Interest Group Density and Policy Change in the States

Interest Group Density and Policy Change in the States Eric R. Hansen ehansen@live.unc.edu Department of Political Science University of North Carolina at Chapel Hill Caroline Carlson carlson8@live.unc.edu Department of Political Science University of North Carolina at Chapel Hill Virginia Gray vagray@email.unc.edu Department of Political Science University of North Carolina at Chapel Hill July 14, 2017

Abstract To what extent do interest groups influence government policymaking decisions? Recent research has found interest groups to be much more influential in preserving the policy status quo than in securing policy change, but this research has focused primarily on individual groups efforts rather than system-level dynamics. We argue that policy change is less likely to occur in more densely populated interest group systems. When more groups are organized and monitoring government activity, the opportunity for groups to mobilize in opposition to unfavorable proposals increases. Because active opposition often sinks the chances of policy proposals moving forward, policy change should be less likely to occur. As evidence, we observe the outcomes of a stratified random sample of 500 bills introduced in 48 state legislatures in 2007. We find evidence that bills introduced in states with more densely populated interest communities are defeated earlier in the legislative process and less likely to become law. These findings imply that as interest groups continue to proliferate in number, policy is more likely to remain static in state legislatures.

Why do governments enact some policies and decline to enact others? While a wide range of factors play into policymakers decisions, the question of interest group influence over those decisions has been of great concern to politicians, activists, and scholars. Popular accounts of contemporary policymaking frequently portray interest groups as disproportionately powerful in the process (e.g. Whitehouse, 2017). But scholarly work tends to take a more nuanced view. While some research demonstrates a positive association between lobbying activity or spending and favorable policy outcomes (e.g. Grasse and Heidbreder, 2011), other work stresses that the effectiveness of lobbying is conditional on a number of factors and is more successful in preserving the status quo than in creating new policy (e.g. Baumgartner et al., 2009). Instead of focusing on how lobbying activity influences the content and outcome of individual policies, as most previous research has done, we turn to a system-level look at how interest group communities as a whole influence the production of public policy. In particular, we study how the density of interest communities at large affects the likelihood of public policy to change through the legislative process. We contend that as interest communities grow in size, policy change is less likely to occur. We assume that a greater supply of interest groups produces greater lobbying activity across all issue areas. When more interests are organized to lobby in a given political environment, more groups are actively monitoring government activity and are ready to mobilize to defeat unfavorable policy proposals. Because interest groups have greater influence in preserving the policy status quo than in putting preferred policies into effect (Baumgartner et al., 2009; Lewis, 2013), a larger supply of interest groups in a political environment should have the overall effect of decreasing the likelihood of any proposal advancing. As a result, bills proposing new policy changes are more likely to draw opposition and stall in early stages of the legislative process, making policy change on many issues less likely. As an empirical test, we study the advancement of 500 bills sampled from 48 state legislatures in 2007. We pair bills with state lobbying group registration data collected from the National Institute on Money in State Politics (e.g. Lowery et al., 2013) in order 1

to compare the effects of interest organization density across states. We measure the effects of interest group community size (which we also refer to as density ) on two dependent variables: bill progress through the legislature and bill enactment. We find that bills on average are defeated earlier in the legislative process and are less likely to pass as the size of the interest community in the state increases. We also test whether differences in the size of lobbying sectors, like business or health, drive the overall results. However, we find that the size of the interest community as a whole, rather than the sizes of individual lobbying sectors, is responsible for the decreasing likelihood of policy change. Our study provides evidence in line with previous findings that interest groups exert the most influence over policy when working to preserve the status quo (Baumgartner et al., 2009). However, our research design allows us to generalize these previous findings outside the context of single political environments like Washington, D.C. It also points to the importance of system-level and contextual factors in helping to determine the outcome of policy decisions. Our findings have implications for the ability of elected leaders to respond to the wide range of pressing policy concerns facing citizens. Interest Group Influence in Policymaking How much influence do interest groups exert over the policies that elected lawmakers write and enact? Answering the question is important to gauging the quality of democratic governance in the United States. If interest groups routinely persuade officials to pass policies that benefit small groups at the expense of public welfare, as many lobbying reform proponents argue, then American democracy may be failing to produce governments that make broadly representative decisions. But if interest groups influence political decisionmaking only under a narrow range of circumstances, then efforts to reform lobbying and spending rules may prove misguided. Popular accounts of lobbying from journalists, activists, and politicians alike tend to portray interest groups as unchecked masters of influence who lobby government to pass favorable legislation. For example, Democratic U.S. Senator Sheldon Whitehouse (2017, 2

xx-xxi) writes: As a bill moves through Congress, corporate lobbyists exploit procedural opportunities to accomplish the industry s purposes out of view of the public... Congress today is working great at helping polluters; it s working great at protecting hedge fund billionaires low tax rates; it s working great at helping corporations offshore jobs, at letting chemicals and genetically modified stuff into your food, at creating tax and safety loopholes for industry. Pundits also frequently point to the seemingly unlimited spending by wealthy interests as evidence that lobbying groups have (a) political goals different than those of average Americans and (b) have the resources and power to translate those goals into public policy. Scholarly work tends to take a more nuanced view of how lobbyists influence government decisions. Lobbying groups do persuade governments to pass new policies they want some of the time perhaps most effectively when writing and propagating model legislation (see Hertel-Fernandez, 2014). But most research in political science over the last two decades has tended to take the viewpoint that interest group influence is not pervasive and is contingent on a number of factors (e.g. Heinz et al., 1993; Baumgartner et al., 2009; Burstein, 2014). Researchers have struggled to provide causal evidence that interest groups assert power in the political process, at least in the way Dahl (1957, 203) defines it: A has power over B to the extent that he can get B to do something that B otherwise would not do. Scholars have also addressed the question of resource bias in lobbying. Though the political goals of organized interests are biased in favor of the economically advantaged (Schattschneider, 1960; Olson, 1965; Schlozman and Tierney, 1986; Kimball et al., 2012), scholars have pointed out that this bias does not necessarily create bias in policy outcomes (Lowery and Gray, 2004). Even in cases where public interests are pitted against wellfunded business interests, public interests most often prevail (Gray et al., 2004; Smith, 2000; Hojnacki et al., 2015; Bauer et al., 1963). In a review essay summarizing a decade s worth of relevant research in both political science and sociology, Burstein and Linton (2002) show that a majority of studies find limited or no evidence that interest groups influence policy outcomes. That being said, scholars do admit the question of interest 3

group influence over public policy is extraordinarily difficult to study researchers have not arrived at a consensus in answering it. Most of this research focuses on how much influence interest groups wield through their lobbying efforts in individual policy decisions. Departing from previous work, we are interested in a related question about the influence that interest group communities have in shaping the entire range of policies that governing institutions produce. By interest group community, we refer to the full set of interests groups organized and lobbying government decisionmaking in a given political environment. And by political environment, we mean the context of where a government s decisions are made, such as Washington, D.C. or a state capital. A Theory of Interest Group Density and Policy Change Our central claim is that as the number of groups organized to lobby on the issues within a political environment increases, the likelihood of any given policy proposal to pass through the legislative process decreases. When more lobbying groups are organized, more groups are actively monitoring legislative proposals and are ready to mobilize to oppose unfavorable actions. Government policy is less likely to change when there is active opposition to that change. As a result, legislatures will tend to keep the status quo in place more often when considering policy changes. Generally speaking, groups are much more effective at torpedoing the policy proposals they dislike than in passing new policy proposals they do like. In a major study tracking Congressional decisionmaking on 98 issues over several years, (Baumgartner et al., 2009) affirm that groups are much more successful when lobbying to protect the status quo. Status quo preservation is an easier task for lobbyists because it requires no coordinated effort on the part of lawmakers. Because new legislation can be blocked at a large number of veto points in the legislative process, lobbyists often only need the support of a small number of elected officials to prevent a bill from becoming law. In contrast, new policy proposals require a majority of legislators (as well as certain key actors, such as committee chairs or chamber leaders) to support the measure. 4

Given interest groups relatively higher chance of success when preserving the status quo, the likelihood of policy change occurring should also be lower when there are more lobbying groups organized in a political environment. In making this connection, we assume that a greater number of organized groups in a political environment translates to increased lobbying activity. Though it is possible that groups organize but remain dormant, we assume that group patrons would not devote resources to maintaining an organization with no expectations that it lobby government, even if only on occasion. The existence of an organization signifies that it is actively working on issues important to its constituents or clients. One of lobbyists primary day-to-day activities is gathering information on new proposals and legislative activities (Kersh, 2002; Nownes and Freeman, 1998). Monitoring the institutions they intend to lobby allows groups to know what new proposals are being considered, what the political dynamics of the institutions are, and who the key actors in a policy debate are. When more lobbying groups are organized, more individuals are employed to monitor the legislature and are dispatched to lobby when threats to the groups interests are detected. The result is that any proposal to change policy is more likely to receive attention and opposition from a group organized and monitoring them. When lobbying groups are not already organized or active, threats to their constituent group are less likely to be detected and acted upon. Without individuals employed on behalf of a constituent group to monitor government activities, it is possible that a policy proposal offensive to them would go unnoticed, especially on low-salience issues likely to escape media attention. Even if the threat is detected, the costs for constituents to organize and coordinate action in reaction would be steep, and perhaps not timely enough to prevent policy change. This is not to say that lobbying groups never form in reaction to specific events, following the disturbance theory of interest group formation put forth by Truman (1951). However, given the costs of group formation and coordinated action (Olson, 1965), it would be much more difficult for a constituent group to mobilize successfully in opposition to a threatening policy proposal without an organization already being in place. 5

Recent studies have investigated the relationship between lobbying activity and policy change, though they have not directly tested our group density hypothesis. As mentioned above, Baumgartner et al. (2009) find in their Washington study that the side protecting the status quo succeeds more often than not; that lobbying group spending correlates with policy success rather modestly; and that groups thought to have a great deal of power in Washington (for example, business groups) are often checked by coalitions of opposing groups. In contrast, other recent work finds a more direct role of interest group influence in the legislative process. Grasse and Heidbreder (2011), in a study of the Wisconsin State Legislature, show that increased spending and lobbying in favor of a proposal increase its chances of passage. But Lewis (2013), using similar data from Wisconsin, finds that greater money and hours spent lobbying pay off in an increased chance of policy victory for interest groups and that the benefits go unevenly in favor of groups lobbying to preserve the status quo. Grossmann and Pyle (2013) study lobbying activity on all bills introduced in two terms of Congress, finding a positive association between bill advancement and lobbying actvity. However, the authors are unable to determine the direction of causality: whether lobbying activity propels bills forward, or whether bills likely to advance for other reasons draw the most lobbying attention. These studies of interest group influence in the lawmaking process examine whether lobbying activity and spending by individual groups result in policy victory or defeat (Baumgartner et al., 2009; Lewis, 2013; Grossmann and Pyle, 2013; Grasse and Heidbreder, 2011). However by studying outcomes in single legislatures (either the U.S. Congress or the Wisconsin State Legislature), these studies neglect characteristics of political environments and interest communities that can potentially affect policy outcomes. Contextual factors such as the size of interest group communities, the capacity of the legislature to consider legislation, and the partisan and ideological leanings of elected officials may determine how likely bills are to pass. In particular, we are interested in how the density of interest group populations or how many groups compete for attention and influence (Gray and Lowery, 1996) affects 6

the likelihood of policy change on key issues across legislatures. Studying density is important because the scope of the interest group system alters the opportunity structure for groups to provide information to lawmakers and to shape policy. A more densely populated system signifies greater opportunity for interest group mobilization and influence than a system with relatively fewer groups competing for attention and sway. Political Environment or Economic Sector? Though our argument focuses specifically on the size of interest group communities as a whole within a political environment, a competing theory might be that differences in the sizes of lobbying communities across economic sectors (like business or health) produce variation in the likelihood of policy change instead. Previous studies have leveraged variation across sectors within a single political environment (usually Washington D.C.) to draw inferences about the relationships between government activity, mobilization, and group spending (e.g. Hansen and Drope, 2005; Mitchell et al., 1997). Groups do not mobilize in response to government activity to the same extent across sectors (Dusso, 2010; Baumgartner and Leech, 2001). Factors such as the size and scale of sectors within economies play a role in the resources available to groups and, as a consequence, the amount of lobbying activity groups can engage in (Lowery and Gray, 2009; Lowery et al., 2005). Business groups, for instance, organize at a high rate and have greater resources at their disposal than other types of groups, such as single-issue groups. An overall finding that interest group density in a state depresses the likelihood of policy change might obscure cross-sector variation. It could be the case that the mobilization of groups in some sectors decreases the likelihood of policy change, but mobilization from other types of groups might not. In short, assuming away differences across sectors while comparing political systems could produce questionable conclusions. However, there are reasons to doubt that large lobbying sectors alone, rather than interest communities at large, depress policy change. Policy battles do not always occur strictly along sector lines. A recurring example in contemporary American politics is the alignment of LGBT+ groups with business groups in debates over religious liberty 7

(e.g. Indiana s Religious Freedom Restoration Act in 2015) or public accommodations (e.g. North Carolina s bathroom bill in 2016). While business groups organizational missions are not necessarily aligned with LGBT+ groups poliical goals, business groups nonetheless have found their own interests to be threatened by such measures. These coalitions of groups that signal unity of opinion among stakeholders to lawmakers often form in response to specific issues and disband as soon as the next comes along (Hojnacki, 1997; Mahoney and Baumgartner, 2015; Heaney and Lorenz, 2013). Coalitions are likely to draw from across sectors, but only likely to draw from groups already organized to lobby in a specific environment. Therefore, larger interest communities in an environment can supply a wide range of groups mobilized to defeat a new policy proposal, while larger sectors may not. Lobbying in State Legislatures To gather supporting evidence for our argument, we observe how interest group density affects the progress of new policy proposals through the legislative process. Specifically, we turn to bills introduced in the fifty American state legislatures. Understanding the factors that enable or hinder policy change in the states is substantively important. In the federal system of government, states are responsible for setting a wide range of economic and social policies that affect the day-to-day lives of citizens. States are also the primary policymakers for key issue areas like education, insurance, and criminal justice. States also provide a useful comparative context in which to assess the influence of interest groups on policy (Lowery and Gray, 2009). States vary in the size and scope of their interest group systems, but remain similar enough in political cultures and institutional structures to warrant comparison. Most studies of interest group influence and lobbying focus on federal government activity in Washington. While lobbying in this context is undeniably important to study in its own right, scholars do not have a good sense if the findings from studies of Washington generalize to other policymaking contexts. Washington stands apart from most state capitals and even from many other national capitals in the professionalization of its political environment, the amount of 8

money spent on politics by interest groups, and the amount of media attention received by political figures, among other factors. On the specific topic of lobbying activity and policy change, others have studied single contexts outside of Washington (namely Madison, Wisconsin see Grasse and Heidbreder, 2011; Lewis, 2013), but this research has not focused on policy change in multiple contexts using a comparative framework. Our study advances our understanding of interest group influence over policy change by observing the contextual factors that vary across states that may strengthen or weaken group influence. We introduce interest group density as a contextual factor that affects the likelihood that a given policy proposal will advance through the legislative process. Our study also compares states to determine whether findings about the role of interest groups in the policymaking process generalizes across political environments. Unfortunately, by comparing states, we cannot provide detailed evidence on specific lobbying activities undertaken by interest groups. States vary widely on their ethics and disclosure laws. We cannot observe positions taken, money spent, or lobbying hours logged by groups on each of our bills because such data on lobbying activity simply do not exist in most states. In our study, we assume that greater interest group density translates into greater lobbying activity, both for and against a range of proposals. Though we sacrifice in-depth analysis of the connection between specific lobbying activities and legislative outcomes, we gain a comparison of system-level processes occurring in the states. Research Design and Data In this study, we were interested in learning how interest group density in a state is related to the likelihood of policy change. We followed Burstein (2014) in measuring policy change by observing the advancement and ultimate success or defeat of discrete policy proposals, with individual bills as the unit of analysis. In line with a number of recent studies that rely on random samples of bills or issues as a source of data (Baumgartner et al., 2009; Dusso, 2010; Burstein, 2014; Lewis, 2013), we sampled 500 bills from a population of the 115,598 bills introduced in the regular sessions of state legislatures 9

in calendar year 2007. Though a single year of observations can only provide limited evidence, we chose 2007 because it was only year for which we could gather concurrent data for bill histories and our measure of interest group density, described further below. Bill texts and histories were gathered from the public webpages of 48 state legislatures. 1 In order to include observations from every state while reflecting the variation in the volume of policymaking activity across states, we used a stratified random sample of bills. We randomly drew at least four bills introduced within each state legislature (n=192) and drew randomly from the population of all state bills for the remainder of the sample (n=308). Also following burstein2014 example, we only sought bills that proposed substantive policy changes. We excluded all bills from the sample that made technical changes to existing legislation (i.e. correcting or updating section numbering, clarifying language) or that appropriated funds. 2 We also excluded resolutions, which are usually symbolic measures and not substantive policy changes, and bills considered during special sessions, which are not used uniformly across states and which often entail atypical policymaking scenarios. We drew a sample of legislation tackling a very wide range of topics. Our bills addressed nationally salient issues, such as a Nevada bill requiring voters to show photo identification at the polls, to arcane issues, such as a Louisiana bill expanding tax credits for rehabilitating historic structures. The vast majority of bills dealt with mundane yet important state policies: for example, criminal penalties, school safety, state employee benefits, and highway construction regulations. The dependent variable we used to measure how far each bill advanced through the legislative process is Bill Progress, a count of the number of stages of the legislative process through which the bill advanced (see Lewis, 2013). The variable is ordinal and coded 1 if the bill died in committee, 2 if the bill died in the originating chamber after being reported from committee, 3 if the bill died in the opposite chamber, 4 if the bill 1 Massachusetts did not make bill texts available for 2007. Kansas did not make bill histories available for 2007. 2 We estimate that a small percentage (< 10%) of all introduced bills fit these categories. Appropriations bills were randomly drawn, discarded, and replaced in fewer than 40 instances. Technical bills were even more infrequently drawn, and came only from more professionalized legislatures like Illinois and New York. 10

was vetoed by the governor, and 5 if the bill ultimately became law. Of these bills, the majority (301) died in committee. For the remaining bills, 31 were reported from committee without receiving a floor vote, 43 passed only in their originating chamber, 11 were approved by the legislature as a whole but vetoed by the governor, and 114 became law. We also conducted separate analyses using Bill Enactment as a binary dependent variable capturing whether a bill was ultimately passed into law (1) or not (0). Descriptive statistics for all subsequent variables are presented in Table A1 in the appendix. Our principal independent variable is Density. The variable is a count of all interest groups registered to lobby in the state where the bill was sponsored. Like the federal government, states do not require registration for every act of lobbying. Rather, states require groups to register if they engage in some minimum amount of lobbying activity throughout the year. As a result, group counts likely provide a conservative estimate of interest group activity in each state, setting up a difficult empirical test of our expectations. Data from 2007 come from Lowery et al. (2013), who collected the data from the National Institute on Money in State Politics (NIMSP). Interest group populations vary widely; only 279 groups registered to lobby in South Dakota that year, while 3,335 groups registered to lobby in Florida. We included a number of control variables to capture variation across state political contexts that could aid or hinder bill progress. As a first step, we controlled for the demands made on legislators time and attention. We included the variable Legislative Professionalism. More professional legislatures meet in longer sessions, consider a greater number of bills, and have more resources available to them in terms of staff support. They also tend to encourage the formation of more densely populated interest group systems (Kattelman, 2014). Controlling for this variable is crucial to distinguish the effects of interest group density from legislative capacity in determining whether bills advance or not. 3 Data for this variable come from a measure of legislative professionalism calculated 3 Our measures of density and legislative professionalism are strongly and positively correlated (r = 0.73), suggesting a possibility of multicollinearity in the model. As a robustness check, we estimate both bivariate and multiple regression models below. The coefficient estimate for the density variable is roughly similar in both models. In any case, including a control for legislative professionalism should create a problem of efficiency rather than biased estimates; inflated standard errors resulting from multicollinearity would provide a harder test of our hypothesis. 11

by Bowen and Greene (2014). We also controlled for a number of institutional variables demonstrated in the legislative politics literature as important in determining the fate of bills. First, we control for the party of each bill s sponsor. Majority Party Sponsor is a dummy variable with values of one indicating that the bill sponsor was a member of the majority party in his or her chamber. We do not expect bills sponsored by members of the minority to advance as far as bills sponsored by majority members (Cox and McCubbins, 2005). Second, we controlled for Cosponsors, a count of the number of cosponsors a bill has. More cosponsors should predict greater bill success, since it signals to non-sponsor legislators that the bill has a greater base of support (Browne, 1985). Third, we include a measure of party control of state governments. Bills will be more likely to pass when the same party controls both chambers of the legislature and the governor s office. We use Unified Government, a dummy variable for which values of one indicate unified party control. Fourth, we control for the ideological orientation of each originating chamber, under the expectation that liberal governments are more willing to pass changes to policy than conservative governments. Government Liberalism is measured as the ideal point of the median member of the chamber where each bill originated, using data from Shor and McCarty (2011). Finally, we sought to control for issue salience on these proposals, under the assumption that bills would advance further on salient issues with broad public support or die quickly on salient issues with broad public opposition. Following Burstein (2014), we selected a random sample of state bills to observe the influence of interest group populations in routine policymaking situations, with the expectation of observing some salient and some non-salient policies. However, random sampling yielded vanishingly few bills addressing salient issues. Like Burstein (2014), we considered an issue salient if a poll measured public opinion on it. We compared the topics of our bills to the 39 policy issues for which Lax and Phillips (2012) obtained state-level public opinion estimates. By our count, 16 of the 500 bills (3.2%) in our model could be considered as addressing salient issues. We consider this a high estimate; we counted a bill as salient even if it 12

had only a tangential relationship to an issue for which a poll was available. For the vast majority of the bills we examined, public opinion estimates were simply not available. We suspect that most people simply do not hold strong opinions on most of our bills due to the technical or mundane subjects that they address: the licensure requirements for speech language pathologists, the disposal of agricultural waste, and the sale of influenza vaccines to state health agencies. Nonetheless, we included the binary variable Salient Issue, with a value of 1 indicating the bill addresses an issue with a public opinion poll conducted and a value of 0 indicating otherwise. Results We begin the statistical analysis by examining how interest group density relates to each bill s progress through the legislature. We expect that as interest group density increases, policy change proposals will be defeated earlier in the legislative process. To test this expectation, we estimate the following model: Bill Progress = β 0 + β 1 Density + β 2 Controls + ɛ If our argument is correct, we should expect to see a negatively signed coefficient estimate associated with our density variable. Because our dependent variable is ordinal, we estimate the models using ordered logistic regression. Table 1 presents our initial results. We find support for our expectation that bills are less likely to progress when interest systems are more densely populated. The coefficient estimate for the independent variable density in Model 1 is signed in the expected negative direction. The estimate is statistically significant at the 0.05 level of confidence. Using the bivariate results, we calculated the predicted probabilities of a bill advancing through each of the five stages of the legislative process given the size of the state s interest group community. Figure 1 displays the results, with probability on the vertical axis and density on the horizontal axis. Each of the five lines represents a differing end stage for the bill. The figure shows that as the density of groups increases from its minimum to its 13

Table 1: Density and Bill Progress Dependent variable: Bill Progress (1) (2) Density (in 100s) -0.21-0.28 + (0.09) (0.16) Legislative Professionalism 0.02 (0.09) Majority Party Sponsor 0.81 (0.22) Cosponsors 0.01 + (0.01) Unified Government 0.74 (0.22) Government Liberalism -0.12 (0.17) Salient Issue -0.05 (0.51) Cut 1 0.12 1.00 (0.15) (0.26) Cut 2 0.39 1.29 (0.15) (0.26) Cut 3 0.81 1.73 (0.16) (0.27) Cut 4 0.93 1.86 (0.16) (0.27) Observations 500 500 AIC 1114.40 1095.17 BIC 1135.47 1141.53 Note: + p<0.1; p<0.05. Coefficient estimates obtained using ordered logistic regression. Significance tests are two-tailed. maximum, the probability of an introduced policy proposal failing to pass from committee increases from 0.53 to 0.70. At the same time, as the density increases from its minimum 14

Figure 1: Predicted Probability of Bill Progress by Density Predicted Probability 0.2.4.6.8 0 500 1000 1500 2000 2500 3000 3500 No. of Registered Lobbying Groups Pr(Die in Committee) Pr(Pass Orig. Chamber) Pr(Signed by Governor) Pr(Pass Committee) Pr(Pass Legislature) to its maximum, the probability of an introduced policy proposal being signed into law by the governor decreases from 0.28 to 0.16. The probability of a proposal advancing to any intermediate stage of the legislative process remain fairly constant across values of density. This figure provides evidence in line with expectations that policy change proposals are less likely to progress through the legislature in states with more densely populated interest group communities. The relationship between density and bill progress may be confounded by other factors such as professionalism, a bill s internal support from legislators, or party control of government. To make sure our finding is robust to these factors, we estimate a second model including controls in Table 1. Even accounting for the other variables, we still find that proposals sponsored in states with greater interest group density do not advance as far in the legislative process, as indicated by the negatively signed coefficient estimate. However, the estimate is significant at the.10 level confidence. Among the control variables, the results show that proposals sponsored by members of the majority 15

Table 2: Density and Bill Enactment Dependent variable: Bill Enactment (1) (2) Density (in 100s) -0.27-0.45 (0.12) (0.20) Legislative Professionalism 0.10 (0.10) Majority Party Sponsor 0.93 (0.29) Cosponsors 0.00 (0.01) Unified Government 1.00 (0.27) Government Liberalism -0.16 (0.21) Salient Issue -0.16 (0.61) Constant -0.86-1.89 (0.18) (0.34) Observations 500 500 AIC 534.93 518.57 BIC 543.36 552.28 Note: + p<0.1; p<0.05. Coefficient estimates obtained using logistic regression. Significance tests are two-tailed. party advance significantly further in the process, as well as bills sponsored in states with unified party control. A greater number of bill cosponsors also positively predicts proposal advancement, but the coefficient estimate is significant at the modest.10 level of confidence. Legislative professionalism, government liberalism, and issue salience do not appear to have any appreciable association with bill advancement. 16

Of course, interest groups are interested in more than how far a bill advances in the process they ultimately care about proposals passing into law or failing. We also replicated the models from Table 1 using a binary dependent variable indicating whether a policy proposal passed into law. The results of the models using this dependent variable are presented in Table 2. The results of Model 1 show a negative association between interest group density and bills passing into law as indicated by the negatively signed, statistically significant coefficient estimate. Adding the same set of controls as used previously to Model 2, we still find a negative and significant association. We calculated a predicted probability of a proposal passing into law given density while holding the continuous control variables at their means and the categorical controls at their medians. The results are presented in Figure 2, with probability of bill passage on the vertical axis and density on the horizontal axis. In line with the results of the ordered logit above, the probability of a policy change being enacted is 0.29 at the minimum value of density, but decreases to 0.08 at the maximum value of density, controlling for other variables in the model. This alternate specification provides further evidence that the likelihood of policy change occurring meaningfully decreases in political environments where more interest groups are present. Among the control variables in Model 2 Table 2, majority party sponsorship and unified party control of government are positively and significantly associated with bill passage. However, none of the remaining controls exhibit a significant association with bill passage, including the number of cosponsors. Given the results of both tables, our findings suggest a greater number of cosponsors might help a policy proposal advance farther through the legislature but does not necessarily aid in its ultimate passage. 4 Overall, the results indicate that the number of interest groups is negatively associated with the likelihood of any single policy proposal becoming law. As the number of interest groups increases, bills are more likely to die in committee and less likely to advance to the governor s desk for signature. One possibility that our analysis so far does not address 4 As a robustness check, we estimated the models above using multilevel regression with random intercepts for states. The results, presented in Tables A2 and A3, continue to indicate a negative association between density and bill progress and enactment. However, the coefficient estimates are not significant at the.05 level of confidence. 17

Figure 2: Predicted Probability of Bill Enactment by Density Predicted Probability of Bill Enactment 0.1.2.3.4 0 500 1000 1500 2000 2500 3000 3500 No. of Registered Lobbying Groups is whether certain types of interest groups are more effective at killing policy change proposals than others. We address this possibility in the following section. Cross-Sector Variation The negative association between interest group density in the state and policy change could be driven by certain economic sectors with larger numbers of interest groups (e.g. business) finding greater success in defeating new policy proposals. To make sure this was not the case, we tested whether differences in the number of lobbying groups across sectors resulted in differences in the association between density and policy change. We relied upon a coding scheme of 11 economic sectors defined in NIMSP s original group count data: communications, education, energy and natural resources, finance and real estate, government organizations, health, ideological and single-issue groups, manufacturing and production, nonprofits, retail and business, and transportation. NIMSP categorized 18

about half the interest groups in their data set in 2007; Lowery et al. (2013) categorized the remaining groups in accordance with NIMSP s scheme and provided details on coding procedures and intercoder reliability. We then categorized our 500 bills to match the 11 economic sectors. Two coders were provided a one-sentence summary and the full text of each bill and asked to assign it to the sector most likely to lobby on it regardless of the side it might take. We chose to match each bill to only one sector, rather than all possible sectors that might lobby on the bill, to provide a hard distinction between whether the sector size or the size of the community at large is more influential in determining a bill s fate. As an example of how bills were coded, an Illinois bill requiring businesses to refund gift card balances under certain circumstances was assigned to the Retail and Business Services sector. A Texas bill changing the salaries paid to correctional officers was assigned to the Government Organizations sector. Bills that did not fit into any sector were coded into an unknown category. Generally, unknown bills were particularistic bills for which it was unclear that any broader community of organizations would have interest in lobbying on the bill: for example, a bill in New York that retroactively extended pension benefits to a single retired state employee. The coders agreed in 69% of cases. The coders discussed discrepancies in person and agreed upon a final categorization for each bill. Having matched bills to sectors, we created the independent variable Sector Density, which is the count of interest groups lobbying within the sector in the bill s state. Table 3 provides a count of bill topics by sector. The bills are distributed across all sectors. By far, the most common sector that bill provisions affected was government agencies, with 38.4% of our sample. These bills usually dealt with matters of local governments, law enforcement, and criminal justice. The remaining bills were fairly evenly spread across the remaining sectors, with the fewest numbers of bills affecting the communications sector. We also calculated the mean number of lobbying groups associated with each sector in the fourth column from the left in Table 3. The number of groups in each sector varies quite a bit; the average state has 339 registered groups that lobby primarily on matters of retail and business, while it only has 16 ideological or 19

Table 3: Bills and Lobbying Groups by Sector Sector No. of Bills % Bills Average No. of Lobbying Groups Government Organizations 192 38.4% 115 Health 44 8.8% 191 Education 44 8.8% 140 Finance, Insurance, & Real Estate 39 7.8% 182 Retail & Business 38 7.6% 339 Energy & Natural Resources 29 5.8% 96 Manufacturing & Production 27 5.4% 168 Transportation 25 5.0% 32 Nonprofit 24 4.8% 64 Ideology & Single-Issue Groups 22 4.4% 16 Communications 7 1.2% 36 Unknown 10 2.0 % 0 single-issue lobbying groups. 5 We conducted two tests to assess the role of economic sectors in the policymaking process. First, we replicated our models from Table 1 above but included fixed effects for the sectors of each bill in our statistical model. If mobilization within certain sectors is primarily responsible for the negative association between overall density and bill progress, then we should expect to see a change in the size and significance of the coefficient estimate for the density variable when fixed effects are included. Second, we replaced the independent variable of interest group density with sector density, measured as the count of interest groups within the bill s lobbying sector in its state of origin. If it is the case that the size of the lobbying community within the relevant sector is responsible for depressing the likelihood of policy change on bills the community cares about, we should expect to see a negatively signed, statistically significant coefficient estimate for the sector density variable. The results of these tests are presented in Table 4. The dependent variable in these models is Bill Progress, a count of the legislative stages a bill proceeds through. 6 First, 5 The original data provide counts of unknown interest groups from their lobbying registration data, the density of which sometimes reaches into the hundreds. However, for the purposes of this analysis, the density of unknown groups is assigned a value of 0 because our unknown bills are considered to have no broader sector of groups interested in lobbying on the proposal. 6 We estimated all models in Table 4 using the binary dependent variable Bill Enactment as well, and report those results in Table A4 of the appendix. The results from this analysis do not show any meaningful differences in the findings. 20

Table 4: Density and Bill Progress across Economic Sectors Density -0.19-0.30 + Dependent variable: Bill Progress (1) (2) (3) (4) (0.09) (0.17) Sector Density -0.00-0.00 (0.00) (0.00) Legislative Professionalism 0.05-0.07 (0.09) (0.06) Majority Party Sponsor 0.90 0.81 (0.23) (0.22) Cosponsors 0.02 0.01 + (0.01) (0.01) Unified Government 0.79 0.61 (0.23) (0.21) Government Liberalism -0.16-0.04 (0.18) (0.16) Salient Issue 0.03-0.07 (0.53) (0.50) Sector Fixed Effects Yes Yes No No Cut 1-0.77-0.24 0.28 1.18 (0.62) (0.66) (0.13) (0.23) Cut 2-0.50 0.05 0.55 1.47 (0.62) (0.66) (0.13) (0.23) Cut 3-0.07 0.51 0.97 1.91 (0.62) (0.66) (0.13) (0.24) Cut 4 0.05 0.64 1.09 2.04 (0.62) (0.66) (1.09) (0.24) Observations 500 500 500 500 AIC 1124.84 1100.66 1117.54 1097.28 BIC 1192.28 1193.39 1138.62 1143.64 Note: + p<0.1; p<0.05. Coefficient estimates obtained using ordered logistic regression. Significance tests are two-tailed. 21

we replicated the two models in Table 1 in Models 1 and 2 of Table 4 and included fixed effects for the 11 economic sectors plus unknown bills. However, the coefficient estimate for the density variable in the first and second models remains signed in the negative direction and statistically significant (at the.05 level in the first model and at the.10 level in the second). This finding provides evidence that, even when accounting for differences across sectors, interest group density in a political environment as a whole decreases the likelihood of policy change. For our second test, we estimated a model using the independent variable Sector Density instead of our measure of statewide group density. Models 3 and 4 in Table 4 provide no evidence in support of the expectation that sector density is associated with a bill s progress through the legislature. The coefficient estimate in the third model is substantively small and not statistically different from zero. Adding the full set of controls in Model 4 yields the same results. Among the control variables, majority party sponsorship, unified party control of government, and having a greater number of cosponsors continue to positively and significantly influence the progress of the bill through the legislative process. Taken together, these models demonstrate that accounting for the economic sector a policy proposal affects does not meaningfully change the earlier finding that more interest groups in a state result in a higher likelihood that a bill will pass into law. In other words, the size of a state s interest group population as a whole, rather than the type of industry the bill affects, influences policy change. Groups from various sectors might lobby on any given bill, increasing the likelihood of opposition and defeat. But the overall supply of groups in an environment seems to be what matters most in affecting policy change. Discussion Our analysis provides support for our expectations that the increased density of interest groups in a state inhibits the progress any given policy change proposal will make through the legislative process. The results show that proposed policy changes tend to be defeated earlier in the process and are ultimately less likely to pass into law in states with densely 22

populated interest communities. Further analysis shows that this relationship is not complicated by differences in interest community sizes across economic sectors. Rather, the size of interest communities organized and available to lobby within a state capital influences the likelihood of policy change occurring, since any group might reasonably choose to lobby on any bill. The findings we present build upon research demonstrating that interest groups are quite effective at preserving the status quo (Baumgartner et al., 2009; Gray and Lowery, 1995). Our study contributes to this literature by observing that characteristics of broader group systems play a role in determining policy outcomes, complementing literature examining the actions and spending habits of individual groups (Grossmann and Pyle, 2013; Lewis, 2013). However, we expand the scope of study to multiple political environments and show that system-level dynamics like group density also influence the probability of policy change. A limitation of this study is that we are unable to observe the exact mechanism by which larger group populations reduce the chance of successful policy change. It could be that a larger number of groups leads to greater negative power of the lobbying community over policy outcomes (Bachrach and Baratz, 1962; Lowery, 2013). But it could also be that the increased lobbying activity accompanying a larger number of groups leads to information overload and increased uncertainty among lawmakers. A research design bringing more detailed lobbying disclosure data about the sides and extent of interest group involvement on policies in multiple legislative contexts could be a good step towards distinguishing between these two mechanisms. Our study is also constrained by time, since we only observe outcomes in 2007. Observations of policy change over multiple years, as in studies by Baumgartner et al. (2009) and?, could give a more complete picture of the long-term influence that group populations wield. This work has substantive implications for the capacity of political institutions to respond to changes in society that require policy solutions. Interest group populations in Washington, D.C. (Wilson, 2015) and the states (Lowery et al., 2013; Lowery et al., 2015) are continuing to grow in size. According to our research, part of the explanation for 23