Congressional Processes and Public Approval of New Laws. James M. Curry Department of Political Science University of Utah

Transcription

1 Congressional Processes and Public Approval of New Laws James M. Curry Department of Political Science University of Utah March 29, 2017 Prepared for presentation at the 2017 Annual Meeting of the Midwest Political Science Association, April 7, This research was supported by the Betty Glad Memorial Fund at the University of Utah.

2 Voters generally care about ends, not means; they judge government by results and are generally ignorant of or indifferent about the methods by which the results are achieved. Samuel Popkin (1994, 99) I don t think procedural stuff really resonates with most Americans. It may add generally to their cynicism, but it is accomplishment or lack of it that matters much more." Tom Daschle, former Senate majority leader 1 Laws, like sausages, cease to inspire respect in proportion as we know how they are made. John Godfrey Saxe 2 Does how Congress makes law affect Americans support for those laws? Since the 1970s, partisan conflict has been on the rise in Congress (Barber and McCarty 2015), and concomitant to this rise has been an increased use of unorthodox and aggressive legislative procedures to make law. Once characterized as a decentralized and norm-driven institution that typically adhered to regular order legislative procedures, the contemporary Congress frequently bypasses traditional deliberative processes when considering legislation, and today s congressional leaders are willing to use any and all procedural tools at their disposal to achieve their goals (Sinclair 2016, Smith 2014; Hanson 2014). Recent congressional majorities have been willing to advance their proposals by any procedural means necessary and congressional minorities have been willing to go to similar lengths to obstruct. The public expresses a general preference for civil and bipartisan policymaking, and there is ample empirical evidence that partisan conflict and legislative politicking can damage public approval of Congress (Harbridge and Malhotra 2011; Jones 2013; Ramirez 2009) But does procedural wrangling influence public approval of the laws being wrangled, too? This 1 Quoted in Adam NaGourney, Procedural Maneuvering and Public Opinion, New York Times, March 20, On the origin of this quote see, Fred R. Shapiro, Quote Misquote, New York Times, July 21,

3 question has been little studied. Harbridge, Malhotra, and Harrison (2014) find that more partisan lawmaking has little impact on public approval of laws passed by Congress, but what about the processes used, irrespective of the levels of partisanship? When Congress resorts to nefarious-sounding processes such as omnibus appropriating, budget reconciliation, selfexecution, fast-tracking, and the like is the public more skeptical of the new law passed, perhaps seeing the process by which it was passed as illegitimate or corrupt? Most existing scholarship suggests the public s approval of a new policy is primarily driven by its congruence with the preferences and ideological mood of the public (Downs 1957; Enelow and Hinich 1984; Ramirez 2013; Wlezien 1995), its perceived effects (Erikson 1989; Nadeau and Lewis-Beck 2001; Tufte 1975), and at a micro-level, the partisan attachments and biases of individual citizens (Campbell et al 1960; Green, Palmquist, and Schickler 2002; Harbridge, Malhotra, and Harrison 2014). As the epigraphs suggest, scholars and practitioners alike dismiss procedural-driven explanations of public support for laws. Yet, we know people dislike sausage-making (Hibbing and Theiss-Morse 2002), have a sense of procedural justice (Tyler 1994), and prefer clean and decisive policymaking action (Lebo, McGlynn, and Koger 2007; Lebo and O Green 2011). It may be that unorthodox approaches to lawmaking suppress approval of not just Congress, but also laws. I assess the impact of lawmaking procedures on public approval of new laws through two survey experiments, the first embedded in a module of the 2014 CCES, and the second fielded using Amazon s Mechanical Turk (MTurk). Both experiments ask respondents to read descriptions of hypothetical new policies passed by Congress, modeled on the language used in newspaper accounts of congressional action. Respondents in control groups just read about the law, while those in the experimental groups are told about processes used to enact the law, or the 2

4 amount of partisan conflict present as the law was considered and passed. The results show that reading about the use of procedural tactics reduces support for new laws, and it does so more consistently and often to a larger degree than when respondents are told about the presence of partisan conflict. These findings have serious implications for public support for governmental policies in an era when Congress frequently relies on unorthodox procedures to make law. RISING CONGRESSIONAL UNORTHODOXY Once characterized as a decentralized legislature governed by institutional norms and an adherence to traditional regular-order legislative processes, the contemporary Congress is much more centralized and frequently bypasses traditional deliberative processes, and its members are more willing to employ any and all procedural tools and their disposal to achieve their legislative goals. Barbara Sinclair (2016, 5) termed this new mode of lawmaking unorthodox lawmaking, as the legislative process on major legislation is now regularly characterized by a variety of what were once unorthodox practices and procedures. What manner of changes have occurred? More legislation is bypassing the committee stages of the legislative process ahead of floor consideration and passage (Bendix 2016; Curry 2016), and fewer bills are subject to committee hearings and testimony (Lewallen, Theriault, and Jones 2016). More bills are considered under restrictive special rules than in the past, which allow the majority party leadership to structure the terms of debate and limit the ability of the members to offer amendments (Bach and Smith 1988; Finocchiaro and Rohde 2008; Sinclair 1992). Legislative decision-making has moved behind the scenes, with party leaders assuming the responsibilities for cutting deals and crafting legislative packages that were once possessed by committee chairs (Hanson 2014; Sinclair 2016). Congressional resources have been centralized in party leaders (Aldrich and Rohde 2000; Rohde 1991) allowing leaders to keep 3

5 even their own rank-and-file in the dark as legislative language is being negotiated and written (Curry 2015). Minority parties, for their part, have responded to these changes by being more obstructionist, including more frequent filibustering in the Senate (Koger 2010; Wawro and Schickler 2013), and as a result both parties are now willing to use every procedural advantage at their disposal during the consideration of legislation (Smith 2014). Importantly, these changes have occurred in response to changes in the broader political environment. Centralized, restrictive, and behind-the-scenes processes help congressional leaders overcome legislative obstacles and build broad support for policy proposals in an era of intensely divided parties, small congressional majorities, and frequent divided government (Curry and Lee 2016). Old-style, decentralized processes provide ample opportunities for dedicated opponents to obstruct and embarrass the majority (Rohde 1991; Sinclair 2016). Unorthodox processes empower majorities to avoid obstructionism, and they may also help majorities achieve partisan policy outcomes (Aldrich and Rohde 2000), and protect their members from difficult votes (Cox and McCubbins 2005). While this approach to lawmaking is effective, many scholars and observers view unorthodox lawmaking processes as anti-deliberative, resulting in dysfunction (Lewallen, Theriault, and Jones 2016; Mann and Ornstein 2006; 2012), poor-quality lawmaking (Drutman 2016), and an erosion of Congress s public esteem (Ramirez 2009). Do such attitudes affect how the broader public responds to congressional action? Might unsavory sausage making leave a foul taste in the mouths of a citizenry feasting on a plate of new policies? The next section explores reasons to suspect it does. WHY PROCESS MIGHT AFFECT APPROVAL OF POLICIES Most scholarship suggests people like or dislike a new policy based on their views of the policy 4

6 itself, their partisan attachments, and their judgments of a policy s effects. In the language of Downs (1957), a person s satisfaction with a new law is a function of the law s congruence with his or her policy preferences. In general, liberals are more satisfied with liberal policy outputs and conservatives more satisfied with conservative outputs. Both the Downsian median voter theory and the probabilistic voting model (see, e.g., Coughlin 1992; Burden 1997; Enelow and Hinich 1984) buttress this perspective, and empirical work finds evidence that the public s policy preferences matter for the approval of new laws (e.g., Page and Shapiro 2010; Wlezien 1995). Even if the public is not informed enough to react to the specifics of a new policy, partisan attachments can work as a heuristic, influencing people s evaluations of issues and policies. Simply, Democrats generally react positively to laws pushed by Democrats, and Republicans tend to prefer laws pushed by Republicans. For instance, Shanks and Miller (1991) show that individuals partisanship affected evaluations of the policy direction of the Reagan administration, and subsequently their votes in the 1988 presidential election (see, also, Miller and Shanks 1996). Layman and Carsey (2002) show that the signals of partisan elites lead partisans in the public to adopt more partisan positions across issue dimensions. Bartels (2002) finds that individuals reactions to governmental performance on various issues are influenced by their partisanship. Evidence is found in policy-specific studies as well, including on how public opinion evolved on health care reform over time (Henderson and Hillygus 2011). Scholars also find that public reaction to a policy is based on its effects or consequences, or at least the perception of those effects. A lengthy literature on economic voting shows that vote choices are affected by their views of the state of the economy as a consequence of public policy (e.g., Tufte 1975; Kinder and Kiewiet 1979; Erikson 1989; Gomez and Wilson 2001; Nadeau and Lewis-Beck 2001). More broadly, Erikson, MacKuen, and Stimson (2002) find that 5

7 the public s mood for governmental action is influenced by the perceived effect of public policies on the economy, including levels of unemployment and inflation. Beyond the economy, the concept of retrospective voting broadly supports the idea that the preferences and voting decisions of individuals in the public are based on satisfaction with the outcomes of policies (e.g., Key 1996; Fiorina 1981; Healy and Malhotra 2013). Further, Marsh and Tilley (2010) find that the effects of policy outputs affect overall support for parties in future elections. Clearly, the public s approval of a new policy is influenced by the direction of the policy, the effects of the policy, and partisan support or opposition for the policy. In other words, approval of laws is driven, in part, by policy preferences. However, there are reasons to suspect the public s reactions to new laws is also affected by how those laws are made. The public has process preferences 3 as well, and these preferences can shape approval of new laws. Hibbing and Theiss-Morse s (2001; 2002) research about the public s preference for a stealth democracy underscores this point. Americans generally view politicking and political maneuvering negatively. The individuals in Hibbing and Theiss-Morse s focus groups saw political conflict as driven by the influence of special interests, evidence of a disconnect between lawmakers and ordinary Americans, and a consequence of corruption in Washington. Abstruse legislative procedures should not be necessary for Congress to make a law because the conflict that makes these procedures necessary would not exist were the government more responsive to ordinary citizens. This is in part why the public prefers clean and decisive legislative action to prolonged deliberation and rewards parties for achieving legislative victories in subsequent elections (Lebo, McGlynn, and Koger 2007; Lebo and O Green 2011). A consequence of these 3 This term was coined by Hibbing and Theiss-Morse (2001; 2002) but is used more broadly here to describe the preference people have for the manner of policymaking processes employed by lawmaking elites. 6

8 attitudes is that Americans are likely to be suspicious when they read or hear about Congress taking extraordinary steps or resorting to unusual processes to pass laws. They may, as a result, be less inclined to support the resulting policies. Americans may also have a sense of procedural justice (Tyler 1994). In other words, they may view some governmental processes as inherently more or less fair and this may color their reactions to government action. Indeed, when people view Congress or the Courts as having violated the rules or proceeded in an unfair manner, public support for these institutions declines, and the legitimacy of their decisions is viewed as more suspect (Gangl 2003; Gibson 2001). Over the long run, political processes viewed as unfair may drive down support for a governmental system as a whole, and increase support for broad reforms (Banducci and Karp 1999; Rose and Mishler 2009). Doherty and Wolak (2012) find the effects of procedural fairness are especially sharp when processes are unambiguously fair or unfair. None of these studies directly analyze the impact of perceived procedural fairness on approval for laws, but it doesn t seem like a stretch. Just as people may question the legitimacy of governmental institutions when they view political processes as unfair, they may question the policies produced by these unfair means. 4 The use of unorthodox procedures in Congress can also sharpen partisan attacks against a law, which may undermine support among those identifying with one or the other political party. Opponents, while perhaps unable to stop a bill from passing Congress, can complain that the majority is ramming the bill through, writing the policy in a shoddy way, stifling debate, advancing the bill in an unfair manner, and maybe even cheating to get the law passed. This 4 Citizens probably also do not need to have an incredibly thorough or sharp understanding of legislative processes to be given the impression that things proceeded in an unfair or unusual fashion. They simply need to be given an impression that this is so, which could easily be communicated by the media or by partisan elites. 7

9 kind of messaging was widespread during the consideration of the Affordable Care Act in late 2009 and early At the time, conservative commentators focused some of their sharpest attacks on the processes Democrats used enact the ACA. Brian Darling of the Heritage Foundation argued that the use of the budget reconciliation process to enact the law in its final form constituted cheating. 5 Others, like Michael Barone of the American Enterprise Institute, argued in retrospect that if the Democrats had constructed the bill through more careful deliberation it would probably have had more Republican support. 6 Arguments like these that the majority is cheating or breaking the rules to pass a bill are intended to engender distrust and dissatisfaction, and there is clear evidence that partisan elite cues like these influence opinion formation in the public (Druckman, Peterson, and Slothuus 2013). Altogether, there are compelling reasons to expect that the use of unorthodox lawmaking tactics may generally drive down public support for laws passed by Congress. Of course, reality may also be more nuanced, fluctuating by the policy issue at hand, or with individuals partisan attachments. Indeed, as Harbridge, Malhotra, and Harrison (2014) find, partisans are willing to set aside their general preferences for bipartisanship if it means their party gets a policy win. The same may be true for process. Partisans in the public may approve of their party using whatever procedural tools they can to advance their party s proposals, or block the proposals of the other party, but they may cry foul when the other side takes advantage of the same legislative tools. In other words, partisans may discount their process preferences or their sense of procedural justice if it means they get the policy they want. Further, the amount of outrage (or lack thereof) we see from partisans in the public may be reinforced by the partisan attacks lobbed by partisan elites in 5 Brian Darling, Lying and Cheating to Pass Obamacare. RedState.com, February 28, Michael Barone, If Only Obamacare Had Been Passed with Careful Deliberation, RealClearPolitics.com, October 1,

10 Washington, sending cues to those in the public about whether they should and should not be outraged. In the analyses below, these possibilities are considered. RESEARCH DESIGN AND MEASUREMENT Using direct observational data to assess the influence of congressional processes on popular support for new laws would be quite challenging. With so many intervening variables, identifying the specific impact of legislative maneuvering may be impossible. For that reason, I conducted two survey experiments. Each asks respondents to read a description of a hypothetical new law passed by Congress and describe their level of approval or disapproval. Additional information about partisanship present during consideration of the law, or the use of unorthodox procedures, is provided to those randomly assigned to several treatment groups. To try to ensure the experiments were externally valid, I took several steps. First, the experiments have respondents read about the passage of laws across five different issue areas transportation policy, energy policy, defense policy, abortion policy, and tax policy. This allows me to ensure the results are not a function of choosing a single policy. It also allows me to see if the effects of the treatments vary by policy. Second, I loosely based the description of each policy on an actual bill considered or passed by at least one chamber of Congress during the last two decades, providing a dose of realism to the policy descriptions. However, I was careful to draw on policy proposals that were not particularly prominent, minimizing concerns of pretreatment bias (Druckman and Leeper 2012). Third, I based the language used for the descriptions on the kind of language used in major newspaper reports of congressional action. Again, this places the experiments within in a realm of reality as the information respondents are reading is similar to the kind of information they might read in a short article in the New York Times. The appendix contains all language used for both experiments. 9

11 The first experiment was embedded in a module of the 2014 CCES with 1,000 respondents. The experiment asks respondents to read about and register their approval or disapproval for a hypothetical transportation policy on a four-point scale (strongly disapprove, disapprove, approve, and strongly approve). Respondents are randomly selected into one of four groups, each of which is provided with a slightly altered description. The first group (policy only) is the control group. These respondents were provided a straightforward description of the policy. The second group (partisan) was provided the same description, but was also told about party-line votes in the House and the Senate and unified opposition from congressional Republicans. The third group (procedure) was given the same policy description, but also told about the use of certain unorthodox legislative procedures specifically the use of budget reconciliation to avoid filibusters from the bill s opponents. Untold is the partisan nature of any opposition to the policy and which party used the procedures. The fourth group (combined) was provided with the policy description as well as the details about both the partisan conflict and the use of unorthodox procedural tactics. Having separate partisan and procedure treatments allows me to assess the relative impact of each, and find out if procedural wrangling had an effect independent of partisan conflict. Additionally, the combined treatment allows analyses to assess if the effect of procedure is conditioned by explicit reference to partisanship. Figure 1 provides an overview of the language used in this experiment. The second survey experiment used a sample of 2,009 adult Americans recruited using Amazon s Mechanical Turk (MTurk). 7 Respondents completed the survey in October Use of Amazon s MTurk to recruit participants in political science studies has become increasingly common. Berinsky, Huber, and Lenz (2012) find that samples drawn using MTurk are typically better educated and more liberal than those drawn using national probability samples, but are more representative than the convenience samples used for most experimental studies. 10

12 FIGURE 1 Sample of Language used in the CCES and MTurk Experiments CCES 2014 Language [italics] = Language added for the 'partisan conflict' [bold] = Language added for the 'procedure' [underline] = Language added in the 'combined' 2016 MTurk Language (Tax policy example) [italics] = Language added for the 'partisan conflict' [bold] = Language added for 'Procedure #1' [underline] = Language added for 'Procedure #2' After months of negotiation, Congress passed a major transportation bill on Wednesday. The bill [, which cleared the House by a close party line vote, , and was unanimously opposed by House Republicans,] allots $286 billion for roads, bridges, mass transportation, and safety projects over the next six years, and is expected to create thousands of new jobs [and was opposed by numerous lawmakers dissatisfied with many of the underlying policies. House and Senate leaders had to rely on a rarely used procedure known as budget reconciliation to push the bill through the floor of each chamber. This arcane procedure allowed the Senate to sidestep any filibusters and pass the bill by a simple majority.] According to the House Committee on Transportation and Infrastructure, 80 percent of the money will go to road projects, 18 percent to mass transit, and the remainder to road safety and other projects. In addition, the bill imposes sweeping new auto safety standards, including provisions that raise rollover standards on minivans and other large vehicles, and a provision that requires the crash test ratings of cars and trucks to be posted on window stickers visible to consumers. Perhaps most sweeping are new rules that would withhold federal transportation funds from states that do not pass laws allowing police officers to pull over drivers for not wearing their seatbelts. [Senate Republicans strenuously opposed this policy arguing it violated state sovereignty and represented an overreach by the federal government. The bill passed the Senate by a close party-line vote last month.] [However, the budget reconciliation procedure restricted Republicans from mounting a filibuster and the bill ultimately passed by a close party-line vote.] [Several senators threatened to filibuster the bill over these rules, but due to the reconciliation process were unable to mount an effective opposition.] The bill is expected to be signed by the President by the weekend. Congress passed new tax legislation on Wednesday [after prolonged debate and strenuous opposition from numerous lawmakers]. The bill, which [cleared the House by a close party line vote, , and was unanimously opposed by House Republicans,] cuts taxes modestly for low- and middle-income families while boosting taxes on high-income individuals and businesses, is expected to increase government revenues by $3 trillion over the next decade. Nearly $1 trillion of that total will be achieved by closing tax loopholes for wealthy Americans, including one loophole that allows households making more than $250,000 a year to avoid some taxes by passing investment income through partnerships or businesses. Other revenue boosters include restoring estate tax rates and thresholds to 2009 levels, imposing new fees on financial institutions, and setting a new minimum tax rate of 30 percent on people who earn more than $1 million. [Senate Republicans strenuously opposed these provisions, arguing that the new taxes were job killers, and another example of Democrats trying to grow the size of the government on the backs of American tax payers.] In addition, several proposals aim to ease the tax burden on lower- and middle-income Americans, including a provision that triples the maximum allowable child care tax credit, and another that strengthens the earned-income tax credit for workers not raising children. [House and Senate leaders had to rely on a rarely used procedure known as budget reconciliation to push the bill through each chamber. This arcane procedure allowed Senate leaders to sidestep attempts by numerous senators to filibuster the bill.] [Congressional leaders used a procedure known as self-execution to ease the passage of the bill. This procedure made various changes to the bill without any debate, and immediately before passage. As a result lawmakers opposed to these new provisions were unable to voice their concerns.] The bill is expected to be signed by the President over the weekend. 11

13 Similar to the CCES experiment, respondents were randomly assigned into a control group and three treatment groups, but they were also randomly assigned to read about one of four different hypothetical policies an energy policy, a defense policy, an abortion policy, and a tax policy. The treatments groups are also slightly different from the CCES experiment. Again, the control group (policy only) was only provided a description of the policy. The second group (partisan) added language about explicit conflict between the parties and opposition from one party or the other, varying by policy. Republicans were described as unanimously opposed to the energy and tax policies, and Democrats as unanimously opposed to the abortion and the defense policies. The final two treatment groups provided respondents with additional information about unorthodox procedures used to pass the bill. The first (procedure #1) added language about the use of budget reconciliation in order to sidestep attempted filibusters in the Senate. The second (procedure #2) added language about the use of a self-executing rule 8 to change the contents of the bill while avoiding debate. Again, respondents were asked to register their approval or disapproval for the policy on a four-point scale. Figure 1 provides an example of the language added for each treatment group for one policy (tax policy). The five policies represented in the study provide some variety to the setting of the experiment. These policies include a distributive policy (transportation), a regulatory policy (energy), a social/cultural policy (abortion), policies that divide the parties along traditional lines (tax, energy, abortion), and those that sometimes are partisan and sometimes bipartisan (transportation and defense). Each policy was written to be more favorable to one side of the aisle or the other. Specifically, the tax and energy policies were written to be viewed favorably 8 Self-executing rules are special rules used by the House majority leadership to make lastminute changes to legislation by a procedural vote, and are an increasingly common part of the lawmaking process in the House. 12

14 by Democrats, and the defense and abortion policies were written to be viewed favorably by Republicans (see the appendix for more details). The transportation policy in the CCES survey was written to be as neutral as possible and serves as a comparison. These partisan differences across issues allows the analyses to assess if the effect procedural maneuvering on the support for new laws is conditioned by partisanship. Partisans inclined to dislike a policy may be more upset by the use of unorthodox or unfair procedures than those inclined to support it. In both the CCES and MTurk surveys, several other measures serve as statistical controls in multivariate analyses. These include a traditional party identification measure that includes partisan leaners, and a seven-point measure of each respondent s self-reported ideology (collapsed into an ideological extremity measure for some analyses). These measures control for respondents political predispositions regarding each policy. Also included is a five-point measure of each respondent s presidential job approval for President Obama. Those approving of the president should be more likely to approve of the policy because the write-up indicates the president is expected to sign the law. Additionally, each respondent s self-reported age, gender (female), annual income, level of education, and race/ethnicity (non-white) was measured. EXPECTATIONS What are the specific expectations for the analyses? The below hypotheses follow from the above theoretical discussion. H1: The procedure treatments will reduce the likelihood a respondent supports the law. One expectation is that each procedure treatment, across all issues, will decrease the likelihood that individuals approve of the new policy. Because Americans process preferences and sense of procedural justice make them generally predisposed to dislike legislating maneuvering and politicking, unorthodox procedures may broadly reduce support for the policies 13

15 described as passed by Congress. H2: The procedure treatments will reduce likelihoods of support among partisans inclined to oppose the policy, but not partisan inclined to support the policy. Another expectation is that procedure effects are conditioned by partisanship. As described above, several of the policies are written to be favorable to one group of partisans and draw opposition from the other. Among these, I expect the procedure treatments will only reduce the likelihood of support among partisans already inclined to oppose the policy. In this way, unorthodox approaches to lawmaking serve to erode support for new laws, but only on one side of the aisle. Specifically, I expect procedure treatments will reduce support among Democrats for the abortion and defense policies, and reduce support among Republicans for the tax and energy policies. With the neutral transportation policy asked about in the CCES survey, I expect that the combined treatment group will result in a similar effect, with Republican support for the policy dropping after they are told about the use of unorthodox procedures over the objections of their party in Congress. H3: The procedure treatments will reduce likelihoods of support among independents. Among independents, procedure treatments may broadly reduce support. Due to their distaste for both major political parties, and generally heightened skepticism about the political system, independents may be more troubled by procedural maneuvering than partisans. H4: The procedure #2 treatment (self-executing rules) will reduce likelihoods of support for the new policies more than the procedure #1 treatment (budget reconciliation). H5: The procedure #1 treatment (budget reconciliation) will increase likelihoods of support among partisan groups inclined to support the policy, and decrease likelihoods of support among partisan groups inclined to oppose the policy. It is also important to consider the nuanced differences between the two procedures treatments included in the MTurk survey experiment. Each describes a different procedure used 14

16 to pass the new law, and each may cause a different reaction among respondents. I expect that the procedure #2 treatment will have a uniform effect across issues. The use of a self-executing rule to pass the bill may sound especially devious, with the description indicating that the procedure made various changes to the bill without any debate such that those opposed to the new provisions were unable to voice their concerns. The very clear anti-deliberative nature of this process should stir up respondent s feelings of procedural justice. The procedure #1 treatment may have different effects. With this treatment, respondents are told that House and Senate leaders used budget reconciliation to push the bill through because this allowed Senate leaders to side-step attempts by numerous senators to filibuster the bill. Public support for filibusters tends to fall along party lines, with partisans whose party would benefit from using the filibuster supportive of the institution and its preservation, and those whose party may be blocked by the filibuster less supportive or more willing to endorse reform (Smith and Park 2013). Consistent with these attitudes, I expect this treatment will only reduce support for a new law among partisans inclined to oppose the policy (and thus would want the policy filibustered). Conversely, I expect it may actually increase support among partisans inclined to approve of the policy as the use of budget reconciliation will allow the policy to pass over objectionable filibusters. RESULTS The results of the survey experiments and subsequent analyses are presented across several tables and figures below. Before discussing the results in detail, a brief summary is in order: In both survey experiments the procedure treatments generally reduced support for the hypothetical policies, providing support for H1, however, the treatment effects vary when analyzed across issues. 15

17 Partisanship conditioned the effects of the procedure treatments as expected in H2, but again the effects vary across issues. In the MTurk survey, procedure #2 had the largest overall affect, providing support for H4. However, with some issues that effect is conditioned by the partisanship of the respondent. In the MTurk survey, the effects of procedure #1 were somewhat conditioned by partisanship, but this varies from issue to issue. This provides some limited support for H5. Independents were not uniformly affected by the treatments. Generally, the treatments reduced independents support for the policies, but for some policies the effect was larger than for others, and sometimes the treatments increased independents support. Figure 2 presents bivariate results from each survey experiment, showing the percent of respondents who indicated they either approved or strongly approved of the hypothetical policy. Across all issues, those exposed to the treatments registered less support than those in the policy only control groups. The very top of the figure shows this for all the issues combined in the MTurk survey: 65% of respondents in the policy only group were supportive of the policy they read about, on average. Among those exposed to the partisan treatment, 60% were in support, a five percentage-point drop. The procedure treatments also reduced support, particular among those exposed to the procedure #2 treatment. Among these respondents, who were told about the use of a self-executing rule to pass the law just 56% were supportive of the policy almost 10 percentage-points fewer than those in the policy only group. The procedure #1 treatment had a more modest effect, which is a result consistent with the expectations above. Across the different issues, the results are generally similar, though overall levels of support and the size of the treatment effects varied. For instance, most respondents were relatively supportive of the tax policy 85% of respondents in the policy only group indicated 16

18 CCES 2014 MTurk 2016 Transportation Energy Defense Abortion Taxes All issues FIGURE 2 Support for New Laws by Treatment Groups (CCES 2014 & MTurk 2016) Procedure #2 Procedure #1 Partisan Policy only Procedure #2 Procedure #1 Partisan Policy only 55.9% 62.6% 60.1% 65.0% 76.2% 77.0% 81.0% 84.9% Procedure #2 Procedure #1 Partisan Policy only Procedure #2 Procedure #1 Partisan Policy only Procedure #2 Procedure #1 Partisan Policy only Combined Procedure Partisan Policy only 34.1% 44.9% 45.2% 47.6% 44.4% 52.9% 41.3% 53.2% 68.8% 75.4% 72.8% 74.0% 62.5% 60.3% 62.1% 69.4% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 17

19 they were supportive but support was 8-9 percentage-points lower among those in the procedure treatment groups. The least overall support was found for the abortion policy (48% in the policy only group), and the effects of the procedure #1 and partisan treatment groups were modest. However, procedure #2 treatment reduced support by over 13 percentage points. The defense policy found moderate overall support (53% among those in the policy only group), but those exposed to the partisan and procedure #2 treatments were 12 and 9 percentage-points less supportive, respectively. The energy policy was broadly liked (74% among the policy only group) and only the procedure #2 treatment had even a modest effect, reducing support by six percentage-points. The transportation policy in the CCES survey was also broadly supported (69% in the policy only group). Respondents in each treatment group were between seven and nine percentage-points more opposed. Tables 1 and 2 show the results of multivariate logistic and ordered logistic regression analyses that look at the impact of the treatments in the CCES and MTurk surveys controlling for other important variables. The dependent variable for the logistic regressions combines approve and strongly approve, and disapprove and strongly disapprove, to create a dichotomous measure of support. The dependent variable for the ordered logistic regression is the four-point measure of support ranging from strongly disapprove (0) to strongly approve (3). This measure can be thought of as the intensity of support/opposition expressed by the respondents. Table 1 shows the results for the CCES survey experiment. Columns 1 and 2 show the results of the logistic regressions, and the results confirm that the treatment effects reduce the likelihood that a respondent will support the described policy. Columns 3 and 4 show the ordered logistic regressions assessing the impact of the treatments on the intensity of support or opposition. When control variables are included (column 4), it is clear that only the combined 18

20 TABLE 1 CCES Survey Experiments (Fall 2014), Transportation Policy Dichotomous Support (Logit) Intensity of Support (Ordered Logit) (1) (2) (3) (4) Partisan group * (0.192) (0.221) (0.168) (0.178) Procedure group ** * * (0.191) (0.218) (0.169) (0.178) Combined group * * (0.194) (0.226) (0.171) (0.180) Democrats * ** (0.237) (0.204) Republicans (0.221) (0.196) Ideology *** *** (0.060) (0.049) Presidential approval *** *** (0.073) (0.063) Age (0.005) (0.004) Female *** ** (0.153) (0.125) Education (0.056) (0.046) Income (0.027) (0.022) Non-white (0.195) (0.155) constant 0.820*** 0.854** (0.143) (0.412) cut 1, constant *** *** (0.146) (0.347) cut 2, constant *** * cut 3, constant *** 2.170*** N 1, , Note: * p<.10 **p<.05 ***p<.01. Standard errors shown in parentheses. treatment significantly reduces the intensity of supprt for the described policy. Thus, the results in Table 1 show that the treatment effects reduce overall support for the policy, but not necessarily how strongly the respondents support or oppose it. Panel A in Figure 3 presents predicted effects from the analyses in Table 1 (specifically those in column 2 of Table 1), showing the substantive size of the treatment effects controlling 19

21 TABLE 2 M-Turk Survey Experiments (Fall 2016) -- All Issues Dichotomous Support (Logit) Intensity of Support (Ordered Logit) (1) (2) (3) (4) (5) (6) (7) (8) Partisan group * * (0.138) (0.138) (0.141) (0.142) (0.119) (0.119) (0.121) (0.121) Either procedure group ** ** * * -- (0.120) (0.122) (0.102) (0.104) Procedure group # (0.138) (0.142) (0.118) (0.120) Procedure group # *** *** ** ** (0.137) (0.140) (0.118) (0.119) Democrats ** 0.334** ** 0.333** (0.162) (0.163) (0.138) (0.138) Republicans *** 0.687*** *** 0.511*** (0.164) (0.164) (0.140) (0.140) Ideological extremity ** ** * (0.078) (0.078) (0.068) (0.068) Presidential approval *** 0.199*** *** 0.174*** (0.048) (0.048) (0.041) (0.041) Age (0.004) (0.004) (0.004) (0.004) Female (0.101) (0.101) (0.087) (0.087) Education (0.042) (0.042) (0.036) (0.036) Income (0.015) (0.015) (0.013) (0.013) Non-white (0.122) (0.122) (0.103) (0.103) Issue group dummies Not shown Not shown Not shown Not shown Not shown Not shown Not shown Not shown constant ** ** (0.123) (0.123) (0.308) (0.308) cut 1, constant *** *** *** *** (0.119) (0.119) (0.268) (0.268) cut 2, constant (0.107) (0.107) (0.264) (0.264) cut 3, constant *** 2.362*** 2.923*** 2.924*** (0.124) (0.124) (0.275) (0.275) N 2,009 2,009 1,961 1,961 2,009 2,009 1,961 1,961 Note: * p<.10 **p<.05 ***p<.01. Standard errors in parentheses. for other important variables. All three treatments reduce the likelihood an individual will support the policy by about 10 percentage-points. Specifically, while respondents in the policy only group had, on average, a 68% likelihood of supporting the policy, those in the combined group had just a 57% likelihood of expressing support, on average. Similarly, those in the procedure group had on average a 58% likelihood of expressing support. Table 2 shows the regression results for the MTurk survey experiments, combining the results across issues. Again, there are both logistic and ordered logistics results shown. 20

22 FIGURE 3 Predicted Effects of Treatment Groups from Analyses in Tables 2 & 3 75% (a) From CCES 2014 (Table 1, column 2) 70% 65% 60% 55% 50% Procedure Partisan Combined Policy only 70% (b) From MTurk 2016 (Table 2, column 3) 65% 60% 55% 50% Either procedure group Partisan Policy only 70% (c) From MTurk 2016 (Table 2, column 4) 65% 60% 55% 50% 45% Procedure #1 Procedure #2 Partisan Policy only 21

23 Additionally, the analyses were conducted first with a variable indicating if respondents were included in either procedure group and then with the two procedure groups indicated separately. Across the eight regressions, the results show that the procedure treatments reduced both the likelihood a respondent will express support for the policy, and the strength of support or opposition the respondent is likely to express. Specifically, the either procedure group coefficient is negative and statistically significant in every analysis. However, the analyses with the procedure treatments separated demonstrates that procedure #2 had the larger and statistically significant effect, with the coefficient for procedure #1 always negative but never statistically significant. Figure 3 shows the substantive effects of the treatments. Respondents in the policy only group had on average a 65% likelihood of supporting the policy they read about. As shown in panel B, the combined procedure treatments reduce that likelihood to 58%, on average. As shown in panel C, the effect is larger for procedure #2 and more modest for procedure #1. Respondents exposed to the procedure #2 treatment had, on average, a 54% likelihood of supporting the policy 11 percentage-points lower than those in the policy only group. By comparison, those exposed to the procedure #1 treatment had on average a 59% likelihood of expressing support. Notably, across all the treatment effects presented in Figure 3, the effects of the procedure treatments are always similar to or larger than the effects of the partisan treatments. Information about Congress s use of unorthodox lawmaking procedures reduces support for new laws as much if not more than information about partisan conflict. The results in Tables 1 and 2 hold controlling for several variables that helped predict respondent s likelihood of supporting the policy, including their partisanship, their ideology, and their approval of President Obama. These variables had large effects on respondents levels of 22

24 the support for the policies. Other demographic controls including age, gender, race, education, and income, did not have statistically significant effects. Combined, the above analyses indicate that when individuals are made aware of the use of unorthodox lawmaking procedures in Congress to pass laws, they are as a result less supportive of those new laws. This provides broad support for H1. In the MTurk experiments, the analyses find that the procedure #2 treatment generally had the largest effect, also providing strong support for H4. The next set of analyses assess the remaining hypotheses, looking at treatment effects across issues and by party. The regression analyses in Table 3 analyze the effect of the treatments broken down by each of the five policy issues, and include interactions between the Democratic and Republican party identification indicators and each of the procedure treatments. These analyses are logistic regressions like those in Table 2 and so assess dichotomous approval. Since interpreting interaction effects from coefficients is difficult, the predicted effects by party and treatment are presented in Figure 4. Generally the issue and party specific effects presented in Figure 4 suggest that the effect of legislative procedures on support for new laws can vary quite a bit from policy to policy, and interactive partisan effects are sometimes predictable and sometimes not. Across issues, there is some general support for H2 that the use of unorthodox procedures primarily reduces support among partisans already inclined to oppose the policy. For instance, with the energy policy, the procedure #2 treatment has only a modest effect on Democrats, who should support the policy on substantive grounds, but both procedure treatments have negative effects on Republicans, reducing their likelihoods of support from 66% on average (in the policy only group) to just 53% and 55% on average among those in the procedure #1 and procedure #2 treatment group, respectively. The analysis of the transportation policy in the 23

25 TABLE 3 All Survey Experiments (CCES 2014 & MTurk 2016), By Issue with Party Interactions Transportation Energy Defense Abortion Taxes Partisan group (both CCES and MTurk) (0.221) (0.336) (0.271) (0.291) (0.378) Procedure group (CCES, only) (0.423) Procedure group # (MTurk, only) (0.685) (0.652) (0.648) (0.627) Procedure group # (MTurk, only) (0.663) (0.620) (0.595) (0.620) Combined group (CCES, only) (0.434) Democrats 0.718** 1.126** ** (0.326) (0.490) (0.439) (0.411) (0.549) Republicans ** 1.413*** (0.305) (0.450) (0.456) (0.431) (0.485) Procedure group * Democrats (CCES, only) (0.492) Procedure group * Republicans (CCES, only) (0.490) Combined group * Democrats (CCES, only) (0.536) Combined group * Republicans (CCES, only) (0.513) Procedure group #1 * Democrats (MTurk, only) (1.028) (0.707) (0.716) (0.848) Procedure group #1 * Republicans (MTurk, only) (0.796) (0.793) (0.794) (0.772) Procedure group #2 * Democrats (MTurk, only) (0.786) (0.685) (0.681) (0.811) Procedure group #2 * Republicans (MTurk, only) (0.778) (0.741) (0.745) (0.750) Ideological extremity *** ** (0.072) (0.192) (0.149) (0.168) (0.211) Presidential approval 0.356*** 0.564*** *** (0.071) (0.103) (0.097) (0.100) (0.117) Age * (0.005) (0.011) (0.009) (0.009) (0.012) Female 0.462*** *** * (0.154) (0.249) (0.197) (0.206) (0.269) Education (0.056) (0.102) (0.080) (0.087) (0.114) Income * ** (0.027) (0.039) (0.031) (0.031) (0.039) Non-white *** * (0.197) (0.299) (0.234) (0.241) (0.348) constant 1.063** *** (0.436) (0.680) (0.625) (0.677) (0.755) N Note: * p<.10 **p<.05 ***p<.01. Coefficients estimated using logistic regression. Standard errors in parentheses. 24

26 FIGURE 4 Predicted Effects of Party Interactions by Issues (CCES 2014 & MTurk 2016) (a) Energy Policy (MTurk) Procedure #2 Procedure #1 Policy only Republicans Independents Democrats 55.4% 53.0% 66.0% 65.4% 77.3% 66.4% 79.1% 93.8% 82.5% 0% 20% 40% 60% 80% 100% (b) Defense Policy (MTurk) Procedure #2 Procedure #1 Policy only Republicans Independents Democrats 61.1% 72.5% 71.0% 30.8% 36.8% 44.6% 38.4% 40.8% 41.4% 0% 20% 40% 60% 80% 100% (c) Abortion Policy (MTurk) Procedure #2 Procedure #1 Policy only Republicans Independents Democrats 54.9% 67.3% 66.2% 22.0% 35.2% 34.1% 28.0% 32.5% 33.1% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 25