 2007 wa equal to nearly 10% of GDP. To undertand the impact of interet group and nonprofit on ocial welfare, it i ueful to undertand what caue interet group to form and grow. Thi thei contribute to that undertanding. In particular, I theoretically and empirically analyze the formation of new environmental interet group United State ince the 1950. I provide evidence that new law caue new interet group to form. Thi finding contrat with conventional widom, which eem to take a given the notion that group form to

2 influence the paage of law. My finding that law caue lobbying ha important implication for how we think about the role of interet group and nonprofit organization in modern, regulated economie. If the primary role of interet group i to get new legilation paed, it i reaonable to expect that there hould be more new interet group jut before the paage of a law than there are after. Thi theory ugget that interet group form in repone to changing public entiment and then puh for new legilation that embodie thi entiment. Once the law i paed, group are le likely to form becaue the demand for a particular policy outcome will have been appeaed. The hypothei that interet group primarily lobby to pa new law ugget that there hould be fewer new interet group forming jut after a major legilative victory than before. A an example, if the primary reaon for wildlife interet group formation i to eek wildlife protection law, then we would expect to ee very few, if any, wildlife group form jut after the paage of the Endangered Specie Act. To many, the incluion of pork barrel and graft in federal legilation offer proof poitive that pecial interet group exert influence on new legilation. On the urface, it eem that there would be relatively little left for interet group to accomplih once a major legilative victory had been won. The increaingly open-ended nature of legilation may actually mean that interet group expect to gain more jut after a major law i paed than jut before. Interet group certainly exert preure to pa new legilation, but they may play an even larger role in the interpretation, implementation, and enforcement of newlypaed law. While attempt to influence the paage of legilation mut collect enough

3 vote to win the day, the bureaucratic apparatu ultimately reponible for implementing new law may provide much more fertile ground for influence. The Economit (2012) decribe jut how complex thi new playing field can be, decribing the recently-paed Dodd-Frank financial regulation, At 848 page, it i 23 time longer than Gla-Steagall, the reform that followed the Wall Street crah of 1929. Wore, every other page demand that regulator fill in further detail. Some of thee clarification are hundred of page long. Jut one bit, the Volcker Rule, which aim to curb riky proprietary trading by bank, include 383 quetion that break down in to 1,420 ubquetion of the 400 rule it mandate, only 93 have been finalized. (2/18/12, 9) When uch a ignificant portion of a policy ubtantive content i determined after it leave the legilature and i igned into law, it i plauible that group may form after major law are paed in order to help hape their ultimate implementation. In the context of environmental regulation, Gregg Eaterbrook (2012) ha argued, Becaue carbon capand-trade ytem are inherently uper-complex, they are nearly certain to be gamed defeated by gimmick, litigation, and pecial-favor lobbying ( Why We Need a Carbon Tax ). The popular pre ha already begun reporting on dramatic increae in lobbying by healthcare interet group ince the paage of the Patient Protection and Affordable Care Act and by group repreenting Wall Street in the wake of Dodd-Frank. 1 Though thi evidence i purely anecdotal, it lend credence to the idea that major legilation may actually increae the rate of interet group formation. While any ignificant legilative hift could plauibly create incentive for group formation, environmental law are epecially likely to do o. Environmental legilation 1 See www.wahingtontime.com/new/2011/jan/5/bet-health-care-political-pull-can-buy/ and http://dealbook.nytime.com/2011/08/01/wall-treet-continue-to-pend-big-on-lobbying/

4 ha high take; many wildlife population and ecoytem are endangered, while environmental protection can create ignificant cot for private firm and individual. Environmental regulation often require the undertanding of cientific reearch, uually at a level of detail beyond the general knowledge of legilator. Legilator may pa trong but vague environmental protection law to atify their contituent, while leaving the actual regulatory detail for later enumeration. Thi approach afford cientit the opportunity to work with bureaucrat and inform their deciion and imultaneouly open the door to interet group preure. Figure 1 how the natural log of the number of environmental interet group exiting in year t and the natural log of the number of major federal environmental law on the book in year t-1, taken from the Animal Law Center at Michigan State Univerity. It i evident that environmental law and environmental interet group have followed a imilar trend over the lat fifty year. The imple correlation coefficient between Environmental_Group and Environmental_Law_Lagged i 0.9745. Thi relationhip could upport either the conventional widom or the hypothei of thi thei. It may be that the increaing number of environmental regulation i the reult of an increae in interet group and/or that the increaing number of environmental law ha caued more interet group to form.

5 Figure 1: Environmental Interet Group and Environmental Law 1950-2008 Economit have tudied the effect of interet group for decade, beginning with a narrow focu on indutry lobbyit (Stigler 1971) and expanding to conider topic uch a the effect of interet group on liting under the Endangered Specie Act (Ando 1999, 2001, 2003). Although there i a large literature within economic on the effect of interet group, their origin have largely been overlooked. 2 The preumption eem to be that group primary motivation in forming i to lobby for new legilation, but thi ha not been teted by careful empirical tudy. Several author have uggeted theoretical 2 McGuire (1976), Crain et al. (1991) Poole and Roenthal (2002), Benneden and Feldmann (2002), Hamlin and Jenning (2004), and Gray et al. (2005) develop theoretical model of interet group formation, but do not tet them empirically and generally do not conider how pecific law might affect group incentive.

6 reaon why group may form in repone to legilation, or why legilator may even deign legilation to incite group formation, but none of thee theorie have been teted empirically. What economit have etablihed empirically i that interet group can have a ignificant effect on policy, and that the expected benefit of lobbying activitie affect interet group choice, at the margin. Thi thei eek to unite thee thread in the empirical literature with the theoretical literature that ugget why group may form after law rather than before. To craft the variou finding from the empirical literature on interet group into a tetable hypothei about the role of new law, I develop a theoretical model of interet group formation. According to the model, interet group will form if the expected benefit of interet group activity exceed the fixed cot of formation. The paage of a new environmental law may decreae the abolute magnitude of benefit becaue any new policy outcome will be a maller improvement over the tatu quo when compared to the pre-law tate of the world, but a new environmental law create a bureaucratic apparatu that may make it eaier for group to influence government. The relative magnitude of thee contrating effect will ultimately determine whether or not a new law increae the expected benefit of interet group activity. Thi model allow law with different attribute to affect interet group differently. Uing data on environmental interet group formation at the tate and national level from 1950-2008, I examine whether interet group are more likely to form jut after a law affecting their area of concern i paed than jut before. The ditribution of the data neceitate the ue of a count model. Variation in the focu of different group

7 allow for the ue of a difference-in-difference trategy at the national level. National regreion how a ubtantial increae in wildlife group formation after the paage of the Endangered Specie Act and a imilar increae in pollution group formation after the Clean Air Act; however, both model alo how an increae in group formation before the law. Exploiting variation in the timing of different tate policie, a difference-indifference-in-difference trategy i employed at the tate level. State-level regreion how an increae in wildlife group formation after tate Endangered Specie Act, but are not concluive for tate air pollution control law or renewable portfolio tandard. Robutne check (placebo regreion) confirm the tate-level reult. The ret of thi thei i organized a follow: Chapter 2 contain a ummary of the economic literature relating to interet group formation, environmental regulation, and broader theorie of lobbying within the realm of public choice. The theoretical model and the condition under which a new law would caue group formation are preented in Chapter 3. In Chapter 4 I decribe the data and dicu the reult of regreion run at the national level. I decribe the data further and preent the tate-level reult in Chapter 5. In Chapter 6, I offer robutne check and anecdotal evidence in upport of the main hypothei. I offer concluding remark and uggetion for future reearch in Chapter 7.

8 THE LITERATURE There i little economic literature that empirically examine interet group formation. The theoretical model of Becker (1983), Suntein (1993), and Hamlin and Jenning (2004) all define a role for the expected benefit of lobbying effort in the deciion calculu of whether or not to form a group in the firt place. The actual ability of interet group to influence policy outcome i perhap the central concern for thinking about the expected benefit of policy. Therefore, undertanding the extent to which interet group have a real effect on policie, and under what condition, i eential to predicting whether or not group will form under given circumtance. There i a rich literature on the effect of interet group on policie that inform an analyi of the expected benefit of lobbying activity. Daniel McFadden (1976) tet variou hypothee about deciion rule ued by policy-maker and find that bureaucrat are driven motly by the benefit-cot ratio of a given policy, and that interet group do not have a ignificant effect. Friedrich Schneider and Jorg Naumann (1982) tudy interet group in Switzerland from 1960 to 1978 and find that interet group have a ignificant effect on the probability of a ye vote on a referendum and on how funding i allocated by government. Teke (1991) explore how interet group behaved during the deregulation of the telecom indutry in the mid-1980, finding that interet group have a ignificant effect on further deregulation and that group adjut their level of preure and overall approach to lobbying the exiting level of regulation. Borge et al. (1995) how that interet group actually low the peed of diburement of public fund at the local level by competing for carce reource and

9 complicating the government deciion-making proce. Brook et al. (1998) et up a three equation ytem to model and tet the effect of PAC on U.S. legilator upport of the U.S. ugar program and find that financial contribution from PAC do influence legilator vote for ugar policie. They alo find that PAC pend more money to lobby congremen who have a higher propenity to vote for favorable policie. Tanguay et al. (2004) ue OLS on a pooled cro ection from 22 OECD countrie to etimate the effect interet group have on the everity of environmental policy, trying to undertand whether legilator eek to maximize public benefit or repond to peronal incentive. Uing a generalized leat quare approach, Tanguay et al. find that interet group preure ha a much larger effect on the everity of recycling policy than doe the expected public benefit. Cropper et al. (1992) find that interet group comment have a large effect on the EPA probability of banning peticide, and that group are more likely to exert preure on propoal favoring their goal than on propoal inconitent with their interet. Crone and Tchirhart (1998) find that interet group comment have a ignificant effect on the deignation of wilderne area by the National Foret Service. Uing roll call vote from 1981 and 1988, Hird (1993) find that ideology, legilator elf-interet, and the preence of environmental lobbyit all have a tatitically ignificant effect on the election of Superfund ite. Confirming thee reult, Sigman (2001) ue a hazard model with fixed effect and find that the EPA prioritize Superfund project baed on interet group preure and that public health concern aociated with a ite do not have a ignificant effect.

10 Amy Ando ha a erie of paper that examine interet group activity related to the federal Endangered Specie Act. For each of the three tudie, the data come from interet group comment on endangered pecie liting between 1989 and 1994. Ando (1999) find that public upport (in-favor comment by group and individual) increae the peed with which a pecie move through the liting proce and the probability that the pecie i ultimately lited. Ando (2001) find economie of cope in endangered pecie protection. In particular, her reult how that there i generally more upport than oppoition for liting a given pecie and that peronal income doe not affect upport for liting. Ando (2003) tet whether group for and againt the liting of pecie engage in trategic behavior, with three important finding: interet group upport for a pecie liting greatly increae the probability that the pecie will be lited, interet group exert more preure when the expected benefit of conervation i greater, and interet group do not directly compete or engage in obervable trategic behavior. Ando reult provide upport for the idea that the expected benefit of interet group preure play a ignificant role in haping group deciion; however her data only look at a ubet of lited pecie over a ix year period. To my knowledge, there are no empirical analye of interet group formation within the United State. The literature i not totally ilent on group formation, however. A erie of paper examine interet group formation on a cro-country bai, teting variou prediction about what type of broad ocio-economic and intitutional factor make group more likely to form in ome countrie than in other. Notably, the literature on national-level group formation i devoid of any formal model, intead teting an ad

11 hoc lit of hypothei about macroeconomic variable that might affect interet group. Much of thi literature relie on cro-ectional data, or, at bet, very limited panel. Though cro-country comparion of group formation are intereting, they are not particularly ueful for guiding policy within the United State. If interet group are to be encouraged or avoided, pecific intitutional factor cauing group formation mut be examined. Thi thei expand upon the literature by examining the effect of pecific legilation on group formation in an extended panel context. The firt empirical work to tudy the determinant of interet group formation wa Peter Murrell 1984 paper, An examination of the factor affecting the formation of interet group in the OECD countrie. Murrell outline and tet eleven ditinct hypothee uing a ingle year of obervation acro 24 OECD countrie. After uing a variety of functional form and ingle and joint hypothei tet, Murrell conclude that higher population, more decentralized government, and political tability are each ignificant determinant of interet group formation. At the ame time, Murrell reject the hypothei that pecific feature of the political ytem have an impact on whether or not group form. Murrell reult are inconcluive a to whether or not ocioeconomic development, number of indutrie, and diverity are important determinant of interet group formation. Building on Murrell work and eeking to provide alternate meaure for ome explanatory variable, Ivo Bichoff (2003) conduct hi own tudy of interet group formation. Bichoff tudy contain data from only 21 OECD countrie, and though he tet many of the ame hypothee a Murrell, he meaure everal key variable

12 differently. Wherea Murrell meaure political tability a the length of time ince the onet of political and economic freedom, Bichoff meaure tability a the number of year ince the mot recent turmoil, which include war and foreign occupation. In mot other repect, the tudie are quite imilar. Bichoff generalized leat quare approach yield reult that ugget markedly different finding from Murrell. Bichoff conclude that political tability doe not have an impact on interet group formation, favoring intead the importance of economic development a a determinant. Coate et al. (2007) offer a reconciliation of the prior finding of Bichoff and Murrell, attributing the lack of agreement between Bichoff and Murell reult to two factor. Firt, they cite the relatively mall ample ize ued in both tudie. Coate et al. argue that data from 21 to 24 countrie for a ingle year provide an inufficient number of obervation to conduct the kind of multiple hypothei tet that Murrell explore. Secondly, Coate et al. point out the aforementioned difference in how Murrell and Bichoff meaure political intability, arguing that there are in fact two ditinct variable at work. Coate et al. argue that Bichoff turmoil variable i not necearily meauring political tability. They point out hitorical example of countrie that have either been engaged in foreign war (like the United State recent occupation of Iraq) or occupied by foreign countrie (uch a the American colonie) and yet remained politically table. They alo offer example of the oppoite cae, uch a the Jim Crow law in the 1960, where intitutional change occurred without violent turmoil. Coate et al. go on to criticize Murrell meaure a well, citing the example of Albania whoe initial takeoff date i 1919, obcuring the fact that they converted to communim in 1944.

13 Coate et al. analyi coalece into a rejection of both Murrell and Bichoff variable a tand-alone meaure for political tability. After offering a critique of previou tudie of interet group formation, Coate et al. conduct their own empirical etimation. To improve upon the mall ample ued by their predeceor, Coate et al. ue a data et of 140 countrie over 5 year. They argue that thi data et provide ufficient obervation to conduct multiple hypothei tet with tatitical rigor. To reolve the meaurement error problem that plagued previou tudie, the author develop a variable that meaure the length of time ince the mot recent regime change, contitutional hift, or political independence. In addition, the author include Murrell takeoff variable and Bichoff turmoil variable. Other variable included in their regreion are income, agricultural hare, urban hare, import hare, newpaper circulation per capita, number of telephone per capita, amount of mail per capita, a democracy index, number of political partie, number of indutrie, population, government pending (a a fraction of GDP), linguitic fractionalization, and religiou fractionalization. The author are alo concerned about the endogeneity of income, government pending, and import hare, and accordingly intrument with lagged life expectancy, lagged hare of government expenditure, and the hare of import in GDP, repectively. Coate et al. find that the number of interet group increae with economic development, regime tability, population, and the ize of government. John Hanen (1985) conduct a more micro-level analyi of what induce individual to join and form group. Hanen emphaize the ubjective nature of the expected cot and benefit of interet group activity. He model memberhip a a

14 function of expected benefit, expected cot, the probability of a public good being provided without memberhip, information, available reource, and rik attitude. Uing memberhip data from the American Farm Bureau Federation, the League of Women Voter, and The National Aociation of Home-Builder, Hanen how that perceived political benefit are the driving factor in the deciion to form or join a group. He alo how that individual are more reponive to change in other parameter during what he call threatening time; when the probability of a good being available outide memberhip i very low. In hi concluion he ugget that thee factor could alo influence the upply of fund available to form interet group by affecting the willingne of donor to contribute. Anecdotal and theoretical contribution to the literature ugget that new law hould be included a determinant of interet group formation, depite the fact that the extant empirical literature doe not conider the effect of pecific legilative change. A ubet of the public choice literature focuing on the relationhip between legilator and bureaucrat find that legilator may actually deign law to incite group formation, for a variety of reaon. Anecdotal obervation by economit and environmentalit alike lend credence to the idea that environmental interet group often form in repone to major environmental legilation. Thi literature provide the bai for the theoretical model and aociated prediction in thi thei. McCubbin and Schwartz (1984) develop a theory of fire-alarm overight, where legilator rely on interet group to alert them to bureaucratic mitep. McCubbin and Schwartz explain, Congre etablihe a ytem of rule, procedure,

15 and informal practice that enable individual citizen and organized interet group to examine adminitrative deciion, to charge executive agencie with violating congreional goal, and to eek remedie from agencie, court, and Congre itelf (166). The author pecifically cite environmental law a an area where thi phenomenon i likely to occur. Fred McCheney (1997) point out that politician can extract rent from group by threatening regulation or harmful legilation, only withholding putative regulation once payment i received. In order for thi cheme to work, intereted group mut exit in the firt place. Legilator may deign law which caue group to form o that they can later extract rent from thoe ame group. Both of thee theorie imply that legilator will intentionally craft law in uch a way that increae potential interet group formation incentive. Elliot et al. (1995) conider more benevolent reaon why legilator may create incentive for group formation. Elliot et al. argue that Congre pae intentionally vague law to auage public concern over environmental quality, relying on interet group to inform the bureaucracy in etting actual policy. Given the lack of direct Congreional control over agencie in the executive branch charged with carrying out their mandate, Congre ha a trong incentive to deign law facilitating the formation of New_Group by contituent to encourage policie that are in line with original legilative intent. Elliot et al. develop a ix tage model of law formation, where interet group help determine final policy, and then ue the Clean Air Act a an illutrative example of their theory. Schoenbrod (2005) offer upport for thi idea, emphaizing legilator lack of cientific knowledge. In effect, Congre create a property right to a

16 certain ubet of the policy pace, inviting interet group to come and claim it. Tober (1989) offer anecdotal evidence to upport thee idea, citing group that formed in repone to legilation affecting bobcat in one cae, and condor in another. The literature provide inight into the effectivene of interet group activity, potential influence on group deciion-making, and the legilative demand for interet group formation, but remain ilent on the effect of legilator action on group formation. The literature ugget that interet group can hape regulatory outcome by exerting preure on legilator, and in particular, bureaucrat. Empirical evidence alo ugget that interet group formation i a function of ocio-economic factor on a crocountry level and that group adjut the level of preure exerted in repone to expected benefit of lobbying at the group level. The literature alo argue that legilator may have an underlying demand for interet group formation, for reaon of bureaucratic overight, rent extraction, or information gathering. Baed on thee finding, it i plauible that new environmental law, which create broad regulatory power, give incentive for interet group to form by increaing the expected benefit from lobbying activity.

17 THEORY The theoretical model in thi thei i deigned to formalize the idea that group may have greater incentive to form after major legilation i paed when compared to etting lacking legilation. An individual will chooe to form a group if the expected benefit of group formation are greater than the cot incurred to that individual. The model aume that the expected benefit are determined by the expected benefit of lobbying activity through the group; there are not other benefit of forming a group. While the tructure of the model implie that the exitence of a law will almot alway increae the expected benefit of lobbying, it alo pecifie condition under which the expected benefit are higher without a new law. Equation 1 give the baic tatement of the model; group will form if the expected benefit of formation,, exceed the fixed cot, C. Equation 2 define the expected benefit of lobbying a the probability of ucce (P) multiplied by the oughtafter benefit (B). (1) (2) The potential benefit of formation depend primarily on the ize of the deired policy change. The potential level of environmental regulation can be thought of a exiting in a two-dimenional policy pace, or pectrum. The potential policy pace in quetion range from zero to one, where zero indicate no regulation and one indicate full regulation/protection. In the context of endangered pecie, full protection would imply that it i illegal to kill or capture a given pecie. In a pollution control context, full

18 protection would imply a governmental mandate that air quality be equal to whatever level the group in quetion deire. The benefit of ucceful lobbying can then be repreented by equation 3. (3) The variable L i binary. It i equal to one if a law currently exit in a given focu area (e.g., wildlife protection, air quality, etc.), but zero otherwie. The variable locate the exiting law in the policy pace. Thu, the upper-bound potential benefit of lobbying i the magnitude of the change between the tatu quo and the deired tate multiplied by the variable, which i the intenity of the individual preference. Figure 2 give a graphical repreentation of the policy pace with a potential law. The location of indicate the trength of the exiting law and determine the total magnitude of the benefit aociated with moving to full protection. The better protected a pecie i in the tatu quo, the maller the added benefit of achieve full protection for that pecie. Figure 2: The Policy Space

19 The probability of ucceful lobbying, P, depend upon the ize of the potential policy change, the malleability of the bureaucracy ( ), and the need to achieve of legilative majority ( ), repreented by equation 4. (4) The ize of the policy change i repreented a a cubic in the denominator of P o that the probability of ucceful lobbying decreae at an increaing rate a the deired policy change grow larger. Thi mean that advocating a mall change i much more likely to meet with ucce than puhing for a large change. The variable encapulate the amount of the policy pace within that can be captured by lobbying; it indexe the proportion of an exiting law (if a law exit) that i left to interpret, implement, and enforce at the time of it paage and the degree to which bureaucrat can be influenced. 3 Value of cloe to one indicate a vague law and/or eaily-influenced bureaucrat, implying that group can have a large impact on how the law i actually implemented. Value of cloe to zero indicate that the enacted law i highly-pecific in nature and/or that bureaucrat cannot be eaily influenced. For example, a pollution control law that detail acceptable level of pollution and punihment for non-compliance would have a low value of. The variable i the number of legilator required to obtain a ufficient majority to pa legilation. When legilation i being ought and a legilative majority mut be achieved, i high. The invere of i additive in P becaue it repreent an extra 3 To ue Barzel (1997) lexicon, indicate how much of the policy pace i in the public domain.

hurdle that mut be overcome to achieve a ucceful lobbying outcome. When a law 20 already exit and lobbying i focued on particular bureaucrat, doe not affect the probability of formation, hence the expreion goe to zero when L i equal to one. Combining equation 1-4, equation 5 give the new condition for group formation. (5) When no law exit, the expected benefit of lobbying are given by equation 6. (6) When a law doe exit, the expected benefit of lobbying are given by equation 7. (7) The paage of a law will increae the expected benefit of lobbying via a new interet group if the inequality in equation 8 hold. (8) The inequality in equation will hold if, implying that the paage of a law increae the expected benefit of lobbying and may incite formation. If a law i perfectly delineated or bureaucrat cannot be influenced at all, then the expected benefit of lobbying would actually be higher before the law than after (becaue i equal to zero). The expected benefit of lobbying before the paage of a law may alo be higher than after the paage if both and are relatively mall. Thi model doen t give precie prediction about when the expected benefit will necearily exceed the

21 formation and operation cot, but it doe how that a new law will increae the benefit, and thu make formation more likely. The model preented above indicate that the paage of a new law will increae the expected benefit of lobbying under mot condition, but the particular characteritic of the law in quetion can affect thi reult. Equation 9 through 11 give the alient comparative tatic, taken when L i equal to one. (9) (10) (11) The comparative tatic prediction for i poitive. Ceteri paribu, larger value of increae the probability of ucce becaue new law that are cloer to perfect protection of the environment make movement cloer to perfect protection more likely (advocating for a mall policy change i eaier than advocating for a large one). The comparative tatic prediction for i poitive, o le precie and pedantic law which facilitate influence on bureaucrat are more likely to encourage group formation. Increae in the intenity of preference, have a poitive effect on the expected benefit of formation. Not all law are created equal, and ome law are more likely to caue group formation than other. In particular, preciely enumerated law with low value of will increae the expected benefit of lobbying bureaucracie to a leer extent than law with high value of. For example, Endangered Specie Act have high value of becaue they et up an apparatu for protecting pecie but leave the actual liting for later, and

are more likely to caue group formation than regulatory law which allow for le 22 uncertainty. In contrat, pollution control law have lower value of becaue they tend to leave little actual policy to be crafted after the paage of the law, and thu may not increae the expected benefit of lobbying at all. If individual concern for environmental iue are auaged by the paage of a law, then will fall and the expected benefit of lobbying will be low. Thi lulling effect i more likely to occur in the cae of broad policy iue like pollution and global warming. For group that care about particular pecie though, will remain high and the benefit of having a general pecie-protecting infratructure will not diminih their concern for protecting their particular pecie of concern. Thi model attempt to decribe an underlying continuum of expected benefit that ultimately guide a dicrete choice: whether or not to form an interet group to lobby legilature and/or bureaucracie. The model doe not definitively predict that a new law will caue more group to form. Rather, the model how that the paage of new law increae the expected benefit of lobbying for a variety of reaon, thu making formation more likely, but not certain. The ize of particular parameter dictate the extent to which new law will increae the expected benefit of lobbying. Baed on the tructure of the model and the characteritic of environmental law, I offer the following prediction. There will be more new interet group after the paage of Endangered Specie Act, which have a high and hould not affect the ize of.the paage of air pollution law and renewable portfolio tandard will have a negligible or negative effect

23 on the likelihood of group formation due to low value of and a potential drop in from the paage of thee law.

24 NATIONAL LEGISLATION Data on interet group formation come from three ource: The National Wildlife Federation Conervation Directory (2008), The Encyclopedia of Aociation (2010), and The National Directory of Nonprofit Organization (2010). After eliminating duplicate, the founding date and primary focu given for each environmental/conervation organization and the addree of the organization headquarter were ued to generate a variable that i a meaure of the number of new interet group, by focu, in each tate in each year from 1950 to 2008. To conider the effect of national legilation on interet group formation, group are aggregated up to the year level. Uing directorie from the pat four year fail to meaure the formation of group that no longer exited by 2008. If the mortality rate of thee group i contant over time, then I am underetimating the number of new group to a greater extent in the earlier year in the data. Tober (1989) ugget that group tend to perit and broaden their focu, which hould help mitigate thi ource of meaurement error. 4 Table 1 give decription of the variable. 4 Whenever poible, I include year fixed effect, which further mitigate any meaurement error that occur by year.

25 Table 1: Variable Decription Variable Name Log_Environmental_Group Decription Total number of environmental group in the United State in year t Log_Environmental_Law_Lagged Total number of federal environmental law in the United State in year t-1 New_Goup Number of new environmental group in tate in year t with focu i ESA Binary equal to 1 for each year after an Endangered Specie Act i Paed ESA_tYr_Pot Binary equal to 1 for t year after an Endangered Specie Act i paed ESA_tYr_Pre Binary equal to 1 for t year before an Endangered Specie Act i paed Air Air_tYr_Pot Binary equal 1 for each year after an air pollution law i paed (for national regreion, the Clean Air Act) Binary equal to 1 for t year after an air pollution law i paed Air_tYr_Pre Binary equal to 1 for t year before an air pollution law i paed RPS Binary equal to 1 for each year after a renewable portfolio tandard i enacted RPS_tYr_Pot Binary equal to 1 for t year after a renewable portfolio tandard i enacted RPS_tYr_Pre Binary equal to 1 for t year before a renewable portfolio tandard i enacted Wildlife Binary equal to 1 for wildlife group Pollution Binary equal to 1 for pollution group Clean Binary equal to 1 for pollution group and natural reource group LNPCI Natural log of per capita income by tate by year LNPOP Natural log of population by tate by year Trend Linear time trend Trend2 Quadratic time trend

26 Figure 1 how the natural log of the total number of environmental interet group and total federal environmental law per year from 1950 to 2008. Environmental law are lagged by one year to help viually examine the relationhip between law and ubequent interet group formation; the figure depict the number of group in year t and the number of law in year t-1. There ha been a teady increae in the number of federal environmental law ince 1950, a Congre ha paed legilation acro a broad pectrum of environmental concern including wildlife protection, air pollution, water pollution, wilderne preervation, and hazardou wate cleanup. At the ame time, the number of environmental interet group ha rien. It i unclear from the figure whether there i any caual relationhip between the number of law and the number of interet group, though the two have certainly followed a imilar trend. To the extent that a caual relationhip exit, it i plauible either that more law lead to more group or that more group lead to more law. The exitence of a long-run relationhip between Environmental_Group and Environmental_Law_Lagged preented in figure 1 doe not etablih a caual relationhip between thee two variable. If new law directly caue new group to form, then the relevant regreion pecification i one that etimate the effect of new law on the number of new group. Even if law do give rie to new interet group, it i very likely that ome group form to purue the paage of new legilation. Public concern for the environment, change in cientific knowledge, and attitude about the role of government may be correlated with both group formation and the paage of legilation. Failing to account for the endogeneity of environmental legilation and for omitted

27 variable will bia the reult of imple etimation of the relationhip between law and group. Moreover, the latter half of the twentieth century aw landmark legilation acro multiple dimenion of environmental policy, including the Endangered Specie Act, the Clean Air Act, the Clean Water Act, the Comprehenive Environmental Repone, Compenation, and Liability Act (Superfund), and other. Some year aw everal law paed, while other year aw none. Potential environmental group may focu on a variety of iue, and not all environmental law will affect them in the ame way. In order to actually meaure the treatment effect of new environmental law, the focu of the law and of any new group mut be included in the analyi. Group have been coded a either focuing on wildlife, pollution, natural reource, or broader conervation effort. Each group in the National Directory of Nonprofit Organization i categorized according to an IRS activity code, uch a wildlife preervation. The fact that each group reported only one activity allow me to clearly aign a focu to each group. For thi reaon, focue for group appearing in multiple directorie were drawn from the National Directory of Nonprofit Organization, where poible, then the Encyclopedia of Aociation, then the Conervation Directory. The group in the Encyclopedia of Aociation each have a primary ubject lited, mot of which are unique. For group that lited more than one ubject, the firt lited ubject wa ued to determine their focu. 5 Group from the Conervation Directory lit multiple focue, o I conulted each group name and webite to determine their primary area of concern. Table 2 contain ummary tatitic for variable ued in national regreion. 5 Group that lit more than one ubject typically lit more refined focue ubequent to the firt entry: e.g. a wildlife protection group might pecify the particular pecie they eek to protect by liting them under Subject.

28 Table 2: National Summary Statitic Variable Ob Mean Std. Dev. Min Max All New_Group 59 83.72881 57.8842 12 266 Wildlife New_Group 59 13.54237 10.29784 0 40 Pollution New_Group 59 4.084746 4.457803 0 21 Conervation New_Group 59 30 20.1768 3 82 Log_Environmental_Group 59 2898.881 1839.02 580 5732 Log_Environmental_Law 59 50.54237 29.66672 5 87 LNPCI 59 9.794552 0.394502 9.130864 10.41021 LNPOP 59 12.31314 0.197545 11.92955 12.62751 Many environmental concern overlap, and ome group focu on a broad range of iue, but thoe group focuing primarily on either pecie protection or pollution prevention are relatively eay to ditinguih, making them ideal treatment group. In particular, natural reource group are harder to ditinguih from general conervation group than are wildlife and pollution group. Wildlife group and pollution group are alo the eaiet to ditinguih from one another, further uggeting their ue a treatment group. The natural choice of treatment for wildlife group i the Endangered Specie Act, paed in 1973. The Clean Air Act, paed in 1970, i ued a the treatment for pollution group. Thee law repreent two of the mot ignificant piece of federal environmental legilation enacted in the twentieth century. Both law alo have tatelevel counterpart that can be clearly ditinguihed, which will aid in the analyi in the next chapter. The Endangered Specie Act dramatically increaed the federal government role in wildlife protection. Though the act wa preceded by the Endangered Specie Preervation

29 Act in 1966 and the Endangered Specie Conervation Act in 1969, it wa the firt act to give the federal government authority over the tate in the name of intertate commerce to enforce it claification of pecie a endangered. The act alo left a ignificant amount of actual policy to be determined ex pot; it et up an apparatu for protecting pecie but did not create a definitive lit of which pecie would be protected. Thi dramatically decreaed the cot of influence for group and potential group concerned with protecting individual pecie, which now only needed to advocate adding a pecie to the lit, rather than eparate legilation addreing each particular pecie. The Clean Air Act of 1970 alo repreented a ubtantial increae in the role of the federal government for protecting the environment. Though many law relating to pollution explicitly et emiion goal and tandard and leave little room for group influence, the Clean Air Act created at leat two opportunitie for interet group to have influence. Firt, the Act handed over broad power for enumeration and enforcement of pecific tandard to the newly-created Environmental Protection Agency. Still in it infant tage, the agency wa ripe for interet group influence. The Clean Air Act alo placed a heavy enforcement burden on tate and required them to develop and ubmit implementation plan within three year. Thi created a pace for maller group to form and focu on how pollution regulation would be enforced on a local level. Unlike pecie protection, pollution regulation may have a lulling effect on contribution to interet group. Although there i ignificant room for private individual and group to take proactive tep to help endangered pecie, nearly all air pollution abatement mut come through government edict. It i theoretically poible that individual may have felt that

30 air pollution wa in eence taken care of after the Clean Air Act, cauing fewer group to form. The model doe not provide a clear prediction for the effect of the Clean Air Act on pollution group, depending intead on the relative magnitude of the contrating effect. If legilation doe caue interet group to form, it i likely that the effect i not indefinite. The prediction that interet group form after a law i paed ret partially on the uppoition that there are regulatory detail to be orted out jut after the law i paed. Once the detail of interpretation and enforcement have become entrenched, the capacity for influence decline. Allowing the length of the treatment window after the paage of each law to vary give ome idea of how long the effect may lat. The baeline regreion on the number of new interet group include the following treatment variable: the variable ESA i equal to zero through 1973, and one after. The variable Air i equal to zero through 1970, and one after. The variable ESA_tYr_Pot and Air_tYr_Pot are equal to one only for t year after the paage of each law, while ESA_tYr_Pre and Air_tYr_Pre are equal to one for t year before the paage of each law. In each cae, the year the law wa paed i not conidered part of the treatment period, due to a lack of information about how long it take an interet group to form. 6 Equation 12 give the empirical model: Xt i a vector of control variable, Trend i a linear time trend, and Trend2 i a quadratic time trend. 7 Table 3 diplay the reult of the ESA regreion and table 4 how the reult of the Clean Air Act regreion. 6 The reult reported here are robut to alternate pecification that include the year of a law paage a part of the treatment window. 7 I do not etimate the effect of amendment on group formation becaue the theory doe not provide any reaonable prediction for what effect amendment might have. Given the number of amendment to both

31 (12) Table 3: Wildlife Group and National ESA (Baeline) VARIABLES (1) (2) (3) (4) (5) (6) (7) New_Grou New_Grou New_Grou New_Grou New_Grou p p p p p New_Grou p ESA -5.285 (3.499) ESA_3Yr_Pot 1.911 1.292 (1.264) (1.404) ESA_5Yr_Yr_P ot 4.808 4.748 New_Grou p (3.594) (3.566) ESA_10Yr_Pot -3.771* -1.985 (1.928) (2.353) ESA_3Yr_Pre -3.266 (2.403) ESA_5YrYr_Pre 0.624 (2.745) ESA_10Yr_Pre 4.399* (2.601) LNPCI -61.02** -60.24** -73.75** -50.54** -51.81* -75.63** -64.70** (24.12) (24.99) (29.07) (23.22) (27.28) (31.11) (26.69) LNPOP -479.3*** -405.9*** -483.0*** -409.7*** -386.0*** -488.1*** -497.8*** (91.37) (65.97) (95.03) (63.67) (68.16) (99.39) (84.50) Trend 10.86*** 9.494*** 11.16*** 9.453*** 8.999*** 11.28*** 10.99*** (1.945) (1.489) (2.148) (1.401) (1.589) (2.261) (1.792) Trend2-0.0617*** -0.0554*** -0.0612*** -0.0577*** -0.0542*** -0.0615*** -0.0607*** (0.00794) (0.00579) (0.00773) (0.00600) (0.00601) (0.00799) (0.00639) Contant 6,264*** 5,382*** 6,419*** 5,339*** 5,069*** 6,497*** 6,519*** (1,215) (917.8) (1,323) (864.0) (968.6) (1,396) (1,172) Obervation 59 59 59 59 59 59 59 R-quared 0.796 0.789 0.794 0.802 0.793 0.794 0.811 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1 The reult from thee highly implified regreion ugget at mot a very limited effect of national legilation on interet group formation. After controlling for a nonlinear time trend, almot none of the ESA treatment variable are ignificant. In column 4, ESA_10Yr_Pot i negative and ignificant at the 90% level, but it loe ignificance law, it i unclear which would be conidered treatment and which would not, abent an ex-pot evaluation of which amendment had a large effect, biaing the election of treatment. Moreover, if law do caue more group to form, then later amendment are likely to be highly endogenou to the activity of group that formed in repone to the law.

32 when a ymmetric ESA_10Yr_Pre poitive and ignificant at the 90% level i included. The Clean Air reult (table 4) are omewhat different; Air_5Yr_Pot i poitive and ignificant in column 3 and 5. Thee reult ugget that the ESA had a negligible effect on wildlife group formation and may have even been the reult of an increae in the number of group. The Clean Air Act appear to have caued a mall but not immediate increae in the number of pollution group that did not perit over time. Both pollution group and wildlife group exhibit an increaing but non-linear time trend. After controlling for thi trend, it appear that group are le likely to form when per capita income and population increae. Thee reult are merely a firt tep; they are only baed on 59 obervation and they do not control for hock that would influence the alience of environmental iue and thu the formation of new group. Conidering the effect of pecific law on group with a related focu help clean up ome of the noie generated by the complex political economy of environmental legilation, but the legilation itelf i till likely to be endogenou to the formation of group. If pecific event or general change in public concern for the environment affect both group formation and legilation, the reult preented above are biaed. In an attempt to reolve thi endogeneity, a difference-in-difference trategy i employed uing the number of new conervation group in each year a a control group.

33 Table 4: Pollution Group and National Clean Air Act (Baeline) VARIABLES (1) (2) (3) (4) (5) (6) (7) New_Group New_Group New_Group New_Group New_Group New_Group Air 1.067 (1.492) Air_3Yr_Pot 2.310 2.547 (1.470) (1.771) Air_5Yr_Pot 3.189*** 4.290** (1.092) (1.698) Air_10Yr_Po t New_Group 1.314 1.368 (1.125) (1.426) Air_3Yr_Pre 0.768 (1.193) Air_5Yr_Pre 2.395 (1.510) Air_10Yr_Pre 0.126 (1.324) LNPCI -38.52** -43.46** -49.83** -44.16** -45.39** -59.05** -44.09** (16.15) (19.38) (18.99) (18.85) (22.01) (23.94) (18.80) LNPOP -182.0*** -205.2*** -215.7*** -201.2*** -211.7*** -251.8*** -203.9*** (48.86) (47.01) (46.04) (43.84) (55.41) (65.03) (63.38) Trend 4.257*** 4.775*** 5.077*** 4.708*** 4.917*** 5.830*** 4.743*** (1.176) (1.238) (1.208) (1.168) (1.428) (1.608) (1.388) Trend2-0.0203*** -0.0221*** -0.0227*** -0.0215*** -0.0225*** -0.0247*** -0.0216*** (0.00491) (0.00469) (0.00454) (0.00444) (0.00514) (0.00543) (0.00486) Contant 2,518*** 2,840*** 3,022*** 2,798*** 2,934*** 3,537*** 2,830*** (695.4) (720.7) (705.3) (675.2) (847.0) (982.5) (891.8) Obervation 59 59 59 59 59 59 59 R-quared 0.587 0.595 0.613 0.592 0.596 0.626 0.592 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1 Thi control group hould reolve endogeneity reulting from broad change in environmental entiment, to the extent that the formation of conervation group capture that entiment, and hould alo iolate the effect of each law on the treatment group only. Equation 13 give the general form of the empirical model. The treatment effect i meaured by the difference-in-difference coefficient:. Table 4 and 5 give the reult, while figure 3 and 4 how plot of each treatment group relative to the control over time.

34 Figure 3: Wildlife Group and Conervation Group Figure 4: Pollution Group and Conervation Group In thi model, the dependent variable i the number of new interet group in the United State in year t with focu i. LAW i equal to 1 during the treatment window and zero otherwie, FOCUS i a fixed effect for the treated focu (either wildlife or pollution), LAW*Focu i the interaction of LAW and Focu, X i a vector of control (13) variable, Trend i a linear time trend, and Trend2 i a quadratic time trend. give the

35 treatment effect, which i defined a: Thi etimator identifie the average treatment effect on the treated (ATT) uing a natural experiment approach. Thi etimator relie on the aumption that both the treatment and control group follow a imilar time trend and that neither undergo any tructural change over the coure of the analyi. The identification of the treatment effect come from meauring the difference comparing the change in new wildlife group before and after treatment with the change in new conervation group before and after treatment. By ubtracting out the change in conervation group, which hould not be affected by the paage of the ESA or the Clean Air Act, a difference-in-difference approach control for any broad change in environmental entiment that may be correlated with the paage of the law and for any other change that would make any potential interet group more or le likely to form. Thi provide a cleaner etimate of the effect of pecific law becaue it remove variation due to change in other factor affecting all environmental group. The difference-in-difference model how an increae in the formation of wildlife group after the paage of the Endangered Specie Act. The treatment variable i inignificant in the baic treated v. untreated framework, but thi i unurpriing (table 5). Whatever advantage there are to forming a group after a law i paed, it i unlikely that they lat forever. Model 2 through 4 how tatitically ignificant increae in wildlife group formation in the three, five, and ten year after the paage of the ESA, compared to conervation group. The effect doe not appear to be contant; group

36 formation increae between 3 and 5 year after the law, but then decline omewhat by the time 10 year have gone by. The average etimated increae in the 5 year after the ESA i 15.15 group when the pre-law window i not controlled for and 17.67 after conidering poible increae in group formation before the ESA. Thi i about a 1.5 tandard deviation increae in the number of new wildlife group. Only the Wild*ESA_10Yr_Pre variable attain ignificance. It may be that broad trend in concern for wildlife caued more group to form, which in turn contributed to the paage of the ESA. The inignificance of Wild*ESA_3Yr_Pre and Wild*ESA_5Yr_Pre implie that the law wa not paed a a direct reult of increae group formation, however. Given the prominent role that interet group have played, through comment ubmitted and lawuit filed ince the ESA paed in 1973, a 1.5 tandard deviation increae in the number of group i an economically ignificant finding. The reult for the Clean Air Act (table 6) are imilar to the reult for the ESA, however more pre-law group formation i apparent. Like the Endangered Specie Act, the Clean Air Act doe not appear to have caued a permanent increae in the number of new pollution group formed. Unlike the wildlife reult, it appear that the treatment effect grow over time; the coefficient on Pollute*Air_3yr_Pot, Pollute*Air_5Yr_Pot, and Pollute*Air_10Yr_Pot are 7.4, 12.5, and 14.7, repectively.

37 Table 5: National ESA Difference-in-Difference. Treated = Wildlife Group. Control = Conervation Group VARIABLES (1) (2) (3) (4) (5) (6) (7) New_Group New_Group New_Group New_Group New_Group New_Group New_Group Wildlife 2.033 4.159-15.87** 6.017 4.097-19.06** 5.700 (3.708) (3.663) (7.531) (4.079) (3.686) (7.445) (3.573) ESA -15.50*** (4.085) ESA_3Yr_Pot -8.928*** -7.846*** (2.855) (2.986) ESA_5Yr_Pot 1.487-0.476 (6.389) (5.809) ESA_10Yr_Pot -14.19*** -13.52*** (2.848) (3.521) ESA_3Yr_Pre 6.408** (2.635) ESA_5Yr_Pre 10.55*** (2.652) ESA_10Yr_Pre 2.152 (3.291) Wild*ESA -3.786 (4.839) Wild*ESA_3Yr_Pot 12.67*** 12.25*** (3.315) (3.516) Wild*Ea5 15.15** 17.67*** (5.900) (5.942) Wild*Ea_10Yr_Pot 9.994*** 14.92*** (3.654) (4.764) Wild*ESA_3Yr_Pre -3.602 (3.618) Wild*ESA_5Yr_Pre -6.459 (4.024) Wild*ESA_10Yr_Pre 11.13*** (3.760) LNPCI -132.0*** -115.6*** -150.7*** -103.2*** -127.5*** -172.7*** -128.1*** (27.63) (29.03) (34.34) (26.28) (32.71) (37.59) (32.79) LNPOP -935.4*** -676.7*** -833.9*** -699.9*** -704.8*** -893.6*** -854.5*** (109.1) (74.68) (117.0) (71.26) (83.02) (123.2) (104.2) Trend 21.33*** 16.46*** 19.69*** 16.94*** 17.14*** 21.00*** 19.84*** (2.407) (1.825) (2.705) (1.674) (2.048) (2.861) (2.300) Trend2-0.112*** -0.0933*** -0.103*** -0.102*** -0.0946*** -0.105*** -0.111*** (0.0103) (0.00839) (0.0102) (0.00846) (0.00890) (0.0104) (0.00953) Wild*Trend -1.133*** -1.462*** -0.896*** -1.737*** -1.426*** -0.742** -2.122*** (0.329) (0.336) (0.303) (0.387) (0.352) (0.307) (0.440) Wild*Trend2 0.0149*** 0.0190*** 0.0158*** 0.0235*** 0.0183*** 0.0140*** 0.0312*** (0.00460) (0.00533) (0.00466) (0.00606) (0.00566) (0.00473) (0.00728) Contant 12,343*** 9,113*** 11,305*** 9,274*** 9,555*** 12,219*** 11,345*** (1,475) (1,065) (1,644) (977.6) (1,206) (1,752) (1,463) Year FE X X X X X X X Obervation 118 118 118 118 118 118 118 R-quared 0.834 0.812 0.823 0.841 0.815 0.835 0.859 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1

38 Thee effect are magnified after controlling for pre-clean Air Act pollution group formation; the coefficient are then 8.2, 14.3, and 21.1, repectively. 8 Model 5 through 7 alo how a ignificant increae in new pollution group relative to conervation group in the three, five, and ten year window before the Clean Air Act. The treatment effect of the Clean Air Act i much larger relative to the mean number of pollution group in comparion to the Endangered Specie Act. The mean number of new pollution group, 4.08 for Pollute*Air_5Yr_Pot, for example, i an increae equal to between 2 and 3 time the average number of new pollution group. The more immediate effect of the Clean Air Act on private buinee and individual, in addition to the requirement for tate implementation plan, may explain thi larger etimate. The difference-in-difference trategy employed here i a major improvement over the imple model relating the number of environmental law to the number of group. Examining interet group by focu help control for ome of the unobervable factor preent in national-level regreion, baed on underlying preference toward environmental quality and government action. Analyi of the Endangered Specie Act and the Clean Air Act clearly demontrate the exitence of a relationhip between major environmental legilation and the propenity of related group to form, both before and after the paage of the legilation. Table 7 through 10 preent the reult of placebo regreion, which tet the effect of the Endangered Specie Act on pollution group and the effect of the Clean Air Act on wildlife group. 8 Thi may be related to ubequent amendment to the Clean Air Act which trengthened the Act.

39 Table 6: National Clean Air Act Difference-In-Difference. Treated = Pollution Group. Control = Conervation Group VARIABLES (1) (2) (3) (4) (5) (6) (7) New_Group New_Group New_Group New_Group New_Group New_Group New_Group Pollution 7.087 8.459** 8.995** 11.00** 8.251** 8.327** 7.011* (4.559) (3.984) (4.032) (4.359) (3.934) (3.831) (3.730) Air -6.046 (4.511) Air_3Yr_Pot 3.122 4.731 (2.568) (2.981) Air_5Yr_Pot -3.463-0.432 (3.800) (4.712) Air_10Yr_Pot -10.57*** -12.96*** (3.156) (3.916) Air_3Yr_Pre 2.844 (3.891) Air_5Yr_Pre 3.679 (3.466) Air_10Yr_Pre Pollute*Air -2.972 (5.151) Pollute*Air_3Yr_Pot 7.446** 8.215*** (2.878) (3.055) Pollute*Air_5Yr_Pot 12.56*** 14.34*** (4.011) (4.555) Pollute*Air_10Yr_Po t -6.349 (3.939) 14.77*** 21.12*** (3.752) (4.475) Pollute*Air_3Yr_Pre 7.262* (3.855) Pollute*Air_5Yr_Pre 9.704*** (3.014) Pollute*Air_10Yr_Pre 16.43*** (3.302) LNPCI -88.83*** -131.9*** -121.3*** -88.25*** -148.2*** -154.2*** -87.35*** (29.20) (33.48) (33.46) (33.55) (36.91) (40.11) (31.56) LNPOP -639.6*** -618.2*** -597.9*** -551.3*** -672.5*** -726.6*** -591.2*** (94.81) (85.14) (83.35) (77.50) (98.17) (111.9) (115.4) Trend 15.40*** 15.94*** 15.43*** 14.16*** 17.17*** 18.17*** 14.82*** (2.064) (2.097) (2.061) (1.933) (2.378) (2.646) (2.217) Trend2-0.0904*** -0.0895*** -0.0891*** -0.0895*** -0.0932*** -0.0973*** -0.0943*** (0.00977) (0.00902) (0.00897) (0.00887) (0.00957) (0.00994) (0.00929) Pollute*Trend -2.463*** -2.706*** -2.851*** -3.206*** -2.755*** -2.954*** -3.490*** (0.430) (0.350) (0.359) (0.405) (0.354) (0.358) (0.400) Pollute*Trend2 0.0360*** 0.0390*** 0.0416*** 0.0477*** 0.0401*** 0.0440*** 0.0550*** (0.00575) (0.00561) (0.00572) (0.00638) (0.00572) (0.00582) (0.00664) Contant 8,425*** 8,562*** 8,223*** 7,366*** 9,357*** 10,057*** 7,836*** (1,252) (1,237) (1,211) (1,121) (1,432) (1,633) (1,505) Year FE X X X X X X X Obervation 118 118 118 118 118 118 118 R-quared 0.836 0.838 0.840 0.849 0.843 0.853 0.868 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1 The difference-in-difference reult (table 9 and 10) how a poitive effect of the Endangered Specie Act on pollution group and a poitive effect of the Clean Air Act on

40 wildlife group. Thi ugget that the difference-in-difference model fail to fully addre the endogeneity of national environmental law. Given that there i no theoretical reaon to expect an effect of wildlife law on pollution group or an effect of pollution law on wildlife group, the trong finding in the placebo regreion highlight the empirical problem preented by the clutering of major federal environmental legilation in the 1970. Even with the difference-in-difference trategy, the analyi at the federal level till ha everal limitation. The regreion reported above till rely on only 118 obervation to identify the treatment effect, making year fixed effect infeaible. The regreion model reported here cannot account for hock to environmental quality or nonparametric hift in opinion on pecific iue like pollution or wildlife protection they can only control for general entiment by differencing out new conervation group. It i till poible that public concern for environmental iue i cauing both legilation and group formation. It i alo poible that public concern caue group formation, and that the two jointly caue legilation. In an attempt to further reolve thee potential ource of endogeneity, the next ection turn to an analyi of tate-level environmental regulation, which exploit time-erie and cro-ectional variation to arrive at a more precie etimate of the effect of environmental legilation on interet group formation. While the clear economic ignificance of law like the Endangered Specie Act and the Clean Air Act i an advantage of national-level regreion, the inability to fully reolve the endogeneity of the law make a tate-level analyi more appealing.

41 Table 7: Placebo Regreion of ESA on Pollution Group (1) (2) (3) (4) (5) (6) (7) VARIABLES New_Group New_Group New_Group New_Group New_Group New_Group New_Group ESA -1.744 (2.417) ESA_3Yr_Pot 1.874* 2.234* (1.028) (1.153) ESA_5Yr_Pot 6.489*** 6.394*** (1.905) (1.851) ESA_10Yr_Pot -0.479-0.211 (1.122) (1.245) ESA_3Yr_Pre 1.902 (1.182) ESA_5Yr_Pre 0.988 (0.999) ESA_10Yr_Pre 0.660 (1.377) LNPCI -36.92** -38.59** -57.92*** -34.83** -43.50** -60.89*** -36.95* (17.45) (17.45) (20.03) (15.86) (20.14) (22.33) (19.65) LNPOP -216.4*** -194.6*** -300.0*** -192.0*** -206.1*** -308.1*** -205.2*** (69.70) (42.79) (64.29) (43.99) (47.41) (68.66) (62.24) Trend 4.856*** 4.481*** 6.770*** 4.377*** 4.769*** 6.957*** 4.608*** (1.599) (1.124) (1.561) (1.118) (1.256) (1.676) (1.458) Trend2-0.0232*** -0.0211*** -0.0291*** -0.0214*** -0.0218*** -0.0295*** -0.0218*** (0.00646) (0.00443) (0.00584) (0.00484) (0.00467) (0.00608) (0.00534) Contant 2,914*** 2,669*** 4,094*** 2,603*** 2,851*** 4,217*** 2,780*** (968.4) (650.7) (930.8) (646.9) (731.5) (1,004) (907.6) Obervation 59 59 59 59 59 59 59 R-quared 0.590 0.593 0.644 0.586 0.600 0.647 0.587 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1

42 Table 8: Placebo Regreion of Clean Air Act on Wildlife Group (OLS) (1) (2) (3) (4) (5) (6) (7) VARIABLES New_Group New_Group New_Group New_Group New_Group New_Group New_Group Air -5.188 (3.929) Air_3Yr_Pot 0.177 2.289 (2.510) (2.799) Air_5Yr_Pot 1.451 5.818** (2.129) (2.197) Air_10Yr_Pot -2.169 2.185 (2.227) (2.258) Air_3Yr_Pre 6.859*** (1.853) Air_5Yr_Pre 9.498*** (2.116) Air_10Yr_Pre 10.18*** (2.136) LNPCI -43.50-57.87* -63.71** -43.29-75.14** -100.3*** -38.40 (28.55) (29.20) (29.71) (31.53) (30.42) (29.60) (28.99) LNPOP -446.0*** -403.3*** -413.4*** -385.4*** -460.8*** -556.8*** -603.4*** (71.79) (71.78) (72.20) (65.13) (80.95) (86.40) (78.89) Trend 9.917*** 9.409*** 9.700*** 8.816*** 10.68*** 12.69*** 11.65*** (1.487) (1.706) (1.709) (1.604) (1.868) (1.891) (1.682) Trend2-0.0592*** -0.0554*** -0.0561*** -0.0546*** -0.0588*** -0.0639*** -0.0605*** (0.00649) (0.00626) (0.00613) (0.00581) (0.00662) (0.00655) (0.00589) Contant 5,707*** 5,330*** 5,504*** 4,983*** 6,172*** 7,546*** 7,540*** (924.1) (1,035) (1,043) (957.3) (1,159) (1,208) (1,082) Obervation 59 59 59 59 59 59 59 R-quared 0.795 0.788 0.789 0.791 0.805 0.828 0.828 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1

Table 9: Placebo Regreion of ESA on Pollution Group. Control = Conervation VARIABLES New_Group 43 (1) (2) (3) (4) (5) (6) (7) New_Group New_Group New_Group New_Group New_Group New_Group Pollution 5.004 9.020** -20.58*** 11.49** 9.088** -23.13*** 11.07** (3.920) (4.077) (7.323) (4.621) (4.094) (7.268) (4.267) ESA -11.30** (4.475) ESA_3Yr_Pot -9.379*** -8.242*** (2.898) (3.027) ESA_5Yr_Pot -1.404-3.134 (6.379) (5.886) ESA_10Yr_Pot -13.84*** -14.68*** (2.931) (3.730) ESA_3Yr_Pre 5.238** (2.617) ESA_5Yr_Pre 10.09*** (2.581) ESA_10Yr_Pre Pollute*ESA -8.635* (4.787) Pollute*ESA_3Yr_Pot 13.54*** 13.99*** (3.525) (3.713) Pollute*ESA_5Yr_Pot 22.61*** 24.63*** (5.751) (5.782) Pollute*ESA_10Yr_Po t -1.430 (3.757) 12.57*** 19.01*** (3.621) (4.732) Pollute*ESA_3Yr_Pre 3.907 (2.942) Pollute*ESA_5Yr_Pre -5.174* (3.000) Pollute*ESA_10Yr_Pre 14.56*** (3.733) LNPCI -119.9*** -104.8*** -142.8*** -95.34*** -123.3*** -165.3*** -114.2*** (28.00) (29.77) (34.19) (27.58) (33.75) (37.61) (35.98) LNPCI -803.9*** -571.1*** -742.4*** -591.1*** -614.8*** -803.6*** -708.2*** (118.3) (77.02) (114.6) (75.36) (84.53) (120.9) (117.2) Trend 18.87*** 14.62*** 18.05*** 15.11*** 15.73*** 19.41*** 17.42*** (2.586) (1.878) (2.678) (1.757) (2.092) (2.848) (2.523) Trend2-0.103*** -0.0869*** -0.0974*** -0.0949*** -0.0899*** -0.100*** -0.104*** (0.0111) (0.00877) (0.0101) (0.00892) (0.00921) (0.0104) (0.0101) Pollute*Trend -2.221*** -2.792*** -2.009*** -3.166*** -2.832*** -1.886*** -3.669*** (0.329) (0.360) (0.292) (0.424) (0.372) (0.296) (0.470) Pollute*Trend2 0.0343*** 0.0405*** 0.0368*** 0.0466*** 0.0412*** 0.0354*** 0.0566*** (0.00476) (0.00571) (0.00468) (0.00663) (0.00596) (0.00474) (0.00774) Contant 10,666*** 7,754*** 10,145*** 7,904*** 8,444*** 11,082*** 9,473*** (1,594) (1,094) (1,621) (1,031) (1,227) (1,732) (1,635) Year FE X X X X X X X Obervation 118 118 118 118 118 118 118 R-quared 0.853 0.838 0.860 0.858 0.843 0.869 0.876 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1

44 Table 10: Placebo Regreion of Clean Air Act on Wildlife Group. Control = Conervation. (1) (2) (3) (4) (5) (6) (7) VARIABLES New_Group New_Group New_Group New_Group New_Group New_Group New_Group Wildlife 2.731 3.534 3.964 5.454 3.285 3.279 1.889 (4.077) (3.598) (3.621) (3.869) (3.496) (3.291) (3.022) Air -9.742** (4.549) Air_3Yr_Pot 4.437 6.906** (2.675) (3.031) Air_5Yr_Pot -2.415 2.225 (4.008) (4.968) Air_10Yr_Pot -10.33*** -10.23** (3.260) (4.034) Air_3Yr_Pre 5.157 (3.824) Air_5Yr_Pre 7.101** (3.414) Air_10Yr_Pre -0.453 (3.574) Wild*Air -1.836 (5.300) Wild*Air_3Yr_Pot 2.684 3.608 (3.250) (3.407) Wild*Air_5Yr_Pot 8.727** 10.56** (4.194) (4.772) Wild*Air_10Yr_Pot 10.80*** 16.48*** (3.882) (4.624) Wild*Air_3Yr_Pot 8.728** (3.725) Wild*Air_5Yr_Pot 9.964*** (2.960) Wild*Air_10Yr_Pot 14.69*** (3.353) LNPCI -91.32*** -139.1*** -128.3*** -87.82** -163.0*** -174.8*** -84.51*** (29.84) (33.68) (33.88) (34.09) (35.99) (38.14) (32.08) LNPOP -771.6*** -717.3*** -696.8*** -643.4*** -797.1*** -879.1*** -791.0*** (88.31) (84.72) (83.18) (75.51) (96.69) (106.9) (100.7) Trend 17.60*** 17.58*** 17.05*** 15.47*** 19.37*** 20.91*** 17.52*** (1.971) (2.098) (2.072) (1.931) (2.323) (2.503) (2.075) Trend2-0.0995*** -0.0950*** -0.0945*** -0.0941*** -0.100*** -0.106*** -0.101*** (0.00914) (0.00878) (0.00875) (0.00865) (0.00913) (0.00924) (0.00873) Wild*Trend -1.208*** -1.340*** -1.465*** -1.732*** -1.399*** -1.570*** -1.987*** (0.408) (0.327) (0.338) (0.388) (0.325) (0.326) (0.385) Wild*Trend2 0.0153*** 0.0169*** 0.0191*** 0.0237*** 0.0181*** 0.0215*** 0.0302*** (0.00538) (0.00525) (0.00542) (0.00616) (0.00526) (0.00533) (0.00643) Contant 10,021*** 9,810*** 9,467*** 8,461*** 10,979*** 12,064*** 10,194*** (1,178) (1,240) (1,220) (1,110) (1,409) (1,555) (1,352) Year Fe X X X X X X X Obervation 118 118 118 118 118 118 118 R-quared 0.815 0.809 0.810 0.821 0.822 0.836 0.843 Robut tandard error in parenthee *** p<0.01, ** p<0.05, * p<0.1

45 STATE LEGISLATION State environmental policy i certainly le comprehenive than federal policy, but a focu on tate law provide ditinct empirical advantage. Becaue general concern about the environment increae the probability of both environmental legilation and group formation, the paage of environmental legilation i highly endogenou. The fact that different tate paed law at different time facilitate a difference-in-difference approach that help control for ource of endogeneity in the model. Federal environmental law ha evolved incrementally over the pat half-century, which create eriou pecification problem in teting for the effect of legilation at the federal level. The Endangered Specie Act of 1973 i actually an expanion of the Endangered Specie Conervation Act and the Endangered Specie Preervation Act; it i unclear which of thee act hould be conidered the firt true endangered pecie act to affect group incentive. Similarly, the Clean Air Act of 1970 wa technically a et of amendment to an earlier, much le ignificant law. In contrat, mot tate pa only one major piece of endangered pecie legilation and one piece of major air quality legilation in the ample period. State endangered pecie act, though weaker than the federal Act, till increae the expected benefit of formation for tate-level interet group for three reaon. State law allow for the liting of pecie that are not covered on the federal lit, which would enable group to ecure protection for a pecie without having to lobby at the national level (where there i more competition). Additionally, ome pecie that do not warrant federal endangered tatu may be eligible for protection in particular tate baed on local

46 wildlife population. Finally, the ex-pot obervation that tate law lack teeth abtract away from the information available to potential group at the time when the law i paed. Many tate law have turned out to be weak, however, when a law i paed there i coniderable uncertainty a to it implementation. Thi uncertainty i what pur group to act. Moreover, mot tate ESA mandate that tate agencie conult intereted peron or organization when deciding which pecie to lit, opening the door for maller group to exert greater influence than at the national level. Unlike tate wildlife regulation, tate pollution regulation i not likely to increae interet group formation. State air quality law were primarily crafted in repone to the Federal Clean Air Act requirement that each tate develop a tate implementation plan. Thee implementation plan were quite pecific and contained a relatively large proportion of the regulatory detail in the actual legilation, leaving little for bureaucrat to accomplih. Thi, in conjunction with the potential lulling effect of pollution legilation dicued in the previou ection, make it unlikely that group would form in repone to the paage of tate pollution regulation. The paage of tate air pollution law i clutered around the paage of the Clean Air Act in 1970, which create ome problem for empirical etimation. More recently-paed tate renewable portfolio tandard, which mandate a certain percentage of the tate energy be generated by renewable ource, provide another tet of highly pecific legilation. Renewable portfolio tandard alo tend to contain a high proportion of adminitrative detail within the actual law, o they hould caue no increae in group formation. Figure 5 how how many tate-level law of each type were paed in each year.

47 Figure 5: Timing of State Law The data for the tate regreion come from the ame ource a the national ngo variable, excluding group that were lited a operating only at the national level. In thi ection, the dependent variable, New_Group, i the number of new interet group in tate in year t with focu i. Table 11 give the ummary tatitic for New_Group. There i unlikely to be a poitive amount of new group with each focu in each tate in each year, which give rie to a large number of zeroe in the data. Figure 6 through 9 how hitogram of New_Group by focu. The prevalence of zeroe in the data and the ditribution of the non-zero obervation ugget that a count model i neceary. The Poion etimator i the mot widely ued count model, but it relie on the aumption that the variance and the mean of the dependent variable are equal. The variable New_Group doe not atify thi condition (that i, it exhibit overdiperion), which

ugget that a negative binomial etimator i more appropriate. 9 I chooe not to ue a zero-inflated Poiion or zero-inflated negative binomial baed on theoretical concern 48 and tet for goodne of fit. 1011 If the data do indeed have a negative binomial ditribution, then the OLS reult will be biaed becaue they are baed on an erroneou ditributional aumption. In addition to controlling for variation between tate, tandard error have been clutered by tate becaue the obervation idioyncratic error are likely to be correlated within each tate. Table 11: State-Level Summary Statitic Variable Ob Mean Std. Dev. Min Max All New_Group 2950 2.317966 3.665745 0 34 Wildlife New_Group 2950 0.26678 0.630696 0 6 Pollution New_Group 2950 0.067458 0.295539 0 4 Clean New_Group 2950 0.781017 1.404172 0 13 Conervation New_Group 2950 0.402712 0.818069 0 7 LNPCI 2950 9.305651 0.380413 8.061526 10.21297 LNPOP 2950 14.81277 1.050939 11.81303 17.41502 9 See Hauman et al. (1984) 10 Zero-inflated model rely on the aumption that the exce zeroe in the data are determined by a fundamentally different underlying proce than the other obervation. The dependent variable in thi analyi repreent a count of how many group formed, which determine both the number of zeroe and the value of the nonzero obervation. 11 I employed the countfit package in Stata, which run the pecified model uing a Poion, a zeroinflated Poion, a negative binomial, and a zero-inflated negative binomial. Countfit report the BIC, AIC, and Voung tatitic for each model, in addition to teting for over-diperion and providing a plot of the reidual from each model. See Appendix for etimation reult uing OLS, a Poion, a zero-inflated Poion, and a zero-inflated negative binomial.

49 Figure 6: Hitogram of Wildlife New_Group Figure 7: Hitogram of Pollution New_Group